diff options
Diffstat (limited to 'doc/documentation/chapters/developer.texi')
-rw-r--r-- | doc/documentation/chapters/developer.texi | 8304 |
1 files changed, 8304 insertions, 0 deletions
diff --git a/doc/documentation/chapters/developer.texi b/doc/documentation/chapters/developer.texi new file mode 100644 index 000000000..996474359 --- /dev/null +++ b/doc/documentation/chapters/developer.texi | |||
@@ -0,0 +1,8304 @@ | |||
1 | @c *********************************************************************** | ||
2 | @node GNUnet Developer Handbook | ||
3 | @chapter GNUnet Developer Handbook | ||
4 | |||
5 | This book is intended to be an introduction for programmers that want to | ||
6 | extend the GNUnet framework. GNUnet is more than a simple peer-to-peer | ||
7 | application. For developers, GNUnet is: | ||
8 | |||
9 | @itemize @bullet | ||
10 | @item Free software under the GNU General Public License, with a community | ||
11 | that believes in the GNU philosophy | ||
12 | @item A set of standards, including coding conventions and | ||
13 | architectural rules | ||
14 | @item A set of layered protocols, both specifying the communication | ||
15 | between peers as well as the communication between components | ||
16 | of a single peer | ||
17 | @item A set of libraries with well-defined APIs suitable for | ||
18 | writing extensions | ||
19 | @end itemize | ||
20 | |||
21 | In particular, the architecture specifies that a peer consists of many | ||
22 | processes communicating via protocols. Processes can be written in almost | ||
23 | any language. C and Java APIs exist for accessing existing services and | ||
24 | for writing extensions. It is possible to write extensions in other | ||
25 | languages by implementing the necessary IPC protocols. | ||
26 | |||
27 | GNUnet can be extended and improved along many possible dimensions, and | ||
28 | anyone interested in Free Software and Freedom-enhancing Networking is | ||
29 | welcome to join the effort. This Developer Handbook attempts to provide | ||
30 | an initial introduction to some of the key design choices and central | ||
31 | components of the system. This part of the GNUNet documentation | ||
32 | is far from complete, and we welcome informed contributions, | ||
33 | be it in the form of new chapters or insightful comments. | ||
34 | |||
35 | @menu | ||
36 | * Developer Introduction:: | ||
37 | * Code overview:: | ||
38 | * System Architecture:: | ||
39 | * Subsystem stability:: | ||
40 | * Naming conventions and coding style guide:: | ||
41 | * Build-system:: | ||
42 | * Developing extensions for GNUnet using the gnunet-ext template:: | ||
43 | * Writing testcases:: | ||
44 | * GNUnet's TESTING library:: | ||
45 | * Performance regression analysis with Gauger:: | ||
46 | * GNUnet's TESTBED Subsystem:: | ||
47 | * libgnunetutil:: | ||
48 | * The Automatic Restart Manager (ARM):: | ||
49 | * GNUnet's TRANSPORT Subsystem:: | ||
50 | * NAT library:: | ||
51 | * Distance-Vector plugin:: | ||
52 | * SMTP plugin:: | ||
53 | * Bluetooth plugin:: | ||
54 | * WLAN plugin:: | ||
55 | * The ATS Subsystem:: | ||
56 | * GNUnet's CORE Subsystem:: | ||
57 | * GNUnet's CADET subsystem:: | ||
58 | * GNUnet's NSE subsystem:: | ||
59 | * GNUnet's HOSTLIST subsystem:: | ||
60 | * GNUnet's IDENTITY subsystem:: | ||
61 | * GNUnet's NAMESTORE Subsystem:: | ||
62 | * GNUnet's PEERINFO subsystem:: | ||
63 | * GNUnet's PEERSTORE subsystem:: | ||
64 | * GNUnet's SET Subsystem:: | ||
65 | * GNUnet's STATISTICS subsystem:: | ||
66 | * GNUnet's Distributed Hash Table (DHT):: | ||
67 | * The GNU Name System (GNS):: | ||
68 | * The GNS Namecache:: | ||
69 | * The REVOCATION Subsystem:: | ||
70 | * GNUnet's File-sharing (FS) Subsystem:: | ||
71 | * GNUnet's REGEX Subsystem:: | ||
72 | @end menu | ||
73 | |||
74 | @node Developer Introduction | ||
75 | @section Developer Introduction | ||
76 | |||
77 | This Developer Handbook is intended as first introduction to GNUnet for | ||
78 | new developers that want to extend the GNUnet framework. After the | ||
79 | introduction, each of the GNUnet subsystems (directories in the | ||
80 | @file{src/} tree) is (supposed to be) covered in its own chapter. In | ||
81 | addition to this documentation, GNUnet developers should be aware of the | ||
82 | services available on the GNUnet server to them. | ||
83 | |||
84 | New developers can have a look a the GNUnet tutorials for C and java | ||
85 | available in the @file{src/} directory of the repository or under the | ||
86 | following links: | ||
87 | |||
88 | @c ** FIXME: Link to files in source, not online. | ||
89 | @c ** FIXME: Where is the Java tutorial? | ||
90 | @itemize @bullet | ||
91 | @item @uref{https://gnunet.org/git/gnunet.git/plain/doc/gnunet-c-tutorial.pdf, GNUnet C tutorial} | ||
92 | @item GNUnet Java tutorial | ||
93 | @end itemize | ||
94 | |||
95 | In addition to this book, the GNUnet server contains various resources for | ||
96 | GNUnet developers. They are all conveniently reachable via the "Developer" | ||
97 | entry in the navigation menu. Some additional tools (such as static | ||
98 | analysis reports) require a special developer access to perform certain | ||
99 | operations. If you feel you need access, you should contact | ||
100 | @uref{http://grothoff.org/christian/, Christian Grothoff}, | ||
101 | GNUnet's maintainer. | ||
102 | |||
103 | The public subsystems on the GNUnet server that help developers are: | ||
104 | |||
105 | @itemize @bullet | ||
106 | |||
107 | @item The Version control system (git) keeps our code and enables | ||
108 | distributed development. | ||
109 | Only developers with write access can commit code, everyone else is | ||
110 | encouraged to submit patches to the | ||
111 | @uref{https://lists.gnu.org/mailman/listinfo/gnunet-developers, GNUnet-developers mailinglist} | ||
112 | . | ||
113 | |||
114 | @item The GNUnet bugtracking system (Mantis) is used to track | ||
115 | feature requests, open bug reports and their resolutions. | ||
116 | Anyone can report bugs, only developers can claim to have fixed them. | ||
117 | |||
118 | @item A site installation of the CI system "Buildbot" is used to check | ||
119 | GNUnet builds automatically on a range of platforms. | ||
120 | Builds are triggered automatically after 30 minutes of no changes to Git. | ||
121 | |||
122 | @item The current quality of our automated test suite is assessed using | ||
123 | Code coverage analysis. This analysis is run daily; however the webpage | ||
124 | is only updated if all automated tests pass at that time. Testcases that | ||
125 | improve our code coverage are always welcome. | ||
126 | |||
127 | @item We try to automatically find bugs using a static analysis scan. | ||
128 | This scan is run daily; however the webpage is only updated if all | ||
129 | automated tests pass at the time. Note that not everything that is | ||
130 | flagged by the analysis is a bug, sometimes even good code can be marked | ||
131 | as possibly problematic. Nevertheless, developers are encouraged to at | ||
132 | least be aware of all issues in their code that are listed. | ||
133 | |||
134 | @item We use Gauger for automatic performance regression visualization. | ||
135 | Details on how to use Gauger are here. | ||
136 | |||
137 | @item We use @uref{http://junit.org/, junit} to automatically test | ||
138 | @command{gnunet-java}. | ||
139 | Automatically generated, current reports on the test suite are here. | ||
140 | |||
141 | @item We use Cobertura to generate test coverage reports for gnunet-java. | ||
142 | Current reports on test coverage are here. | ||
143 | |||
144 | @end itemize | ||
145 | |||
146 | |||
147 | |||
148 | @c *********************************************************************** | ||
149 | @menu | ||
150 | * Project overview:: | ||
151 | @end menu | ||
152 | |||
153 | @node Project overview | ||
154 | @subsection Project overview | ||
155 | |||
156 | The GNUnet project consists at this point of several sub-projects. This | ||
157 | section is supposed to give an initial overview about the various | ||
158 | sub-projects. Note that this description also lists projects that are far | ||
159 | from complete, including even those that have literally not a single line | ||
160 | of code in them yet. | ||
161 | |||
162 | GNUnet sub-projects in order of likely relevance are currently: | ||
163 | |||
164 | @table @asis | ||
165 | |||
166 | @item gnunet | ||
167 | Core of the P2P framework, including file-sharing, VPN and | ||
168 | chat applications; this is what the developer handbook covers mostly | ||
169 | @item gnunet-gtk Gtk+-based user interfaces, including gnunet-fs-gtk | ||
170 | (file-sharing), gnunet-statistics-gtk (statistics over time), | ||
171 | gnunet-peerinfo-gtk (information about current connections and known | ||
172 | peers), gnunet-chat-gtk (chat GUI) and gnunet-setup (setup tool for | ||
173 | "everything") | ||
174 | @item gnunet-fuse | ||
175 | Mounting directories shared via GNUnet's file-sharing | ||
176 | on Linux | ||
177 | @item gnunet-update | ||
178 | Installation and update tool | ||
179 | @item gnunet-ext | ||
180 | Template for starting 'external' GNUnet projects | ||
181 | @item gnunet-java | ||
182 | Java APIs for writing GNUnet services and applications | ||
183 | @c ** FIXME: Point to new website repository once we have it: | ||
184 | @c ** @item svn/gnunet-www/ Code and media helping drive the GNUnet | ||
185 | @c website | ||
186 | @item eclectic | ||
187 | Code to run GNUnet nodes on testbeds for research, development, | ||
188 | testing and evaluation | ||
189 | @c ** FIXME: Solve the status and location of gnunet-qt | ||
190 | @item gnunet-qt | ||
191 | Qt-based GNUnet GUI (dead?) | ||
192 | @item gnunet-cocoa | ||
193 | cocoa-based GNUnet GUI (dead?) | ||
194 | |||
195 | @end table | ||
196 | |||
197 | We are also working on various supporting libraries and tools: | ||
198 | @c ** FIXME: What about gauger, and what about libmwmodem? | ||
199 | |||
200 | @table @asis | ||
201 | @item libextractor | ||
202 | GNU libextractor (meta data extraction) | ||
203 | @item libmicrohttpd | ||
204 | GNU libmicrohttpd (embedded HTTP(S) server library) | ||
205 | @item gauger | ||
206 | Tool for performance regression analysis | ||
207 | @item monkey | ||
208 | Tool for automated debugging of distributed systems | ||
209 | @item libmwmodem | ||
210 | Library for accessing satellite connection quality | ||
211 | reports | ||
212 | @item libgnurl | ||
213 | gnURL (feature restricted variant of cURL/libcurl) | ||
214 | @end table | ||
215 | |||
216 | Finally, there are various external projects (see links for a list of | ||
217 | those that have a public website) which build on top of the GNUnet | ||
218 | framework. | ||
219 | |||
220 | @c *********************************************************************** | ||
221 | @node Code overview | ||
222 | @section Code overview | ||
223 | |||
224 | This section gives a brief overview of the GNUnet source code. | ||
225 | Specifically, we sketch the function of each of the subdirectories in | ||
226 | the @file{gnunet/src/} directory. The order given is roughly bottom-up | ||
227 | (in terms of the layers of the system). | ||
228 | |||
229 | @table @asis | ||
230 | @item @file{util/} --- libgnunetutil | ||
231 | Library with general utility functions, all | ||
232 | GNUnet binaries link against this library. Anything from memory | ||
233 | allocation and data structures to cryptography and inter-process | ||
234 | communication. The goal is to provide an OS-independent interface and | ||
235 | more 'secure' or convenient implementations of commonly used primitives. | ||
236 | The API is spread over more than a dozen headers, developers should study | ||
237 | those closely to avoid duplicating existing functions. | ||
238 | @item @file{hello/} --- libgnunethello | ||
239 | HELLO messages are used to | ||
240 | describe under which addresses a peer can be reached (for example, | ||
241 | protocol, IP, port). This library manages parsing and generating of HELLO | ||
242 | messages. | ||
243 | @item @file{block/} --- libgnunetblock | ||
244 | The DHT and other components of GNUnet | ||
245 | store information in units called 'blocks'. Each block has a type and the | ||
246 | type defines a particular format and how that binary format is to be | ||
247 | linked to a hash code (the key for the DHT and for databases). The block | ||
248 | library is a wapper around block plugins which provide the necessary | ||
249 | functions for each block type. | ||
250 | @item @file{statistics/} | ||
251 | The statistics service enables associating | ||
252 | values (of type uint64_t) with a componenet name and a string. The main | ||
253 | uses is debugging (counting events), performance tracking and user | ||
254 | entertainment (what did my peer do today?). | ||
255 | @item @file{arm/} --- Automatic Restart Manager (ARM) | ||
256 | The automatic-restart-manager (ARM) service | ||
257 | is the GNUnet master service. Its role is to start gnunet-services, to | ||
258 | re-start them when they crashed and finally to shut down the system when | ||
259 | requested. | ||
260 | @item @file{peerinfo/} | ||
261 | The peerinfo service keeps track of which peers are known | ||
262 | to the local peer and also tracks the validated addresses for each peer | ||
263 | (in the form of a HELLO message) for each of those peers. The peer is not | ||
264 | necessarily connected to all peers known to the peerinfo service. | ||
265 | Peerinfo provides persistent storage for peer identities --- peers are | ||
266 | not forgotten just because of a system restart. | ||
267 | @item @file{datacache/} --- libgnunetdatacache | ||
268 | The datacache library provides (temporary) block storage for the DHT. | ||
269 | Existing plugins can store blocks in Sqlite, Postgres or MySQL databases. | ||
270 | All data stored in the cache is lost when the peer is stopped or | ||
271 | restarted (datacache uses temporary tables). | ||
272 | @item @file{datastore/} | ||
273 | The datastore service stores file-sharing blocks in | ||
274 | databases for extended periods of time. In contrast to the datacache, data | ||
275 | is not lost when peers restart. However, quota restrictions may still | ||
276 | cause old, expired or low-priority data to be eventually discarded. | ||
277 | Existing plugins can store blocks in Sqlite, Postgres or MySQL databases. | ||
278 | @item @file{template/} | ||
279 | Template for writing a new service. Does nothing. | ||
280 | @item @file{ats/} --- Automatic Transport Selection | ||
281 | The automatic transport | ||
282 | selection (ATS) service is responsible for deciding which address (i.e. | ||
283 | which transport plugin) should be used for communication with other peers, | ||
284 | and at what bandwidth. | ||
285 | @item @file{nat/} --- libgnunetnat | ||
286 | Library that provides basic functions for NAT traversal. | ||
287 | The library supports NAT traversal with | ||
288 | manual hole-punching by the user, UPnP and ICMP-based autonomous NAT | ||
289 | traversal. The library also includes an API for testing if the current | ||
290 | configuration works and the @code{gnunet-nat-server} which provides an | ||
291 | external service to test the local configuration. | ||
292 | @item @file{fragmentation/} --- libgnunetfragmentation | ||
293 | Some transports (UDP and WLAN, mostly) have restrictions on the maximum | ||
294 | transfer unit (MTU) for packets. The fragmentation library can be used to | ||
295 | break larger packets into chunks of at most 1k and transmit the resulting | ||
296 | fragments reliabily (with acknowledgement, retransmission, timeouts, | ||
297 | etc.). | ||
298 | @item @file{transport/} | ||
299 | The transport service is responsible for managing the | ||
300 | basic P2P communication. It uses plugins to support P2P communication | ||
301 | over TCP, UDP, HTTP, HTTPS and other protocols.The transport service | ||
302 | validates peer addresses, enforces bandwidth restrictions, limits the | ||
303 | total number of connections and enforces connectivity restrictions (i.e. | ||
304 | friends-only). | ||
305 | @item @file{peerinfo-tool/} | ||
306 | This directory contains the gnunet-peerinfo binary which can be used to | ||
307 | inspect the peers and HELLOs known to the peerinfo service. | ||
308 | @item @file{core/} | ||
309 | The core service is responsible for establishing encrypted, authenticated | ||
310 | connections with other peers, encrypting and decrypting messages and | ||
311 | forwarding messages to higher-level services that are interested in them. | ||
312 | @item @file{testing/} --- libgnunettesting | ||
313 | The testing library allows starting (and stopping) peers | ||
314 | for writing testcases. | ||
315 | It also supports automatic generation of configurations for peers | ||
316 | ensuring that the ports and paths are disjoint. libgnunettesting is also | ||
317 | the foundation for the testbed service | ||
318 | @item @file{testbed/} | ||
319 | The testbed service is used for creating small or large scale deployments | ||
320 | of GNUnet peers for evaluation of protocols. | ||
321 | It facilitates peer depolyments on multiple | ||
322 | hosts (for example, in a cluster) and establishing varous network | ||
323 | topologies (both underlay and overlay). | ||
324 | @item @file{nse/} --- Network Size Estimation | ||
325 | The network size estimation (NSE) service | ||
326 | implements a protocol for (securely) estimating the current size of the | ||
327 | P2P network. | ||
328 | @item @file{dht/} --- distributed hash table | ||
329 | The distributed hash table (DHT) service provides a | ||
330 | distributed implementation of a hash table to store blocks under hash | ||
331 | keys in the P2P network. | ||
332 | @item @file{hostlist/} | ||
333 | The hostlist service allows learning about | ||
334 | other peers in the network by downloading HELLO messages from an HTTP | ||
335 | server, can be configured to run such an HTTP server and also implements | ||
336 | a P2P protocol to advertise and automatically learn about other peers | ||
337 | that offer a public hostlist server. | ||
338 | @item @file{topology/} | ||
339 | The topology service is responsible for | ||
340 | maintaining the mesh topology. It tries to maintain connections to friends | ||
341 | (depending on the configuration) and also tries to ensure that the peer | ||
342 | has a decent number of active connections at all times. If necessary, new | ||
343 | connections are added. All peers should run the topology service, | ||
344 | otherwise they may end up not being connected to any other peer (unless | ||
345 | some other service ensures that core establishes the required | ||
346 | connections). The topology service also tells the transport service which | ||
347 | connections are permitted (for friend-to-friend networking) | ||
348 | @item @file{fs/} --- file-sharing | ||
349 | The file-sharing (FS) service implements GNUnet's | ||
350 | file-sharing application. Both anonymous file-sharing (using gap) and | ||
351 | non-anonymous file-sharing (using dht) are supported. | ||
352 | @item @file{cadet/} | ||
353 | The CADET service provides a general-purpose routing abstraction to create | ||
354 | end-to-end encrypted tunnels in mesh networks. We wrote a paper | ||
355 | documenting key aspects of the design. | ||
356 | @item @file{tun/} --- libgnunettun | ||
357 | Library for building IPv4, IPv6 packets and creating | ||
358 | checksums for UDP, TCP and ICMP packets. The header | ||
359 | defines C structs for common Internet packet formats and in particular | ||
360 | structs for interacting with TUN (virtual network) interfaces. | ||
361 | @item @file{mysql/} --- libgnunetmysql | ||
362 | Library for creating and executing prepared MySQL | ||
363 | statements and to manage the connection to the MySQL database. | ||
364 | Essentially a lightweight wrapper for the interaction between GNUnet | ||
365 | components and libmysqlclient. | ||
366 | @item @file{dns/} | ||
367 | Service that allows intercepting and modifying DNS requests of | ||
368 | the local machine. Currently used for IPv4-IPv6 protocol translation | ||
369 | (DNS-ALG) as implemented by "pt/" and for the GNUnet naming system. The | ||
370 | service can also be configured to offer an exit service for DNS traffic. | ||
371 | @item @file{vpn/} | ||
372 | The virtual public network (VPN) service provides a virtual | ||
373 | tunnel interface (VTUN) for IP routing over GNUnet. | ||
374 | Needs some other peers to run an "exit" service to work. | ||
375 | Can be activated using the "gnunet-vpn" tool or integrated with DNS using | ||
376 | the "pt" daemon. | ||
377 | @item @file{exit/} | ||
378 | Daemon to allow traffic from the VPN to exit this | ||
379 | peer to the Internet or to specific IP-based services of the local peer. | ||
380 | Currently, an exit service can only be restricted to IPv4 or IPv6, not to | ||
381 | specific ports and or IP address ranges. If this is not acceptable, | ||
382 | additional firewall rules must be added manually. exit currently only | ||
383 | works for normal UDP, TCP and ICMP traffic; DNS queries need to leave the | ||
384 | system via a DNS service. | ||
385 | @item @file{pt/} | ||
386 | protocol translation daemon. This daemon enables 4-to-6, | ||
387 | 6-to-4, 4-over-6 or 6-over-4 transitions for the local system. It | ||
388 | essentially uses "DNS" to intercept DNS replies and then maps results to | ||
389 | those offered by the VPN, which then sends them using mesh to some daemon | ||
390 | offering an appropriate exit service. | ||
391 | @item @file{identity/} | ||
392 | Management of egos (alter egos) of a user; identities are | ||
393 | essentially named ECC private keys and used for zones in the GNU name | ||
394 | system and for namespaces in file-sharing, but might find other uses later | ||
395 | @item @file{revocation/} | ||
396 | Key revocation service, can be used to revoke the | ||
397 | private key of an identity if it has been compromised | ||
398 | @item @file{namecache/} | ||
399 | Cache for resolution results for the GNU name system; | ||
400 | data is encrypted and can be shared among users, | ||
401 | loss of the data should ideally only result in a | ||
402 | performance degradation (persistence not required) | ||
403 | @item @file{namestore/} | ||
404 | Database for the GNU name system with per-user private information, | ||
405 | persistence required | ||
406 | @item @file{gns/} | ||
407 | GNU name system, a GNU approach to DNS and PKI. | ||
408 | @item @file{dv/} | ||
409 | A plugin for distance-vector (DV)-based routing. | ||
410 | DV consists of a service and a transport plugin to provide peers | ||
411 | with the illusion of a direct P2P connection for connections | ||
412 | that use multiple (typically up to 3) hops in the actual underlay network. | ||
413 | @item @file{regex/} | ||
414 | Service for the (distributed) evaluation of regular expressions. | ||
415 | @item @file{scalarproduct/} | ||
416 | The scalar product service offers an API to perform a secure multiparty | ||
417 | computation which calculates a scalar product between two peers | ||
418 | without exposing the private input vectors of the peers to each other. | ||
419 | @item @file{consensus/} | ||
420 | The consensus service will allow a set of peers to agree | ||
421 | on a set of values via a distributed set union computation. | ||
422 | @item @file{rest/} | ||
423 | The rest API allows access to GNUnet services using RESTful interaction. | ||
424 | The services provide plugins that can exposed by the rest server. | ||
425 | @item @file{experimentation/} | ||
426 | The experimentation daemon coordinates distributed | ||
427 | experimentation to evaluate transport and ATS properties. | ||
428 | @end table | ||
429 | |||
430 | @c *********************************************************************** | ||
431 | @node System Architecture | ||
432 | @section System Architecture | ||
433 | |||
434 | GNUnet developers like LEGOs. The blocks are indestructible, can be | ||
435 | stacked together to construct complex buildings and it is generally easy | ||
436 | to swap one block for a different one that has the same shape. GNUnet's | ||
437 | architecture is based on LEGOs: | ||
438 | |||
439 | @c images here | ||
440 | |||
441 | This chapter documents the GNUnet LEGO system, also known as GNUnet's | ||
442 | system architecture. | ||
443 | |||
444 | The most common GNUnet component is a service. Services offer an API (or | ||
445 | several, depending on what you count as "an API") which is implemented as | ||
446 | a library. The library communicates with the main process of the service | ||
447 | using a service-specific network protocol. The main process of the service | ||
448 | typically doesn't fully provide everything that is needed --- it has holes | ||
449 | to be filled by APIs to other services. | ||
450 | |||
451 | A special kind of component in GNUnet are user interfaces and daemons. | ||
452 | Like services, they have holes to be filled by APIs of other services. | ||
453 | Unlike services, daemons do not implement their own network protocol and | ||
454 | they have no API: | ||
455 | |||
456 | The GNUnet system provides a range of services, daemons and user | ||
457 | interfaces, which are then combined into a layered GNUnet instance (also | ||
458 | known as a peer). | ||
459 | |||
460 | Note that while it is generally possible to swap one service for another | ||
461 | compatible service, there is often only one implementation. However, | ||
462 | during development we often have a "new" version of a service in parallel | ||
463 | with an "old" version. While the "new" version is not working, developers | ||
464 | working on other parts of the service can continue their development by | ||
465 | simply using the "old" service. Alternative design ideas can also be | ||
466 | easily investigated by swapping out individual components. This is | ||
467 | typically achieved by simply changing the name of the "BINARY" in the | ||
468 | respective configuration section. | ||
469 | |||
470 | Key properties of GNUnet services are that they must be separate | ||
471 | processes and that they must protect themselves by applying tight error | ||
472 | checking against the network protocol they implement (thereby achieving a | ||
473 | certain degree of robustness). | ||
474 | |||
475 | On the other hand, the APIs are implemented to tolerate failures of the | ||
476 | service, isolating their host process from errors by the service. If the | ||
477 | service process crashes, other services and daemons around it should not | ||
478 | also fail, but instead wait for the service process to be restarted by | ||
479 | ARM. | ||
480 | |||
481 | |||
482 | @c *********************************************************************** | ||
483 | @node Subsystem stability | ||
484 | @section Subsystem stability | ||
485 | |||
486 | This section documents the current stability of the various GNUnet | ||
487 | subsystems. Stability here describes the expected degree of compatibility | ||
488 | with future versions of GNUnet. For each subsystem we distinguish between | ||
489 | compatibility on the P2P network level (communication protocol between | ||
490 | peers), the IPC level (communication between the service and the service | ||
491 | library) and the API level (stability of the API). P2P compatibility is | ||
492 | relevant in terms of which applications are likely going to be able to | ||
493 | communicate with future versions of the network. IPC communication is | ||
494 | relevant for the implementation of language bindings that re-implement the | ||
495 | IPC messages. Finally, API compatibility is relevant to developers that | ||
496 | hope to be able to avoid changes to applications build on top of the APIs | ||
497 | of the framework. | ||
498 | |||
499 | The following table summarizes our current view of the stability of the | ||
500 | respective protocols or APIs: | ||
501 | |||
502 | @multitable @columnfractions .20 .20 .20 .20 | ||
503 | @headitem Subsystem @tab P2P @tab IPC @tab C API | ||
504 | @item util @tab n/a @tab n/a @tab stable | ||
505 | @item arm @tab n/a @tab stable @tab stable | ||
506 | @item ats @tab n/a @tab unstable @tab testing | ||
507 | @item block @tab n/a @tab n/a @tab stable | ||
508 | @item cadet @tab testing @tab testing @tab testing | ||
509 | @item consensus @tab experimental @tab experimental @tab experimental | ||
510 | @item core @tab stable @tab stable @tab stable | ||
511 | @item datacache @tab n/a @tab n/a @tab stable | ||
512 | @item datastore @tab n/a @tab stable @tab stable | ||
513 | @item dht @tab stable @tab stable @tab stable | ||
514 | @item dns @tab stable @tab stable @tab stable | ||
515 | @item dv @tab testing @tab testing @tab n/a | ||
516 | @item exit @tab testing @tab n/a @tab n/a | ||
517 | @item fragmentation @tab stable @tab n/a @tab stable | ||
518 | @item fs @tab stable @tab stable @tab stable | ||
519 | @item gns @tab stable @tab stable @tab stable | ||
520 | @item hello @tab n/a @tab n/a @tab testing | ||
521 | @item hostlist @tab stable @tab stable @tab n/a | ||
522 | @item identity @tab stable @tab stable @tab n/a | ||
523 | @item multicast @tab experimental @tab experimental @tab experimental | ||
524 | @item mysql @tab stable @tab n/a @tab stable | ||
525 | @item namestore @tab n/a @tab stable @tab stable | ||
526 | @item nat @tab n/a @tab n/a @tab stable | ||
527 | @item nse @tab stable @tab stable @tab stable | ||
528 | @item peerinfo @tab n/a @tab stable @tab stable | ||
529 | @item psyc @tab experimental @tab experimental @tab experimental | ||
530 | @item pt @tab n/a @tab n/a @tab n/a | ||
531 | @item regex @tab stable @tab stable @tab stable | ||
532 | @item revocation @tab stable @tab stable @tab stable | ||
533 | @item social @tab experimental @tab experimental @tab experimental | ||
534 | @item statistics @tab n/a @tab stable @tab stable | ||
535 | @item testbed @tab n/a @tab testing @tab testing | ||
536 | @item testing @tab n/a @tab n/a @tab testing | ||
537 | @item topology @tab n/a @tab n/a @tab n/a | ||
538 | @item transport @tab stable @tab stable @tab stable | ||
539 | @item tun @tab n/a @tab n/a @tab stable | ||
540 | @item vpn @tab testing @tab n/a @tab n/a | ||
541 | @end multitable | ||
542 | |||
543 | Here is a rough explanation of the values: | ||
544 | |||
545 | @table @samp | ||
546 | @item stable | ||
547 | No incompatible changes are planned at this time; for IPC/APIs, if | ||
548 | there are incompatible changes, they will be minor and might only require | ||
549 | minimal changes to existing code; for P2P, changes will be avoided if at | ||
550 | all possible for the 0.10.x-series | ||
551 | |||
552 | @item testing | ||
553 | No incompatible changes are | ||
554 | planned at this time, but the code is still known to be in flux; so while | ||
555 | we have no concrete plans, our expectation is that there will still be | ||
556 | minor modifications; for P2P, changes will likely be extensions that | ||
557 | should not break existing code | ||
558 | |||
559 | @item unstable | ||
560 | Changes are planned and will happen; however, they | ||
561 | will not be totally radical and the result should still resemble what is | ||
562 | there now; nevertheless, anticipated changes will break protocol/API | ||
563 | compatibility | ||
564 | |||
565 | @item experimental | ||
566 | Changes are planned and the result may look nothing like | ||
567 | what the API/protocol looks like today | ||
568 | |||
569 | @item unknown | ||
570 | Someone should think about where this subsystem headed | ||
571 | |||
572 | @item n/a | ||
573 | This subsystem does not have an API/IPC-protocol/P2P-protocol | ||
574 | @end table | ||
575 | |||
576 | @c *********************************************************************** | ||
577 | @node Naming conventions and coding style guide | ||
578 | @section Naming conventions and coding style guide | ||
579 | |||
580 | Here you can find some rules to help you write code for GNUnet. | ||
581 | |||
582 | @c *********************************************************************** | ||
583 | @menu | ||
584 | * Naming conventions:: | ||
585 | * Coding style:: | ||
586 | @end menu | ||
587 | |||
588 | @node Naming conventions | ||
589 | @subsection Naming conventions | ||
590 | |||
591 | |||
592 | @c *********************************************************************** | ||
593 | @menu | ||
594 | * include files:: | ||
595 | * binaries:: | ||
596 | * logging:: | ||
597 | * configuration:: | ||
598 | * exported symbols:: | ||
599 | * private (library-internal) symbols (including structs and macros):: | ||
600 | * testcases:: | ||
601 | * performance tests:: | ||
602 | * src/ directories:: | ||
603 | @end menu | ||
604 | |||
605 | @node include files | ||
606 | @subsubsection include files | ||
607 | |||
608 | @itemize @bullet | ||
609 | @item _lib: library without need for a process | ||
610 | @item _service: library that needs a service process | ||
611 | @item _plugin: plugin definition | ||
612 | @item _protocol: structs used in network protocol | ||
613 | @item exceptions: | ||
614 | @itemize @bullet | ||
615 | @item gnunet_config.h --- generated | ||
616 | @item platform.h --- first included | ||
617 | @item plibc.h --- external library | ||
618 | @item gnunet_common.h --- fundamental routines | ||
619 | @item gnunet_directories.h --- generated | ||
620 | @item gettext.h --- external library | ||
621 | @end itemize | ||
622 | @end itemize | ||
623 | |||
624 | @c *********************************************************************** | ||
625 | @node binaries | ||
626 | @subsubsection binaries | ||
627 | |||
628 | @itemize @bullet | ||
629 | @item gnunet-service-xxx: service process (has listen socket) | ||
630 | @item gnunet-daemon-xxx: daemon process (no listen socket) | ||
631 | @item gnunet-helper-xxx[-yyy]: SUID helper for module xxx | ||
632 | @item gnunet-yyy: command-line tool for end-users | ||
633 | @item libgnunet_plugin_xxx_yyy.so: plugin for API xxx | ||
634 | @item libgnunetxxx.so: library for API xxx | ||
635 | @end itemize | ||
636 | |||
637 | @c *********************************************************************** | ||
638 | @node logging | ||
639 | @subsubsection logging | ||
640 | |||
641 | @itemize @bullet | ||
642 | @item services and daemons use their directory name in | ||
643 | @code{GNUNET_log_setup} (i.e. 'core') and log using | ||
644 | plain 'GNUNET_log'. | ||
645 | @item command-line tools use their full name in | ||
646 | @code{GNUNET_log_setup} (i.e. 'gnunet-publish') and log using | ||
647 | plain 'GNUNET_log'. | ||
648 | @item service access libraries log using | ||
649 | '@code{GNUNET_log_from}' and use '@code{DIRNAME-api}' for the | ||
650 | component (i.e. 'core-api') | ||
651 | @item pure libraries (without associated service) use | ||
652 | '@code{GNUNET_log_from}' with the component set to their | ||
653 | library name (without lib or '@file{.so}'), | ||
654 | which should also be their directory name (i.e. '@file{nat}') | ||
655 | @item plugins should use '@code{GNUNET_log_from}' | ||
656 | with the directory name and the plugin name combined to produce | ||
657 | the component name (i.e. 'transport-tcp'). | ||
658 | @item logging should be unified per-file by defining a | ||
659 | @code{LOG} macro with the appropriate arguments, | ||
660 | along these lines: | ||
661 | |||
662 | @example | ||
663 | #define LOG(kind,...) | ||
664 | GNUNET_log_from (kind, "example-api",__VA_ARGS__) | ||
665 | @end example | ||
666 | |||
667 | @end itemize | ||
668 | |||
669 | @c *********************************************************************** | ||
670 | @node configuration | ||
671 | @subsubsection configuration | ||
672 | |||
673 | @itemize @bullet | ||
674 | @item paths (that are substituted in all filenames) are in PATHS | ||
675 | (have as few as possible) | ||
676 | @item all options for a particular module (@file{src/MODULE}) | ||
677 | are under @code{[MODULE]} | ||
678 | @item options for a plugin of a module | ||
679 | are under @code{[MODULE-PLUGINNAME]} | ||
680 | @end itemize | ||
681 | |||
682 | @c *********************************************************************** | ||
683 | @node exported symbols | ||
684 | @subsubsection exported symbols | ||
685 | |||
686 | @itemize @bullet | ||
687 | @item must start with "@code{GNUNET_modulename_}" and be defined in | ||
688 | "@file{modulename.c}" | ||
689 | @item exceptions: those defined in @file{gnunet_common.h} | ||
690 | @end itemize | ||
691 | |||
692 | @c *********************************************************************** | ||
693 | @node private (library-internal) symbols (including structs and macros) | ||
694 | @subsubsection private (library-internal) symbols (including structs and macros) | ||
695 | |||
696 | @itemize @bullet | ||
697 | @item must NOT start with any prefix | ||
698 | @item must not be exported in a way that linkers could use them or@ other | ||
699 | libraries might see them via headers; they must be either | ||
700 | declared/defined in C source files or in headers that are in the | ||
701 | respective directory under @file{src/modulename/} and NEVER be declared | ||
702 | in @file{src/include/}. | ||
703 | @end itemize | ||
704 | |||
705 | @node testcases | ||
706 | @subsubsection testcases | ||
707 | |||
708 | @itemize @bullet | ||
709 | @item must be called "@file{test_module-under-test_case-description.c}" | ||
710 | @item "case-description" maybe omitted if there is only one test | ||
711 | @end itemize | ||
712 | |||
713 | @c *********************************************************************** | ||
714 | @node performance tests | ||
715 | @subsubsection performance tests | ||
716 | |||
717 | @itemize @bullet | ||
718 | @item must be called "@file{perf_module-under-test_case-description.c}" | ||
719 | @item "case-description" maybe omitted if there is only one performance | ||
720 | test | ||
721 | @item Must only be run if @code{HAVE_BENCHMARKS} is satisfied | ||
722 | @end itemize | ||
723 | |||
724 | @c *********************************************************************** | ||
725 | @node src/ directories | ||
726 | @subsubsection src/ directories | ||
727 | |||
728 | @itemize @bullet | ||
729 | @item gnunet-NAME: end-user applications (i.e., gnunet-search, gnunet-arm) | ||
730 | @item gnunet-service-NAME: service processes with accessor library (i.e., | ||
731 | gnunet-service-arm) | ||
732 | @item libgnunetNAME: accessor library (_service.h-header) or standalone | ||
733 | library (_lib.h-header) | ||
734 | @item gnunet-daemon-NAME: daemon process without accessor library (i.e., | ||
735 | gnunet-daemon-hostlist) and no GNUnet management port | ||
736 | @item libgnunet_plugin_DIR_NAME: loadable plugins (i.e., | ||
737 | libgnunet_plugin_transport_tcp) | ||
738 | @end itemize | ||
739 | |||
740 | @cindex Coding style | ||
741 | @node Coding style | ||
742 | @subsection Coding style | ||
743 | |||
744 | @itemize @bullet | ||
745 | @item We follow the GNU Coding Standards (@pxref{Top, The GNU Coding Standards,, standards, The GNU Coding Standards}); | ||
746 | @item Indentation is done with spaces, two per level, no tabs; | ||
747 | @item C99 struct initialization is fine; | ||
748 | @item declare only one variable per line, for example: | ||
749 | |||
750 | @noindent | ||
751 | instead of | ||
752 | |||
753 | @example | ||
754 | int i,j; | ||
755 | @end example | ||
756 | |||
757 | @noindent | ||
758 | write: | ||
759 | |||
760 | @example | ||
761 | int i; | ||
762 | int j; | ||
763 | @end example | ||
764 | |||
765 | @c TODO: include actual example from a file in source | ||
766 | |||
767 | @noindent | ||
768 | This helps keep diffs small and forces developers to think precisely about | ||
769 | the type of every variable. | ||
770 | Note that @code{char *} is different from @code{const char*} and | ||
771 | @code{int} is different from @code{unsigned int} or @code{uint32_t}. | ||
772 | Each variable type should be chosen with care. | ||
773 | |||
774 | @item While @code{goto} should generally be avoided, having a | ||
775 | @code{goto} to the end of a function to a block of clean up | ||
776 | statements (free, close, etc.) can be acceptable. | ||
777 | |||
778 | @item Conditions should be written with constants on the left (to avoid | ||
779 | accidental assignment) and with the 'true' target being either the | ||
780 | 'error' case or the significantly simpler continuation. For example: | ||
781 | |||
782 | @example | ||
783 | if (0 != stat ("filename," &sbuf)) @{ | ||
784 | error(); | ||
785 | @} | ||
786 | else @{ | ||
787 | /* handle normal case here */ | ||
788 | @} | ||
789 | @end example | ||
790 | |||
791 | @noindent | ||
792 | instead of | ||
793 | |||
794 | @example | ||
795 | if (stat ("filename," &sbuf) == 0) @{ | ||
796 | /* handle normal case here */ | ||
797 | @} else @{ | ||
798 | error(); | ||
799 | @} | ||
800 | @end example | ||
801 | |||
802 | @noindent | ||
803 | If possible, the error clause should be terminated with a 'return' (or | ||
804 | 'goto' to some cleanup routine) and in this case, the 'else' clause | ||
805 | should be omitted: | ||
806 | |||
807 | @example | ||
808 | if (0 != stat ("filename," &sbuf)) @{ | ||
809 | error(); | ||
810 | return; | ||
811 | @} | ||
812 | /* handle normal case here */ | ||
813 | @end example | ||
814 | |||
815 | This serves to avoid deep nesting. The 'constants on the left' rule | ||
816 | applies to all constants (including. @code{GNUNET_SCHEDULER_NO_TASK}), | ||
817 | NULL, and enums). With the two above rules (constants on left, errors in | ||
818 | 'true' branch), there is only one way to write most branches correctly. | ||
819 | |||
820 | @item Combined assignments and tests are allowed if they do not hinder | ||
821 | code clarity. For example, one can write: | ||
822 | |||
823 | @example | ||
824 | if (NULL == (value = lookup_function())) @{ | ||
825 | error(); | ||
826 | return; | ||
827 | @} | ||
828 | @end example | ||
829 | |||
830 | @item Use @code{break} and @code{continue} wherever possible to avoid | ||
831 | deep(er) nesting. Thus, we would write: | ||
832 | |||
833 | @example | ||
834 | next = head; | ||
835 | while (NULL != (pos = next)) @{ | ||
836 | next = pos->next; | ||
837 | if (! should_free (pos)) | ||
838 | continue; | ||
839 | GNUNET_CONTAINER_DLL_remove (head, tail, pos); | ||
840 | GNUNET_free (pos); | ||
841 | @} | ||
842 | @end example | ||
843 | |||
844 | instead of | ||
845 | |||
846 | @example | ||
847 | next = head; while (NULL != (pos = next)) @{ | ||
848 | next = pos->next; | ||
849 | if (should_free (pos)) @{ | ||
850 | /* unnecessary nesting! */ | ||
851 | GNUNET_CONTAINER_DLL_remove (head, tail, pos); | ||
852 | GNUNET_free (pos); | ||
853 | @} | ||
854 | @} | ||
855 | @end example | ||
856 | |||
857 | @item We primarily use @code{for} and @code{while} loops. | ||
858 | A @code{while} loop is used if the method for advancing in the loop is | ||
859 | not a straightforward increment operation. In particular, we use: | ||
860 | |||
861 | @example | ||
862 | next = head; | ||
863 | while (NULL != (pos = next)) | ||
864 | @{ | ||
865 | next = pos->next; | ||
866 | if (! should_free (pos)) | ||
867 | continue; | ||
868 | GNUNET_CONTAINER_DLL_remove (head, tail, pos); | ||
869 | GNUNET_free (pos); | ||
870 | @} | ||
871 | @end example | ||
872 | |||
873 | to free entries in a list (as the iteration changes the structure of the | ||
874 | list due to the free; the equivalent @code{for} loop does no longer | ||
875 | follow the simple @code{for} paradigm of @code{for(INIT;TEST;INC)}). | ||
876 | However, for loops that do follow the simple @code{for} paradigm we do | ||
877 | use @code{for}, even if it involves linked lists: | ||
878 | |||
879 | @example | ||
880 | /* simple iteration over a linked list */ | ||
881 | for (pos = head; | ||
882 | NULL != pos; | ||
883 | pos = pos->next) | ||
884 | @{ | ||
885 | use (pos); | ||
886 | @} | ||
887 | @end example | ||
888 | |||
889 | |||
890 | @item The first argument to all higher-order functions in GNUnet must be | ||
891 | declared to be of type @code{void *} and is reserved for a closure. We do | ||
892 | not use inner functions, as trampolines would conflict with setups that | ||
893 | use non-executable stacks. | ||
894 | The first statement in a higher-order function, which unusually should | ||
895 | be part of the variable declarations, should assign the | ||
896 | @code{cls} argument to the precise expected type. For example: | ||
897 | |||
898 | @example | ||
899 | int callback (void *cls, char *args) @{ | ||
900 | struct Foo *foo = cls; | ||
901 | int other_variables; | ||
902 | |||
903 | /* rest of function */ | ||
904 | @} | ||
905 | @end example | ||
906 | |||
907 | |||
908 | @item It is good practice to write complex @code{if} expressions instead | ||
909 | of using deeply nested @code{if} statements. However, except for addition | ||
910 | and multiplication, all operators should use parens. This is fine: | ||
911 | |||
912 | @example | ||
913 | if ( (1 == foo) || ((0 == bar) && (x != y)) ) | ||
914 | return x; | ||
915 | @end example | ||
916 | |||
917 | |||
918 | However, this is not: | ||
919 | |||
920 | @example | ||
921 | if (1 == foo) | ||
922 | return x; | ||
923 | if (0 == bar && x != y) | ||
924 | return x; | ||
925 | @end example | ||
926 | |||
927 | @noindent | ||
928 | Note that splitting the @code{if} statement above is debateable as the | ||
929 | @code{return x} is a very trivial statement. However, once the logic after | ||
930 | the branch becomes more complicated (and is still identical), the "or" | ||
931 | formulation should be used for sure. | ||
932 | |||
933 | @item There should be two empty lines between the end of the function and | ||
934 | the comments describing the following function. There should be a single | ||
935 | empty line after the initial variable declarations of a function. If a | ||
936 | function has no local variables, there should be no initial empty line. If | ||
937 | a long function consists of several complex steps, those steps might be | ||
938 | separated by an empty line (possibly followed by a comment describing the | ||
939 | following step). The code should not contain empty lines in arbitrary | ||
940 | places; if in doubt, it is likely better to NOT have an empty line (this | ||
941 | way, more code will fit on the screen). | ||
942 | @end itemize | ||
943 | |||
944 | @c *********************************************************************** | ||
945 | @node Build-system | ||
946 | @section Build-system | ||
947 | |||
948 | If you have code that is likely not to compile or build rules you might | ||
949 | want to not trigger for most developers, use @code{if HAVE_EXPERIMENTAL} | ||
950 | in your @file{Makefile.am}. | ||
951 | Then it is OK to (temporarily) add non-compiling (or known-to-not-port) | ||
952 | code. | ||
953 | |||
954 | If you want to compile all testcases but NOT run them, run configure with | ||
955 | the @code{--enable-test-suppression} option. | ||
956 | |||
957 | If you want to run all testcases, including those that take a while, run | ||
958 | configure with the @code{--enable-expensive-testcases} option. | ||
959 | |||
960 | If you want to compile and run benchmarks, run configure with the | ||
961 | @code{--enable-benchmarks} option. | ||
962 | |||
963 | If you want to obtain code coverage results, run configure with the | ||
964 | @code{--enable-coverage} option and run the @file{coverage.sh} script in | ||
965 | the @file{contrib/} directory. | ||
966 | |||
967 | @cindex gnunet-ext | ||
968 | @node Developing extensions for GNUnet using the gnunet-ext template | ||
969 | @section Developing extensions for GNUnet using the gnunet-ext template | ||
970 | |||
971 | For developers who want to write extensions for GNUnet we provide the | ||
972 | gnunet-ext template to provide an easy to use skeleton. | ||
973 | |||
974 | gnunet-ext contains the build environment and template files for the | ||
975 | development of GNUnet services, command line tools, APIs and tests. | ||
976 | |||
977 | First of all you have to obtain gnunet-ext from git: | ||
978 | |||
979 | @example | ||
980 | git clone https://gnunet.org/git/gnunet-ext.git | ||
981 | @end example | ||
982 | |||
983 | The next step is to bootstrap and configure it. For configure you have to | ||
984 | provide the path containing GNUnet with | ||
985 | @code{--with-gnunet=/path/to/gnunet} and the prefix where you want the | ||
986 | install the extension using @code{--prefix=/path/to/install}: | ||
987 | |||
988 | @example | ||
989 | ./bootstrap | ||
990 | ./configure --prefix=/path/to/install --with-gnunet=/path/to/gnunet | ||
991 | @end example | ||
992 | |||
993 | When your GNUnet installation is not included in the default linker search | ||
994 | path, you have to add @code{/path/to/gnunet} to the file | ||
995 | @file{/etc/ld.so.conf} and run @code{ldconfig} or your add it to the | ||
996 | environmental variable @code{LD_LIBRARY_PATH} by using | ||
997 | |||
998 | @example | ||
999 | export LD_LIBRARY_PATH=/path/to/gnunet/lib | ||
1000 | @end example | ||
1001 | |||
1002 | @cindex writing testcases | ||
1003 | @node Writing testcases | ||
1004 | @section Writing testcases | ||
1005 | |||
1006 | Ideally, any non-trivial GNUnet code should be covered by automated | ||
1007 | testcases. Testcases should reside in the same place as the code that is | ||
1008 | being tested. The name of source files implementing tests should begin | ||
1009 | with "@code{test_}" followed by the name of the file | ||
1010 | that contains the code that is being tested. | ||
1011 | |||
1012 | Testcases in GNUnet should be integrated with the autotools build system. | ||
1013 | This way, developers and anyone building binary packages will be able to | ||
1014 | run all testcases simply by running @code{make check}. The final | ||
1015 | testcases shipped with the distribution should output at most some brief | ||
1016 | progress information and not display debug messages by default. The | ||
1017 | success or failure of a testcase must be indicated by returning zero | ||
1018 | (success) or non-zero (failure) from the main method of the testcase. | ||
1019 | The integration with the autotools is relatively straightforward and only | ||
1020 | requires modifications to the @file{Makefile.am} in the directory | ||
1021 | containing the testcase. For a testcase testing the code in @file{foo.c} | ||
1022 | the @file{Makefile.am} would contain the following lines: | ||
1023 | |||
1024 | @example | ||
1025 | check_PROGRAMS = test_foo | ||
1026 | TESTS = $(check_PROGRAMS) | ||
1027 | test_foo_SOURCES = test_foo.c | ||
1028 | test_foo_LDADD = $(top_builddir)/src/util/libgnunetutil.la | ||
1029 | @end example | ||
1030 | |||
1031 | Naturally, other libraries used by the testcase may be specified in the | ||
1032 | @code{LDADD} directive as necessary. | ||
1033 | |||
1034 | Often testcases depend on additional input files, such as a configuration | ||
1035 | file. These support files have to be listed using the @code{EXTRA_DIST} | ||
1036 | directive in order to ensure that they are included in the distribution. | ||
1037 | |||
1038 | Example: | ||
1039 | |||
1040 | @example | ||
1041 | EXTRA_DIST = test_foo_data.conf | ||
1042 | @end example | ||
1043 | |||
1044 | Executing @code{make check} will run all testcases in the current | ||
1045 | directory and all subdirectories. Testcases can be compiled individually | ||
1046 | by running @code{make test_foo} and then invoked directly using | ||
1047 | @code{./test_foo}. Note that due to the use of plugins in GNUnet, it is | ||
1048 | typically necessary to run @code{make install} before running any | ||
1049 | testcases. Thus the canonical command @code{make check install} has to be | ||
1050 | changed to @code{make install check} for GNUnet. | ||
1051 | |||
1052 | @cindex TESTING library | ||
1053 | @node GNUnet's TESTING library | ||
1054 | @section GNUnet's TESTING library | ||
1055 | |||
1056 | The TESTING library is used for writing testcases which involve starting a | ||
1057 | single or multiple peers. While peers can also be started by testcases | ||
1058 | using the ARM subsystem, using TESTING library provides an elegant way to | ||
1059 | do this. The configurations of the peers are auto-generated from a given | ||
1060 | template to have non-conflicting port numbers ensuring that peers' | ||
1061 | services do not run into bind errors. This is achieved by testing ports' | ||
1062 | availability by binding a listening socket to them before allocating them | ||
1063 | to services in the generated configurations. | ||
1064 | |||
1065 | An another advantage while using TESTING is that it shortens the testcase | ||
1066 | startup time as the hostkeys for peers are copied from a pre-computed set | ||
1067 | of hostkeys instead of generating them at peer startup which may take a | ||
1068 | considerable amount of time when starting multiple peers or on an embedded | ||
1069 | processor. | ||
1070 | |||
1071 | TESTING also allows for certain services to be shared among peers. This | ||
1072 | feature is invaluable when testing with multiple peers as it helps to | ||
1073 | reduce the number of services run per each peer and hence the total | ||
1074 | number of processes run per testcase. | ||
1075 | |||
1076 | TESTING library only handles creating, starting and stopping peers. | ||
1077 | Features useful for testcases such as connecting peers in a topology are | ||
1078 | not available in TESTING but are available in the TESTBED subsystem. | ||
1079 | Furthermore, TESTING only creates peers on the localhost, however by | ||
1080 | using TESTBED testcases can benefit from creating peers across multiple | ||
1081 | hosts. | ||
1082 | |||
1083 | @menu | ||
1084 | * API:: | ||
1085 | * Finer control over peer stop:: | ||
1086 | * Helper functions:: | ||
1087 | * Testing with multiple processes:: | ||
1088 | @end menu | ||
1089 | |||
1090 | @cindex TESTING API | ||
1091 | @node API | ||
1092 | @subsection API | ||
1093 | |||
1094 | TESTING abstracts a group of peers as a TESTING system. All peers in a | ||
1095 | system have common hostname and no two services of these peers have a | ||
1096 | same port or a UNIX domain socket path. | ||
1097 | |||
1098 | TESTING system can be created with the function | ||
1099 | @code{GNUNET_TESTING_system_create()} which returns a handle to the | ||
1100 | system. This function takes a directory path which is used for generating | ||
1101 | the configurations of peers, an IP address from which connections to the | ||
1102 | peers' services should be allowed, the hostname to be used in peers' | ||
1103 | configuration, and an array of shared service specifications of type | ||
1104 | @code{struct GNUNET_TESTING_SharedService}. | ||
1105 | |||
1106 | The shared service specification must specify the name of the service to | ||
1107 | share, the configuration pertaining to that shared service and the | ||
1108 | maximum number of peers that are allowed to share a single instance of | ||
1109 | the shared service. | ||
1110 | |||
1111 | TESTING system created with @code{GNUNET_TESTING_system_create()} chooses | ||
1112 | ports from the default range @code{12000} - @code{56000} while | ||
1113 | auto-generating configurations for peers. | ||
1114 | This range can be customised with the function | ||
1115 | @code{GNUNET_TESTING_system_create_with_portrange()}. This function is | ||
1116 | similar to @code{GNUNET_TESTING_system_create()} except that it take 2 | ||
1117 | additional parameters --- the start and end of the port range to use. | ||
1118 | |||
1119 | A TESTING system is destroyed with the funciton | ||
1120 | @code{GNUNET_TESTING_system_destory()}. This function takes the handle of | ||
1121 | the system and a flag to remove the files created in the directory used | ||
1122 | to generate configurations. | ||
1123 | |||
1124 | A peer is created with the function | ||
1125 | @code{GNUNET_TESTING_peer_configure()}. This functions takes the system | ||
1126 | handle, a configuration template from which the configuration for the peer | ||
1127 | is auto-generated and the index from where the hostkey for the peer has to | ||
1128 | be copied from. When successfull, this function returs a handle to the | ||
1129 | peer which can be used to start and stop it and to obtain the identity of | ||
1130 | the peer. If unsuccessful, a NULL pointer is returned with an error | ||
1131 | message. This function handles the generated configuration to have | ||
1132 | non-conflicting ports and paths. | ||
1133 | |||
1134 | Peers can be started and stopped by calling the functions | ||
1135 | @code{GNUNET_TESTING_peer_start()} and @code{GNUNET_TESTING_peer_stop()} | ||
1136 | respectively. A peer can be destroyed by calling the function | ||
1137 | @code{GNUNET_TESTING_peer_destroy}. When a peer is destroyed, the ports | ||
1138 | and paths in allocated in its configuration are reclaimed for usage in new | ||
1139 | peers. | ||
1140 | |||
1141 | @c *********************************************************************** | ||
1142 | @node Finer control over peer stop | ||
1143 | @subsection Finer control over peer stop | ||
1144 | |||
1145 | Using @code{GNUNET_TESTING_peer_stop()} is normally fine for testcases. | ||
1146 | However, calling this function for each peer is inefficient when trying to | ||
1147 | shutdown multiple peers as this function sends the termination signal to | ||
1148 | the given peer process and waits for it to terminate. It would be faster | ||
1149 | in this case to send the termination signals to the peers first and then | ||
1150 | wait on them. This is accomplished by the functions | ||
1151 | @code{GNUNET_TESTING_peer_kill()} which sends a termination signal to the | ||
1152 | peer, and the function @code{GNUNET_TESTING_peer_wait()} which waits on | ||
1153 | the peer. | ||
1154 | |||
1155 | Further finer control can be achieved by choosing to stop a peer | ||
1156 | asynchronously with the function @code{GNUNET_TESTING_peer_stop_async()}. | ||
1157 | This function takes a callback parameter and a closure for it in addition | ||
1158 | to the handle to the peer to stop. The callback function is called with | ||
1159 | the given closure when the peer is stopped. Using this function | ||
1160 | eliminates blocking while waiting for the peer to terminate. | ||
1161 | |||
1162 | An asynchronous peer stop can be cancelled by calling the function | ||
1163 | @code{GNUNET_TESTING_peer_stop_async_cancel()}. Note that calling this | ||
1164 | function does not prevent the peer from terminating if the termination | ||
1165 | signal has already been sent to it. It does, however, cancels the | ||
1166 | callback to be called when the peer is stopped. | ||
1167 | |||
1168 | @c *********************************************************************** | ||
1169 | @node Helper functions | ||
1170 | @subsection Helper functions | ||
1171 | |||
1172 | Most of the testcases can benefit from an abstraction which configures a | ||
1173 | peer and starts it. This is provided by the function | ||
1174 | @code{GNUNET_TESTING_peer_run()}. This function takes the testing | ||
1175 | directory pathname, a configuration template, a callback and its closure. | ||
1176 | This function creates a peer in the given testing directory by using the | ||
1177 | configuration template, starts the peer and calls the given callback with | ||
1178 | the given closure. | ||
1179 | |||
1180 | The function @code{GNUNET_TESTING_peer_run()} starts the ARM service of | ||
1181 | the peer which starts the rest of the configured services. A similar | ||
1182 | function @code{GNUNET_TESTING_service_run} can be used to just start a | ||
1183 | single service of a peer. In this case, the peer's ARM service is not | ||
1184 | started; instead, only the given service is run. | ||
1185 | |||
1186 | @c *********************************************************************** | ||
1187 | @node Testing with multiple processes | ||
1188 | @subsection Testing with multiple processes | ||
1189 | |||
1190 | When testing GNUnet, the splitting of the code into a services and clients | ||
1191 | often complicates testing. The solution to this is to have the testcase | ||
1192 | fork @code{gnunet-service-arm}, ask it to start the required server and | ||
1193 | daemon processes and then execute appropriate client actions (to test the | ||
1194 | client APIs or the core module or both). If necessary, multiple ARM | ||
1195 | services can be forked using different ports (!) to simulate a network. | ||
1196 | However, most of the time only one ARM process is needed. Note that on | ||
1197 | exit, the testcase should shutdown ARM with a @code{TERM} signal (to give | ||
1198 | it the chance to cleanly stop its child processes). | ||
1199 | |||
1200 | The following code illustrates spawning and killing an ARM process from a | ||
1201 | testcase: | ||
1202 | |||
1203 | @example | ||
1204 | static void run (void *cls, | ||
1205 | char *const *args, | ||
1206 | const char *cfgfile, | ||
1207 | const struct GNUNET_CONFIGURATION_Handle *cfg) @{ | ||
1208 | struct GNUNET_OS_Process *arm_pid; | ||
1209 | arm_pid = GNUNET_OS_start_process (NULL, | ||
1210 | NULL, | ||
1211 | "gnunet-service-arm", | ||
1212 | "gnunet-service-arm", | ||
1213 | "-c", | ||
1214 | cfgname, | ||
1215 | NULL); | ||
1216 | /* do real test work here */ | ||
1217 | if (0 != GNUNET_OS_process_kill (arm_pid, SIGTERM)) | ||
1218 | GNUNET_log_strerror | ||
1219 | (GNUNET_ERROR_TYPE_WARNING, "kill"); | ||
1220 | GNUNET_assert (GNUNET_OK == GNUNET_OS_process_wait (arm_pid)); | ||
1221 | GNUNET_OS_process_close (arm_pid); @} | ||
1222 | |||
1223 | GNUNET_PROGRAM_run (argc, argv, | ||
1224 | "NAME-OF-TEST", | ||
1225 | "nohelp", | ||
1226 | options, | ||
1227 | &run, | ||
1228 | cls); | ||
1229 | @end example | ||
1230 | |||
1231 | |||
1232 | An alternative way that works well to test plugins is to implement a | ||
1233 | mock-version of the environment that the plugin expects and then to | ||
1234 | simply load the plugin directly. | ||
1235 | |||
1236 | @c *********************************************************************** | ||
1237 | @node Performance regression analysis with Gauger | ||
1238 | @section Performance regression analysis with Gauger | ||
1239 | |||
1240 | To help avoid performance regressions, GNUnet uses Gauger. Gauger is a | ||
1241 | simple logging tool that allows remote hosts to send performance data to | ||
1242 | a central server, where this data can be analyzed and visualized. Gauger | ||
1243 | shows graphs of the repository revisions and the performace data recorded | ||
1244 | for each revision, so sudden performance peaks or drops can be identified | ||
1245 | and linked to a specific revision number. | ||
1246 | |||
1247 | In the case of GNUnet, the buildbots log the performance data obtained | ||
1248 | during the tests after each build. The data can be accesed on GNUnet's | ||
1249 | Gauger page. | ||
1250 | |||
1251 | The menu on the left allows to select either the results of just one | ||
1252 | build bot (under "Hosts") or review the data from all hosts for a given | ||
1253 | test result (under "Metrics"). In case of very different absolute value | ||
1254 | of the results, for instance arm vs. amd64 machines, the option | ||
1255 | "Normalize" on a metric view can help to get an idea about the | ||
1256 | performance evolution across all hosts. | ||
1257 | |||
1258 | Using Gauger in GNUnet and having the performance of a module tracked over | ||
1259 | time is very easy. First of course, the testcase must generate some | ||
1260 | consistent metric, which makes sense to have logged. Highly volatile or | ||
1261 | random dependant metrics probably are not ideal candidates for meaningful | ||
1262 | regression detection. | ||
1263 | |||
1264 | To start logging any value, just include @code{gauger.h} in your testcase | ||
1265 | code. Then, use the macro @code{GAUGER()} to make the Buildbots log | ||
1266 | whatever value is of interest for you to @code{gnunet.org}'s Gauger | ||
1267 | server. No setup is necessary as most Buildbots have already everything | ||
1268 | in place and new metrics are created on demand. To delete a metric, you | ||
1269 | need to contact a member of the GNUnet development team (a file will need | ||
1270 | to be removed manually from the respective directory). | ||
1271 | |||
1272 | The code in the test should look like this: | ||
1273 | |||
1274 | @example | ||
1275 | [other includes] | ||
1276 | #include <gauger.h> | ||
1277 | |||
1278 | int main (int argc, char *argv[]) @{ | ||
1279 | |||
1280 | [run test, generate data] | ||
1281 | GAUGER("YOUR_MODULE", | ||
1282 | "METRIC_NAME", | ||
1283 | (float)value, | ||
1284 | "UNIT"); @} | ||
1285 | @end example | ||
1286 | |||
1287 | Where: | ||
1288 | |||
1289 | @table @asis | ||
1290 | |||
1291 | @item @strong{YOUR_MODULE} is a category in the gauger page and should be | ||
1292 | the name of the module or subsystem like "Core" or "DHT" | ||
1293 | @item @strong{METRIC} is | ||
1294 | the name of the metric being collected and should be concise and | ||
1295 | descriptive, like "PUT operations in sqlite-datastore". | ||
1296 | @item @strong{value} is the value | ||
1297 | of the metric that is logged for this run. | ||
1298 | @item @strong{UNIT} is the unit in | ||
1299 | which the value is measured, for instance "kb/s" or "kb of RAM/node". | ||
1300 | @end table | ||
1301 | |||
1302 | If you wish to use Gauger for your own project, you can grab a copy of the | ||
1303 | latest stable release or check out Gauger's Subversion repository. | ||
1304 | |||
1305 | @cindex TESTBED Subsystem | ||
1306 | @node GNUnet's TESTBED Subsystem | ||
1307 | @section GNUnet's TESTBED Subsystem | ||
1308 | |||
1309 | The TESTBED subsystem facilitates testing and measuring of multi-peer | ||
1310 | deployments on a single host or over multiple hosts. | ||
1311 | |||
1312 | The architecture of the testbed module is divided into the following: | ||
1313 | @itemize @bullet | ||
1314 | |||
1315 | @item Testbed API: An API which is used by the testing driver programs. It | ||
1316 | provides with functions for creating, destroying, starting, stopping | ||
1317 | peers, etc. | ||
1318 | |||
1319 | @item Testbed service (controller): A service which is started through the | ||
1320 | Testbed API. This service handles operations to create, destroy, start, | ||
1321 | stop peers, connect them, modify their configurations. | ||
1322 | |||
1323 | @item Testbed helper: When a controller has to be started on a host, the | ||
1324 | testbed API starts the testbed helper on that host which in turn starts | ||
1325 | the controller. The testbed helper receives a configuration for the | ||
1326 | controller through its stdin and changes it to ensure the controller | ||
1327 | doesn't run into any port conflict on that host. | ||
1328 | @end itemize | ||
1329 | |||
1330 | |||
1331 | The testbed service (controller) is different from the other GNUnet | ||
1332 | services in that it is not started by ARM and is not supposed to be run | ||
1333 | as a daemon. It is started by the testbed API through a testbed helper. | ||
1334 | In a typical scenario involving multiple hosts, a controller is started | ||
1335 | on each host. Controllers take up the actual task of creating peers, | ||
1336 | starting and stopping them on the hosts they run. | ||
1337 | |||
1338 | While running deployments on a single localhost the testbed API starts the | ||
1339 | testbed helper directly as a child process. When running deployments on | ||
1340 | remote hosts the testbed API starts Testbed Helpers on each remote host | ||
1341 | through remote shell. By default testbed API uses SSH as a remote shell. | ||
1342 | This can be changed by setting the environmental variable | ||
1343 | GNUNET_TESTBED_RSH_CMD to the required remote shell program. This | ||
1344 | variable can also contain parameters which are to be passed to the remote | ||
1345 | shell program. For e.g: | ||
1346 | |||
1347 | @example | ||
1348 | export GNUNET_TESTBED_RSH_CMD="ssh -o BatchMode=yes \ | ||
1349 | -o NoHostAuthenticationForLocalhost=yes %h"@ | ||
1350 | @end example | ||
1351 | |||
1352 | Substitutions are allowed int the above command string also allows for | ||
1353 | substitions. through placemarks which begin with a `%'. At present the | ||
1354 | following substitutions are supported | ||
1355 | |||
1356 | @itemize @bullet | ||
1357 | @item | ||
1358 | %h: hostname | ||
1359 | @item | ||
1360 | %u: username | ||
1361 | @item | ||
1362 | %p: port | ||
1363 | @end itemize | ||
1364 | |||
1365 | Note that the substitution placemark is replaced only when the | ||
1366 | corresponding field is available and only once. Specifying | ||
1367 | @example | ||
1368 | %u@atchar{}%h | ||
1369 | @end example | ||
1370 | doesn't work either. | ||
1371 | If you want to user username substitutions for SSH | ||
1372 | use the argument @code{-l} before the username substitution. | ||
1373 | For exmaple: | ||
1374 | @example | ||
1375 | ssh -l %u -p %p %h | ||
1376 | @end example | ||
1377 | |||
1378 | The testbed API and the helper communicate through the helpers stdin and | ||
1379 | stdout. As the helper is started through a remote shell on remote hosts | ||
1380 | any output messages from the remote shell interfere with the communication | ||
1381 | and results in a failure while starting the helper. For this reason, it is | ||
1382 | suggested to use flags to make the remote shells produce no output | ||
1383 | messages and to have password-less logins. The default remote shell, SSH, | ||
1384 | the default options are: | ||
1385 | |||
1386 | @example | ||
1387 | -o BatchMode=yes -o NoHostBasedAuthenticationForLocalhost=yes" | ||
1388 | @end example | ||
1389 | |||
1390 | Password-less logins should be ensured by using SSH keys. | ||
1391 | |||
1392 | Since the testbed API executes the remote shell as a non-interactive | ||
1393 | shell, certain scripts like .bashrc, .profiler may not be executed. If | ||
1394 | this is the case testbed API can be forced to execute an interactive | ||
1395 | shell by setting up the environmental variable | ||
1396 | @code{GNUNET_TESTBED_RSH_CMD_SUFFIX} to a shell program. | ||
1397 | |||
1398 | An example could be: | ||
1399 | |||
1400 | @example | ||
1401 | export GNUNET_TESTBED_RSH_CMD_SUFFIX="sh -lc" | ||
1402 | @end example | ||
1403 | |||
1404 | The testbed API will then execute the remote shell program as: | ||
1405 | |||
1406 | @example | ||
1407 | $GNUNET_TESTBED_RSH_CMD -p $port $dest $GNUNET_TESTBED_RSH_CMD_SUFFIX \ | ||
1408 | gnunet-helper-testbed | ||
1409 | @end example | ||
1410 | |||
1411 | On some systems, problems may arise while starting testbed helpers if | ||
1412 | GNUnet is installed into a custom location since the helper may not be | ||
1413 | found in the standard path. This can be addressed by setting the variable | ||
1414 | `@code{HELPER_BINARY_PATH}' to the path of the testbed helper. | ||
1415 | Testbed API will then use this path to start helper binaries both | ||
1416 | locally and remotely. | ||
1417 | |||
1418 | Testbed API can accessed by including the | ||
1419 | "@file{gnunet_testbed_service.h}" file and linking with -lgnunettestbed. | ||
1420 | |||
1421 | @c *********************************************************************** | ||
1422 | @menu | ||
1423 | * Supported Topologies:: | ||
1424 | * Hosts file format:: | ||
1425 | * Topology file format:: | ||
1426 | * Testbed Barriers:: | ||
1427 | * Automatic large-scale deployment in the PlanetLab testbed:: | ||
1428 | * TESTBED Caveats:: | ||
1429 | @end menu | ||
1430 | |||
1431 | @node Supported Topologies | ||
1432 | @subsection Supported Topologies | ||
1433 | |||
1434 | While testing multi-peer deployments, it is often needed that the peers | ||
1435 | are connected in some topology. This requirement is addressed by the | ||
1436 | function @code{GNUNET_TESTBED_overlay_connect()} which connects any given | ||
1437 | two peers in the testbed. | ||
1438 | |||
1439 | The API also provides a helper function | ||
1440 | @code{GNUNET_TESTBED_overlay_configure_topology()} to connect a given set | ||
1441 | of peers in any of the following supported topologies: | ||
1442 | |||
1443 | @itemize @bullet | ||
1444 | |||
1445 | @item @code{GNUNET_TESTBED_TOPOLOGY_CLIQUE}: All peers are connected with | ||
1446 | each other | ||
1447 | |||
1448 | @item @code{GNUNET_TESTBED_TOPOLOGY_LINE}: Peers are connected to form a | ||
1449 | line | ||
1450 | |||
1451 | @item @code{GNUNET_TESTBED_TOPOLOGY_RING}: Peers are connected to form a | ||
1452 | ring topology | ||
1453 | |||
1454 | @item @code{GNUNET_TESTBED_TOPOLOGY_2D_TORUS}: Peers are connected to | ||
1455 | form a 2 dimensional torus topology. The number of peers may not be a | ||
1456 | perfect square, in that case the resulting torus may not have the uniform | ||
1457 | poloidal and toroidal lengths | ||
1458 | |||
1459 | @item @code{GNUNET_TESTBED_TOPOLOGY_ERDOS_RENYI}: Topology is generated | ||
1460 | to form a random graph. The number of links to be present should be given | ||
1461 | |||
1462 | @item @code{GNUNET_TESTBED_TOPOLOGY_SMALL_WORLD}: Peers are connected to | ||
1463 | form a 2D Torus with some random links among them. The number of random | ||
1464 | links are to be given | ||
1465 | |||
1466 | @item @code{GNUNET_TESTBED_TOPOLOGY_SMALL_WORLD_RING}: Peers are | ||
1467 | connected to form a ring with some random links among them. The number of | ||
1468 | random links are to be given | ||
1469 | |||
1470 | @item @code{GNUNET_TESTBED_TOPOLOGY_SCALE_FREE}: Connects peers in a | ||
1471 | topology where peer connectivity follows power law - new peers are | ||
1472 | connected with high probabililty to well connected peers. | ||
1473 | @footnote{See Emergence of Scaling in Random Networks. Science 286, | ||
1474 | 509-512, 1999 | ||
1475 | (@uref{https://gnunet.org/git/bibliography.git/plain/docs/emergence_of_scaling_in_random_networks__barabasi_albert_science_286__1999.pdf, pdf})} | ||
1476 | |||
1477 | @item @code{GNUNET_TESTBED_TOPOLOGY_FROM_FILE}: The topology information | ||
1478 | is loaded from a file. The path to the file has to be given. | ||
1479 | @xref{Topology file format}, for the format of this file. | ||
1480 | |||
1481 | @item @code{GNUNET_TESTBED_TOPOLOGY_NONE}: No topology | ||
1482 | @end itemize | ||
1483 | |||
1484 | |||
1485 | The above supported topologies can be specified respectively by setting | ||
1486 | the variable @code{OVERLAY_TOPOLOGY} to the following values in the | ||
1487 | configuration passed to Testbed API functions | ||
1488 | @code{GNUNET_TESTBED_test_run()} and | ||
1489 | @code{GNUNET_TESTBED_run()}: | ||
1490 | |||
1491 | @itemize @bullet | ||
1492 | @item @code{CLIQUE} | ||
1493 | @item @code{RING} | ||
1494 | @item @code{LINE} | ||
1495 | @item @code{2D_TORUS} | ||
1496 | @item @code{RANDOM} | ||
1497 | @item @code{SMALL_WORLD} | ||
1498 | @item @code{SMALL_WORLD_RING} | ||
1499 | @item @code{SCALE_FREE} | ||
1500 | @item @code{FROM_FILE} | ||
1501 | @item @code{NONE} | ||
1502 | @end itemize | ||
1503 | |||
1504 | |||
1505 | Topologies @code{RANDOM}, @code{SMALL_WORLD} and @code{SMALL_WORLD_RING} | ||
1506 | require the option @code{OVERLAY_RANDOM_LINKS} to be set to the number of | ||
1507 | random links to be generated in the configuration. The option will be | ||
1508 | ignored for the rest of the topologies. | ||
1509 | |||
1510 | Topology @code{SCALE_FREE} requires the options | ||
1511 | @code{SCALE_FREE_TOPOLOGY_CAP} to be set to the maximum number of peers | ||
1512 | which can connect to a peer and @code{SCALE_FREE_TOPOLOGY_M} to be set to | ||
1513 | how many peers a peer should be atleast connected to. | ||
1514 | |||
1515 | Similarly, the topology @code{FROM_FILE} requires the option | ||
1516 | @code{OVERLAY_TOPOLOGY_FILE} to contain the path of the file containing | ||
1517 | the topology information. This option is ignored for the rest of the | ||
1518 | topologies. @xref{Topology file format}, for the format of this file. | ||
1519 | |||
1520 | @c *********************************************************************** | ||
1521 | @node Hosts file format | ||
1522 | @subsection Hosts file format | ||
1523 | |||
1524 | The testbed API offers the function | ||
1525 | @code{GNUNET_TESTBED_hosts_load_from_file()} to load from a given file | ||
1526 | details about the hosts which testbed can use for deploying peers. | ||
1527 | This function is useful to keep the data about hosts | ||
1528 | separate instead of hard coding them in code. | ||
1529 | |||
1530 | Another helper function from testbed API, @code{GNUNET_TESTBED_run()} | ||
1531 | also takes a hosts file name as its parameter. It uses the above | ||
1532 | function to populate the hosts data structures and start controllers to | ||
1533 | deploy peers. | ||
1534 | |||
1535 | These functions require the hosts file to be of the following format: | ||
1536 | @itemize @bullet | ||
1537 | @item Each line is interpreted to have details about a host | ||
1538 | @item Host details should include the username to use for logging into the | ||
1539 | host, the hostname of the host and the port number to use for the remote | ||
1540 | shell program. All thee values should be given. | ||
1541 | @item These details should be given in the following format: | ||
1542 | @example | ||
1543 | <username>@@<hostname>:<port> | ||
1544 | @end example | ||
1545 | @end itemize | ||
1546 | |||
1547 | Note that having canonical hostnames may cause problems while resolving | ||
1548 | the IP addresses (See this bug). Hence it is advised to provide the hosts' | ||
1549 | IP numerical addresses as hostnames whenever possible. | ||
1550 | |||
1551 | @c *********************************************************************** | ||
1552 | @node Topology file format | ||
1553 | @subsection Topology file format | ||
1554 | |||
1555 | A topology file describes how peers are to be connected. It should adhere | ||
1556 | to the following format for testbed to parse it correctly. | ||
1557 | |||
1558 | Each line should begin with the target peer id. This should be followed by | ||
1559 | a colon(`:') and origin peer ids seperated by `|'. All spaces except for | ||
1560 | newline characters are ignored. The API will then try to connect each | ||
1561 | origin peer to the target peer. | ||
1562 | |||
1563 | For example, the following file will result in 5 overlay connections: | ||
1564 | [2->1], [3->1],[4->3], [0->3], [2->0]@ | ||
1565 | @code{@ 1:2|3@ 3:4| 0@ 0: 2@ } | ||
1566 | |||
1567 | @c *********************************************************************** | ||
1568 | @node Testbed Barriers | ||
1569 | @subsection Testbed Barriers | ||
1570 | |||
1571 | The testbed subsystem's barriers API facilitates coordination among the | ||
1572 | peers run by the testbed and the experiment driver. The concept is | ||
1573 | similar to the barrier synchronisation mechanism found in parallel | ||
1574 | programming or multi-threading paradigms - a peer waits at a barrier upon | ||
1575 | reaching it until the barrier is reached by a predefined number of peers. | ||
1576 | This predefined number of peers required to cross a barrier is also called | ||
1577 | quorum. We say a peer has reached a barrier if the peer is waiting for the | ||
1578 | barrier to be crossed. Similarly a barrier is said to be reached if the | ||
1579 | required quorum of peers reach the barrier. A barrier which is reached is | ||
1580 | deemed as crossed after all the peers waiting on it are notified. | ||
1581 | |||
1582 | The barriers API provides the following functions: | ||
1583 | @itemize @bullet | ||
1584 | @item @strong{@code{GNUNET_TESTBED_barrier_init()}:} function to | ||
1585 | initialse a barrier in the experiment | ||
1586 | @item @strong{@code{GNUNET_TESTBED_barrier_cancel()}:} function to cancel | ||
1587 | a barrier which has been initialised before | ||
1588 | @item @strong{@code{GNUNET_TESTBED_barrier_wait()}:} function to signal | ||
1589 | barrier service that the caller has reached a barrier and is waiting for | ||
1590 | it to be crossed | ||
1591 | @item @strong{@code{GNUNET_TESTBED_barrier_wait_cancel()}:} function to | ||
1592 | stop waiting for a barrier to be crossed | ||
1593 | @end itemize | ||
1594 | |||
1595 | |||
1596 | Among the above functions, the first two, namely | ||
1597 | @code{GNUNET_TESTBED_barrier_init()} and | ||
1598 | @code{GNUNET_TESTBED_barrier_cancel()} are used by experiment drivers. All | ||
1599 | barriers should be initialised by the experiment driver by calling | ||
1600 | @code{GNUNET_TESTBED_barrier_init()}. This function takes a name to | ||
1601 | identify the barrier, the quorum required for the barrier to be crossed | ||
1602 | and a notification callback for notifying the experiment driver when the | ||
1603 | barrier is crossed. @code{GNUNET_TESTBED_barrier_cancel()} cancels an | ||
1604 | initialised barrier and frees the resources allocated for it. This | ||
1605 | function can be called upon a initialised barrier before it is crossed. | ||
1606 | |||
1607 | The remaining two functions @code{GNUNET_TESTBED_barrier_wait()} and | ||
1608 | @code{GNUNET_TESTBED_barrier_wait_cancel()} are used in the peer's | ||
1609 | processes. @code{GNUNET_TESTBED_barrier_wait()} connects to the local | ||
1610 | barrier service running on the same host the peer is running on and | ||
1611 | registers that the caller has reached the barrier and is waiting for the | ||
1612 | barrier to be crossed. Note that this function can only be used by peers | ||
1613 | which are started by testbed as this function tries to access the local | ||
1614 | barrier service which is part of the testbed controller service. Calling | ||
1615 | @code{GNUNET_TESTBED_barrier_wait()} on an uninitialised barrier results | ||
1616 | in failure. @code{GNUNET_TESTBED_barrier_wait_cancel()} cancels the | ||
1617 | notification registered by @code{GNUNET_TESTBED_barrier_wait()}. | ||
1618 | |||
1619 | |||
1620 | @c *********************************************************************** | ||
1621 | @menu | ||
1622 | * Implementation:: | ||
1623 | @end menu | ||
1624 | |||
1625 | @node Implementation | ||
1626 | @subsubsection Implementation | ||
1627 | |||
1628 | Since barriers involve coordination between experiment driver and peers, | ||
1629 | the barrier service in the testbed controller is split into two | ||
1630 | components. The first component responds to the message generated by the | ||
1631 | barrier API used by the experiment driver (functions | ||
1632 | @code{GNUNET_TESTBED_barrier_init()} and | ||
1633 | @code{GNUNET_TESTBED_barrier_cancel()}) and the second component to the | ||
1634 | messages generated by barrier API used by peers (functions | ||
1635 | @code{GNUNET_TESTBED_barrier_wait()} and | ||
1636 | @code{GNUNET_TESTBED_barrier_wait_cancel()}). | ||
1637 | |||
1638 | Calling @code{GNUNET_TESTBED_barrier_init()} sends a | ||
1639 | @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_INIT} message to the master | ||
1640 | controller. The master controller then registers a barrier and calls | ||
1641 | @code{GNUNET_TESTBED_barrier_init()} for each its subcontrollers. In this | ||
1642 | way barrier initialisation is propagated to the controller hierarchy. | ||
1643 | While propagating initialisation, any errors at a subcontroller such as | ||
1644 | timeout during further propagation are reported up the hierarchy back to | ||
1645 | the experiment driver. | ||
1646 | |||
1647 | Similar to @code{GNUNET_TESTBED_barrier_init()}, | ||
1648 | @code{GNUNET_TESTBED_barrier_cancel()} propagates | ||
1649 | @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_CANCEL} message which causes | ||
1650 | controllers to remove an initialised barrier. | ||
1651 | |||
1652 | The second component is implemented as a separate service in the binary | ||
1653 | `gnunet-service-testbed' which already has the testbed controller service. | ||
1654 | Although this deviates from the gnunet process architecture of having one | ||
1655 | service per binary, it is needed in this case as this component needs | ||
1656 | access to barrier data created by the first component. This component | ||
1657 | responds to @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_WAIT} messages from | ||
1658 | local peers when they call @code{GNUNET_TESTBED_barrier_wait()}. Upon | ||
1659 | receiving @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_WAIT} message, the | ||
1660 | service checks if the requested barrier has been initialised before and | ||
1661 | if it was not initialised, an error status is sent through | ||
1662 | @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_STATUS} message to the local | ||
1663 | peer and the connection from the peer is terminated. If the barrier is | ||
1664 | initialised before, the barrier's counter for reached peers is incremented | ||
1665 | and a notification is registered to notify the peer when the barrier is | ||
1666 | reached. The connection from the peer is left open. | ||
1667 | |||
1668 | When enough peers required to attain the quorum send | ||
1669 | @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_WAIT} messages, the controller | ||
1670 | sends a @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_STATUS} message to its | ||
1671 | parent informing that the barrier is crossed. If the controller has | ||
1672 | started further subcontrollers, it delays this message until it receives | ||
1673 | a similar notification from each of those subcontrollers. Finally, the | ||
1674 | barriers API at the experiment driver receives the | ||
1675 | @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_STATUS} when the barrier is | ||
1676 | reached at all the controllers. | ||
1677 | |||
1678 | The barriers API at the experiment driver responds to the | ||
1679 | @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_STATUS} message by echoing it | ||
1680 | back to the master controller and notifying the experiment controller | ||
1681 | through the notification callback that a barrier has been crossed. The | ||
1682 | echoed @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_STATUS} message is | ||
1683 | propagated by the master controller to the controller hierarchy. This | ||
1684 | propagation triggers the notifications registered by peers at each of the | ||
1685 | controllers in the hierarchy. Note the difference between this downward | ||
1686 | propagation of the @code{GNUNET_MESSAGE_TYPE_TESTBED_BARRIER_STATUS} | ||
1687 | message from its upward propagation --- the upward propagation is needed | ||
1688 | for ensuring that the barrier is reached by all the controllers and the | ||
1689 | downward propagation is for triggering that the barrier is crossed. | ||
1690 | |||
1691 | @cindex PlanetLab testbed | ||
1692 | @node Automatic large-scale deployment in the PlanetLab testbed | ||
1693 | @subsection Automatic large-scale deployment in the PlanetLab testbed | ||
1694 | |||
1695 | PlanetLab is a testbed for computer networking and distributed systems | ||
1696 | research. It was established in 2002 and as of June 2010 was composed of | ||
1697 | 1090 nodes at 507 sites worldwide. | ||
1698 | |||
1699 | To automate the GNUnet we created a set of automation tools to simplify | ||
1700 | the large-scale deployment. We provide you a set of scripts you can use | ||
1701 | to deploy GNUnet on a set of nodes and manage your installation. | ||
1702 | |||
1703 | Please also check @uref{https://gnunet.org/installation-fedora8-svn} and | ||
1704 | @uref{https://gnunet.org/installation-fedora12-svn} to find detailled | ||
1705 | instructions how to install GNUnet on a PlanetLab node. | ||
1706 | |||
1707 | |||
1708 | @c *********************************************************************** | ||
1709 | @menu | ||
1710 | * PlanetLab Automation for Fedora8 nodes:: | ||
1711 | * Install buildslave on PlanetLab nodes running fedora core 8:: | ||
1712 | * Setup a new PlanetLab testbed using GPLMT:: | ||
1713 | * Why do i get an ssh error when using the regex profiler?:: | ||
1714 | @end menu | ||
1715 | |||
1716 | @node PlanetLab Automation for Fedora8 nodes | ||
1717 | @subsubsection PlanetLab Automation for Fedora8 nodes | ||
1718 | |||
1719 | @c *********************************************************************** | ||
1720 | @node Install buildslave on PlanetLab nodes running fedora core 8 | ||
1721 | @subsubsection Install buildslave on PlanetLab nodes running fedora core 8 | ||
1722 | @c ** Actually this is a subsubsubsection, but must be fixed differently | ||
1723 | @c ** as subsubsection is the lowest. | ||
1724 | |||
1725 | Since most of the PlanetLab nodes are running the very old Fedora core 8 | ||
1726 | image, installing the buildslave software is quite some pain. For our | ||
1727 | PlanetLab testbed we figured out how to install the buildslave software | ||
1728 | best. | ||
1729 | |||
1730 | @c This is a vvery terrible way to suggest installing software. | ||
1731 | @c FIXME: Is there an official, safer way instead of blind-piping a | ||
1732 | @c script? | ||
1733 | @c FIXME: Use newer pypi URLs below. | ||
1734 | Install Distribute for Python: | ||
1735 | |||
1736 | @example | ||
1737 | curl http://python-distribute.org/distribute_setup.py | sudo python | ||
1738 | @end example | ||
1739 | |||
1740 | Install Distribute for zope.interface <= 3.8.0 (4.0 and 4.0.1 will not | ||
1741 | work): | ||
1742 | |||
1743 | @example | ||
1744 | export PYPI=@value{PYPI-URL} | ||
1745 | wget $PYPI/z/zope.interface/zope.interface-3.8.0.tar.gz | ||
1746 | tar zvfz zope.interface-3.8.0.tar.gz | ||
1747 | cd zope.interface-3.8.0 | ||
1748 | sudo python setup.py install | ||
1749 | @end example | ||
1750 | |||
1751 | Install the buildslave software (0.8.6 was the latest version): | ||
1752 | |||
1753 | @example | ||
1754 | export GCODE="http://buildbot.googlecode.com/files" | ||
1755 | wget $GCODE/buildbot-slave-0.8.6p1.tar.gz | ||
1756 | tar xvfz buildbot-slave-0.8.6p1.tar.gz | ||
1757 | cd buildslave-0.8.6p1 | ||
1758 | sudo python setup.py install | ||
1759 | @end example | ||
1760 | |||
1761 | The setup will download the matching twisted package and install it. | ||
1762 | It will also try to install the latest version of zope.interface which | ||
1763 | will fail to install. Buildslave will work anyway since version 3.8.0 | ||
1764 | was installed before! | ||
1765 | |||
1766 | @c *********************************************************************** | ||
1767 | @node Setup a new PlanetLab testbed using GPLMT | ||
1768 | @subsubsection Setup a new PlanetLab testbed using GPLMT | ||
1769 | |||
1770 | @itemize @bullet | ||
1771 | @item Get a new slice and assign nodes | ||
1772 | Ask your PlanetLab PI to give you a new slice and assign the nodes you | ||
1773 | need | ||
1774 | @item Install a buildmaster | ||
1775 | You can stick to the buildbot documentation:@ | ||
1776 | @uref{http://buildbot.net/buildbot/docs/current/manual/installation.html} | ||
1777 | @item Install the buildslave software on all nodes | ||
1778 | To install the buildslave on all nodes assigned to your slice you can use | ||
1779 | the tasklist @code{install_buildslave_fc8.xml} provided with GPLMT: | ||
1780 | |||
1781 | @example | ||
1782 | ./gplmt.py -c contrib/tumple_gnunet.conf -t \ | ||
1783 | contrib/tasklists/install_buildslave_fc8.xml -a -p <planetlab password> | ||
1784 | @end example | ||
1785 | |||
1786 | @item Create the buildmaster configuration and the slave setup commands | ||
1787 | |||
1788 | The master and the and the slaves have need to have credentials and the | ||
1789 | master has to have all nodes configured. This can be done with the | ||
1790 | @file{create_buildbot_configuration.py} script in the @file{scripts} | ||
1791 | directory. | ||
1792 | |||
1793 | This scripts takes a list of nodes retrieved directly from PlanetLab or | ||
1794 | read from a file and a configuration template and creates: | ||
1795 | |||
1796 | @itemize @bullet | ||
1797 | @item a tasklist which can be executed with gplmt to setup the slaves | ||
1798 | @item a master.cfg file containing a PlanetLab nodes | ||
1799 | @end itemize | ||
1800 | |||
1801 | A configuration template is included in the <contrib>, most important is | ||
1802 | that the script replaces the following tags in the template: | ||
1803 | |||
1804 | %GPLMT_BUILDER_DEFINITION :@ GPLMT_BUILDER_SUMMARY@ GPLMT_SLAVES@ | ||
1805 | %GPLMT_SCHEDULER_BUILDERS | ||
1806 | |||
1807 | Create configuration for all nodes assigned to a slice: | ||
1808 | |||
1809 | @example | ||
1810 | ./create_buildbot_configuration.py -u <planetlab username> \ | ||
1811 | -p <planetlab password> -s <slice> -m <buildmaster+port> \ | ||
1812 | -t <template> | ||
1813 | @end example | ||
1814 | |||
1815 | Create configuration for some nodes in a file: | ||
1816 | |||
1817 | @example | ||
1818 | ./create_buildbot_configuration.p -f <node_file> \ | ||
1819 | -m <buildmaster+port> -t <template> | ||
1820 | @end example | ||
1821 | |||
1822 | @item Copy the @file{master.cfg} to the buildmaster and start it | ||
1823 | Use @code{buildbot start <basedir>} to start the server | ||
1824 | @item Setup the buildslaves | ||
1825 | @end itemize | ||
1826 | |||
1827 | @c *********************************************************************** | ||
1828 | @node Why do i get an ssh error when using the regex profiler? | ||
1829 | @subsubsection Why do i get an ssh error when using the regex profiler? | ||
1830 | |||
1831 | Why do i get an ssh error "Permission denied (publickey,password)." when | ||
1832 | using the regex profiler although passwordless ssh to localhost works | ||
1833 | using publickey and ssh-agent? | ||
1834 | |||
1835 | You have to generate a public/private-key pair with no password:@ | ||
1836 | @code{ssh-keygen -t rsa -b 4096 -f ~/.ssh/id_localhost}@ | ||
1837 | and then add the following to your ~/.ssh/config file: | ||
1838 | |||
1839 | @code{Host 127.0.0.1@ IdentityFile ~/.ssh/id_localhost} | ||
1840 | |||
1841 | now make sure your hostsfile looks like | ||
1842 | |||
1843 | @example | ||
1844 | [USERNAME]@@127.0.0.1:22@ | ||
1845 | [USERNAME]@@127.0.0.1:22 | ||
1846 | @end example | ||
1847 | |||
1848 | You can test your setup by running @code{ssh 127.0.0.1} in a | ||
1849 | terminal and then in the opened session run it again. | ||
1850 | If you were not asked for a password on either login, | ||
1851 | then you should be good to go. | ||
1852 | |||
1853 | @cindex TESTBED Caveats | ||
1854 | @node TESTBED Caveats | ||
1855 | @subsection TESTBED Caveats | ||
1856 | |||
1857 | This section documents a few caveats when using the GNUnet testbed | ||
1858 | subsystem. | ||
1859 | |||
1860 | @c *********************************************************************** | ||
1861 | @menu | ||
1862 | * CORE must be started:: | ||
1863 | * ATS must want the connections:: | ||
1864 | @end menu | ||
1865 | |||
1866 | @node CORE must be started | ||
1867 | @subsubsection CORE must be started | ||
1868 | |||
1869 | A simple issue is #3993: Your configuration MUST somehow ensure that for | ||
1870 | each peer the CORE service is started when the peer is setup, otherwise | ||
1871 | TESTBED may fail to connect peers when the topology is initialized, as | ||
1872 | TESTBED will start some CORE services but not necessarily all (but it | ||
1873 | relies on all of them running). The easiest way is to set | ||
1874 | 'FORCESTART = YES' in the '[core]' section of the configuration file. | ||
1875 | Alternatively, having any service that directly or indirectly depends on | ||
1876 | CORE being started with FORCESTART will also do. This issue largely arises | ||
1877 | if users try to over-optimize by not starting any services with | ||
1878 | FORCESTART. | ||
1879 | |||
1880 | @c *********************************************************************** | ||
1881 | @node ATS must want the connections | ||
1882 | @subsubsection ATS must want the connections | ||
1883 | |||
1884 | When TESTBED sets up connections, it only offers the respective HELLO | ||
1885 | information to the TRANSPORT service. It is then up to the ATS service to | ||
1886 | @strong{decide} to use the connection. The ATS service will typically | ||
1887 | eagerly establish any connection if the number of total connections is | ||
1888 | low (relative to bandwidth). Details may further depend on the | ||
1889 | specific ATS backend that was configured. If ATS decides to NOT establish | ||
1890 | a connection (even though TESTBED provided the required information), then | ||
1891 | that connection will count as failed for TESTBED. Note that you can | ||
1892 | configure TESTBED to tolerate a certain number of connection failures | ||
1893 | (see '-e' option of gnunet-testbed-profiler). This issue largely arises | ||
1894 | for dense overlay topologies, especially if you try to create cliques | ||
1895 | with more than 20 peers. | ||
1896 | |||
1897 | @cindex libgnunetutil | ||
1898 | @node libgnunetutil | ||
1899 | @section libgnunetutil | ||
1900 | |||
1901 | libgnunetutil is the fundamental library that all GNUnet code builds upon. | ||
1902 | Ideally, this library should contain most of the platform dependent code | ||
1903 | (except for user interfaces and really special needs that only few | ||
1904 | applications have). It is also supposed to offer basic services that most | ||
1905 | if not all GNUnet binaries require. The code of libgnunetutil is in the | ||
1906 | @file{src/util/} directory. The public interface to the library is in the | ||
1907 | gnunet_util.h header. The functions provided by libgnunetutil fall | ||
1908 | roughly into the following categories (in roughly the order of importance | ||
1909 | for new developers): | ||
1910 | |||
1911 | @itemize @bullet | ||
1912 | @item logging (common_logging.c) | ||
1913 | @item memory allocation (common_allocation.c) | ||
1914 | @item endianess conversion (common_endian.c) | ||
1915 | @item internationalization (common_gettext.c) | ||
1916 | @item String manipulation (string.c) | ||
1917 | @item file access (disk.c) | ||
1918 | @item buffered disk IO (bio.c) | ||
1919 | @item time manipulation (time.c) | ||
1920 | @item configuration parsing (configuration.c) | ||
1921 | @item command-line handling (getopt*.c) | ||
1922 | @item cryptography (crypto_*.c) | ||
1923 | @item data structures (container_*.c) | ||
1924 | @item CPS-style scheduling (scheduler.c) | ||
1925 | @item Program initialization (program.c) | ||
1926 | @item Networking (network.c, client.c, server*.c, service.c) | ||
1927 | @item message queueing (mq.c) | ||
1928 | @item bandwidth calculations (bandwidth.c) | ||
1929 | @item Other OS-related (os*.c, plugin.c, signal.c) | ||
1930 | @item Pseudonym management (pseudonym.c) | ||
1931 | @end itemize | ||
1932 | |||
1933 | It should be noted that only developers that fully understand this entire | ||
1934 | API will be able to write good GNUnet code. | ||
1935 | |||
1936 | Ideally, porting GNUnet should only require porting the gnunetutil | ||
1937 | library. More testcases for the gnunetutil APIs are therefore a great | ||
1938 | way to make porting of GNUnet easier. | ||
1939 | |||
1940 | @menu | ||
1941 | * Logging:: | ||
1942 | * Interprocess communication API (IPC):: | ||
1943 | * Cryptography API:: | ||
1944 | * Message Queue API:: | ||
1945 | * Service API:: | ||
1946 | * Optimizing Memory Consumption of GNUnet's (Multi-) Hash Maps:: | ||
1947 | * The CONTAINER_MDLL API:: | ||
1948 | @end menu | ||
1949 | |||
1950 | @cindex Logging | ||
1951 | @cindex log levels | ||
1952 | @node Logging | ||
1953 | @subsection Logging | ||
1954 | |||
1955 | GNUnet is able to log its activity, mostly for the purposes of debugging | ||
1956 | the program at various levels. | ||
1957 | |||
1958 | @file{gnunet_common.h} defines several @strong{log levels}: | ||
1959 | @table @asis | ||
1960 | |||
1961 | @item ERROR for errors (really problematic situations, often leading to | ||
1962 | crashes) | ||
1963 | @item WARNING for warnings (troubling situations that might have | ||
1964 | negative consequences, although not fatal) | ||
1965 | @item INFO for various information. | ||
1966 | Used somewhat rarely, as GNUnet statistics is used to hold and display | ||
1967 | most of the information that users might find interesting. | ||
1968 | @item DEBUG for debugging. | ||
1969 | Does not produce much output on normal builds, but when extra logging is | ||
1970 | enabled at compile time, a staggering amount of data is outputted under | ||
1971 | this log level. | ||
1972 | @end table | ||
1973 | |||
1974 | |||
1975 | Normal builds of GNUnet (configured with @code{--enable-logging[=yes]}) | ||
1976 | are supposed to log nothing under DEBUG level. The | ||
1977 | @code{--enable-logging=verbose} configure option can be used to create a | ||
1978 | build with all logging enabled. However, such build will produce large | ||
1979 | amounts of log data, which is inconvenient when one tries to hunt down a | ||
1980 | specific problem. | ||
1981 | |||
1982 | To mitigate this problem, GNUnet provides facilities to apply a filter to | ||
1983 | reduce the logs: | ||
1984 | @table @asis | ||
1985 | |||
1986 | @item Logging by default When no log levels are configured in any other | ||
1987 | way (see below), GNUnet will default to the WARNING log level. This | ||
1988 | mostly applies to GNUnet command line utilities, services and daemons; | ||
1989 | tests will always set log level to WARNING or, if | ||
1990 | @code{--enable-logging=verbose} was passed to configure, to DEBUG. The | ||
1991 | default level is suggested for normal operation. | ||
1992 | @item The -L option Most GNUnet executables accept an "-L loglevel" or | ||
1993 | "--log=loglevel" option. If used, it makes the process set a global log | ||
1994 | level to "loglevel". Thus it is possible to run some processes | ||
1995 | with -L DEBUG, for example, and others with -L ERROR to enable specific | ||
1996 | settings to diagnose problems with a particular process. | ||
1997 | @item Configuration files. Because GNUnet | ||
1998 | service and deamon processes are usually launched by gnunet-arm, it is not | ||
1999 | possible to pass different custom command line options directly to every | ||
2000 | one of them. The options passed to @code{gnunet-arm} only affect | ||
2001 | gnunet-arm and not the rest of GNUnet. However, one can specify a | ||
2002 | configuration key "OPTIONS" in the section that corresponds to a service | ||
2003 | or a daemon, and put a value of "-L loglevel" there. This will make the | ||
2004 | respective service or daemon set its log level to "loglevel" (as the | ||
2005 | value of OPTIONS will be passed as a command-line argument). | ||
2006 | |||
2007 | To specify the same log level for all services without creating separate | ||
2008 | "OPTIONS" entries in the configuration for each one, the user can specify | ||
2009 | a config key "GLOBAL_POSTFIX" in the [arm] section of the configuration | ||
2010 | file. The value of GLOBAL_POSTFIX will be appended to all command lines | ||
2011 | used by the ARM service to run other services. It can contain any option | ||
2012 | valid for all GNUnet commands, thus in particular the "-L loglevel" | ||
2013 | option. The ARM service itself is, however, unaffected by GLOBAL_POSTFIX; | ||
2014 | to set log level for it, one has to specify "OPTIONS" key in the [arm] | ||
2015 | section. | ||
2016 | @item Environment variables. | ||
2017 | Setting global per-process log levels with "-L loglevel" does not offer | ||
2018 | sufficient log filtering granularity, as one service will call interface | ||
2019 | libraries and supporting libraries of other GNUnet services, potentially | ||
2020 | producing lots of debug log messages from these libraries. Also, changing | ||
2021 | the config file is not always convenient (especially when running the | ||
2022 | GNUnet test suite).@ To fix that, and to allow GNUnet to use different | ||
2023 | log filtering at runtime without re-compiling the whole source tree, the | ||
2024 | log calls were changed to be configurable at run time. To configure them | ||
2025 | one has to define environment variables "GNUNET_FORCE_LOGFILE", | ||
2026 | "GNUNET_LOG" and/or "GNUNET_FORCE_LOG": | ||
2027 | @itemize @bullet | ||
2028 | |||
2029 | @item "GNUNET_LOG" only affects the logging when no global log level is | ||
2030 | configured by any other means (that is, the process does not explicitly | ||
2031 | set its own log level, there are no "-L loglevel" options on command line | ||
2032 | or in configuration files), and can be used to override the default | ||
2033 | WARNING log level. | ||
2034 | |||
2035 | @item "GNUNET_FORCE_LOG" will completely override any other log | ||
2036 | configuration options given. | ||
2037 | |||
2038 | @item "GNUNET_FORCE_LOGFILE" will completely override the location of the | ||
2039 | file to log messages to. It should contain a relative or absolute file | ||
2040 | name. Setting GNUNET_FORCE_LOGFILE is equivalent to passing | ||
2041 | "--log-file=logfile" or "-l logfile" option (see below). It supports "[]" | ||
2042 | format in file names, but not "@{@}" (see below). | ||
2043 | @end itemize | ||
2044 | |||
2045 | |||
2046 | Because environment variables are inherited by child processes when they | ||
2047 | are launched, starting or re-starting the ARM service with these | ||
2048 | variables will propagate them to all other services. | ||
2049 | |||
2050 | "GNUNET_LOG" and "GNUNET_FORCE_LOG" variables must contain a specially | ||
2051 | formatted @strong{logging definition} string, which looks like this:@ | ||
2052 | |||
2053 | @c FIXME: Can we close this with [/component] instead? | ||
2054 | @example | ||
2055 | [component];[file];[function];[from_line[-to_line]];loglevel[/component...] | ||
2056 | @end example | ||
2057 | |||
2058 | That is, a logging definition consists of definition entries, separated by | ||
2059 | slashes ('/'). If only one entry is present, there is no need to add a | ||
2060 | slash to its end (although it is not forbidden either).@ All definition | ||
2061 | fields (component, file, function, lines and loglevel) are mandatory, but | ||
2062 | (except for the loglevel) they can be empty. An empty field means | ||
2063 | "match anything". Note that even if fields are empty, the semicolon (';') | ||
2064 | separators must be present.@ The loglevel field is mandatory, and must | ||
2065 | contain one of the log level names (ERROR, WARNING, INFO or DEBUG).@ | ||
2066 | The lines field might contain one non-negative number, in which case it | ||
2067 | matches only one line, or a range "from_line-to_line", in which case it | ||
2068 | matches any line in the interval [from_line;to_line] (that is, including | ||
2069 | both start and end line).@ GNUnet mostly defaults component name to the | ||
2070 | name of the service that is implemented in a process ('transport', | ||
2071 | 'core', 'peerinfo', etc), but logging calls can specify custom component | ||
2072 | names using @code{GNUNET_log_from}.@ File name and function name are | ||
2073 | provided by the compiler (__FILE__ and __FUNCTION__ built-ins). | ||
2074 | |||
2075 | Component, file and function fields are interpreted as non-extended | ||
2076 | regular expressions (GNU libc regex functions are used). Matching is | ||
2077 | case-sensitive, "^" and "$" will match the beginning and the end of the | ||
2078 | text. If a field is empty, its contents are automatically replaced with | ||
2079 | a ".*" regular expression, which matches anything. Matching is done in | ||
2080 | the default way, which means that the expression matches as long as it's | ||
2081 | contained anywhere in the string. Thus "GNUNET_" will match both | ||
2082 | "GNUNET_foo" and "BAR_GNUNET_BAZ". Use '^' and/or '$' to make sure that | ||
2083 | the expression matches at the start and/or at the end of the string. | ||
2084 | The semicolon (';') can't be escaped, and GNUnet will not use it in | ||
2085 | component names (it can't be used in function names and file names | ||
2086 | anyway). | ||
2087 | |||
2088 | @end table | ||
2089 | |||
2090 | |||
2091 | Every logging call in GNUnet code will be (at run time) matched against | ||
2092 | the log definitions passed to the process. If a log definition fields are | ||
2093 | matching the call arguments, then the call log level is compared the the | ||
2094 | log level of that definition. If the call log level is less or equal to | ||
2095 | the definition log level, the call is allowed to proceed. Otherwise the | ||
2096 | logging call is forbidden, and nothing is logged. If no definitions | ||
2097 | matched at all, GNUnet will use the global log level or (if a global log | ||
2098 | level is not specified) will default to WARNING (that is, it will allow | ||
2099 | the call to proceed, if its level is less or equal to the global log | ||
2100 | level or to WARNING). | ||
2101 | |||
2102 | That is, definitions are evaluated from left to right, and the first | ||
2103 | matching definition is used to allow or deny the logging call. Thus it is | ||
2104 | advised to place narrow definitions at the beginning of the logdef | ||
2105 | string, and generic definitions - at the end. | ||
2106 | |||
2107 | Whether a call is allowed or not is only decided the first time this | ||
2108 | particular call is made. The evaluation result is then cached, so that | ||
2109 | any attempts to make the same call later will be allowed or disallowed | ||
2110 | right away. Because of that runtime log level evaluation should not | ||
2111 | significantly affect the process performance. | ||
2112 | Log definition parsing is only done once, at the first call to | ||
2113 | GNUNET_log_setup () made by the process (which is usually done soon after | ||
2114 | it starts). | ||
2115 | |||
2116 | At the moment of writing there is no way to specify logging definitions | ||
2117 | from configuration files, only via environment variables. | ||
2118 | |||
2119 | At the moment GNUnet will stop processing a log definition when it | ||
2120 | encounters an error in definition formatting or an error in regular | ||
2121 | expression syntax, and will not report the failure in any way. | ||
2122 | |||
2123 | |||
2124 | @c *********************************************************************** | ||
2125 | @menu | ||
2126 | * Examples:: | ||
2127 | * Log files:: | ||
2128 | * Updated behavior of GNUNET_log:: | ||
2129 | @end menu | ||
2130 | |||
2131 | @node Examples | ||
2132 | @subsubsection Examples | ||
2133 | |||
2134 | @table @asis | ||
2135 | |||
2136 | @item @code{GNUNET_FORCE_LOG=";;;;DEBUG" gnunet-arm -s} Start GNUnet | ||
2137 | process tree, running all processes with DEBUG level (one should be | ||
2138 | careful with it, as log files will grow at alarming rate!) | ||
2139 | @item @code{GNUNET_FORCE_LOG="core;;;;DEBUG" gnunet-arm -s} Start GNUnet | ||
2140 | process tree, running the core service under DEBUG level (everything else | ||
2141 | will use configured or default level). | ||
2142 | |||
2143 | @item Start GNUnet process tree, allowing any logging calls from | ||
2144 | gnunet-service-transport_validation.c (everything else will use | ||
2145 | configured or default level). | ||
2146 | |||
2147 | @example | ||
2148 | GNUNET_FORCE_LOG=";gnunet-service-transport_validation.c;;; DEBUG" \ | ||
2149 | gnunet-arm -s | ||
2150 | @end example | ||
2151 | |||
2152 | @item Start GNUnet process tree, allowing any logging calls from | ||
2153 | gnunet-gnunet-service-fs_push.c (everything else will use configured or | ||
2154 | default level). | ||
2155 | |||
2156 | @example | ||
2157 | GNUNET_FORCE_LOG="fs;gnunet-service-fs_push.c;;;DEBUG" gnunet-arm -s | ||
2158 | @end example | ||
2159 | |||
2160 | @item Start GNUnet process tree, allowing any logging calls from the | ||
2161 | GNUNET_NETWORK_socket_select function (everything else will use | ||
2162 | configured or default level). | ||
2163 | |||
2164 | @example | ||
2165 | GNUNET_FORCE_LOG=";;GNUNET_NETWORK_socket_select;;DEBUG" gnunet-arm -s | ||
2166 | @end example | ||
2167 | |||
2168 | @item Start GNUnet process tree, allowing any logging calls from the | ||
2169 | components that have "transport" in their names, and are made from | ||
2170 | function that have "send" in their names. Everything else will be allowed | ||
2171 | to be logged only if it has WARNING level. | ||
2172 | |||
2173 | @example | ||
2174 | GNUNET_FORCE_LOG="transport.*;;.*send.*;;DEBUG/;;;;WARNING" gnunet-arm -s | ||
2175 | @end example | ||
2176 | |||
2177 | @end table | ||
2178 | |||
2179 | |||
2180 | On Windows, one can use batch files to run GNUnet processes with special | ||
2181 | environment variables, without affecting the whole system. Such batch | ||
2182 | file will look like this: | ||
2183 | |||
2184 | @example | ||
2185 | set GNUNET_FORCE_LOG=;;do_transmit;;DEBUG@ gnunet-arm -s | ||
2186 | @end example | ||
2187 | |||
2188 | (note the absence of double quotes in the environment variable definition, | ||
2189 | as opposed to earlier examples, which use the shell). | ||
2190 | Another limitation, on Windows, GNUNET_FORCE_LOGFILE @strong{MUST} be set | ||
2191 | in order to GNUNET_FORCE_LOG to work. | ||
2192 | |||
2193 | |||
2194 | @cindex Log files | ||
2195 | @node Log files | ||
2196 | @subsubsection Log files | ||
2197 | |||
2198 | GNUnet can be told to log everything into a file instead of stderr (which | ||
2199 | is the default) using the "--log-file=logfile" or "-l logfile" option. | ||
2200 | This option can also be passed via command line, or from the "OPTION" and | ||
2201 | "GLOBAL_POSTFIX" configuration keys (see above). The file name passed | ||
2202 | with this option is subject to GNUnet filename expansion. If specified in | ||
2203 | "GLOBAL_POSTFIX", it is also subject to ARM service filename expansion, | ||
2204 | in particular, it may contain "@{@}" (left and right curly brace) | ||
2205 | sequence, which will be replaced by ARM with the name of the service. | ||
2206 | This is used to keep logs from more than one service separate, while only | ||
2207 | specifying one template containing "@{@}" in GLOBAL_POSTFIX. | ||
2208 | |||
2209 | As part of a secondary file name expansion, the first occurrence of "[]" | ||
2210 | sequence ("left square brace" followed by "right square brace") in the | ||
2211 | file name will be replaced with a process identifier or the process when | ||
2212 | it initializes its logging subsystem. As a result, all processes will log | ||
2213 | into different files. This is convenient for isolating messages of a | ||
2214 | particular process, and prevents I/O races when multiple processes try to | ||
2215 | write into the file at the same time. This expansion is done | ||
2216 | independently of "@{@}" expansion that ARM service does (see above). | ||
2217 | |||
2218 | The log file name that is specified via "-l" can contain format characters | ||
2219 | from the 'strftime' function family. For example, "%Y" will be replaced | ||
2220 | with the current year. Using "basename-%Y-%m-%d.log" would include the | ||
2221 | current year, month and day in the log file. If a GNUnet process runs for | ||
2222 | long enough to need more than one log file, it will eventually clean up | ||
2223 | old log files. Currently, only the last three log files (plus the current | ||
2224 | log file) are preserved. So once the fifth log file goes into use (so | ||
2225 | after 4 days if you use "%Y-%m-%d" as above), the first log file will be | ||
2226 | automatically deleted. Note that if your log file name only contains "%Y", | ||
2227 | then log files would be kept for 4 years and the logs from the first year | ||
2228 | would be deleted once year 5 begins. If you do not use any date-related | ||
2229 | string format codes, logs would never be automatically deleted by GNUnet. | ||
2230 | |||
2231 | |||
2232 | @c *********************************************************************** | ||
2233 | |||
2234 | @node Updated behavior of GNUNET_log | ||
2235 | @subsubsection Updated behavior of GNUNET_log | ||
2236 | |||
2237 | It's currently quite common to see constructions like this all over the | ||
2238 | code: | ||
2239 | |||
2240 | @example | ||
2241 | #if MESH_DEBUG | ||
2242 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "MESH: client disconnected\n"); | ||
2243 | #endif | ||
2244 | @end example | ||
2245 | |||
2246 | The reason for the #if is not to avoid displaying the message when | ||
2247 | disabled (GNUNET_ERROR_TYPE takes care of that), but to avoid the | ||
2248 | compiler including it in the binary at all, when compiling GNUnet for | ||
2249 | platforms with restricted storage space / memory (MIPS routers, | ||
2250 | ARM plug computers / dev boards, etc). | ||
2251 | |||
2252 | This presents several problems: the code gets ugly, hard to write and it | ||
2253 | is very easy to forget to include the #if guards, creating non-consistent | ||
2254 | code. A new change in GNUNET_log aims to solve these problems. | ||
2255 | |||
2256 | @strong{This change requires to @file{./configure} with at least | ||
2257 | @code{--enable-logging=verbose} to see debug messages.} | ||
2258 | |||
2259 | Here is an example of code with dense debug statements: | ||
2260 | |||
2261 | @example | ||
2262 | switch (restrict_topology) @{ | ||
2263 | case GNUNET_TESTING_TOPOLOGY_CLIQUE:#if VERBOSE_TESTING | ||
2264 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, _("Blacklisting all but clique | ||
2265 | topology\n")); #endif unblacklisted_connections = create_clique (pg, | ||
2266 | &remove_connections, BLACKLIST, GNUNET_NO); break; case | ||
2267 | GNUNET_TESTING_TOPOLOGY_SMALL_WORLD_RING: #if VERBOSE_TESTING GNUNET_log | ||
2268 | (GNUNET_ERROR_TYPE_DEBUG, _("Blacklisting all but small world (ring) | ||
2269 | topology\n")); #endif unblacklisted_connections = create_small_world_ring | ||
2270 | (pg,&remove_connections, BLACKLIST); break; | ||
2271 | @end example | ||
2272 | |||
2273 | |||
2274 | Pretty hard to follow, huh? | ||
2275 | |||
2276 | From now on, it is not necessary to include the #if / #endif statements to | ||
2277 | achieve the same behavior. The GNUNET_log and GNUNET_log_from macros take | ||
2278 | care of it for you, depending on the configure option: | ||
2279 | |||
2280 | @itemize @bullet | ||
2281 | @item If @code{--enable-logging} is set to @code{no}, the binary will | ||
2282 | contain no log messages at all. | ||
2283 | @item If @code{--enable-logging} is set to @code{yes}, the binary will | ||
2284 | contain no DEBUG messages, and therefore running with -L DEBUG will have | ||
2285 | no effect. Other messages (ERROR, WARNING, INFO, etc) will be included. | ||
2286 | @item If @code{--enable-logging} is set to @code{verbose}, or | ||
2287 | @code{veryverbose} the binary will contain DEBUG messages (still, it will | ||
2288 | be neccessary to run with -L DEBUG or set the DEBUG config option to show | ||
2289 | them). | ||
2290 | @end itemize | ||
2291 | |||
2292 | |||
2293 | If you are a developer: | ||
2294 | @itemize @bullet | ||
2295 | @item please make sure that you @code{./configure | ||
2296 | --enable-logging=@{verbose,veryverbose@}}, so you can see DEBUG messages. | ||
2297 | @item please remove the @code{#if} statements around @code{GNUNET_log | ||
2298 | (GNUNET_ERROR_TYPE_DEBUG, ...)} lines, to improve the readibility of your | ||
2299 | code. | ||
2300 | @end itemize | ||
2301 | |||
2302 | Since now activating DEBUG automatically makes it VERBOSE and activates | ||
2303 | @strong{all} debug messages by default, you probably want to use the | ||
2304 | https://gnunet.org/logging functionality to filter only relevant messages. | ||
2305 | A suitable configuration could be: | ||
2306 | |||
2307 | @example | ||
2308 | $ export GNUNET_FORCE_LOG="^YOUR_SUBSYSTEM$;;;;DEBUG/;;;;WARNING" | ||
2309 | @end example | ||
2310 | |||
2311 | Which will behave almost like enabling DEBUG in that subsytem before the | ||
2312 | change. Of course you can adapt it to your particular needs, this is only | ||
2313 | a quick example. | ||
2314 | |||
2315 | @cindex Interprocess communication API | ||
2316 | @cindex ICP | ||
2317 | @node Interprocess communication API (IPC) | ||
2318 | @subsection Interprocess communication API (IPC) | ||
2319 | |||
2320 | In GNUnet a variety of new message types might be defined and used in | ||
2321 | interprocess communication, in this tutorial we use the | ||
2322 | @code{struct AddressLookupMessage} as a example to introduce how to | ||
2323 | construct our own message type in GNUnet and how to implement the message | ||
2324 | communication between service and client. | ||
2325 | (Here, a client uses the @code{struct AddressLookupMessage} as a request | ||
2326 | to ask the server to return the address of any other peer connecting to | ||
2327 | the service.) | ||
2328 | |||
2329 | |||
2330 | @c *********************************************************************** | ||
2331 | @menu | ||
2332 | * Define new message types:: | ||
2333 | * Define message struct:: | ||
2334 | * Client - Establish connection:: | ||
2335 | * Client - Initialize request message:: | ||
2336 | * Client - Send request and receive response:: | ||
2337 | * Server - Startup service:: | ||
2338 | * Server - Add new handles for specified messages:: | ||
2339 | * Server - Process request message:: | ||
2340 | * Server - Response to client:: | ||
2341 | * Server - Notification of clients:: | ||
2342 | * Conversion between Network Byte Order (Big Endian) and Host Byte Order:: | ||
2343 | @end menu | ||
2344 | |||
2345 | @node Define new message types | ||
2346 | @subsubsection Define new message types | ||
2347 | |||
2348 | First of all, you should define the new message type in | ||
2349 | @file{gnunet_protocols.h}: | ||
2350 | |||
2351 | @example | ||
2352 | // Request to look addresses of peers in server. | ||
2353 | #define GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_LOOKUP 29 | ||
2354 | // Response to the address lookup request. | ||
2355 | #define GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_REPLY 30 | ||
2356 | @end example | ||
2357 | |||
2358 | @c *********************************************************************** | ||
2359 | @node Define message struct | ||
2360 | @subsubsection Define message struct | ||
2361 | |||
2362 | After the type definition, the specified message structure should also be | ||
2363 | described in the header file, e.g. transport.h in our case. | ||
2364 | |||
2365 | @example | ||
2366 | struct AddressLookupMessage @{ | ||
2367 | struct GNUNET_MessageHeader header; | ||
2368 | int32_t numeric_only GNUNET_PACKED; | ||
2369 | struct GNUNET_TIME_AbsoluteNBO timeout; | ||
2370 | uint32_t addrlen GNUNET_PACKED; | ||
2371 | /* followed by 'addrlen' bytes of the actual address, then | ||
2372 | followed by the 0-terminated name of the transport */ @}; | ||
2373 | GNUNET_NETWORK_STRUCT_END | ||
2374 | @end example | ||
2375 | |||
2376 | |||
2377 | Please note @code{GNUNET_NETWORK_STRUCT_BEGIN} and @code{GNUNET_PACKED} | ||
2378 | which both ensure correct alignment when sending structs over the network. | ||
2379 | |||
2380 | @menu | ||
2381 | @end menu | ||
2382 | |||
2383 | @c *********************************************************************** | ||
2384 | @node Client - Establish connection | ||
2385 | @subsubsection Client - Establish connection | ||
2386 | @c %**end of header | ||
2387 | |||
2388 | |||
2389 | At first, on the client side, the underlying API is employed to create a | ||
2390 | new connection to a service, in our example the transport service would be | ||
2391 | connected. | ||
2392 | |||
2393 | @example | ||
2394 | struct GNUNET_CLIENT_Connection *client; | ||
2395 | client = GNUNET_CLIENT_connect ("transport", cfg); | ||
2396 | @end example | ||
2397 | |||
2398 | @c *********************************************************************** | ||
2399 | @node Client - Initialize request message | ||
2400 | @subsubsection Client - Initialize request message | ||
2401 | @c %**end of header | ||
2402 | |||
2403 | When the connection is ready, we initialize the message. In this step, | ||
2404 | all the fields of the message should be properly initialized, namely the | ||
2405 | size, type, and some extra user-defined data, such as timeout, name of | ||
2406 | transport, address and name of transport. | ||
2407 | |||
2408 | @example | ||
2409 | struct AddressLookupMessage *msg; | ||
2410 | size_t len = sizeof (struct AddressLookupMessage) | ||
2411 | + addressLen | ||
2412 | + strlen (nameTrans) | ||
2413 | + 1; | ||
2414 | msg->header->size = htons (len); | ||
2415 | msg->header->type = htons | ||
2416 | (GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_LOOKUP); | ||
2417 | msg->timeout = GNUNET_TIME_absolute_hton (abs_timeout); | ||
2418 | msg->addrlen = htonl (addressLen); | ||
2419 | char *addrbuf = (char *) &msg[1]; | ||
2420 | memcpy (addrbuf, address, addressLen); | ||
2421 | char *tbuf = &addrbuf[addressLen]; | ||
2422 | memcpy (tbuf, nameTrans, strlen (nameTrans) + 1); | ||
2423 | @end example | ||
2424 | |||
2425 | Note that, here the functions @code{htonl}, @code{htons} and | ||
2426 | @code{GNUNET_TIME_absolute_hton} are applied to convert little endian | ||
2427 | into big endian, about the usage of the big/small edian order and the | ||
2428 | corresponding conversion function please refer to Introduction of | ||
2429 | Big Endian and Little Endian. | ||
2430 | |||
2431 | @c *********************************************************************** | ||
2432 | @node Client - Send request and receive response | ||
2433 | @subsubsection Client - Send request and receive response | ||
2434 | @c %**end of header | ||
2435 | |||
2436 | @b{FIXME: This is very outdated, see the tutorial for the current API!} | ||
2437 | |||
2438 | Next, the client would send the constructed message as a request to the | ||
2439 | service and wait for the response from the service. To accomplish this | ||
2440 | goal, there are a number of API calls that can be used. In this example, | ||
2441 | @code{GNUNET_CLIENT_transmit_and_get_response} is chosen as the most | ||
2442 | appropriate function to use. | ||
2443 | |||
2444 | @example | ||
2445 | GNUNET_CLIENT_transmit_and_get_response | ||
2446 | (client, msg->header, timeout, GNUNET_YES, &address_response_processor, | ||
2447 | arp_ctx); | ||
2448 | @end example | ||
2449 | |||
2450 | the argument @code{address_response_processor} is a function with | ||
2451 | @code{GNUNET_CLIENT_MessageHandler} type, which is used to process the | ||
2452 | reply message from the service. | ||
2453 | |||
2454 | @node Server - Startup service | ||
2455 | @subsubsection Server - Startup service | ||
2456 | |||
2457 | After receiving the request message, we run a standard GNUnet service | ||
2458 | startup sequence using @code{GNUNET_SERVICE_run}, as follows, | ||
2459 | |||
2460 | @example | ||
2461 | int main(int argc, char**argv) @{ | ||
2462 | GNUNET_SERVICE_run(argc, argv, "transport" | ||
2463 | GNUNET_SERVICE_OPTION_NONE, &run, NULL)); @} | ||
2464 | @end example | ||
2465 | |||
2466 | @c *********************************************************************** | ||
2467 | @node Server - Add new handles for specified messages | ||
2468 | @subsubsection Server - Add new handles for specified messages | ||
2469 | @c %**end of header | ||
2470 | |||
2471 | in the function above the argument @code{run} is used to initiate | ||
2472 | transport service,and defined like this: | ||
2473 | |||
2474 | @example | ||
2475 | static void run (void *cls, | ||
2476 | struct GNUNET_SERVER_Handle *serv, | ||
2477 | const struct GNUNET_CONFIGURATION_Handle *cfg) @{ | ||
2478 | GNUNET_SERVER_add_handlers (serv, handlers); @} | ||
2479 | @end example | ||
2480 | |||
2481 | |||
2482 | Here, @code{GNUNET_SERVER_add_handlers} must be called in the run | ||
2483 | function to add new handlers in the service. The parameter | ||
2484 | @code{handlers} is a list of @code{struct GNUNET_SERVER_MessageHandler} | ||
2485 | to tell the service which function should be called when a particular | ||
2486 | type of message is received, and should be defined in this way: | ||
2487 | |||
2488 | @example | ||
2489 | static struct GNUNET_SERVER_MessageHandler handlers[] = @{ | ||
2490 | @{&handle_start, | ||
2491 | NULL, | ||
2492 | GNUNET_MESSAGE_TYPE_TRANSPORT_START, | ||
2493 | 0@}, | ||
2494 | @{&handle_send, | ||
2495 | NULL, | ||
2496 | GNUNET_MESSAGE_TYPE_TRANSPORT_SEND, | ||
2497 | 0@}, | ||
2498 | @{&handle_try_connect, | ||
2499 | NULL, | ||
2500 | GNUNET_MESSAGE_TYPE_TRANSPORT_TRY_CONNECT, | ||
2501 | sizeof (struct TryConnectMessage) | ||
2502 | @}, | ||
2503 | @{&handle_address_lookup, | ||
2504 | NULL, | ||
2505 | GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_LOOKUP, | ||
2506 | 0@}, | ||
2507 | @{NULL, | ||
2508 | NULL, | ||
2509 | 0, | ||
2510 | 0@} | ||
2511 | @}; | ||
2512 | @end example | ||
2513 | |||
2514 | |||
2515 | As shown, the first member of the struct in the first area is a callback | ||
2516 | function, which is called to process the specified message types, given | ||
2517 | as the third member. The second parameter is the closure for the callback | ||
2518 | function, which is set to @code{NULL} in most cases, and the last | ||
2519 | parameter is the expected size of the message of this type, usually we | ||
2520 | set it to 0 to accept variable size, for special cases the exact size of | ||
2521 | the specified message also can be set. In addition, the terminator sign | ||
2522 | depicted as @code{@{NULL, NULL, 0, 0@}} is set in the last aera. | ||
2523 | |||
2524 | @c *********************************************************************** | ||
2525 | @node Server - Process request message | ||
2526 | @subsubsection Server - Process request message | ||
2527 | @c %**end of header | ||
2528 | |||
2529 | After the initialization of transport service, the request message would | ||
2530 | be processed. Before handling the main message data, the validity of this | ||
2531 | message should be checked out, e.g., to check whether the size of message | ||
2532 | is correct. | ||
2533 | |||
2534 | @example | ||
2535 | size = ntohs (message->size); | ||
2536 | if (size < sizeof (struct AddressLookupMessage)) @{ | ||
2537 | GNUNET_break_op (0); | ||
2538 | GNUNET_SERVER_receive_done (client, GNUNET_SYSERR); | ||
2539 | return; @} | ||
2540 | @end example | ||
2541 | |||
2542 | |||
2543 | Note that, opposite to the construction method of the request message in | ||
2544 | the client, in the server the function @code{nothl} and @code{ntohs} | ||
2545 | should be employed during the extraction of the data from the message, so | ||
2546 | that the data in big endian order can be converted back into little | ||
2547 | endian order. See more in detail please refer to Introduction of | ||
2548 | Big Endian and Little Endian. | ||
2549 | |||
2550 | Moreover in this example, the name of the transport stored in the message | ||
2551 | is a 0-terminated string, so we should also check whether the name of the | ||
2552 | transport in the received message is 0-terminated: | ||
2553 | |||
2554 | @example | ||
2555 | nameTransport = (const char *) &address[addressLen]; | ||
2556 | if (nameTransport[size - sizeof | ||
2557 | (struct AddressLookupMessage) | ||
2558 | - addressLen - 1] != '\0') @{ | ||
2559 | GNUNET_break_op (0); | ||
2560 | GNUNET_SERVER_receive_done (client, | ||
2561 | GNUNET_SYSERR); | ||
2562 | return; @} | ||
2563 | @end example | ||
2564 | |||
2565 | Here, @code{GNUNET_SERVER_receive_done} should be called to tell the | ||
2566 | service that the request is done and can receive the next message. The | ||
2567 | argument @code{GNUNET_SYSERR} here indicates that the service didn't | ||
2568 | understand the request message, and the processing of this request would | ||
2569 | be terminated. | ||
2570 | |||
2571 | In comparison to the aforementioned situation, when the argument is equal | ||
2572 | to @code{GNUNET_OK}, the service would continue to process the requst | ||
2573 | message. | ||
2574 | |||
2575 | @c *********************************************************************** | ||
2576 | @node Server - Response to client | ||
2577 | @subsubsection Server - Response to client | ||
2578 | @c %**end of header | ||
2579 | |||
2580 | Once the processing of current request is done, the server should give the | ||
2581 | response to the client. A new @code{struct AddressLookupMessage} would be | ||
2582 | produced by the server in a similar way as the client did and sent to the | ||
2583 | client, but here the type should be | ||
2584 | @code{GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_REPLY} rather than | ||
2585 | @code{GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_LOOKUP} in client. | ||
2586 | @example | ||
2587 | struct AddressLookupMessage *msg; | ||
2588 | size_t len = sizeof (struct AddressLookupMessage) | ||
2589 | + addressLen | ||
2590 | + strlen (nameTrans) + 1; | ||
2591 | msg->header->size = htons (len); | ||
2592 | msg->header->type = htons | ||
2593 | (GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_REPLY); | ||
2594 | |||
2595 | // ... | ||
2596 | |||
2597 | struct GNUNET_SERVER_TransmitContext *tc; | ||
2598 | tc = GNUNET_SERVER_transmit_context_create (client); | ||
2599 | GNUNET_SERVER_transmit_context_append_data | ||
2600 | (tc, | ||
2601 | NULL, | ||
2602 | 0, | ||
2603 | GNUNET_MESSAGE_TYPE_TRANSPORT_ADDRESS_REPLY); | ||
2604 | GNUNET_SERVER_transmit_context_run (tc, rtimeout); | ||
2605 | @end example | ||
2606 | |||
2607 | |||
2608 | Note that, there are also a number of other APIs provided to the service | ||
2609 | to send the message. | ||
2610 | |||
2611 | @c *********************************************************************** | ||
2612 | @node Server - Notification of clients | ||
2613 | @subsubsection Server - Notification of clients | ||
2614 | @c %**end of header | ||
2615 | |||
2616 | Often a service needs to (repeatedly) transmit notifications to a client | ||
2617 | or a group of clients. In these cases, the client typically has once | ||
2618 | registered for a set of events and then needs to receive a message | ||
2619 | whenever such an event happens (until the client disconnects). The use of | ||
2620 | a notification context can help manage message queues to clients and | ||
2621 | handle disconnects. Notification contexts can be used to send | ||
2622 | individualized messages to a particular client or to broadcast messages | ||
2623 | to a group of clients. An individualized notification might look like | ||
2624 | this: | ||
2625 | |||
2626 | @example | ||
2627 | GNUNET_SERVER_notification_context_unicast(nc, | ||
2628 | client, | ||
2629 | msg, | ||
2630 | GNUNET_YES); | ||
2631 | @end example | ||
2632 | |||
2633 | |||
2634 | Note that after processing the original registration message for | ||
2635 | notifications, the server code still typically needs to call | ||
2636 | @code{GNUNET_SERVER_receive_done} so that the client can transmit further | ||
2637 | messages to the server. | ||
2638 | |||
2639 | @c *********************************************************************** | ||
2640 | @node Conversion between Network Byte Order (Big Endian) and Host Byte Order | ||
2641 | @subsubsection Conversion between Network Byte Order (Big Endian) and Host Byte Order | ||
2642 | @c %** subsub? it's a referenced page on the ipc document. | ||
2643 | @c %**end of header | ||
2644 | |||
2645 | Here we can simply comprehend big endian and little endian as Network Byte | ||
2646 | Order and Host Byte Order respectively. What is the difference between | ||
2647 | both two? | ||
2648 | |||
2649 | Usually in our host computer we store the data byte as Host Byte Order, | ||
2650 | for example, we store a integer in the RAM which might occupies 4 Byte, | ||
2651 | as Host Byte Order the higher Byte would be stored at the lower address | ||
2652 | of RAM, and the lower Byte would be stored at the higher address of RAM. | ||
2653 | However, contrast to this, Network Byte Order just take the totally | ||
2654 | opposite way to store the data, says, it will store the lower Byte at the | ||
2655 | lower address, and the higher Byte will stay at higher address. | ||
2656 | |||
2657 | For the current communication of network, we normally exchange the | ||
2658 | information by surveying the data package, every two host wants to | ||
2659 | communicate with each other must send and receive data package through | ||
2660 | network. In order to maintain the identity of data through the | ||
2661 | transmission in the network, the order of the Byte storage must changed | ||
2662 | before sending and after receiving the data. | ||
2663 | |||
2664 | There ten convenient functions to realize the conversion of Byte Order in | ||
2665 | GNUnet, as following: | ||
2666 | |||
2667 | @table @asis | ||
2668 | |||
2669 | @item uint16_t htons(uint16_t hostshort) Convert host byte order to net | ||
2670 | byte order with short int | ||
2671 | @item uint32_t htonl(uint32_t hostlong) Convert host byte | ||
2672 | order to net byte order with long int | ||
2673 | @item uint16_t ntohs(uint16_t netshort) | ||
2674 | Convert net byte order to host byte order with short int | ||
2675 | @item uint32_t | ||
2676 | ntohl(uint32_t netlong) Convert net byte order to host byte order with | ||
2677 | long int | ||
2678 | @item unsigned long long GNUNET_ntohll (unsigned long long netlonglong) | ||
2679 | Convert net byte order to host byte order with long long int | ||
2680 | @item unsigned long long GNUNET_htonll (unsigned long long hostlonglong) | ||
2681 | Convert host byte order to net byte order with long long int | ||
2682 | @item struct GNUNET_TIME_RelativeNBO GNUNET_TIME_relative_hton | ||
2683 | (struct GNUNET_TIME_Relative a) Convert relative time to network byte | ||
2684 | order. | ||
2685 | @item struct GNUNET_TIME_Relative GNUNET_TIME_relative_ntoh | ||
2686 | (struct GNUNET_TIME_RelativeNBO a) Convert relative time from network | ||
2687 | byte order. | ||
2688 | @item struct GNUNET_TIME_AbsoluteNBO GNUNET_TIME_absolute_hton | ||
2689 | (struct GNUNET_TIME_Absolute a) Convert relative time to network byte | ||
2690 | order. | ||
2691 | @item struct GNUNET_TIME_Absolute GNUNET_TIME_absolute_ntoh | ||
2692 | (struct GNUNET_TIME_AbsoluteNBO a) Convert relative time from network | ||
2693 | byte order. | ||
2694 | @end table | ||
2695 | |||
2696 | @cindex Cryptography API | ||
2697 | @node Cryptography API | ||
2698 | @subsection Cryptography API | ||
2699 | @c %**end of header | ||
2700 | |||
2701 | The gnunetutil APIs provides the cryptographic primitives used in GNUnet. | ||
2702 | GNUnet uses 2048 bit RSA keys for the session key exchange and for signing | ||
2703 | messages by peers and most other public-key operations. Most researchers | ||
2704 | in cryptography consider 2048 bit RSA keys as secure and practically | ||
2705 | unbreakable for a long time. The API provides functions to create a fresh | ||
2706 | key pair, read a private key from a file (or create a new file if the | ||
2707 | file does not exist), encrypt, decrypt, sign, verify and extraction of | ||
2708 | the public key into a format suitable for network transmission. | ||
2709 | |||
2710 | For the encryption of files and the actual data exchanged between peers | ||
2711 | GNUnet uses 256-bit AES encryption. Fresh, session keys are negotiated | ||
2712 | for every new connection.@ Again, there is no published technique to | ||
2713 | break this cipher in any realistic amount of time. The API provides | ||
2714 | functions for generation of keys, validation of keys (important for | ||
2715 | checking that decryptions using RSA succeeded), encryption and decryption. | ||
2716 | |||
2717 | GNUnet uses SHA-512 for computing one-way hash codes. The API provides | ||
2718 | functions to compute a hash over a block in memory or over a file on disk. | ||
2719 | |||
2720 | The crypto API also provides functions for randomizing a block of memory, | ||
2721 | obtaining a single random number and for generating a permuation of the | ||
2722 | numbers 0 to n-1. Random number generation distinguishes between WEAK and | ||
2723 | STRONG random number quality; WEAK random numbers are pseudo-random | ||
2724 | whereas STRONG random numbers use entropy gathered from the operating | ||
2725 | system. | ||
2726 | |||
2727 | Finally, the crypto API provides a means to deterministically generate a | ||
2728 | 1024-bit RSA key from a hash code. These functions should most likely not | ||
2729 | be used by most applications; most importantly, | ||
2730 | GNUNET_CRYPTO_rsa_key_create_from_hash does not create an RSA-key that | ||
2731 | should be considered secure for traditional applications of RSA. | ||
2732 | |||
2733 | @cindex Message Queue API | ||
2734 | @node Message Queue API | ||
2735 | @subsection Message Queue API | ||
2736 | @c %**end of header | ||
2737 | |||
2738 | @strong{ Introduction }@ | ||
2739 | Often, applications need to queue messages that | ||
2740 | are to be sent to other GNUnet peers, clients or services. As all of | ||
2741 | GNUnet's message-based communication APIs, by design, do not allow | ||
2742 | messages to be queued, it is common to implement custom message queues | ||
2743 | manually when they are needed. However, writing very similar code in | ||
2744 | multiple places is tedious and leads to code duplication. | ||
2745 | |||
2746 | MQ (for Message Queue) is an API that provides the functionality to | ||
2747 | implement and use message queues. We intend to eventually replace all of | ||
2748 | the custom message queue implementations in GNUnet with MQ. | ||
2749 | |||
2750 | @strong{ Basic Concepts }@ | ||
2751 | The two most important entities in MQ are queues and envelopes. | ||
2752 | |||
2753 | Every queue is backed by a specific implementation (e.g. for mesh, stream, | ||
2754 | connection, server client, etc.) that will actually deliver the queued | ||
2755 | messages. For convenience,@ some queues also allow to specify a list of | ||
2756 | message handlers. The message queue will then also wait for incoming | ||
2757 | messages and dispatch them appropriately. | ||
2758 | |||
2759 | An envelope holds the the memory for a message, as well as metadata | ||
2760 | (Where is the envelope queued? What should happen after it has been | ||
2761 | sent?). Any envelope can only be queued in one message queue. | ||
2762 | |||
2763 | @strong{ Creating Queues }@ | ||
2764 | The following is a list of currently available message queues. Note that | ||
2765 | to avoid layering issues, message queues for higher level APIs are not | ||
2766 | part of @code{libgnunetutil}, but@ the respective API itself provides the | ||
2767 | queue implementation. | ||
2768 | |||
2769 | @table @asis | ||
2770 | |||
2771 | @item @code{GNUNET_MQ_queue_for_connection_client} | ||
2772 | Transmits queued messages over a @code{GNUNET_CLIENT_Connection} handle. | ||
2773 | Also supports receiving with message handlers. | ||
2774 | |||
2775 | @item @code{GNUNET_MQ_queue_for_server_client} | ||
2776 | Transmits queued messages over a @code{GNUNET_SERVER_Client} handle. Does | ||
2777 | not support incoming message handlers. | ||
2778 | |||
2779 | @item @code{GNUNET_MESH_mq_create} Transmits queued messages over a | ||
2780 | @code{GNUNET_MESH_Tunnel} handle. Does not support incoming message | ||
2781 | handlers. | ||
2782 | |||
2783 | @item @code{GNUNET_MQ_queue_for_callbacks} This is the most general | ||
2784 | implementation. Instead of delivering and receiving messages with one of | ||
2785 | GNUnet's communication APIs, implementation callbacks are called. Refer to | ||
2786 | "Implementing Queues" for a more detailed explanation. | ||
2787 | @end table | ||
2788 | |||
2789 | |||
2790 | @strong{ Allocating Envelopes }@ | ||
2791 | A GNUnet message (as defined by the GNUNET_MessageHeader) has three | ||
2792 | parts: The size, the type, and the body. | ||
2793 | |||
2794 | MQ provides macros to allocate an envelope containing a message | ||
2795 | conveniently, automatically setting the size and type fields of the | ||
2796 | message. | ||
2797 | |||
2798 | Consider the following simple message, with the body consisting of a | ||
2799 | single number value. | ||
2800 | @c why the empy code function? | ||
2801 | @code{} | ||
2802 | |||
2803 | @example | ||
2804 | struct NumberMessage @{ | ||
2805 | /** Type: GNUNET_MESSAGE_TYPE_EXAMPLE_1 */ | ||
2806 | struct GNUNET_MessageHeader header; | ||
2807 | uint32_t number GNUNET_PACKED; | ||
2808 | @}; | ||
2809 | @end example | ||
2810 | |||
2811 | An envelope containing an instance of the NumberMessage can be | ||
2812 | constructed like this: | ||
2813 | |||
2814 | @example | ||
2815 | struct GNUNET_MQ_Envelope *ev; | ||
2816 | struct NumberMessage *msg; | ||
2817 | ev = GNUNET_MQ_msg (msg, GNUNET_MESSAGE_TYPE_EXAMPLE_1); | ||
2818 | msg->number = htonl (42); | ||
2819 | @end example | ||
2820 | |||
2821 | In the above code, @code{GNUNET_MQ_msg} is a macro. The return value is | ||
2822 | the newly allocated envelope. The first argument must be a pointer to some | ||
2823 | @code{struct} containing a @code{struct GNUNET_MessageHeader header} | ||
2824 | field, while the second argument is the desired message type, in host | ||
2825 | byte order. | ||
2826 | |||
2827 | The @code{msg} pointer now points to an allocated message, where the | ||
2828 | message type and the message size are already set. The message's size is | ||
2829 | inferred from the type of the @code{msg} pointer: It will be set to | ||
2830 | 'sizeof(*msg)', properly converted to network byte order. | ||
2831 | |||
2832 | If the message body's size is dynamic, the the macro | ||
2833 | @code{GNUNET_MQ_msg_extra} can be used to allocate an envelope whose | ||
2834 | message has additional space allocated after the @code{msg} structure. | ||
2835 | |||
2836 | If no structure has been defined for the message, | ||
2837 | @code{GNUNET_MQ_msg_header_extra} can be used to allocate additional space | ||
2838 | after the message header. The first argument then must be a pointer to a | ||
2839 | @code{GNUNET_MessageHeader}. | ||
2840 | |||
2841 | @strong{Envelope Properties}@ | ||
2842 | A few functions in MQ allow to set additional properties on envelopes: | ||
2843 | |||
2844 | @table @asis | ||
2845 | |||
2846 | @item @code{GNUNET_MQ_notify_sent} Allows to specify a function that will | ||
2847 | be called once the envelope's message has been sent irrevocably. | ||
2848 | An envelope can be canceled precisely up to the@ point where the notify | ||
2849 | sent callback has been called. | ||
2850 | |||
2851 | @item @code{GNUNET_MQ_disable_corking} No corking will be used when | ||
2852 | sending the message. Not every@ queue supports this flag, per default, | ||
2853 | envelopes are sent with corking.@ | ||
2854 | |||
2855 | @end table | ||
2856 | |||
2857 | |||
2858 | @strong{Sending Envelopes}@ | ||
2859 | Once an envelope has been constructed, it can be queued for sending with | ||
2860 | @code{GNUNET_MQ_send}. | ||
2861 | |||
2862 | Note that in order to avoid memory leaks, an envelope must either be sent | ||
2863 | (the queue will free it) or destroyed explicitly with | ||
2864 | @code{GNUNET_MQ_discard}. | ||
2865 | |||
2866 | @strong{Canceling Envelopes}@ | ||
2867 | An envelope queued with @code{GNUNET_MQ_send} can be canceled with | ||
2868 | @code{GNUNET_MQ_cancel}. Note that after the notify sent callback has | ||
2869 | been called, canceling a message results in undefined behavior. | ||
2870 | Thus it is unsafe to cancel an envelope that does not have a notify sent | ||
2871 | callback. When canceling an envelope, it is not necessary@ to call | ||
2872 | @code{GNUNET_MQ_discard}, and the envelope can't be sent again. | ||
2873 | |||
2874 | @strong{ Implementing Queues }@ | ||
2875 | @code{TODO} | ||
2876 | |||
2877 | @cindex Service API | ||
2878 | @node Service API | ||
2879 | @subsection Service API | ||
2880 | @c %**end of header | ||
2881 | |||
2882 | Most GNUnet code lives in the form of services. Services are processes | ||
2883 | that offer an API for other components of the system to build on. Those | ||
2884 | other components can be command-line tools for users, graphical user | ||
2885 | interfaces or other services. Services provide their API using an IPC | ||
2886 | protocol. For this, each service must listen on either a TCP port or a | ||
2887 | UNIX domain socket; for this, the service implementation uses the server | ||
2888 | API. This use of server is exposed directly to the users of the service | ||
2889 | API. Thus, when using the service API, one is usually also often using | ||
2890 | large parts of the server API. The service API provides various | ||
2891 | convenience functions, such as parsing command-line arguments and the | ||
2892 | configuration file, which are not found in the server API. | ||
2893 | The dual to the service/server API is the client API, which can be used to | ||
2894 | access services. | ||
2895 | |||
2896 | The most common way to start a service is to use the | ||
2897 | @code{GNUNET_SERVICE_run} function from the program's main function. | ||
2898 | @code{GNUNET_SERVICE_run} will then parse the command line and | ||
2899 | configuration files and, based on the options found there, | ||
2900 | start the server. It will then give back control to the main | ||
2901 | program, passing the server and the configuration to the | ||
2902 | @code{GNUNET_SERVICE_Main} callback. @code{GNUNET_SERVICE_run} | ||
2903 | will also take care of starting the scheduler loop. | ||
2904 | If this is inappropriate (for example, because the scheduler loop | ||
2905 | is already running), @code{GNUNET_SERVICE_start} and | ||
2906 | related functions provide an alternative to @code{GNUNET_SERVICE_run}. | ||
2907 | |||
2908 | When starting a service, the service_name option is used to determine | ||
2909 | which sections in the configuration file should be used to configure the | ||
2910 | service. A typical value here is the name of the @file{src/} | ||
2911 | sub-directory, for example "@file{statistics}". | ||
2912 | The same string would also be given to | ||
2913 | @code{GNUNET_CLIENT_connect} to access the service. | ||
2914 | |||
2915 | Once a service has been initialized, the program should use the | ||
2916 | @code{GNUNET_SERVICE_Main} callback to register message handlers | ||
2917 | using @code{GNUNET_SERVER_add_handlers}. | ||
2918 | The service will already have registered a handler for the | ||
2919 | "TEST" message. | ||
2920 | |||
2921 | @fnindex GNUNET_SERVICE_Options | ||
2922 | The option bitfield (@code{enum GNUNET_SERVICE_Options}) | ||
2923 | determines how a service should behave during shutdown. | ||
2924 | There are three key strategies: | ||
2925 | |||
2926 | @table @asis | ||
2927 | |||
2928 | @item instant (@code{GNUNET_SERVICE_OPTION_NONE}) | ||
2929 | Upon receiving the shutdown | ||
2930 | signal from the scheduler, the service immediately terminates the server, | ||
2931 | closing all existing connections with clients. | ||
2932 | @item manual (@code{GNUNET_SERVICE_OPTION_MANUAL_SHUTDOWN}) | ||
2933 | The service does nothing by itself | ||
2934 | during shutdown. The main program will need to take the appropriate | ||
2935 | action by calling GNUNET_SERVER_destroy or GNUNET_SERVICE_stop (depending | ||
2936 | on how the service was initialized) to terminate the service. This method | ||
2937 | is used by gnunet-service-arm and rather uncommon. | ||
2938 | @item soft (@code{GNUNET_SERVICE_OPTION_SOFT_SHUTDOWN}) | ||
2939 | Upon receiving the shutdown signal from the scheduler, | ||
2940 | the service immediately tells the server to stop | ||
2941 | listening for incoming clients. Requests from normal existing clients are | ||
2942 | still processed and the server/service terminates once all normal clients | ||
2943 | have disconnected. Clients that are not expected to ever disconnect (such | ||
2944 | as clients that monitor performance values) can be marked as 'monitor' | ||
2945 | clients using GNUNET_SERVER_client_mark_monitor. Those clients will | ||
2946 | continue to be processed until all 'normal' clients have disconnected. | ||
2947 | Then, the server will terminate, closing the monitor connections. | ||
2948 | This mode is for example used by 'statistics', allowing existing 'normal' | ||
2949 | clients to set (possibly persistent) statistic values before terminating. | ||
2950 | |||
2951 | @end table | ||
2952 | |||
2953 | @c *********************************************************************** | ||
2954 | @node Optimizing Memory Consumption of GNUnet's (Multi-) Hash Maps | ||
2955 | @subsection Optimizing Memory Consumption of GNUnet's (Multi-) Hash Maps | ||
2956 | @c %**end of header | ||
2957 | |||
2958 | A commonly used data structure in GNUnet is a (multi-)hash map. It is most | ||
2959 | often used to map a peer identity to some data structure, but also to map | ||
2960 | arbitrary keys to values (for example to track requests in the distributed | ||
2961 | hash table or in file-sharing). As it is commonly used, the DHT is | ||
2962 | actually sometimes responsible for a large share of GNUnet's overall | ||
2963 | memory consumption (for some processes, 30% is not uncommon). The | ||
2964 | following text documents some API quirks (and their implications for | ||
2965 | applications) that were recently introduced to minimize the footprint of | ||
2966 | the hash map. | ||
2967 | |||
2968 | |||
2969 | @c *********************************************************************** | ||
2970 | @menu | ||
2971 | * Analysis:: | ||
2972 | * Solution:: | ||
2973 | * Migration:: | ||
2974 | * Conclusion:: | ||
2975 | * Availability:: | ||
2976 | @end menu | ||
2977 | |||
2978 | @node Analysis | ||
2979 | @subsubsection Analysis | ||
2980 | @c %**end of header | ||
2981 | |||
2982 | The main reason for the "excessive" memory consumption by the hash map is | ||
2983 | that GNUnet uses 512-bit cryptographic hash codes --- and the | ||
2984 | (multi-)hash map also uses the same 512-bit 'struct GNUNET_HashCode'. As | ||
2985 | a result, storing just the keys requires 64 bytes of memory for each key. | ||
2986 | As some applications like to keep a large number of entries in the hash | ||
2987 | map (after all, that's what maps are good for), 64 bytes per hash is | ||
2988 | significant: keeping a pointer to the value and having a linked list for | ||
2989 | collisions consume between 8 and 16 bytes, and 'malloc' may add about the | ||
2990 | same overhead per allocation, putting us in the 16 to 32 byte per entry | ||
2991 | ballpark. Adding a 64-byte key then triples the overall memory | ||
2992 | requirement for the hash map. | ||
2993 | |||
2994 | To make things "worse", most of the time storing the key in the hash map | ||
2995 | is not required: it is typically already in memory elsewhere! In most | ||
2996 | cases, the values stored in the hash map are some application-specific | ||
2997 | struct that _also_ contains the hash. Here is a simplified example: | ||
2998 | |||
2999 | @example | ||
3000 | struct MyValue @{ | ||
3001 | struct GNUNET_HashCode key; | ||
3002 | unsigned int my_data; @}; | ||
3003 | |||
3004 | // ... | ||
3005 | val = GNUNET_malloc (sizeof (struct MyValue)); | ||
3006 | val->key = key; | ||
3007 | val->my_data = 42; | ||
3008 | GNUNET_CONTAINER_multihashmap_put (map, &key, val, ...); | ||
3009 | @end example | ||
3010 | |||
3011 | This is a common pattern as later the entries might need to be removed, | ||
3012 | and at that time it is convenient to have the key immediately at hand: | ||
3013 | |||
3014 | @example | ||
3015 | GNUNET_CONTAINER_multihashmap_remove (map, &val->key, val); | ||
3016 | @end example | ||
3017 | |||
3018 | |||
3019 | Note that here we end up with two times 64 bytes for the key, plus maybe | ||
3020 | 64 bytes total for the rest of the 'struct MyValue' and the map entry in | ||
3021 | the hash map. The resulting redundant storage of the key increases | ||
3022 | overall memory consumption per entry from the "optimal" 128 bytes to 192 | ||
3023 | bytes. This is not just an extreme example: overheads in practice are | ||
3024 | actually sometimes close to those highlighted in this example. This is | ||
3025 | especially true for maps with a significant number of entries, as there | ||
3026 | we tend to really try to keep the entries small. | ||
3027 | |||
3028 | @c *********************************************************************** | ||
3029 | @node Solution | ||
3030 | @subsubsection Solution | ||
3031 | @c %**end of header | ||
3032 | |||
3033 | The solution that has now been implemented is to @strong{optionally} | ||
3034 | allow the hash map to not make a (deep) copy of the hash but instead have | ||
3035 | a pointer to the hash/key in the entry. This reduces the memory | ||
3036 | consumption for the key from 64 bytes to 4 to 8 bytes. However, it can | ||
3037 | also only work if the key is actually stored in the entry (which is the | ||
3038 | case most of the time) and if the entry does not modify the key (which in | ||
3039 | all of the code I'm aware of has been always the case if there key is | ||
3040 | stored in the entry). Finally, when the client stores an entry in the | ||
3041 | hash map, it @strong{must} provide a pointer to the key within the entry, | ||
3042 | not just a pointer to a transient location of the key. If | ||
3043 | the client code does not meet these requirements, the result is a dangling | ||
3044 | pointer and undefined behavior of the (multi-)hash map API. | ||
3045 | |||
3046 | @c *********************************************************************** | ||
3047 | @node Migration | ||
3048 | @subsubsection Migration | ||
3049 | @c %**end of header | ||
3050 | |||
3051 | To use the new feature, first check that the values contain the respective | ||
3052 | key (and never modify it). Then, all calls to | ||
3053 | @code{GNUNET_CONTAINER_multihashmap_put} on the respective map must be | ||
3054 | audited and most likely changed to pass a pointer into the value's struct. | ||
3055 | For the initial example, the new code would look like this: | ||
3056 | |||
3057 | @example | ||
3058 | struct MyValue @{ | ||
3059 | struct GNUNET_HashCode key; | ||
3060 | unsigned int my_data; @}; | ||
3061 | |||
3062 | // ... | ||
3063 | val = GNUNET_malloc (sizeof (struct MyValue)); | ||
3064 | val->key = key; val->my_data = 42; | ||
3065 | GNUNET_CONTAINER_multihashmap_put (map, &val->key, val, ...); | ||
3066 | @end example | ||
3067 | |||
3068 | |||
3069 | Note that @code{&val} was changed to @code{&val->key} in the argument to | ||
3070 | the @code{put} call. This is critical as often @code{key} is on the stack | ||
3071 | or in some other transient data structure and thus having the hash map | ||
3072 | keep a pointer to @code{key} would not work. Only the key inside of | ||
3073 | @code{val} has the same lifetime as the entry in the map (this must of | ||
3074 | course be checked as well). Naturally, @code{val->key} must be | ||
3075 | intiialized before the @code{put} call. Once all @code{put} calls have | ||
3076 | been converted and double-checked, you can change the call to create the | ||
3077 | hash map from | ||
3078 | |||
3079 | @example | ||
3080 | map = | ||
3081 | GNUNET_CONTAINER_multihashmap_create (SIZE, GNUNET_NO); | ||
3082 | @end example | ||
3083 | |||
3084 | to | ||
3085 | |||
3086 | @example | ||
3087 | map = GNUNET_CONTAINER_multihashmap_create (SIZE, GNUNET_YES); | ||
3088 | @end example | ||
3089 | |||
3090 | If everything was done correctly, you now use about 60 bytes less memory | ||
3091 | per entry in @code{map}. However, if now (or in the future) any call to | ||
3092 | @code{put} does not ensure that the given key is valid until the entry is | ||
3093 | removed from the map, undefined behavior is likely to be observed. | ||
3094 | |||
3095 | @c *********************************************************************** | ||
3096 | @node Conclusion | ||
3097 | @subsubsection Conclusion | ||
3098 | @c %**end of header | ||
3099 | |||
3100 | The new optimization can is often applicable and can result in a | ||
3101 | reduction in memory consumption of up to 30% in practice. However, it | ||
3102 | makes the code less robust as additional invariants are imposed on the | ||
3103 | multi hash map client. Thus applications should refrain from enabling the | ||
3104 | new mode unless the resulting performance increase is deemed significant | ||
3105 | enough. In particular, it should generally not be used in new code (wait | ||
3106 | at least until benchmarks exist). | ||
3107 | |||
3108 | @c *********************************************************************** | ||
3109 | @node Availability | ||
3110 | @subsubsection Availability | ||
3111 | @c %**end of header | ||
3112 | |||
3113 | The new multi hash map code was committed in SVN 24319 (will be in GNUnet | ||
3114 | 0.9.4). Various subsystems (transport, core, dht, file-sharing) were | ||
3115 | previously audited and modified to take advantage of the new capability. | ||
3116 | In particular, memory consumption of the file-sharing service is expected | ||
3117 | to drop by 20-30% due to this change. | ||
3118 | |||
3119 | |||
3120 | @cindex CONTAINER_MDLL API | ||
3121 | @node The CONTAINER_MDLL API | ||
3122 | @subsection The CONTAINER_MDLL API | ||
3123 | @c %**end of header | ||
3124 | |||
3125 | This text documents the GNUNET_CONTAINER_MDLL API. The | ||
3126 | GNUNET_CONTAINER_MDLL API is similar to the GNUNET_CONTAINER_DLL API in | ||
3127 | that it provides operations for the construction and manipulation of | ||
3128 | doubly-linked lists. The key difference to the (simpler) DLL-API is that | ||
3129 | the MDLL-version allows a single element (instance of a "struct") to be | ||
3130 | in multiple linked lists at the same time. | ||
3131 | |||
3132 | Like the DLL API, the MDLL API stores (most of) the data structures for | ||
3133 | the doubly-linked list with the respective elements; only the 'head' and | ||
3134 | 'tail' pointers are stored "elsewhere" --- and the application needs to | ||
3135 | provide the locations of head and tail to each of the calls in the | ||
3136 | MDLL API. The key difference for the MDLL API is that the "next" and | ||
3137 | "previous" pointers in the struct can no longer be simply called "next" | ||
3138 | and "prev" --- after all, the element may be in multiple doubly-linked | ||
3139 | lists, so we cannot just have one "next" and one "prev" pointer! | ||
3140 | |||
3141 | The solution is to have multiple fields that must have a name of the | ||
3142 | format "next_XX" and "prev_XX" where "XX" is the name of one of the | ||
3143 | doubly-linked lists. Here is a simple example: | ||
3144 | |||
3145 | @example | ||
3146 | struct MyMultiListElement @{ | ||
3147 | struct MyMultiListElement *next_ALIST; | ||
3148 | struct MyMultiListElement *prev_ALIST; | ||
3149 | struct MyMultiListElement *next_BLIST; | ||
3150 | struct MyMultiListElement *prev_BLIST; | ||
3151 | void | ||
3152 | *data; | ||
3153 | @}; | ||
3154 | @end example | ||
3155 | |||
3156 | |||
3157 | Note that by convention, we use all-uppercase letters for the list names. | ||
3158 | In addition, the program needs to have a location for the head and tail | ||
3159 | pointers for both lists, for example: | ||
3160 | |||
3161 | @example | ||
3162 | static struct MyMultiListElement *head_ALIST; | ||
3163 | static struct MyMultiListElement *tail_ALIST; | ||
3164 | static struct MyMultiListElement *head_BLIST; | ||
3165 | static struct MyMultiListElement *tail_BLIST; | ||
3166 | @end example | ||
3167 | |||
3168 | |||
3169 | Using the MDLL-macros, we can now insert an element into the ALIST: | ||
3170 | |||
3171 | @example | ||
3172 | GNUNET_CONTAINER_MDLL_insert (ALIST, head_ALIST, tail_ALIST, element); | ||
3173 | @end example | ||
3174 | |||
3175 | |||
3176 | Passing "ALIST" as the first argument to MDLL specifies which of the | ||
3177 | next/prev fields in the 'struct MyMultiListElement' should be used. The | ||
3178 | extra "ALIST" argument and the "_ALIST" in the names of the | ||
3179 | next/prev-members are the only differences between the MDDL and DLL-API. | ||
3180 | Like the DLL-API, the MDLL-API offers functions for inserting (at head, | ||
3181 | at tail, after a given element) and removing elements from the list. | ||
3182 | Iterating over the list should be done by directly accessing the | ||
3183 | "next_XX" and/or "prev_XX" members. | ||
3184 | |||
3185 | @cindex Automatic Restart Manager | ||
3186 | @cindex ARM | ||
3187 | @node The Automatic Restart Manager (ARM) | ||
3188 | @section The Automatic Restart Manager (ARM) | ||
3189 | @c %**end of header | ||
3190 | |||
3191 | GNUnet's Automated Restart Manager (ARM) is the GNUnet service responsible | ||
3192 | for system initialization and service babysitting. ARM starts and halts | ||
3193 | services, detects configuration changes and restarts services impacted by | ||
3194 | the changes as needed. It's also responsible for restarting services in | ||
3195 | case of crashes and is planned to incorporate automatic debugging for | ||
3196 | diagnosing service crashes providing developers insights about crash | ||
3197 | reasons. The purpose of this document is to give GNUnet developer an idea | ||
3198 | about how ARM works and how to interact with it. | ||
3199 | |||
3200 | @menu | ||
3201 | * Basic functionality:: | ||
3202 | * Key configuration options:: | ||
3203 | * ARM - Availability:: | ||
3204 | * Reliability:: | ||
3205 | @end menu | ||
3206 | |||
3207 | @c *********************************************************************** | ||
3208 | @node Basic functionality | ||
3209 | @subsection Basic functionality | ||
3210 | @c %**end of header | ||
3211 | |||
3212 | @itemize @bullet | ||
3213 | @item ARM source code can be found under "src/arm".@ Service processes are | ||
3214 | managed by the functions in "gnunet-service-arm.c" which is controlled | ||
3215 | with "gnunet-arm.c" (main function in that file is ARM's entry point). | ||
3216 | |||
3217 | @item The functions responsible for communicating with ARM , starting and | ||
3218 | stopping services -including ARM service itself- are provided by the | ||
3219 | ARM API "arm_api.c".@ Function: GNUNET_ARM_connect() returns to the caller | ||
3220 | an ARM handle after setting it to the caller's context (configuration and | ||
3221 | scheduler in use). This handle can be used afterwards by the caller to | ||
3222 | communicate with ARM. Functions GNUNET_ARM_start_service() and | ||
3223 | GNUNET_ARM_stop_service() are used for starting and stopping services | ||
3224 | respectively. | ||
3225 | |||
3226 | @item A typical example of using these basic ARM services can be found in | ||
3227 | file test_arm_api.c. The test case connects to ARM, starts it, then uses | ||
3228 | it to start a service "resolver", stops the "resolver" then stops "ARM". | ||
3229 | @end itemize | ||
3230 | |||
3231 | @c *********************************************************************** | ||
3232 | @node Key configuration options | ||
3233 | @subsection Key configuration options | ||
3234 | @c %**end of header | ||
3235 | |||
3236 | Configurations for ARM and services should be available in a .conf file | ||
3237 | (As an example, see test_arm_api_data.conf). When running ARM, the | ||
3238 | configuration file to use should be passed to the command: | ||
3239 | |||
3240 | @example | ||
3241 | $ gnunet-arm -s -c configuration_to_use.conf | ||
3242 | @end example | ||
3243 | |||
3244 | If no configuration is passed, the default configuration file will be used | ||
3245 | (see GNUNET_PREFIX/share/gnunet/defaults.conf which is created from | ||
3246 | contrib/defaults.conf).@ Each of the services is having a section starting | ||
3247 | by the service name between square brackets, for example: "[arm]". | ||
3248 | The following options configure how ARM configures or interacts with the | ||
3249 | various services: | ||
3250 | |||
3251 | @table @asis | ||
3252 | |||
3253 | @item PORT Port number on which the service is listening for incoming TCP | ||
3254 | connections. ARM will start the services should it notice a request at | ||
3255 | this port. | ||
3256 | |||
3257 | @item HOSTNAME Specifies on which host the service is deployed. Note | ||
3258 | that ARM can only start services that are running on the local system | ||
3259 | (but will not check that the hostname matches the local machine name). | ||
3260 | This option is used by the @code{gnunet_client_lib.h} implementation to | ||
3261 | determine which system to connect to. The default is "localhost". | ||
3262 | |||
3263 | @item BINARY The name of the service binary file. | ||
3264 | |||
3265 | @item OPTIONS To be passed to the service. | ||
3266 | |||
3267 | @item PREFIX A command to pre-pend to the actual command, for example, | ||
3268 | running a service with "valgrind" or "gdb" | ||
3269 | |||
3270 | @item DEBUG Run in debug mode (much verbosity). | ||
3271 | |||
3272 | @item AUTOSTART ARM will listen to UNIX domain socket and/or TCP port of | ||
3273 | the service and start the service on-demand. | ||
3274 | |||
3275 | @item FORCESTART ARM will always start this service when the peer | ||
3276 | is started. | ||
3277 | |||
3278 | @item ACCEPT_FROM IPv4 addresses the service accepts connections from. | ||
3279 | |||
3280 | @item ACCEPT_FROM6 IPv6 addresses the service accepts connections from. | ||
3281 | |||
3282 | @end table | ||
3283 | |||
3284 | |||
3285 | Options that impact the operation of ARM overall are in the "[arm]" | ||
3286 | section. ARM is a normal service and has (except for AUTOSTART) all of the | ||
3287 | options that other services do. In addition, ARM has the | ||
3288 | following options: | ||
3289 | |||
3290 | @table @asis | ||
3291 | |||
3292 | @item GLOBAL_PREFIX Command to be pre-pended to all services that are | ||
3293 | going to run. | ||
3294 | |||
3295 | @item GLOBAL_POSTFIX Global option that will be supplied to all the | ||
3296 | services that are going to run. | ||
3297 | |||
3298 | @end table | ||
3299 | |||
3300 | @c *********************************************************************** | ||
3301 | @node ARM - Availability | ||
3302 | @subsection ARM - Availability | ||
3303 | @c %**end of header | ||
3304 | |||
3305 | As mentioned before, one of the features provided by ARM is starting | ||
3306 | services on demand. Consider the example of one service "client" that | ||
3307 | wants to connect to another service a "server". The "client" will ask ARM | ||
3308 | to run the "server". ARM starts the "server". The "server" starts | ||
3309 | listening to incoming connections. The "client" will establish a | ||
3310 | connection with the "server". And then, they will start to communicate | ||
3311 | together.@ One problem with that scheme is that it's slow!@ | ||
3312 | The "client" service wants to communicate with the "server" service at | ||
3313 | once and is not willing wait for it to be started and listening to | ||
3314 | incoming connections before serving its request.@ One solution for that | ||
3315 | problem will be that ARM starts all services as default services. That | ||
3316 | solution will solve the problem, yet, it's not quite practical, for some | ||
3317 | services that are going to be started can never be used or are going to | ||
3318 | be used after a relatively long time.@ | ||
3319 | The approach followed by ARM to solve this problem is as follows: | ||
3320 | |||
3321 | @itemize @bullet | ||
3322 | |||
3323 | @item For each service having a PORT field in the configuration file and | ||
3324 | that is not one of the default services ( a service that accepts incoming | ||
3325 | connections from clients), ARM creates listening sockets for all addresses | ||
3326 | associated with that service. | ||
3327 | |||
3328 | @item The "client" will immediately establish a connection with | ||
3329 | the "server". | ||
3330 | |||
3331 | @item ARM --- pretending to be the "server" --- will listen on the | ||
3332 | respective port and notice the incoming connection from the "client" | ||
3333 | (but not accept it), instead | ||
3334 | |||
3335 | @item Once there is an incoming connection, ARM will start the "server", | ||
3336 | passing on the listen sockets (now, the service is started and can do its | ||
3337 | work). | ||
3338 | |||
3339 | @item Other client services now can directly connect directly to the | ||
3340 | "server". | ||
3341 | |||
3342 | @end itemize | ||
3343 | |||
3344 | @c *********************************************************************** | ||
3345 | @node Reliability | ||
3346 | @subsection Reliability | ||
3347 | |||
3348 | One of the features provided by ARM, is the automatic restart of crashed | ||
3349 | services.@ ARM needs to know which of the running services died. Function | ||
3350 | "gnunet-service-arm.c/maint_child_death()" is responsible for that. The | ||
3351 | function is scheduled to run upon receiving a SIGCHLD signal. The | ||
3352 | function, then, iterates ARM's list of services running and monitors | ||
3353 | which service has died (crashed). For all crashing services, ARM restarts | ||
3354 | them.@ | ||
3355 | Now, considering the case of a service having a serious problem causing it | ||
3356 | to crash each time it's started by ARM. If ARM keeps blindly restarting | ||
3357 | such a service, we are going to have the pattern: | ||
3358 | start-crash-restart-crash-restart-crash and so forth!! Which is of course | ||
3359 | not practical.@ | ||
3360 | For that reason, ARM schedules the service to be restarted after waiting | ||
3361 | for some delay that grows exponentially with each crash/restart of that | ||
3362 | service.@ To clarify the idea, considering the following example: | ||
3363 | |||
3364 | @itemize @bullet | ||
3365 | |||
3366 | @item Service S crashed. | ||
3367 | |||
3368 | @item ARM receives the SIGCHLD and inspects its list of services to find | ||
3369 | the dead one(s). | ||
3370 | |||
3371 | @item ARM finds S dead and schedules it for restarting after "backoff" | ||
3372 | time which is initially set to 1ms. ARM will double the backoff time | ||
3373 | correspondent to S (now backoff(S) = 2ms) | ||
3374 | |||
3375 | @item Because there is a severe problem with S, it crashed again. | ||
3376 | |||
3377 | @item Again ARM receives the SIGCHLD and detects that it's S again that's | ||
3378 | crashed. ARM schedules it for restarting but after its new backoff time | ||
3379 | (which became 2ms), and doubles its backoff time (now backoff(S) = 4). | ||
3380 | |||
3381 | @item and so on, until backoff(S) reaches a certain threshold | ||
3382 | (@code{EXPONENTIAL_BACKOFF_THRESHOLD} is set to half an hour), | ||
3383 | after reaching it, backoff(S) will remain half an hour, | ||
3384 | hence ARM won't be busy for a lot of time trying to restart a | ||
3385 | problematic service. | ||
3386 | @end itemize | ||
3387 | |||
3388 | @cindex TRANSPORT Subsystem | ||
3389 | @node GNUnet's TRANSPORT Subsystem | ||
3390 | @section GNUnet's TRANSPORT Subsystem | ||
3391 | @c %**end of header | ||
3392 | |||
3393 | This chapter documents how the GNUnet transport subsystem works. The | ||
3394 | GNUnet transport subsystem consists of three main components: the | ||
3395 | transport API (the interface used by the rest of the system to access the | ||
3396 | transport service), the transport service itself (most of the interesting | ||
3397 | functions, such as choosing transports, happens here) and the transport | ||
3398 | plugins. A transport plugin is a concrete implementation for how two | ||
3399 | GNUnet peers communicate; many plugins exist, for example for | ||
3400 | communication via TCP, UDP, HTTP, HTTPS and others. Finally, the | ||
3401 | transport subsystem uses supporting code, especially the NAT/UPnP | ||
3402 | library to help with tasks such as NAT traversal. | ||
3403 | |||
3404 | Key tasks of the transport service include: | ||
3405 | |||
3406 | @itemize @bullet | ||
3407 | |||
3408 | @item Create our HELLO message, notify clients and neighbours if our HELLO | ||
3409 | changes (using NAT library as necessary) | ||
3410 | |||
3411 | @item Validate HELLOs from other peers (send PING), allow other peers to | ||
3412 | validate our HELLO's addresses (send PONG) | ||
3413 | |||
3414 | @item Upon request, establish connections to other peers (using address | ||
3415 | selection from ATS subsystem) and maintain them (again using PINGs and | ||
3416 | PONGs) as long as desired | ||
3417 | |||
3418 | @item Accept incoming connections, give ATS service the opportunity to | ||
3419 | switch communication channels | ||
3420 | |||
3421 | @item Notify clients about peers that have connected to us or that have | ||
3422 | been disconnected from us | ||
3423 | |||
3424 | @item If a (stateful) connection goes down unexpectedly (without explicit | ||
3425 | DISCONNECT), quickly attempt to recover (without notifying clients) but do | ||
3426 | notify clients quickly if reconnecting fails | ||
3427 | |||
3428 | @item Send (payload) messages arriving from clients to other peers via | ||
3429 | transport plugins and receive messages from other peers, forwarding | ||
3430 | those to clients | ||
3431 | |||
3432 | @item Enforce inbound traffic limits (using flow-control if it is | ||
3433 | applicable); outbound traffic limits are enforced by CORE, not by us (!) | ||
3434 | |||
3435 | @item Enforce restrictions on P2P connection as specified by the blacklist | ||
3436 | configuration and blacklisting clients | ||
3437 | @end itemize | ||
3438 | |||
3439 | Note that the term "clients" in the list above really refers to the | ||
3440 | GNUnet-CORE service, as CORE is typically the only client of the | ||
3441 | transport service. | ||
3442 | |||
3443 | @menu | ||
3444 | * Address validation protocol:: | ||
3445 | @end menu | ||
3446 | |||
3447 | @node Address validation protocol | ||
3448 | @subsection Address validation protocol | ||
3449 | @c %**end of header | ||
3450 | |||
3451 | This section documents how the GNUnet transport service validates | ||
3452 | connections with other peers. It is a high-level description of the | ||
3453 | protocol necessary to understand the details of the implementation. It | ||
3454 | should be noted that when we talk about PING and PONG messages in this | ||
3455 | section, we refer to transport-level PING and PONG messages, which are | ||
3456 | different from core-level PING and PONG messages (both in implementation | ||
3457 | and function). | ||
3458 | |||
3459 | The goal of transport-level address validation is to minimize the chances | ||
3460 | of a successful man-in-the-middle attack against GNUnet peers on the | ||
3461 | transport level. Such an attack would not allow the adversary to decrypt | ||
3462 | the P2P transmissions, but a successful attacker could at least measure | ||
3463 | traffic volumes and latencies (raising the adversaries capablities by | ||
3464 | those of a global passive adversary in the worst case). The scenarios we | ||
3465 | are concerned about is an attacker, Mallory, giving a HELLO to Alice that | ||
3466 | claims to be for Bob, but contains Mallory's IP address instead of Bobs | ||
3467 | (for some transport). Mallory would then forward the traffic to Bob (by | ||
3468 | initiating a connection to Bob and claiming to be Alice). As a further | ||
3469 | complication, the scheme has to work even if say Alice is behind a NAT | ||
3470 | without traversal support and hence has no address of their own (and thus | ||
3471 | Alice must always initiate the connection to Bob). | ||
3472 | |||
3473 | An additional constraint is that HELLO messages do not contain a | ||
3474 | cryptographic signature since other peers must be able to edit | ||
3475 | (i.e. remove) addresses from the HELLO at any time (this was not true in | ||
3476 | GNUnet 0.8.x). A basic @strong{assumption} is that each peer knows the | ||
3477 | set of possible network addresses that it @strong{might} be reachable | ||
3478 | under (so for example, the external IP address of the NAT plus the LAN | ||
3479 | address(es) with the respective ports). | ||
3480 | |||
3481 | The solution is the following. If Alice wants to validate that a given | ||
3482 | address for Bob is valid (i.e. is actually established @strong{directly} | ||
3483 | with the intended target), it sends a PING message over that connection | ||
3484 | to Bob. Note that in this case, Alice initiated the connection so only | ||
3485 | Alice knows which address was used for sure (Alice maybe behind NAT, so | ||
3486 | whatever address Bob sees may not be an address Alice knows they have). | ||
3487 | Bob | ||
3488 | checks that the address given in the PING is actually one of Bob's | ||
3489 | addresses | ||
3490 | (does not belong to Mallory), and if it is, sends back a PONG (with a | ||
3491 | signature that says that Bob owns/uses the address from the PING). Alice | ||
3492 | checks the signature and is happy if it is valid and the address in the | ||
3493 | PONG is the address Alice used. | ||
3494 | This is similar to the 0.8.x protocol where the HELLO contained a | ||
3495 | signature from Bob for each address used by Bob. | ||
3496 | Here, the purpose code for the signature is | ||
3497 | @code{GNUNET_SIGNATURE_PURPOSE_TRANSPORT_PONG_OWN}. After this, Alice will | ||
3498 | remember Bob's address and consider the address valid for a while (12h in | ||
3499 | the current implementation). Note that after this exchange, Alice only | ||
3500 | considers Bob's address to be valid, the connection itself is not | ||
3501 | considered 'established'. In particular, Alice may have many addresses | ||
3502 | for Bob that Alice considers valid. | ||
3503 | |||
3504 | The PONG message is protected with a nonce/challenge against replay | ||
3505 | attacks and uses an expiration time for the signature (but those are | ||
3506 | almost implementation details). | ||
3507 | |||
3508 | @cindex NAT library | ||
3509 | @node NAT library | ||
3510 | @section NAT library | ||
3511 | @c %**end of header | ||
3512 | |||
3513 | The goal of the GNUnet NAT library is to provide a general-purpose API for | ||
3514 | NAT traversal @strong{without} third-party support. So protocols that | ||
3515 | involve contacting a third peer to help establish a connection between | ||
3516 | two peers are outside of the scope of this API. That does not mean that | ||
3517 | GNUnet doesn't support involving a third peer (we can do this with the | ||
3518 | distance-vector transport or using application-level protocols), it just | ||
3519 | means that the NAT API is not concerned with this possibility. The API is | ||
3520 | written so that it will work for IPv6-NAT in the future as well as | ||
3521 | current IPv4-NAT. Furthermore, the NAT API is always used, even for peers | ||
3522 | that are not behind NAT --- in that case, the mapping provided is simply | ||
3523 | the identity. | ||
3524 | |||
3525 | NAT traversal is initiated by calling @code{GNUNET_NAT_register}. Given a | ||
3526 | set of addresses that the peer has locally bound to (TCP or UDP), the NAT | ||
3527 | library will return (via callback) a (possibly longer) list of addresses | ||
3528 | the peer @strong{might} be reachable under. Internally, depending on the | ||
3529 | configuration, the NAT library will try to punch a hole (using UPnP) or | ||
3530 | just "know" that the NAT was manually punched and generate the respective | ||
3531 | external IP address (the one that should be globally visible) based on | ||
3532 | the given information. | ||
3533 | |||
3534 | The NAT library also supports ICMP-based NAT traversal. Here, the other | ||
3535 | peer can request connection-reversal by this peer (in this special case, | ||
3536 | the peer is even allowed to configure a port number of zero). If the NAT | ||
3537 | library detects a connection-reversal request, it returns the respective | ||
3538 | target address to the client as well. It should be noted that | ||
3539 | connection-reversal is currently only intended for TCP, so other plugins | ||
3540 | @strong{must} pass @code{NULL} for the reversal callback. Naturally, the | ||
3541 | NAT library also supports requesting connection reversal from a remote | ||
3542 | peer (@code{GNUNET_NAT_run_client}). | ||
3543 | |||
3544 | Once initialized, the NAT handle can be used to test if a given address is | ||
3545 | possibly a valid address for this peer (@code{GNUNET_NAT_test_address}). | ||
3546 | This is used for validating our addresses when generating PONGs. | ||
3547 | |||
3548 | Finally, the NAT library contains an API to test if our NAT configuration | ||
3549 | is correct. Using @code{GNUNET_NAT_test_start} @strong{before} binding to | ||
3550 | the respective port, the NAT library can be used to test if the | ||
3551 | configuration works. The test function act as a local client, initialize | ||
3552 | the NAT traversal and then contact a @code{gnunet-nat-server} (running by | ||
3553 | default on @code{gnunet.org}) and ask for a connection to be established. | ||
3554 | This way, it is easy to test if the current NAT configuration is valid. | ||
3555 | |||
3556 | @node Distance-Vector plugin | ||
3557 | @section Distance-Vector plugin | ||
3558 | @c %**end of header | ||
3559 | |||
3560 | The Distance Vector (DV) transport is a transport mechanism that allows | ||
3561 | peers to act as relays for each other, thereby connecting peers that would | ||
3562 | otherwise be unable to connect. This gives a larger connection set to | ||
3563 | applications that may work better with more peers to choose from (for | ||
3564 | example, File Sharing and/or DHT). | ||
3565 | |||
3566 | The Distance Vector transport essentially has two functions. The first is | ||
3567 | "gossiping" connection information about more distant peers to directly | ||
3568 | connected peers. The second is taking messages intended for non-directly | ||
3569 | connected peers and encapsulating them in a DV wrapper that contains the | ||
3570 | required information for routing the message through forwarding peers. Via | ||
3571 | gossiping, optimal routes through the known DV neighborhood are discovered | ||
3572 | and utilized and the message encapsulation provides some benefits in | ||
3573 | addition to simply getting the message from the correct source to the | ||
3574 | proper destination. | ||
3575 | |||
3576 | The gossiping function of DV provides an up to date routing table of | ||
3577 | peers that are available up to some number of hops. We call this a | ||
3578 | fisheye view of the network (like a fish, nearby objects are known while | ||
3579 | more distant ones unknown). Gossip messages are sent only to directly | ||
3580 | connected peers, but they are sent about other knowns peers within the | ||
3581 | "fisheye distance". Whenever two peers connect, they immediately gossip | ||
3582 | to each other about their appropriate other neighbors. They also gossip | ||
3583 | about the newly connected peer to previously | ||
3584 | connected neighbors. In order to keep the routing tables up to date, | ||
3585 | disconnect notifications are propogated as gossip as well (because | ||
3586 | disconnects may not be sent/received, timeouts are also used remove | ||
3587 | stagnant routing table entries). | ||
3588 | |||
3589 | Routing of messages via DV is straightforward. When the DV transport is | ||
3590 | notified of a message destined for a non-direct neighbor, the appropriate | ||
3591 | forwarding peer is selected, and the base message is encapsulated in a DV | ||
3592 | message which contains information about the initial peer and the intended | ||
3593 | recipient. At each forwarding hop, the initial peer is validated (the | ||
3594 | forwarding peer ensures that it has the initial peer in its neighborhood, | ||
3595 | otherwise the message is dropped). Next the base message is | ||
3596 | re-encapsulated in a new DV message for the next hop in the forwarding | ||
3597 | chain (or delivered to the current peer, if it has arrived at the | ||
3598 | destination). | ||
3599 | |||
3600 | Assume a three peer network with peers Alice, Bob and Carol. Assume that | ||
3601 | Alice <-> Bob and Bob <-> Carol are direct (e.g. over TCP or UDP | ||
3602 | transports) connections, but that Alice cannot directly connect to Carol. | ||
3603 | This may be the case due to NAT or firewall restrictions, or perhaps | ||
3604 | based on one of the peers respective configurations. If the Distance | ||
3605 | Vector transport is enabled on all three peers, it will automatically | ||
3606 | discover (from the gossip protocol) that Alice and Carol can connect via | ||
3607 | Bob and provide a "virtual" Alice <-> Carol connection. Routing between | ||
3608 | Alice and Carol happens as follows; Alice creates a message destined for | ||
3609 | Carol and notifies the DV transport about it. The DV transport at Alice | ||
3610 | looks up Carol in the routing table and finds that the message must be | ||
3611 | sent through Bob for Carol. The message is encapsulated setting Alice as | ||
3612 | the initiator and Carol as the destination and sent to Bob. Bob receives | ||
3613 | the messages, verifies both Alice and Carol are known to Bob, and re-wraps | ||
3614 | the message in a new DV message for Carol. The DV transport at Carol | ||
3615 | receives this message, unwraps the original message, and delivers it to | ||
3616 | Carol as though it came directly from Alice. | ||
3617 | |||
3618 | @cindex SMTP plugin | ||
3619 | @node SMTP plugin | ||
3620 | @section SMTP plugin | ||
3621 | @c %**end of header | ||
3622 | |||
3623 | This section describes the new SMTP transport plugin for GNUnet as it | ||
3624 | exists in the 0.7.x and 0.8.x branch. SMTP support is currently not | ||
3625 | available in GNUnet 0.9.x. This page also describes the transport layer | ||
3626 | abstraction (as it existed in 0.7.x and 0.8.x) in more detail and gives | ||
3627 | some benchmarking results. The performance results presented are quite | ||
3628 | old and maybe outdated at this point. | ||
3629 | |||
3630 | @itemize @bullet | ||
3631 | @item Why use SMTP for a peer-to-peer transport? | ||
3632 | @item SMTPHow does it work? | ||
3633 | @item How do I configure my peer? | ||
3634 | @item How do I test if it works? | ||
3635 | @item How fast is it? | ||
3636 | @item Is there any additional documentation? | ||
3637 | @end itemize | ||
3638 | |||
3639 | |||
3640 | @menu | ||
3641 | * Why use SMTP for a peer-to-peer transport?:: | ||
3642 | * How does it work?:: | ||
3643 | * How do I configure my peer?:: | ||
3644 | * How do I test if it works?:: | ||
3645 | * How fast is it?:: | ||
3646 | @end menu | ||
3647 | |||
3648 | @node Why use SMTP for a peer-to-peer transport? | ||
3649 | @subsection Why use SMTP for a peer-to-peer transport? | ||
3650 | @c %**end of header | ||
3651 | |||
3652 | There are many reasons why one would not want to use SMTP: | ||
3653 | |||
3654 | @itemize @bullet | ||
3655 | @item SMTP is using more bandwidth than TCP, UDP or HTTP | ||
3656 | @item SMTP has a much higher latency. | ||
3657 | @item SMTP requires significantly more computation (encoding and decoding | ||
3658 | time) for the peers. | ||
3659 | @item SMTP is significantly more complicated to configure. | ||
3660 | @item SMTP may be abused by tricking GNUnet into sending mail to@ | ||
3661 | non-participating third parties. | ||
3662 | @end itemize | ||
3663 | |||
3664 | So why would anybody want to use SMTP? | ||
3665 | @itemize @bullet | ||
3666 | @item SMTP can be used to contact peers behind NAT boxes (in virtual | ||
3667 | private networks). | ||
3668 | @item SMTP can be used to circumvent policies that limit or prohibit | ||
3669 | peer-to-peer traffic by masking as "legitimate" traffic. | ||
3670 | @item SMTP uses E-mail addresses which are independent of a specific IP, | ||
3671 | which can be useful to address peers that use dynamic IP addresses. | ||
3672 | @item SMTP can be used to initiate a connection (e.g. initial address | ||
3673 | exchange) and peers can then negotiate the use of a more efficient | ||
3674 | protocol (e.g. TCP) for the actual communication. | ||
3675 | @end itemize | ||
3676 | |||
3677 | In summary, SMTP can for example be used to send a message to a peer | ||
3678 | behind a NAT box that has a dynamic IP to tell the peer to establish a | ||
3679 | TCP connection to a peer outside of the private network. Even an | ||
3680 | extraordinary overhead for this first message would be irrelevant in this | ||
3681 | type of situation. | ||
3682 | |||
3683 | @node How does it work? | ||
3684 | @subsection How does it work? | ||
3685 | @c %**end of header | ||
3686 | |||
3687 | When a GNUnet peer needs to send a message to another GNUnet peer that has | ||
3688 | advertised (only) an SMTP transport address, GNUnet base64-encodes the | ||
3689 | message and sends it in an E-mail to the advertised address. The | ||
3690 | advertisement contains a filter which is placed in the E-mail header, | ||
3691 | such that the receiving host can filter the tagged E-mails and forward it | ||
3692 | to the GNUnet peer process. The filter can be specified individually by | ||
3693 | each peer and be changed over time. This makes it impossible to censor | ||
3694 | GNUnet E-mail messages by searching for a generic filter. | ||
3695 | |||
3696 | @node How do I configure my peer? | ||
3697 | @subsection How do I configure my peer? | ||
3698 | @c %**end of header | ||
3699 | |||
3700 | First, you need to configure @code{procmail} to filter your inbound E-mail | ||
3701 | for GNUnet traffic. The GNUnet messages must be delivered into a pipe, for | ||
3702 | example @code{/tmp/gnunet.smtp}. You also need to define a filter that is | ||
3703 | used by @command{procmail} to detect GNUnet messages. You are free to | ||
3704 | choose whichever filter you like, but you should make sure that it does | ||
3705 | not occur in your other E-mail. In our example, we will use | ||
3706 | @code{X-mailer: GNUnet}. The @code{~/.procmailrc} configuration file then | ||
3707 | looks like this: | ||
3708 | |||
3709 | @example | ||
3710 | :0: | ||
3711 | * ^X-mailer: GNUnet | ||
3712 | /tmp/gnunet.smtp | ||
3713 | # where do you want your other e-mail delivered to | ||
3714 | # (default: /var/spool/mail/) | ||
3715 | :0: /var/spool/mail/ | ||
3716 | @end example | ||
3717 | |||
3718 | After adding this file, first make sure that your regular E-mail still | ||
3719 | works (e.g. by sending an E-mail to yourself). Then edit the GNUnet | ||
3720 | configuration. In the section @code{SMTP} you need to specify your E-mail | ||
3721 | address under @code{EMAIL}, your mail server (for outgoing mail) under | ||
3722 | @code{SERVER}, the filter (X-mailer: GNUnet in the example) under | ||
3723 | @code{FILTER} and the name of the pipe under @code{PIPE}.@ The completed | ||
3724 | section could then look like this: | ||
3725 | |||
3726 | @example | ||
3727 | EMAIL = me@@mail.gnu.org MTU = 65000 SERVER = mail.gnu.org:25 FILTER = | ||
3728 | "X-mailer: GNUnet" PIPE = /tmp/gnunet.smtp | ||
3729 | @end example | ||
3730 | |||
3731 | Finally, you need to add @code{smtp} to the list of @code{TRANSPORTS} in | ||
3732 | the @code{GNUNETD} section. GNUnet peers will use the E-mail address that | ||
3733 | you specified to contact your peer until the advertisement times out. | ||
3734 | Thus, if you are not sure if everything works properly or if you are not | ||
3735 | planning to be online for a long time, you may want to configure this | ||
3736 | timeout to be short, e.g. just one hour. For this, set | ||
3737 | @code{HELLOEXPIRES} to @code{1} in the @code{GNUNETD} section. | ||
3738 | |||
3739 | This should be it, but you may probably want to test it first. | ||
3740 | |||
3741 | @node How do I test if it works? | ||
3742 | @subsection How do I test if it works? | ||
3743 | @c %**end of header | ||
3744 | |||
3745 | Any transport can be subjected to some rudimentary tests using the | ||
3746 | @code{gnunet-transport-check} tool. The tool sends a message to the local | ||
3747 | node via the transport and checks that a valid message is received. While | ||
3748 | this test does not involve other peers and can not check if firewalls or | ||
3749 | other network obstacles prohibit proper operation, this is a great | ||
3750 | testcase for the SMTP transport since it tests pretty much nearly all of | ||
3751 | the functionality. | ||
3752 | |||
3753 | @code{gnunet-transport-check} should only be used without running | ||
3754 | @code{gnunetd} at the same time. By default, @code{gnunet-transport-check} | ||
3755 | tests all transports that are specified in the configuration file. But | ||
3756 | you can specifically test SMTP by giving the option | ||
3757 | @code{--transport=smtp}. | ||
3758 | |||
3759 | Note that this test always checks if a transport can receive and send. | ||
3760 | While you can configure most transports to only receive or only send | ||
3761 | messages, this test will only work if you have configured the transport | ||
3762 | to send and receive messages. | ||
3763 | |||
3764 | @node How fast is it? | ||
3765 | @subsection How fast is it? | ||
3766 | @c %**end of header | ||
3767 | |||
3768 | We have measured the performance of the UDP, TCP and SMTP transport layer | ||
3769 | directly and when used from an application using the GNUnet core. | ||
3770 | Measureing just the transport layer gives the better view of the actual | ||
3771 | overhead of the protocol, whereas evaluating the transport from the | ||
3772 | application puts the overhead into perspective from a practical point of | ||
3773 | view. | ||
3774 | |||
3775 | The loopback measurements of the SMTP transport were performed on three | ||
3776 | different machines spanning a range of modern SMTP configurations. We | ||
3777 | used a PIII-800 running RedHat 7.3 with the Purdue Computer Science | ||
3778 | configuration which includes filters for spam. We also used a Xenon 2 GHZ | ||
3779 | with a vanilla RedHat 8.0 sendmail configuration. Furthermore, we used | ||
3780 | qmail on a PIII-1000 running Sorcerer GNU Linux (SGL). The numbers for | ||
3781 | UDP and TCP are provided using the SGL configuration. The qmail benchmark | ||
3782 | uses qmail's internal filtering whereas the sendmail benchmarks relies on | ||
3783 | procmail to filter and deliver the mail. We used the transport layer to | ||
3784 | send a message of b bytes (excluding transport protocol headers) directly | ||
3785 | to the local machine. This way, network latency and packet loss on the | ||
3786 | wire have no impact on the timings. n messages were sent sequentially over | ||
3787 | the transport layer, sending message i+1 after the i-th message was | ||
3788 | received. All messages were sent over the same connection and the time to | ||
3789 | establish the connection was not taken into account since this overhead is | ||
3790 | miniscule in practice --- as long as a connection is used for a | ||
3791 | significant number of messages. | ||
3792 | |||
3793 | @multitable @columnfractions .20 .15 .15 .15 .15 .15 | ||
3794 | @headitem Transport @tab UDP @tab TCP @tab SMTP (Purdue sendmail) | ||
3795 | @tab SMTP (RH 8.0) @tab SMTP (SGL qmail) | ||
3796 | @item 11 bytes @tab 31 ms @tab 55 ms @tab 781 s @tab 77 s @tab 24 s | ||
3797 | @item 407 bytes @tab 37 ms @tab 62 ms @tab 789 s @tab 78 s @tab 25 s | ||
3798 | @item 1,221 bytes @tab 46 ms @tab 73 ms @tab 804 s @tab 78 s @tab 25 s | ||
3799 | @end multitable | ||
3800 | |||
3801 | The benchmarks show that UDP and TCP are, as expected, both significantly | ||
3802 | faster compared with any of the SMTP services. Among the SMTP | ||
3803 | implementations, there can be significant differences depending on the | ||
3804 | SMTP configuration. Filtering with an external tool like procmail that | ||
3805 | needs to re-parse its configuration for each mail can be very expensive. | ||
3806 | Applying spam filters can also significantly impact the performance of | ||
3807 | the underlying SMTP implementation. The microbenchmark shows that SMTP | ||
3808 | can be a viable solution for initiating peer-to-peer sessions: a couple of | ||
3809 | seconds to connect to a peer are probably not even going to be noticed by | ||
3810 | users. The next benchmark measures the possible throughput for a | ||
3811 | transport. Throughput can be measured by sending multiple messages in | ||
3812 | parallel and measuring packet loss. Note that not only UDP but also the | ||
3813 | TCP transport can actually loose messages since the TCP implementation | ||
3814 | drops messages if the @code{write} to the socket would block. While the | ||
3815 | SMTP protocol never drops messages itself, it is often so | ||
3816 | slow that only a fraction of the messages can be sent and received in the | ||
3817 | given time-bounds. For this benchmark we report the message loss after | ||
3818 | allowing t time for sending m messages. If messages were not sent (or | ||
3819 | received) after an overall timeout of t, they were considered lost. The | ||
3820 | benchmark was performed using two Xeon 2 GHZ machines running RedHat 8.0 | ||
3821 | with sendmail. The machines were connected with a direct 100 MBit ethernet | ||
3822 | connection.@ Figures udp1200, tcp1200 and smtp-MTUs show that the | ||
3823 | throughput for messages of size 1,200 octects is 2,343 kbps, 3,310 kbps | ||
3824 | and 6 kbps for UDP, TCP and SMTP respectively. The high per-message | ||
3825 | overhead of SMTP can be improved by increasing the MTU, for example, an | ||
3826 | MTU of 12,000 octets improves the throughput to 13 kbps as figure | ||
3827 | smtp-MTUs shows. Our research paper) has some more details on the | ||
3828 | benchmarking results. | ||
3829 | |||
3830 | @cindex Bluetooth plugin | ||
3831 | @node Bluetooth plugin | ||
3832 | @section Bluetooth plugin | ||
3833 | @c %**end of header | ||
3834 | |||
3835 | This page describes the new Bluetooth transport plugin for GNUnet. The | ||
3836 | plugin is still in the testing stage so don't expect it to work | ||
3837 | perfectly. If you have any questions or problems just post them here or | ||
3838 | ask on the IRC channel. | ||
3839 | |||
3840 | @itemize @bullet | ||
3841 | @item What do I need to use the Bluetooth plugin transport? | ||
3842 | @item BluetoothHow does it work? | ||
3843 | @item What possible errors should I be aware of? | ||
3844 | @item How do I configure my peer? | ||
3845 | @item How can I test it? | ||
3846 | @end itemize | ||
3847 | |||
3848 | @menu | ||
3849 | * What do I need to use the Bluetooth plugin transport?:: | ||
3850 | * How does it work2?:: | ||
3851 | * What possible errors should I be aware of?:: | ||
3852 | * How do I configure my peer2?:: | ||
3853 | * How can I test it?:: | ||
3854 | * The implementation of the Bluetooth transport plugin:: | ||
3855 | @end menu | ||
3856 | |||
3857 | @node What do I need to use the Bluetooth plugin transport? | ||
3858 | @subsection What do I need to use the Bluetooth plugin transport? | ||
3859 | @c %**end of header | ||
3860 | |||
3861 | If you are a Linux user and you want to use the Bluetooth transport plugin | ||
3862 | you should install the BlueZ development libraries (if they aren't already | ||
3863 | installed). For instructions about how to install the libraries you should | ||
3864 | check out the BlueZ site | ||
3865 | (@uref{http://www.bluez.org/, http://www.bluez.org}). If you don't know if | ||
3866 | you have the necesarry libraries, don't worry, just run the GNUnet | ||
3867 | configure script and you will be able to see a notification at the end | ||
3868 | which will warn you if you don't have the necessary libraries. | ||
3869 | |||
3870 | If you are a Windows user you should have installed the | ||
3871 | @emph{MinGW}/@emph{MSys2} with the latest updates (especially the | ||
3872 | @emph{ws2bth} header). If this is your first build of GNUnet on Windows | ||
3873 | you should check out the SBuild repository. It will semi-automatically | ||
3874 | assembles a @emph{MinGW}/@emph{MSys2} installation with a lot of extra | ||
3875 | packages which are needed for the GNUnet build. So this will ease your | ||
3876 | work!@ Finally you just have to be sure that you have the correct drivers | ||
3877 | for your Bluetooth device installed and that your device is on and in a | ||
3878 | discoverable mode. The Windows Bluetooth Stack supports only the RFCOMM | ||
3879 | protocol so we cannot turn on your device programatically! | ||
3880 | |||
3881 | @c FIXME: Change to unique title | ||
3882 | @node How does it work2? | ||
3883 | @subsection How does it work2? | ||
3884 | @c %**end of header | ||
3885 | |||
3886 | The Bluetooth transport plugin uses virtually the same code as the WLAN | ||
3887 | plugin and only the helper binary is different. The helper takes a single | ||
3888 | argument, which represents the interface name and is specified in the | ||
3889 | configuration file. Here are the basic steps that are followed by the | ||
3890 | helper binary used on Linux: | ||
3891 | |||
3892 | @itemize @bullet | ||
3893 | @item it verifies if the name corresponds to a Bluetooth interface name | ||
3894 | @item it verifies if the iterface is up (if it is not, it tries to bring | ||
3895 | it up) | ||
3896 | @item it tries to enable the page and inquiry scan in order to make the | ||
3897 | device discoverable and to accept incoming connection requests | ||
3898 | @emph{The above operations require root access so you should start the | ||
3899 | transport plugin with root privileges.} | ||
3900 | @item it finds an available port number and registers a SDP service which | ||
3901 | will be used to find out on which port number is the server listening on | ||
3902 | and switch the socket in listening mode | ||
3903 | @item it sends a HELLO message with its address | ||
3904 | @item finally it forwards traffic from the reading sockets to the STDOUT | ||
3905 | and from the STDIN to the writing socket | ||
3906 | @end itemize | ||
3907 | |||
3908 | Once in a while the device will make an inquiry scan to discover the | ||
3909 | nearby devices and it will send them randomly HELLO messages for peer | ||
3910 | discovery. | ||
3911 | |||
3912 | @node What possible errors should I be aware of? | ||
3913 | @subsection What possible errors should I be aware of? | ||
3914 | @c %**end of header | ||
3915 | |||
3916 | @emph{This section is dedicated for Linux users} | ||
3917 | |||
3918 | Well there are many ways in which things could go wrong but I will try to | ||
3919 | present some tools that you could use to debug and some scenarios. | ||
3920 | |||
3921 | @itemize @bullet | ||
3922 | |||
3923 | @item @code{bluetoothd -n -d} : use this command to enable logging in the | ||
3924 | foreground and to print the logging messages | ||
3925 | |||
3926 | @item @code{hciconfig}: can be used to configure the Bluetooth devices. | ||
3927 | If you run it without any arguments it will print information about the | ||
3928 | state of the interfaces. So if you receive an error that the device | ||
3929 | couldn't be brought up you should try to bring it manually and to see if | ||
3930 | it works (use @code{hciconfig -a hciX up}). If you can't and the | ||
3931 | Bluetooth address has the form 00:00:00:00:00:00 it means that there is | ||
3932 | something wrong with the D-Bus daemon or with the Bluetooth daemon. Use | ||
3933 | @code{bluetoothd} tool to see the logs | ||
3934 | |||
3935 | @item @code{sdptool} can be used to control and interogate SDP servers. | ||
3936 | If you encounter problems regarding the SDP server (like the SDP server is | ||
3937 | down) you should check out if the D-Bus daemon is running correctly and to | ||
3938 | see if the Bluetooth daemon started correctly(use @code{bluetoothd} tool). | ||
3939 | Also, sometimes the SDP service could work but somehow the device couldn't | ||
3940 | register his service. Use @code{sdptool browse [dev-address]} to see if | ||
3941 | the service is registered. There should be a service with the name of the | ||
3942 | interface and GNUnet as provider. | ||
3943 | |||
3944 | @item @code{hcitool} : another useful tool which can be used to configure | ||
3945 | the device and to send some particular commands to it. | ||
3946 | |||
3947 | @item @code{hcidump} : could be used for low level debugging | ||
3948 | @end itemize | ||
3949 | |||
3950 | @c FIXME: A more unique name | ||
3951 | @node How do I configure my peer2? | ||
3952 | @subsection How do I configure my peer2? | ||
3953 | @c %**end of header | ||
3954 | |||
3955 | On Linux, you just have to be sure that the interface name corresponds to | ||
3956 | the one that you want to use. Use the @code{hciconfig} tool to check that. | ||
3957 | By default it is set to hci0 but you can change it. | ||
3958 | |||
3959 | A basic configuration looks like this: | ||
3960 | |||
3961 | @example | ||
3962 | [transport-bluetooth] | ||
3963 | # Name of the interface (typically hciX) | ||
3964 | INTERFACE = hci0 | ||
3965 | # Real hardware, no testing | ||
3966 | TESTMODE = 0 TESTING_IGNORE_KEYS = ACCEPT_FROM; | ||
3967 | @end example | ||
3968 | |||
3969 | In order to use the Bluetooth transport plugin when the transport service | ||
3970 | is started, you must add the plugin name to the default transport service | ||
3971 | plugins list. For example: | ||
3972 | |||
3973 | @example | ||
3974 | [transport] ... PLUGINS = dns bluetooth ... | ||
3975 | @end example | ||
3976 | |||
3977 | If you want to use only the Bluetooth plugin set | ||
3978 | @emph{PLUGINS = bluetooth} | ||
3979 | |||
3980 | On Windows, you cannot specify which device to use. The only thing that | ||
3981 | you should do is to add @emph{bluetooth} on the plugins list of the | ||
3982 | transport service. | ||
3983 | |||
3984 | @node How can I test it? | ||
3985 | @subsection How can I test it? | ||
3986 | @c %**end of header | ||
3987 | |||
3988 | If you have two Bluetooth devices on the same machine which use Linux you | ||
3989 | must: | ||
3990 | |||
3991 | @itemize @bullet | ||
3992 | |||
3993 | @item create two different file configuration (one which will use the | ||
3994 | first interface (@emph{hci0}) and the other which will use the second | ||
3995 | interface (@emph{hci1})). Let's name them @emph{peer1.conf} and | ||
3996 | @emph{peer2.conf}. | ||
3997 | |||
3998 | @item run @emph{gnunet-peerinfo -c peerX.conf -s} in order to generate the | ||
3999 | peers private keys. The @strong{X} must be replace with 1 or 2. | ||
4000 | |||
4001 | @item run @emph{gnunet-arm -c peerX.conf -s -i=transport} in order to | ||
4002 | start the transport service. (Make sure that you have "bluetooth" on the | ||
4003 | transport plugins list if the Bluetooth transport service doesn't start.) | ||
4004 | |||
4005 | @item run @emph{gnunet-peerinfo -c peer1.conf -s} to get the first peer's | ||
4006 | ID. If you already know your peer ID (you saved it from the first | ||
4007 | command), this can be skipped. | ||
4008 | |||
4009 | @item run @emph{gnunet-transport -c peer2.conf -p=PEER1_ID -s} to start | ||
4010 | sending data for benchmarking to the other peer. | ||
4011 | |||
4012 | @end itemize | ||
4013 | |||
4014 | |||
4015 | This scenario will try to connect the second peer to the first one and | ||
4016 | then start sending data for benchmarking. | ||
4017 | |||
4018 | On Windows you cannot test the plugin functionality using two Bluetooth | ||
4019 | devices from the same machine because after you install the drivers there | ||
4020 | will occur some conflicts between the Bluetooth stacks. (At least that is | ||
4021 | what happend on my machine : I wasn't able to use the Bluesoleil stack and | ||
4022 | the WINDCOMM one in the same time). | ||
4023 | |||
4024 | If you have two different machines and your configuration files are good | ||
4025 | you can use the same scenario presented on the begining of this section. | ||
4026 | |||
4027 | Another way to test the plugin functionality is to create your own | ||
4028 | application which will use the GNUnet framework with the Bluetooth | ||
4029 | transport service. | ||
4030 | |||
4031 | @node The implementation of the Bluetooth transport plugin | ||
4032 | @subsection The implementation of the Bluetooth transport plugin | ||
4033 | @c %**end of header | ||
4034 | |||
4035 | This page describes the implementation of the Bluetooth transport plugin. | ||
4036 | |||
4037 | First I want to remind you that the Bluetooth transport plugin uses | ||
4038 | virtually the same code as the WLAN plugin and only the helper binary is | ||
4039 | different. Also the scope of the helper binary from the Bluetooth | ||
4040 | transport plugin is the same as the one used for the wlan transport | ||
4041 | plugin: it acceses the interface and then it forwards traffic in both | ||
4042 | directions between the Bluetooth interface and stdin/stdout of the | ||
4043 | process involved. | ||
4044 | |||
4045 | The Bluetooth plugin transport could be used both on Linux and Windows | ||
4046 | platforms. | ||
4047 | |||
4048 | @itemize @bullet | ||
4049 | @item Linux functionality | ||
4050 | @item Windows functionality | ||
4051 | @item Pending Features | ||
4052 | @end itemize | ||
4053 | |||
4054 | |||
4055 | |||
4056 | @menu | ||
4057 | * Linux functionality:: | ||
4058 | * THE INITIALIZATION:: | ||
4059 | * THE LOOP:: | ||
4060 | * Details about the broadcast implementation:: | ||
4061 | * Windows functionality:: | ||
4062 | * Pending features:: | ||
4063 | @end menu | ||
4064 | |||
4065 | @node Linux functionality | ||
4066 | @subsubsection Linux functionality | ||
4067 | @c %**end of header | ||
4068 | |||
4069 | In order to implement the plugin functionality on Linux I used the BlueZ | ||
4070 | stack. For the communication with the other devices I used the RFCOMM | ||
4071 | protocol. Also I used the HCI protocol to gain some control over the | ||
4072 | device. The helper binary takes a single argument (the name of the | ||
4073 | Bluetooth interface) and is separated in two stages: | ||
4074 | |||
4075 | @c %** 'THE INITIALIZATION' should be in bigger letters or stand out, not | ||
4076 | @c %** starting a new section? | ||
4077 | @node THE INITIALIZATION | ||
4078 | @subsubsection THE INITIALIZATION | ||
4079 | |||
4080 | @itemize @bullet | ||
4081 | @item first, it checks if we have root privilegies | ||
4082 | (@emph{Remember that we need to have root privilegies in order to be able | ||
4083 | to bring the interface up if it is down or to change its state.}). | ||
4084 | |||
4085 | @item second, it verifies if the interface with the given name exists. | ||
4086 | |||
4087 | @strong{If the interface with that name exists and it is a Bluetooth | ||
4088 | interface:} | ||
4089 | |||
4090 | @item it creates a RFCOMM socket which will be used for listening and call | ||
4091 | the @emph{open_device} method | ||
4092 | |||
4093 | On the @emph{open_device} method: | ||
4094 | @itemize @bullet | ||
4095 | @item creates a HCI socket used to send control events to the the device | ||
4096 | @item searches for the device ID using the interface name | ||
4097 | @item saves the device MAC address | ||
4098 | @item checks if the interface is down and tries to bring it UP | ||
4099 | @item checks if the interface is in discoverable mode and tries to make it | ||
4100 | discoverable | ||
4101 | @item closes the HCI socket and binds the RFCOMM one | ||
4102 | @item switches the RFCOMM socket in listening mode | ||
4103 | @item registers the SDP service (the service will be used by the other | ||
4104 | devices to get the port on which this device is listening on) | ||
4105 | @end itemize | ||
4106 | |||
4107 | @item drops the root privilegies | ||
4108 | |||
4109 | @strong{If the interface is not a Bluetooth interface the helper exits | ||
4110 | with a suitable error} | ||
4111 | @end itemize | ||
4112 | |||
4113 | @c %** Same as for @node entry above | ||
4114 | @node THE LOOP | ||
4115 | @subsubsection THE LOOP | ||
4116 | |||
4117 | The helper binary uses a list where it saves all the connected neighbour | ||
4118 | devices (@emph{neighbours.devices}) and two buffers (@emph{write_pout} and | ||
4119 | @emph{write_std}). The first message which is send is a control message | ||
4120 | with the device's MAC address in order to announce the peer presence to | ||
4121 | the neighbours. Here are a short description of what happens in the main | ||
4122 | loop: | ||
4123 | |||
4124 | @itemize @bullet | ||
4125 | @item Every time when it receives something from the STDIN it processes | ||
4126 | the data and saves the message in the first buffer (@emph{write_pout}). | ||
4127 | When it has something in the buffer, it gets the destination address from | ||
4128 | the buffer, searches the destination address in the list (if there is no | ||
4129 | connection with that device, it creates a new one and saves it to the | ||
4130 | list) and sends the message. | ||
4131 | @item Every time when it receives something on the listening socket it | ||
4132 | accepts the connection and saves the socket on a list with the reading | ||
4133 | sockets. @item Every time when it receives something from a reading | ||
4134 | socket it parses the message, verifies the CRC and saves it in the | ||
4135 | @emph{write_std} buffer in order to be sent later to the STDOUT. | ||
4136 | @end itemize | ||
4137 | |||
4138 | So in the main loop we use the select function to wait until one of the | ||
4139 | file descriptor saved in one of the two file descriptors sets used is | ||
4140 | ready to use. The first set (@emph{rfds}) represents the reading set and | ||
4141 | it could contain the list with the reading sockets, the STDIN file | ||
4142 | descriptor or the listening socket. The second set (@emph{wfds}) is the | ||
4143 | writing set and it could contain the sending socket or the STDOUT file | ||
4144 | descriptor. After the select function returns, we check which file | ||
4145 | descriptor is ready to use and we do what is supposed to do on that kind | ||
4146 | of event. @emph{For example:} if it is the listening socket then we | ||
4147 | accept a new connection and save the socket in the reading list; if it is | ||
4148 | the STDOUT file descriptor, then we write to STDOUT the message from the | ||
4149 | @emph{write_std} buffer. | ||
4150 | |||
4151 | To find out on which port a device is listening on we connect to the local | ||
4152 | SDP server and searche the registered service for that device. | ||
4153 | |||
4154 | @emph{You should be aware of the fact that if the device fails to connect | ||
4155 | to another one when trying to send a message it will attempt one more | ||
4156 | time. If it fails again, then it skips the message.} | ||
4157 | @emph{Also you should know that the transport Bluetooth plugin has | ||
4158 | support for @strong{broadcast messages}.} | ||
4159 | |||
4160 | @node Details about the broadcast implementation | ||
4161 | @subsubsection Details about the broadcast implementation | ||
4162 | @c %**end of header | ||
4163 | |||
4164 | First I want to point out that the broadcast functionality for the CONTROL | ||
4165 | messages is not implemented in a conventional way. Since the inquiry scan | ||
4166 | time is too big and it will take some time to send a message to all the | ||
4167 | discoverable devices I decided to tackle the problem in a different way. | ||
4168 | Here is how I did it: | ||
4169 | |||
4170 | @itemize @bullet | ||
4171 | @item If it is the first time when I have to broadcast a message I make an | ||
4172 | inquiry scan and save all the devices' addresses to a vector. | ||
4173 | @item After the inquiry scan ends I take the first address from the list | ||
4174 | and I try to connect to it. If it fails, I try to connect to the next one. | ||
4175 | If it succeeds, I save the socket to a list and send the message to the | ||
4176 | device. | ||
4177 | @item When I have to broadcast another message, first I search on the list | ||
4178 | for a new device which I'm not connected to. If there is no new device on | ||
4179 | the list I go to the beginning of the list and send the message to the | ||
4180 | old devices. After 5 cycles I make a new inquiry scan to check out if | ||
4181 | there are new discoverable devices and save them to the list. If there | ||
4182 | are no new discoverable devices I reset the cycling counter and go again | ||
4183 | through the old list and send messages to the devices saved in it. | ||
4184 | @end itemize | ||
4185 | |||
4186 | @strong{Therefore}: | ||
4187 | |||
4188 | @itemize @bullet | ||
4189 | @item every time when I have a broadcast message I look up on the list | ||
4190 | for a new device and send the message to it | ||
4191 | @item if I reached the end of the list for 5 times and I'm connected to | ||
4192 | all the devices from the list I make a new inquiry scan. | ||
4193 | @emph{The number of the list's cycles after an inquiry scan could be | ||
4194 | increased by redefining the MAX_LOOPS variable} | ||
4195 | @item when there are no new devices I send messages to the old ones. | ||
4196 | @end itemize | ||
4197 | |||
4198 | Doing so, the broadcast control messages will reach the devices but with | ||
4199 | delay. | ||
4200 | |||
4201 | @emph{NOTICE:} When I have to send a message to a certain device first I | ||
4202 | check on the broadcast list to see if we are connected to that device. If | ||
4203 | not we try to connect to it and in case of success we save the address and | ||
4204 | the socket on the list. If we are already connected to that device we | ||
4205 | simply use the socket. | ||
4206 | |||
4207 | @node Windows functionality | ||
4208 | @subsubsection Windows functionality | ||
4209 | @c %**end of header | ||
4210 | |||
4211 | For Windows I decided to use the Microsoft Bluetooth stack which has the | ||
4212 | advantage of coming standard from Windows XP SP2. The main disadvantage is | ||
4213 | that it only supports the RFCOMM protocol so we will not be able to have | ||
4214 | a low level control over the Bluetooth device. Therefore it is the user | ||
4215 | responsability to check if the device is up and in the discoverable mode. | ||
4216 | Also there are no tools which could be used for debugging in order to read | ||
4217 | the data coming from and going to a Bluetooth device, which obviously | ||
4218 | hindered my work. Another thing that slowed down the implementation of the | ||
4219 | plugin (besides that I wasn't too accomodated with the win32 API) was that | ||
4220 | there were some bugs on MinGW regarding the Bluetooth. Now they are solved | ||
4221 | but you should keep in mind that you should have the latest updates | ||
4222 | (especially the @emph{ws2bth} header). | ||
4223 | |||
4224 | Besides the fact that it uses the Windows Sockets, the Windows | ||
4225 | implemenation follows the same principles as the Linux one: | ||
4226 | |||
4227 | @itemize @bullet | ||
4228 | @item It has a initalization part where it initializes the | ||
4229 | Windows Sockets, creates a RFCOMM socket which will be binded and switched | ||
4230 | to the listening mode and registers a SDP service. In the Microsoft | ||
4231 | Bluetooth API there are two ways to work with the SDP: | ||
4232 | @itemize @bullet | ||
4233 | @item an easy way which works with very simple service records | ||
4234 | @item a hard way which is useful when you need to update or to delete the | ||
4235 | record | ||
4236 | @end itemize | ||
4237 | @end itemize | ||
4238 | |||
4239 | Since I only needed the SDP service to find out on which port the device | ||
4240 | is listening on and that did not change, I decided to use the easy way. | ||
4241 | In order to register the service I used the @emph{WSASetService} function | ||
4242 | and I generated the @emph{Universally Unique Identifier} with the | ||
4243 | @emph{guidgen.exe} Windows's tool. | ||
4244 | |||
4245 | In the loop section the only difference from the Linux implementation is | ||
4246 | that I used the GNUNET_NETWORK library for functions like @emph{accept}, | ||
4247 | @emph{bind}, @emph{connect} or @emph{select}. I decided to use the | ||
4248 | GNUNET_NETWORK library because I also needed to interact with the STDIN | ||
4249 | and STDOUT handles and on Windows the select function is only defined for | ||
4250 | sockets, and it will not work for arbitrary file handles. | ||
4251 | |||
4252 | Another difference between Linux and Windows implementation is that in | ||
4253 | Linux, the Bluetooth address is represented in 48 bits while in Windows is | ||
4254 | represented in 64 bits. Therefore I had to do some changes on | ||
4255 | @emph{plugin_transport_wlan} header. | ||
4256 | |||
4257 | Also, currently on Windows the Bluetooth plugin doesn't have support for | ||
4258 | broadcast messages. When it receives a broadcast message it will skip it. | ||
4259 | |||
4260 | @node Pending features | ||
4261 | @subsubsection Pending features | ||
4262 | @c %**end of header | ||
4263 | |||
4264 | @itemize @bullet | ||
4265 | @item Implement the broadcast functionality on Windows @emph{(currently | ||
4266 | working on)} | ||
4267 | @item Implement a testcase for the helper :@ @emph{The testcase | ||
4268 | consists of a program which emaluates the plugin and uses the helper. It | ||
4269 | will simulate connections, disconnections and data transfers.} | ||
4270 | @end itemize | ||
4271 | |||
4272 | If you have a new idea about a feature of the plugin or suggestions about | ||
4273 | how I could improve the implementation you are welcome to comment or to | ||
4274 | contact me. | ||
4275 | |||
4276 | @node WLAN plugin | ||
4277 | @section WLAN plugin | ||
4278 | @c %**end of header | ||
4279 | |||
4280 | This section documents how the wlan transport plugin works. Parts which | ||
4281 | are not implemented yet or could be better implemented are described at | ||
4282 | the end. | ||
4283 | |||
4284 | @cindex ats subsystem | ||
4285 | @node The ATS Subsystem | ||
4286 | @section The ATS Subsystem | ||
4287 | @c %**end of header | ||
4288 | |||
4289 | ATS stands for "automatic transport selection", and the function of ATS in | ||
4290 | GNUnet is to decide on which address (and thus transport plugin) should | ||
4291 | be used for two peers to communicate, and what bandwidth limits should be | ||
4292 | imposed on such an individual connection. To help ATS make an informed | ||
4293 | decision, higher-level services inform the ATS service about their | ||
4294 | requirements and the quality of the service rendered. The ATS service | ||
4295 | also interacts with the transport service to be appraised of working | ||
4296 | addresses and to communicate its resource allocation decisions. Finally, | ||
4297 | the ATS service's operation can be observed using a monitoring API. | ||
4298 | |||
4299 | The main logic of the ATS service only collects the available addresses, | ||
4300 | their performance characteristics and the applications requirements, but | ||
4301 | does not make the actual allocation decision. This last critical step is | ||
4302 | left to an ATS plugin, as we have implemented (currently three) different | ||
4303 | allocation strategies which differ significantly in their performance and | ||
4304 | maturity, and it is still unclear if any particular plugin is generally | ||
4305 | superior. | ||
4306 | |||
4307 | @cindex core subsystem | ||
4308 | @cindex CORE subsystem | ||
4309 | @node GNUnet's CORE Subsystem | ||
4310 | @section GNUnet's CORE Subsystem | ||
4311 | @c %**end of header | ||
4312 | |||
4313 | The CORE subsystem in GNUnet is responsible for securing link-layer | ||
4314 | communications between nodes in the GNUnet overlay network. CORE builds | ||
4315 | on the TRANSPORT subsystem which provides for the actual, insecure, | ||
4316 | unreliable link-layer communication (for example, via UDP or WLAN), and | ||
4317 | then adds fundamental security to the connections: | ||
4318 | |||
4319 | @itemize @bullet | ||
4320 | @item confidentiality with so-called perfect forward secrecy; we use | ||
4321 | ECDHE@footnote{@uref{http://en.wikipedia.org/wiki/Elliptic_curve_Diffie%E2%80%93Hellman, Elliptic-curve Diffie---Hellman}} | ||
4322 | powered by Curve25519 | ||
4323 | @footnote{@uref{http://cr.yp.to/ecdh.html, Curve25519}} for the key | ||
4324 | exchange and then use symmetric encryption, encrypting with both AES-256 | ||
4325 | @footnote{@uref{http://en.wikipedia.org/wiki/Rijndael, AES-256}} and | ||
4326 | Twofish @footnote{@uref{http://en.wikipedia.org/wiki/Twofish, Twofish}} | ||
4327 | @item @uref{http://en.wikipedia.org/wiki/Authentication, authentication} | ||
4328 | is achieved by signing the ephemeral keys using Ed25519 | ||
4329 | @footnote{@uref{http://ed25519.cr.yp.to/, Ed25519}}, a deterministic | ||
4330 | variant of ECDSA | ||
4331 | @footnote{@uref{http://en.wikipedia.org/wiki/ECDSA, ECDSA}} | ||
4332 | @item integrity protection (using SHA-512 | ||
4333 | @footnote{@uref{http://en.wikipedia.org/wiki/SHA-2, SHA-512}} to do | ||
4334 | encrypt-then-MAC | ||
4335 | @footnote{@uref{http://en.wikipedia.org/wiki/Authenticated_encryption, encrypt-then-MAC}}) | ||
4336 | @item Replay | ||
4337 | @footnote{@uref{http://en.wikipedia.org/wiki/Replay_attack, replay}} | ||
4338 | protection (using nonces, timestamps, challenge-response, | ||
4339 | message counters and ephemeral keys) | ||
4340 | @item liveness (keep-alive messages, timeout) | ||
4341 | @end itemize | ||
4342 | |||
4343 | @menu | ||
4344 | * Limitations:: | ||
4345 | * When is a peer "connected"?:: | ||
4346 | * libgnunetcore:: | ||
4347 | * The CORE Client-Service Protocol:: | ||
4348 | * The CORE Peer-to-Peer Protocol:: | ||
4349 | @end menu | ||
4350 | |||
4351 | @cindex core subsystem limitations | ||
4352 | @node Limitations | ||
4353 | @subsection Limitations | ||
4354 | @c %**end of header | ||
4355 | |||
4356 | CORE does not perform | ||
4357 | @uref{http://en.wikipedia.org/wiki/Routing, routing}; using CORE it is | ||
4358 | only possible to communicate with peers that happen to already be | ||
4359 | "directly" connected with each other. CORE also does not have an | ||
4360 | API to allow applications to establish such "direct" connections --- for | ||
4361 | this, applications can ask TRANSPORT, but TRANSPORT might not be able to | ||
4362 | establish a "direct" connection. The TOPOLOGY subsystem is responsible for | ||
4363 | trying to keep a few "direct" connections open at all times. Applications | ||
4364 | that need to talk to particular peers should use the CADET subsystem, as | ||
4365 | it can establish arbitrary "indirect" connections. | ||
4366 | |||
4367 | Because CORE does not perform routing, CORE must only be used directly by | ||
4368 | applications that either perform their own routing logic (such as | ||
4369 | anonymous file-sharing) or that do not require routing, for example | ||
4370 | because they are based on flooding the network. CORE communication is | ||
4371 | unreliable and delivery is possibly out-of-order. Applications that | ||
4372 | require reliable communication should use the CADET service. Each | ||
4373 | application can only queue one message per target peer with the CORE | ||
4374 | service at any time; messages cannot be larger than approximately | ||
4375 | 63 kilobytes. If messages are small, CORE may group multiple messages | ||
4376 | (possibly from different applications) prior to encryption. If permitted | ||
4377 | by the application (using the @uref{http://baus.net/on-tcp_cork/, cork} | ||
4378 | option), CORE may delay transmissions to facilitate grouping of multiple | ||
4379 | small messages. If cork is not enabled, CORE will transmit the message as | ||
4380 | soon as TRANSPORT allows it (TRANSPORT is responsible for limiting | ||
4381 | bandwidth and congestion control). CORE does not allow flow control; | ||
4382 | applications are expected to process messages at line-speed. If flow | ||
4383 | control is needed, applications should use the CADET service. | ||
4384 | |||
4385 | @cindex when is a peer connected | ||
4386 | @node When is a peer "connected"? | ||
4387 | @subsection When is a peer "connected"? | ||
4388 | @c %**end of header | ||
4389 | |||
4390 | In addition to the security features mentioned above, CORE also provides | ||
4391 | one additional key feature to applications using it, and that is a | ||
4392 | limited form of protocol-compatibility checking. CORE distinguishes | ||
4393 | between TRANSPORT-level connections (which enable communication with other | ||
4394 | peers) and application-level connections. Applications using the CORE API | ||
4395 | will (typically) learn about application-level connections from CORE, and | ||
4396 | not about TRANSPORT-level connections. When a typical application uses | ||
4397 | CORE, it will specify a set of message types | ||
4398 | (from @code{gnunet_protocols.h}) that it understands. CORE will then | ||
4399 | notify the application about connections it has with other peers if and | ||
4400 | only if those applications registered an intersecting set of message | ||
4401 | types with their CORE service. Thus, it is quite possible that CORE only | ||
4402 | exposes a subset of the established direct connections to a particular | ||
4403 | application --- and different applications running above CORE might see | ||
4404 | different sets of connections at the same time. | ||
4405 | |||
4406 | A special case are applications that do not register a handler for any | ||
4407 | message type. | ||
4408 | CORE assumes that these applications merely want to monitor connections | ||
4409 | (or "all" messages via other callbacks) and will notify those applications | ||
4410 | about all connections. This is used, for example, by the | ||
4411 | @code{gnunet-core} command-line tool to display the active connections. | ||
4412 | Note that it is also possible that the TRANSPORT service has more active | ||
4413 | connections than the CORE service, as the CORE service first has to | ||
4414 | perform a key exchange with connecting peers before exchanging information | ||
4415 | about supported message types and notifying applications about the new | ||
4416 | connection. | ||
4417 | |||
4418 | @cindex libgnunetcore | ||
4419 | @node libgnunetcore | ||
4420 | @subsection libgnunetcore | ||
4421 | @c %**end of header | ||
4422 | |||
4423 | The CORE API (defined in @file{gnunet_core_service.h}) is the basic | ||
4424 | messaging API used by P2P applications built using GNUnet. It provides | ||
4425 | applications the ability to send and receive encrypted messages to the | ||
4426 | peer's "directly" connected neighbours. | ||
4427 | |||
4428 | As CORE connections are generally "direct" connections,@ applications must | ||
4429 | not assume that they can connect to arbitrary peers this way, as "direct" | ||
4430 | connections may not always be possible. Applications using CORE are | ||
4431 | notified about which peers are connected. Creating new "direct" | ||
4432 | connections must be done using the TRANSPORT API. | ||
4433 | |||
4434 | The CORE API provides unreliable, out-of-order delivery. While the | ||
4435 | implementation tries to ensure timely, in-order delivery, both message | ||
4436 | losses and reordering are not detected and must be tolerated by the | ||
4437 | application. Most important, the core will NOT perform retransmission if | ||
4438 | messages could not be delivered. | ||
4439 | |||
4440 | Note that CORE allows applications to queue one message per connected | ||
4441 | peer. The rate at which each connection operates is influenced by the | ||
4442 | preferences expressed by local application as well as restrictions | ||
4443 | imposed by the other peer. Local applications can express their | ||
4444 | preferences for particular connections using the "performance" API of the | ||
4445 | ATS service. | ||
4446 | |||
4447 | Applications that require more sophisticated transmission capabilities | ||
4448 | such as TCP-like behavior, or if you intend to send messages to arbitrary | ||
4449 | remote peers, should use the CADET API. | ||
4450 | |||
4451 | The typical use of the CORE API is to connect to the CORE service using | ||
4452 | @code{GNUNET_CORE_connect}, process events from the CORE service (such as | ||
4453 | peers connecting, peers disconnecting and incoming messages) and send | ||
4454 | messages to connected peers using | ||
4455 | @code{GNUNET_CORE_notify_transmit_ready}. Note that applications must | ||
4456 | cancel pending transmission requests if they receive a disconnect event | ||
4457 | for a peer that had a transmission pending; furthermore, queueing more | ||
4458 | than one transmission request per peer per application using the | ||
4459 | service is not permitted. | ||
4460 | |||
4461 | The CORE API also allows applications to monitor all communications of the | ||
4462 | peer prior to encryption (for outgoing messages) or after decryption (for | ||
4463 | incoming messages). This can be useful for debugging, diagnostics or to | ||
4464 | establish the presence of cover traffic (for anonymity). As monitoring | ||
4465 | applications are often not interested in the payload, the monitoring | ||
4466 | callbacks can be configured to only provide the message headers (including | ||
4467 | the message type and size) instead of copying the full data stream to the | ||
4468 | monitoring client. | ||
4469 | |||
4470 | The init callback of the @code{GNUNET_CORE_connect} function is called | ||
4471 | with the hash of the public key of the peer. This public key is used to | ||
4472 | identify the peer globally in the GNUnet network. Applications are | ||
4473 | encouraged to check that the provided hash matches the hash that they are | ||
4474 | using (as theoretically the application may be using a different | ||
4475 | configuration file with a different private key, which would result in | ||
4476 | hard to find bugs). | ||
4477 | |||
4478 | As with most service APIs, the CORE API isolates applications from crashes | ||
4479 | of the CORE service. If the CORE service crashes, the application will see | ||
4480 | disconnect events for all existing connections. Once the connections are | ||
4481 | re-established, the applications will be receive matching connect events. | ||
4482 | |||
4483 | @cindex core clinet-service protocol | ||
4484 | @node The CORE Client-Service Protocol | ||
4485 | @subsection The CORE Client-Service Protocol | ||
4486 | @c %**end of header | ||
4487 | |||
4488 | This section describes the protocol between an application using the CORE | ||
4489 | service (the client) and the CORE service process itself. | ||
4490 | |||
4491 | |||
4492 | @menu | ||
4493 | * Setup2:: | ||
4494 | * Notifications:: | ||
4495 | * Sending:: | ||
4496 | @end menu | ||
4497 | |||
4498 | @node Setup2 | ||
4499 | @subsubsection Setup2 | ||
4500 | @c %**end of header | ||
4501 | |||
4502 | When a client connects to the CORE service, it first sends a | ||
4503 | @code{InitMessage} which specifies options for the connection and a set of | ||
4504 | message type values which are supported by the application. The options | ||
4505 | bitmask specifies which events the client would like to be notified about. | ||
4506 | The options include: | ||
4507 | |||
4508 | @table @asis | ||
4509 | @item GNUNET_CORE_OPTION_NOTHING No notifications | ||
4510 | @item GNUNET_CORE_OPTION_STATUS_CHANGE Peers connecting and disconnecting | ||
4511 | @item GNUNET_CORE_OPTION_FULL_INBOUND All inbound messages (after | ||
4512 | decryption) with full payload | ||
4513 | @item GNUNET_CORE_OPTION_HDR_INBOUND Just the @code{MessageHeader} | ||
4514 | of all inbound messages | ||
4515 | @item GNUNET_CORE_OPTION_FULL_OUTBOUND All outbound | ||
4516 | messages (prior to encryption) with full payload | ||
4517 | @item GNUNET_CORE_OPTION_HDR_OUTBOUND Just the @code{MessageHeader} of all | ||
4518 | outbound messages | ||
4519 | @end table | ||
4520 | |||
4521 | Typical applications will only monitor for connection status changes. | ||
4522 | |||
4523 | The CORE service responds to the @code{InitMessage} with an | ||
4524 | @code{InitReplyMessage} which contains the peer's identity. Afterwards, | ||
4525 | both CORE and the client can send messages. | ||
4526 | |||
4527 | @node Notifications | ||
4528 | @subsubsection Notifications | ||
4529 | @c %**end of header | ||
4530 | |||
4531 | The CORE will send @code{ConnectNotifyMessage}s and | ||
4532 | @code{DisconnectNotifyMessage}s whenever peers connect or disconnect from | ||
4533 | the CORE (assuming their type maps overlap with the message types | ||
4534 | registered by the client). When the CORE receives a message that matches | ||
4535 | the set of message types specified during the @code{InitMessage} (or if | ||
4536 | monitoring is enabled in for inbound messages in the options), it sends a | ||
4537 | @code{NotifyTrafficMessage} with the peer identity of the sender and the | ||
4538 | decrypted payload. The same message format (except with | ||
4539 | @code{GNUNET_MESSAGE_TYPE_CORE_NOTIFY_OUTBOUND} for the message type) is | ||
4540 | used to notify clients monitoring outbound messages; here, the peer | ||
4541 | identity given is that of the receiver. | ||
4542 | |||
4543 | @node Sending | ||
4544 | @subsubsection Sending | ||
4545 | @c %**end of header | ||
4546 | |||
4547 | When a client wants to transmit a message, it first requests a | ||
4548 | transmission slot by sending a @code{SendMessageRequest} which specifies | ||
4549 | the priority, deadline and size of the message. Note that these values | ||
4550 | may be ignored by CORE. When CORE is ready for the message, it answers | ||
4551 | with a @code{SendMessageReady} response. The client can then transmit the | ||
4552 | payload with a @code{SendMessage} message. Note that the actual message | ||
4553 | size in the @code{SendMessage} is allowed to be smaller than the size in | ||
4554 | the original request. A client may at any time send a fresh | ||
4555 | @code{SendMessageRequest}, which then superceeds the previous | ||
4556 | @code{SendMessageRequest}, which is then no longer valid. The client can | ||
4557 | tell which @code{SendMessageRequest} the CORE service's | ||
4558 | @code{SendMessageReady} message is for as all of these messages contain a | ||
4559 | "unique" request ID (based on a counter incremented by the client | ||
4560 | for each request). | ||
4561 | |||
4562 | @cindex CORE Peer-to-Peer Protocol | ||
4563 | @node The CORE Peer-to-Peer Protocol | ||
4564 | @subsection The CORE Peer-to-Peer Protocol | ||
4565 | @c %**end of header | ||
4566 | |||
4567 | |||
4568 | @menu | ||
4569 | * Creating the EphemeralKeyMessage:: | ||
4570 | * Establishing a connection:: | ||
4571 | * Encryption and Decryption:: | ||
4572 | * Type maps:: | ||
4573 | @end menu | ||
4574 | |||
4575 | @cindex EphemeralKeyMessage creation | ||
4576 | @node Creating the EphemeralKeyMessage | ||
4577 | @subsubsection Creating the EphemeralKeyMessage | ||
4578 | @c %**end of header | ||
4579 | |||
4580 | When the CORE service starts, each peer creates a fresh ephemeral (ECC) | ||
4581 | public-private key pair and signs the corresponding | ||
4582 | @code{EphemeralKeyMessage} with its long-term key (which we usually call | ||
4583 | the peer's identity; the hash of the public long term key is what results | ||
4584 | in a @code{struct GNUNET_PeerIdentity} in all GNUnet APIs. The ephemeral | ||
4585 | key is ONLY used for an ECDHE@footnote{@uref{http://en.wikipedia.org/wiki/Elliptic_curve_Diffie%E2%80%93Hellman, Elliptic-curve Diffie---Hellman}} | ||
4586 | exchange by the CORE service to establish symmetric session keys. A peer | ||
4587 | will use the same @code{EphemeralKeyMessage} for all peers for | ||
4588 | @code{REKEY_FREQUENCY}, which is usually 12 hours. After that time, it | ||
4589 | will create a fresh ephemeral key (forgetting the old one) and broadcast | ||
4590 | the new @code{EphemeralKeyMessage} to all connected peers, resulting in | ||
4591 | fresh symmetric session keys. Note that peers independently decide on | ||
4592 | when to discard ephemeral keys; it is not a protocol violation to discard | ||
4593 | keys more often. Ephemeral keys are also never stored to disk; restarting | ||
4594 | a peer will thus always create a fresh ephemeral key. The use of ephemeral | ||
4595 | keys is what provides @uref{http://en.wikipedia.org/wiki/Forward_secrecy, forward secrecy}. | ||
4596 | |||
4597 | Just before transmission, the @code{EphemeralKeyMessage} is patched to | ||
4598 | reflect the current sender_status, which specifies the current state of | ||
4599 | the connection from the point of view of the sender. The possible values | ||
4600 | are: | ||
4601 | |||
4602 | @itemize @bullet | ||
4603 | @item @code{KX_STATE_DOWN} Initial value, never used on the network | ||
4604 | @item @code{KX_STATE_KEY_SENT} We sent our ephemeral key, do not know the | ||
4605 | key of the other peer | ||
4606 | @item @code{KX_STATE_KEY_RECEIVED} This peer has received a valid | ||
4607 | ephemeral key of the other peer, but we are waiting for the other peer to | ||
4608 | confirm it's authenticity (ability to decode) via challenge-response. | ||
4609 | @item @code{KX_STATE_UP} The connection is fully up from the point of | ||
4610 | view of the sender (now performing keep-alives) | ||
4611 | @item @code{KX_STATE_REKEY_SENT} The sender has initiated a rekeying | ||
4612 | operation; the other peer has so far failed to confirm a working | ||
4613 | connection using the new ephemeral key | ||
4614 | @end itemize | ||
4615 | |||
4616 | @node Establishing a connection | ||
4617 | @subsubsection Establishing a connection | ||
4618 | @c %**end of header | ||
4619 | |||
4620 | Peers begin their interaction by sending a @code{EphemeralKeyMessage} to | ||
4621 | the other peer once the TRANSPORT service notifies the CORE service about | ||
4622 | the connection. | ||
4623 | A peer receiving an @code{EphemeralKeyMessage} with a status | ||
4624 | indicating that the sender does not have the receiver's ephemeral key, the | ||
4625 | receiver's @code{EphemeralKeyMessage} is sent in response. | ||
4626 | Additionally, if the receiver has not yet confirmed the authenticity of | ||
4627 | the sender, it also sends an (encrypted)@code{PingMessage} with a | ||
4628 | challenge (and the identity of the target) to the other peer. Peers | ||
4629 | receiving a @code{PingMessage} respond with an (encrypted) | ||
4630 | @code{PongMessage} which includes the challenge. Peers receiving a | ||
4631 | @code{PongMessage} check the challenge, and if it matches set the | ||
4632 | connection to @code{KX_STATE_UP}. | ||
4633 | |||
4634 | @node Encryption and Decryption | ||
4635 | @subsubsection Encryption and Decryption | ||
4636 | @c %**end of header | ||
4637 | |||
4638 | All functions related to the key exchange and encryption/decryption of | ||
4639 | messages can be found in @file{gnunet-service-core_kx.c} (except for the | ||
4640 | cryptographic primitives, which are in @file{util/crypto*.c}). | ||
4641 | Given the key material from ECDHE, a Key derivation function | ||
4642 | @footnote{@uref{https://en.wikipedia.org/wiki/Key_derivation_function, Key derivation function}} | ||
4643 | is used to derive two pairs of encryption and decryption keys for AES-256 | ||
4644 | and TwoFish, as well as initialization vectors and authentication keys | ||
4645 | (for HMAC@footnote{@uref{https://en.wikipedia.org/wiki/HMAC, HMAC}}). | ||
4646 | The HMAC is computed over the encrypted payload. | ||
4647 | Encrypted messages include an iv_seed and the HMAC in the header. | ||
4648 | |||
4649 | Each encrypted message in the CORE service includes a sequence number and | ||
4650 | a timestamp in the encrypted payload. The CORE service remembers the | ||
4651 | largest observed sequence number and a bit-mask which represents which of | ||
4652 | the previous 32 sequence numbers were already used. | ||
4653 | Messages with sequence numbers lower than the largest observed sequence | ||
4654 | number minus 32 are discarded. Messages with a timestamp that is less | ||
4655 | than @code{REKEY_TOLERANCE} off (5 minutes) are also discarded. This of | ||
4656 | course means that system clocks need to be reasonably synchronized for | ||
4657 | peers to be able to communicate. Additionally, as the ephemeral key | ||
4658 | changes every 12 hours, a peer would not even be able to decrypt messages | ||
4659 | older than 12 hours. | ||
4660 | |||
4661 | @node Type maps | ||
4662 | @subsubsection Type maps | ||
4663 | @c %**end of header | ||
4664 | |||
4665 | Once an encrypted connection has been established, peers begin to exchange | ||
4666 | type maps. Type maps are used to allow the CORE service to determine which | ||
4667 | (encrypted) connections should be shown to which applications. A type map | ||
4668 | is an array of 65536 bits representing the different types of messages | ||
4669 | understood by applications using the CORE service. Each CORE service | ||
4670 | maintains this map, simply by setting the respective bit for each message | ||
4671 | type supported by any of the applications using the CORE service. Note | ||
4672 | that bits for message types embedded in higher-level protocols (such as | ||
4673 | MESH) will not be included in these type maps. | ||
4674 | |||
4675 | Typically, the type map of a peer will be sparse. Thus, the CORE service | ||
4676 | attempts to compress its type map using @code{gzip}-style compression | ||
4677 | ("deflate") prior to transmission. However, if the compression fails to | ||
4678 | compact the map, the map may also be transmitted without compression | ||
4679 | (resulting in @code{GNUNET_MESSAGE_TYPE_CORE_COMPRESSED_TYPE_MAP} or | ||
4680 | @code{GNUNET_MESSAGE_TYPE_CORE_BINARY_TYPE_MAP} messages respectively). | ||
4681 | Upon receiving a type map, the respective CORE service notifies | ||
4682 | applications about the connection to the other peer if they support any | ||
4683 | message type indicated in the type map (or no message type at all). | ||
4684 | If the CORE service experience a connect or disconnect event from an | ||
4685 | application, it updates its type map (setting or unsetting the respective | ||
4686 | bits) and notifies its neighbours about the change. | ||
4687 | The CORE services of the neighbours then in turn generate connect and | ||
4688 | disconnect events for the peer that sent the type map for their respective | ||
4689 | applications. As CORE messages may be lost, the CORE service confirms | ||
4690 | receiving a type map by sending back a | ||
4691 | @code{GNUNET_MESSAGE_TYPE_CORE_CONFIRM_TYPE_MAP}. If such a confirmation | ||
4692 | (with the correct hash of the type map) is not received, the sender will | ||
4693 | retransmit the type map (with exponential back-off). | ||
4694 | |||
4695 | @cindex cadet subsystem | ||
4696 | @cindex CADET | ||
4697 | @node GNUnet's CADET subsystem | ||
4698 | @section GNUnet's CADET subsystem | ||
4699 | |||
4700 | The CADET subsystem in GNUnet is responsible for secure end-to-end | ||
4701 | communications between nodes in the GNUnet overlay network. CADET builds | ||
4702 | on the CORE subsystem which provides for the link-layer communication and | ||
4703 | then adds routing, forwarding and additional security to the connections. | ||
4704 | CADET offers the same cryptographic services as CORE, but on an | ||
4705 | end-to-end level. This is done so peers retransmitting traffic on behalf | ||
4706 | of other peers cannot access the payload data. | ||
4707 | |||
4708 | @itemize @bullet | ||
4709 | @item CADET provides confidentiality with so-called perfect forward | ||
4710 | secrecy; we use ECDHE powered by Curve25519 for the key exchange and then | ||
4711 | use symmetric encryption, encrypting with both AES-256 and Twofish | ||
4712 | @item authentication is achieved by signing the ephemeral keys using | ||
4713 | Ed25519, a deterministic variant of ECDSA | ||
4714 | @item integrity protection (using SHA-512 to do encrypt-then-MAC, although | ||
4715 | only 256 bits are sent to reduce overhead) | ||
4716 | @item replay protection (using nonces, timestamps, challenge-response, | ||
4717 | message counters and ephemeral keys) | ||
4718 | @item liveness (keep-alive messages, timeout) | ||
4719 | @end itemize | ||
4720 | |||
4721 | Additional to the CORE-like security benefits, CADET offers other | ||
4722 | properties that make it a more universal service than CORE. | ||
4723 | |||
4724 | @itemize @bullet | ||
4725 | @item CADET can establish channels to arbitrary peers in GNUnet. If a | ||
4726 | peer is not immediately reachable, CADET will find a path through the | ||
4727 | network and ask other peers to retransmit the traffic on its behalf. | ||
4728 | @item CADET offers (optional) reliability mechanisms. In a reliable | ||
4729 | channel traffic is guaranteed to arrive complete, unchanged and in-order. | ||
4730 | @item CADET takes care of flow and congestion control mechanisms, not | ||
4731 | allowing the sender to send more traffic than the receiver or the network | ||
4732 | are able to process. | ||
4733 | @end itemize | ||
4734 | |||
4735 | @menu | ||
4736 | * libgnunetcadet:: | ||
4737 | @end menu | ||
4738 | |||
4739 | @cindex libgnunetcadet | ||
4740 | @node libgnunetcadet | ||
4741 | @subsection libgnunetcadet | ||
4742 | |||
4743 | |||
4744 | The CADET API (defined in @file{gnunet_cadet_service.h}) is the | ||
4745 | messaging API used by P2P applications built using GNUnet. | ||
4746 | It provides applications the ability to send and receive encrypted | ||
4747 | messages to any peer participating in GNUnet. | ||
4748 | The API is heavily base on the CORE API. | ||
4749 | |||
4750 | CADET delivers messages to other peers in "channels". | ||
4751 | A channel is a permanent connection defined by a destination peer | ||
4752 | (identified by its public key) and a port number. | ||
4753 | Internally, CADET tunnels all channels towards a destiantion peer | ||
4754 | using one session key and relays the data on multiple "connections", | ||
4755 | independent from the channels. | ||
4756 | |||
4757 | Each channel has optional paramenters, the most important being the | ||
4758 | reliability flag. | ||
4759 | Should a message get lost on TRANSPORT/CORE level, if a channel is | ||
4760 | created with as reliable, CADET will retransmit the lost message and | ||
4761 | deliver it in order to the destination application. | ||
4762 | |||
4763 | To communicate with other peers using CADET, it is necessary to first | ||
4764 | connect to the service using @code{GNUNET_CADET_connect}. | ||
4765 | This function takes several parameters in form of callbacks, to allow the | ||
4766 | client to react to various events, like incoming channels or channels that | ||
4767 | terminate, as well as specify a list of ports the client wishes to listen | ||
4768 | to (at the moment it is not possible to start listening on further ports | ||
4769 | once connected, but nothing prevents a client to connect several times to | ||
4770 | CADET, even do one connection per listening port). | ||
4771 | The function returns a handle which has to be used for any further | ||
4772 | interaction with the service. | ||
4773 | |||
4774 | To connect to a remote peer a client has to call the | ||
4775 | @code{GNUNET_CADET_channel_create} function. The most important parameters | ||
4776 | given are the remote peer's identity (it public key) and a port, which | ||
4777 | specifies which application on the remote peer to connect to, similar to | ||
4778 | TCP/UDP ports. CADET will then find the peer in the GNUnet network and | ||
4779 | establish the proper low-level connections and do the necessary key | ||
4780 | exchanges to assure and authenticated, secure and verified communication. | ||
4781 | Similar to @code{GNUNET_CADET_connect},@code{GNUNET_CADET_create_channel} | ||
4782 | returns a handle to interact with the created channel. | ||
4783 | |||
4784 | For every message the client wants to send to the remote application, | ||
4785 | @code{GNUNET_CADET_notify_transmit_ready} must be called, indicating the | ||
4786 | channel on which the message should be sent and the size of the message | ||
4787 | (but not the message itself!). Once CADET is ready to send the message, | ||
4788 | the provided callback will fire, and the message contents are provided to | ||
4789 | this callback. | ||
4790 | |||
4791 | Please note the CADET does not provide an explicit notification of when a | ||
4792 | channel is connected. In loosely connected networks, like big wireless | ||
4793 | mesh networks, this can take several seconds, even minutes in the worst | ||
4794 | case. To be alerted when a channel is online, a client can call | ||
4795 | @code{GNUNET_CADET_notify_transmit_ready} immediately after | ||
4796 | @code{GNUNET_CADET_create_channel}. When the callback is activated, it | ||
4797 | means that the channel is online. The callback can give 0 bytes to CADET | ||
4798 | if no message is to be sent, this is ok. | ||
4799 | |||
4800 | If a transmission was requested but before the callback fires it is no | ||
4801 | longer needed, it can be cancelled with | ||
4802 | @code{GNUNET_CADET_notify_transmit_ready_cancel}, which uses the handle | ||
4803 | given back by @code{GNUNET_CADET_notify_transmit_ready}. | ||
4804 | As in the case of CORE, only one message can be requested at a time: a | ||
4805 | client must not call @code{GNUNET_CADET_notify_transmit_ready} again until | ||
4806 | the callback is called or the request is cancelled. | ||
4807 | |||
4808 | When a channel is no longer needed, a client can call | ||
4809 | @code{GNUNET_CADET_channel_destroy} to get rid of it. | ||
4810 | Note that CADET will try to transmit all pending traffic before notifying | ||
4811 | the remote peer of the destruction of the channel, including | ||
4812 | retransmitting lost messages if the channel was reliable. | ||
4813 | |||
4814 | Incoming channels, channels being closed by the remote peer, and traffic | ||
4815 | on any incoming or outgoing channels are given to the client when CADET | ||
4816 | executes the callbacks given to it at the time of | ||
4817 | @code{GNUNET_CADET_connect}. | ||
4818 | |||
4819 | Finally, when an application no longer wants to use CADET, it should call | ||
4820 | @code{GNUNET_CADET_disconnect}, but first all channels and pending | ||
4821 | transmissions must be closed (otherwise CADET will complain). | ||
4822 | |||
4823 | @cindex nse subsystem | ||
4824 | @cindex NSE | ||
4825 | @node GNUnet's NSE subsystem | ||
4826 | @section GNUnet's NSE subsystem | ||
4827 | |||
4828 | |||
4829 | NSE stands for @dfn{Network Size Estimation}. The NSE subsystem provides | ||
4830 | other subsystems and users with a rough estimate of the number of peers | ||
4831 | currently participating in the GNUnet overlay. | ||
4832 | The computed value is not a precise number as producing a precise number | ||
4833 | in a decentralized, efficient and secure way is impossible. | ||
4834 | While NSE's estimate is inherently imprecise, NSE also gives the expected | ||
4835 | range. For a peer that has been running in a stable network for a | ||
4836 | while, the real network size will typically (99.7% of the time) be in the | ||
4837 | range of [2/3 estimate, 3/2 estimate]. We will now give an overview of the | ||
4838 | algorithm used to calculate the estimate; | ||
4839 | all of the details can be found in this technical report. | ||
4840 | |||
4841 | @c FIXME: link to the report. | ||
4842 | |||
4843 | @menu | ||
4844 | * Motivation:: | ||
4845 | * Principle:: | ||
4846 | * libgnunetnse:: | ||
4847 | * The NSE Client-Service Protocol:: | ||
4848 | * The NSE Peer-to-Peer Protocol:: | ||
4849 | @end menu | ||
4850 | |||
4851 | @node Motivation | ||
4852 | @subsection Motivation | ||
4853 | |||
4854 | |||
4855 | Some subsytems, like DHT, need to know the size of the GNUnet network to | ||
4856 | optimize some parameters of their own protocol. The decentralized nature | ||
4857 | of GNUnet makes efficient and securely counting the exact number of peers | ||
4858 | infeasable. Although there are several decentralized algorithms to count | ||
4859 | the number of peers in a system, so far there is none to do so securely. | ||
4860 | Other protocols may allow any malicious peer to manipulate the final | ||
4861 | result or to take advantage of the system to perform | ||
4862 | @dfn{Denial of Service} (DoS) attacks against the network. | ||
4863 | GNUnet's NSE protocol avoids these drawbacks. | ||
4864 | |||
4865 | |||
4866 | |||
4867 | @menu | ||
4868 | * Security:: | ||
4869 | @end menu | ||
4870 | |||
4871 | @cindex NSE security | ||
4872 | @cindex nse security | ||
4873 | @node Security | ||
4874 | @subsubsection Security | ||
4875 | |||
4876 | |||
4877 | The NSE subsystem is designed to be resilient against these attacks. | ||
4878 | It uses @uref{http://en.wikipedia.org/wiki/Proof-of-work_system, proofs of work} | ||
4879 | to prevent one peer from impersonating a large number of participants, | ||
4880 | which would otherwise allow an adversary to artifically inflate the | ||
4881 | estimate. | ||
4882 | The DoS protection comes from the time-based nature of the protocol: | ||
4883 | the estimates are calculated periodically and out-of-time traffic is | ||
4884 | either ignored or stored for later retransmission by benign peers. | ||
4885 | In particular, peers cannot trigger global network communication at will. | ||
4886 | |||
4887 | @cindex NSE principle | ||
4888 | @cindex nse principle | ||
4889 | @node Principle | ||
4890 | @subsection Principle | ||
4891 | |||
4892 | |||
4893 | The algorithm calculates the estimate by finding the globally closest | ||
4894 | peer ID to a random, time-based value. | ||
4895 | |||
4896 | The idea is that the closer the ID is to the random value, the more | ||
4897 | "densely packed" the ID space is, and therefore, more peers are in the | ||
4898 | network. | ||
4899 | |||
4900 | |||
4901 | |||
4902 | @menu | ||
4903 | * Example:: | ||
4904 | * Algorithm:: | ||
4905 | * Target value:: | ||
4906 | * Timing:: | ||
4907 | * Controlled Flooding:: | ||
4908 | * Calculating the estimate:: | ||
4909 | @end menu | ||
4910 | |||
4911 | @node Example | ||
4912 | @subsubsection Example | ||
4913 | |||
4914 | |||
4915 | Suppose all peers have IDs between 0 and 100 (our ID space), and the | ||
4916 | random value is 42. | ||
4917 | If the closest peer has the ID 70 we can imagine that the average | ||
4918 | "distance" between peers is around 30 and therefore the are around 3 | ||
4919 | peers in the whole ID space. On the other hand, if the closest peer has | ||
4920 | the ID 44, we can imagine that the space is rather packed with peers, | ||
4921 | maybe as much as 50 of them. | ||
4922 | Naturally, we could have been rather unlucky, and there is only one peer | ||
4923 | and happens to have the ID 44. Thus, the current estimate is calculated | ||
4924 | as the average over multiple rounds, and not just a single sample. | ||
4925 | |||
4926 | @node Algorithm | ||
4927 | @subsubsection Algorithm | ||
4928 | |||
4929 | |||
4930 | Given that example, one can imagine that the job of the subsystem is to | ||
4931 | efficiently communicate the ID of the closest peer to the target value | ||
4932 | to all the other peers, who will calculate the estimate from it. | ||
4933 | |||
4934 | @node Target value | ||
4935 | @subsubsection Target value | ||
4936 | |||
4937 | @c %**end of header | ||
4938 | |||
4939 | The target value itself is generated by hashing the current time, rounded | ||
4940 | down to an agreed value. If the rounding amount is 1h (default) and the | ||
4941 | time is 12:34:56, the time to hash would be 12:00:00. The process is | ||
4942 | repeated each rouning amount (in this example would be every hour). | ||
4943 | Every repetition is called a round. | ||
4944 | |||
4945 | @node Timing | ||
4946 | @subsubsection Timing | ||
4947 | @c %**end of header | ||
4948 | |||
4949 | The NSE subsystem has some timing control to avoid everybody broadcasting | ||
4950 | its ID all at one. Once each peer has the target random value, it | ||
4951 | compares its own ID to the target and calculates the hypothetical size of | ||
4952 | the network if that peer were to be the closest. | ||
4953 | Then it compares the hypothetical size with the estimate from the previous | ||
4954 | rounds. For each value there is an assiciated point in the period, | ||
4955 | let's call it "broadcast time". If its own hypothetical estimate | ||
4956 | is the same as the previous global estimate, its "broadcast time" will be | ||
4957 | in the middle of the round. If its bigger it will be earlier and if its | ||
4958 | smaller (the most likely case) it will be later. This ensures that the | ||
4959 | peers closests to the target value start broadcasting their ID the first. | ||
4960 | |||
4961 | @node Controlled Flooding | ||
4962 | @subsubsection Controlled Flooding | ||
4963 | |||
4964 | @c %**end of header | ||
4965 | |||
4966 | When a peer receives a value, first it verifies that it is closer than the | ||
4967 | closest value it had so far, otherwise it answers the incoming message | ||
4968 | with a message containing the better value. Then it checks a proof of | ||
4969 | work that must be included in the incoming message, to ensure that the | ||
4970 | other peer's ID is not made up (otherwise a malicious peer could claim to | ||
4971 | have an ID of exactly the target value every round). Once validated, it | ||
4972 | compares the brodcast time of the received value with the current time | ||
4973 | and if it's not too early, sends the received value to its neighbors. | ||
4974 | Otherwise it stores the value until the correct broadcast time comes. | ||
4975 | This prevents unnecessary traffic of sub-optimal values, since a better | ||
4976 | value can come before the broadcast time, rendering the previous one | ||
4977 | obsolete and saving the traffic that would have been used to broadcast it | ||
4978 | to the neighbors. | ||
4979 | |||
4980 | @node Calculating the estimate | ||
4981 | @subsubsection Calculating the estimate | ||
4982 | |||
4983 | @c %**end of header | ||
4984 | |||
4985 | Once the closest ID has been spread across the network each peer gets the | ||
4986 | exact distance betweed this ID and the target value of the round and | ||
4987 | calculates the estimate with a mathematical formula described in the tech | ||
4988 | report. The estimate generated with this method for a single round is not | ||
4989 | very precise. Remember the case of the example, where the only peer is the | ||
4990 | ID 44 and we happen to generate the target value 42, thinking there are | ||
4991 | 50 peers in the network. Therefore, the NSE subsystem remembers the last | ||
4992 | 64 estimates and calculates an average over them, giving a result of which | ||
4993 | usually has one bit of uncertainty (the real size could be half of the | ||
4994 | estimate or twice as much). Note that the actual network size is | ||
4995 | calculated in powers of two of the raw input, thus one bit of uncertainty | ||
4996 | means a factor of two in the size estimate. | ||
4997 | |||
4998 | @cindex libgnunetnse | ||
4999 | @node libgnunetnse | ||
5000 | @subsection libgnunetnse | ||
5001 | |||
5002 | @c %**end of header | ||
5003 | |||
5004 | The NSE subsystem has the simplest API of all services, with only two | ||
5005 | calls: @code{GNUNET_NSE_connect} and @code{GNUNET_NSE_disconnect}. | ||
5006 | |||
5007 | The connect call gets a callback function as a parameter and this function | ||
5008 | is called each time the network agrees on an estimate. This usually is | ||
5009 | once per round, with some exceptions: if the closest peer has a late | ||
5010 | local clock and starts spreading his ID after everyone else agreed on a | ||
5011 | value, the callback might be activated twice in a round, the second value | ||
5012 | being always bigger than the first. The default round time is set to | ||
5013 | 1 hour. | ||
5014 | |||
5015 | The disconnect call disconnects from the NSE subsystem and the callback | ||
5016 | is no longer called with new estimates. | ||
5017 | |||
5018 | |||
5019 | |||
5020 | @menu | ||
5021 | * Results:: | ||
5022 | * libgnunetnse - Examples:: | ||
5023 | @end menu | ||
5024 | |||
5025 | @node Results | ||
5026 | @subsubsection Results | ||
5027 | |||
5028 | @c %**end of header | ||
5029 | |||
5030 | The callback provides two values: the average and the | ||
5031 | @uref{http://en.wikipedia.org/wiki/Standard_deviation, standard deviation} | ||
5032 | of the last 64 rounds. The values provided by the callback function are | ||
5033 | logarithmic, this means that the real estimate numbers can be obtained by | ||
5034 | calculating 2 to the power of the given value (2average). From a | ||
5035 | statistics point of view this means that: | ||
5036 | |||
5037 | @itemize @bullet | ||
5038 | @item 68% of the time the real size is included in the interval | ||
5039 | [(2average-stddev), 2] | ||
5040 | @item 95% of the time the real size is included in the interval | ||
5041 | [(2average-2*stddev, 2^average+2*stddev] | ||
5042 | @item 99.7% of the time the real size is included in the interval | ||
5043 | [(2average-3*stddev, 2average+3*stddev] | ||
5044 | @end itemize | ||
5045 | |||
5046 | The expected standard variation for 64 rounds in a network of stable size | ||
5047 | is 0.2. Thus, we can say that normally: | ||
5048 | |||
5049 | @itemize @bullet | ||
5050 | @item 68% of the time the real size is in the range [-13%, +15%] | ||
5051 | @item 95% of the time the real size is in the range [-24%, +32%] | ||
5052 | @item 99.7% of the time the real size is in the range [-34%, +52%] | ||
5053 | @end itemize | ||
5054 | |||
5055 | As said in the introduction, we can be quite sure that usually the real | ||
5056 | size is between one third and three times the estimate. This can of | ||
5057 | course vary with network conditions. | ||
5058 | Thus, applications may want to also consider the provided standard | ||
5059 | deviation value, not only the average (in particular, if the standard | ||
5060 | veriation is very high, the average maybe meaningless: the network size is | ||
5061 | changing rapidly). | ||
5062 | |||
5063 | @node libgnunetnse - Examples | ||
5064 | @subsubsection libgnunetnse -Examples | ||
5065 | |||
5066 | @c %**end of header | ||
5067 | |||
5068 | Let's close with a couple examples. | ||
5069 | |||
5070 | @table @asis | ||
5071 | |||
5072 | @item Average: 10, std dev: 1 Here the estimate would be | ||
5073 | 2^10 = 1024 peers. @footnote{The range in which we can be 95% sure is: | ||
5074 | [2^8, 2^12] = [256, 4096]. We can be very (>99.7%) sure that the network | ||
5075 | is not a hundred peers and absolutely sure that it is not a million peers, | ||
5076 | but somewhere around a thousand.} | ||
5077 | |||
5078 | @item Average 22, std dev: 0.2 Here the estimate would be | ||
5079 | 2^22 = 4 Million peers. @footnote{The range in which we can be 99.7% sure | ||
5080 | is: [2^21.4, 2^22.6] = [2.8M, 6.3M]. We can be sure that the network size | ||
5081 | is around four million, with absolutely way of it being 1 million.} | ||
5082 | |||
5083 | @end table | ||
5084 | |||
5085 | To put this in perspective, if someone remembers the LHC Higgs boson | ||
5086 | results, were announced with "5 sigma" and "6 sigma" certainties. In this | ||
5087 | case a 5 sigma minimum would be 2 million and a 6 sigma minimum, | ||
5088 | 1.8 million. | ||
5089 | |||
5090 | @node The NSE Client-Service Protocol | ||
5091 | @subsection The NSE Client-Service Protocol | ||
5092 | |||
5093 | @c %**end of header | ||
5094 | |||
5095 | As with the API, the client-service protocol is very simple, only has 2 | ||
5096 | different messages, defined in @code{src/nse/nse.h}: | ||
5097 | |||
5098 | @itemize @bullet | ||
5099 | @item @code{GNUNET_MESSAGE_TYPE_NSE_START}@ This message has no parameters | ||
5100 | and is sent from the client to the service upon connection. | ||
5101 | @item @code{GNUNET_MESSAGE_TYPE_NSE_ESTIMATE}@ This message is sent from | ||
5102 | the service to the client for every new estimate and upon connection. | ||
5103 | Contains a timestamp for the estimate, the average and the standard | ||
5104 | deviation for the respective round. | ||
5105 | @end itemize | ||
5106 | |||
5107 | When the @code{GNUNET_NSE_disconnect} API call is executed, the client | ||
5108 | simply disconnects from the service, with no message involved. | ||
5109 | |||
5110 | @cindex NSE Peer-to-Peer Protocol | ||
5111 | @node The NSE Peer-to-Peer Protocol | ||
5112 | @subsection The NSE Peer-to-Peer Protocol | ||
5113 | |||
5114 | @c %**end of header | ||
5115 | |||
5116 | The NSE subsystem only has one message in the P2P protocol, the | ||
5117 | @code{GNUNET_MESSAGE_TYPE_NSE_P2P_FLOOD} message. | ||
5118 | |||
5119 | This message key contents are the timestamp to identify the round | ||
5120 | (differences in system clocks may cause some peers to send messages way | ||
5121 | too early or way too late, so the timestamp allows other peers to | ||
5122 | identify such messages easily), the | ||
5123 | @uref{http://en.wikipedia.org/wiki/Proof-of-work_system, proof of work} | ||
5124 | used to make it difficult to mount a | ||
5125 | @uref{http://en.wikipedia.org/wiki/Sybil_attack, Sybil attack}, and the | ||
5126 | public key, which is used to verify the signature on the message. | ||
5127 | |||
5128 | Every peer stores a message for the previous, current and next round. The | ||
5129 | messages for the previous and current round are given to peers that | ||
5130 | connect to us. The message for the next round is simply stored until our | ||
5131 | system clock advances to the next round. The message for the current round | ||
5132 | is what we are flooding the network with right now. | ||
5133 | At the beginning of each round the peer does the following: | ||
5134 | |||
5135 | @itemize @bullet | ||
5136 | @item calculates his own distance to the target value | ||
5137 | @item creates, signs and stores the message for the current round (unless | ||
5138 | it has a better message in the "next round" slot which came early in the | ||
5139 | previous round) | ||
5140 | @item calculates, based on the stored round message (own or received) when | ||
5141 | to stard flooding it to its neighbors | ||
5142 | @end itemize | ||
5143 | |||
5144 | Upon receiving a message the peer checks the validity of the message | ||
5145 | (round, proof of work, signature). The next action depends on the | ||
5146 | contents of the incoming message: | ||
5147 | |||
5148 | @itemize @bullet | ||
5149 | @item if the message is worse than the current stored message, the peer | ||
5150 | sends the current message back immediately, to stop the other peer from | ||
5151 | spreading suboptimal results | ||
5152 | @item if the message is better than the current stored message, the peer | ||
5153 | stores the new message and calculates the new target time to start | ||
5154 | spreading it to its neighbors (excluding the one the message came from) | ||
5155 | @item if the message is for the previous round, it is compared to the | ||
5156 | message stored in the "previous round slot", which may then be updated | ||
5157 | @item if the message is for the next round, it is compared to the message | ||
5158 | stored in the "next round slot", which again may then be updated | ||
5159 | @end itemize | ||
5160 | |||
5161 | Finally, when it comes to send the stored message for the current round to | ||
5162 | the neighbors there is a random delay added for each neighbor, to avoid | ||
5163 | traffic spikes and minimize cross-messages. | ||
5164 | |||
5165 | @cindex HOSTLIST subsystem | ||
5166 | @cindex hostlist subsystem | ||
5167 | @node GNUnet's HOSTLIST subsystem | ||
5168 | @section GNUnet's HOSTLIST subsystem | ||
5169 | |||
5170 | @c %**end of header | ||
5171 | |||
5172 | Peers in the GNUnet overlay network need address information so that they | ||
5173 | can connect with other peers. GNUnet uses so called HELLO messages to | ||
5174 | store and exchange peer addresses. | ||
5175 | GNUnet provides several methods for peers to obtain this information: | ||
5176 | |||
5177 | @itemize @bullet | ||
5178 | @item out-of-band exchange of HELLO messages (manually, using for example | ||
5179 | gnunet-peerinfo) | ||
5180 | @item HELLO messages shipped with GNUnet (automatic with distribution) | ||
5181 | @item UDP neighbor discovery in LAN (IPv4 broadcast, IPv6 multicast) | ||
5182 | @item topology gossiping (learning from other peers we already connected | ||
5183 | to), and | ||
5184 | @item the HOSTLIST daemon covered in this section, which is particularly | ||
5185 | relevant for bootstrapping new peers. | ||
5186 | @end itemize | ||
5187 | |||
5188 | New peers have no existing connections (and thus cannot learn from gossip | ||
5189 | among peers), may not have other peers in their LAN and might be started | ||
5190 | with an outdated set of HELLO messages from the distribution. | ||
5191 | In this case, getting new peers to connect to the network requires either | ||
5192 | manual effort or the use of a HOSTLIST to obtain HELLOs. | ||
5193 | |||
5194 | @menu | ||
5195 | * HELLOs:: | ||
5196 | * Overview for the HOSTLIST subsystem:: | ||
5197 | * Interacting with the HOSTLIST daemon:: | ||
5198 | * Hostlist security address validation:: | ||
5199 | * The HOSTLIST daemon:: | ||
5200 | * The HOSTLIST server:: | ||
5201 | * The HOSTLIST client:: | ||
5202 | * Usage:: | ||
5203 | @end menu | ||
5204 | |||
5205 | @node HELLOs | ||
5206 | @subsection HELLOs | ||
5207 | |||
5208 | @c %**end of header | ||
5209 | |||
5210 | The basic information peers require to connect to other peers are | ||
5211 | contained in so called HELLO messages you can think of as a business card. | ||
5212 | Besides the identity of the peer (based on the cryptographic public key) a | ||
5213 | HELLO message may contain address information that specifies ways to | ||
5214 | contact a peer. By obtaining HELLO messages, a peer can learn how to | ||
5215 | contact other peers. | ||
5216 | |||
5217 | @node Overview for the HOSTLIST subsystem | ||
5218 | @subsection Overview for the HOSTLIST subsystem | ||
5219 | |||
5220 | @c %**end of header | ||
5221 | |||
5222 | The HOSTLIST subsystem provides a way to distribute and obtain contact | ||
5223 | information to connect to other peers using a simple HTTP GET request. | ||
5224 | It's implementation is split in three parts, the main file for the daemon | ||
5225 | itself (@file{gnunet-daemon-hostlist.c}), the HTTP client used to download | ||
5226 | peer information (@file{hostlist-client.c}) and the server component used | ||
5227 | to provide this information to other peers (@file{hostlist-server.c}). | ||
5228 | The server is basically a small HTTP web server (based on GNU | ||
5229 | libmicrohttpd) which provides a list of HELLOs known to the local peer for | ||
5230 | download. The client component is basically a HTTP client | ||
5231 | (based on libcurl) which can download hostlists from one or more websites. | ||
5232 | The hostlist format is a binary blob containing a sequence of HELLO | ||
5233 | messages. Note that any HTTP server can theoretically serve a hostlist, | ||
5234 | the build-in hostlist server makes it simply convenient to offer this | ||
5235 | service. | ||
5236 | |||
5237 | |||
5238 | @menu | ||
5239 | * Features:: | ||
5240 | * HOSTLIST - Limitations:: | ||
5241 | @end menu | ||
5242 | |||
5243 | @node Features | ||
5244 | @subsubsection Features | ||
5245 | |||
5246 | @c %**end of header | ||
5247 | |||
5248 | The HOSTLIST daemon can: | ||
5249 | |||
5250 | @itemize @bullet | ||
5251 | @item provide HELLO messages with validated addresses obtained from | ||
5252 | PEERINFO to download for other peers | ||
5253 | @item download HELLO messages and forward these message to the TRANSPORT | ||
5254 | subsystem for validation | ||
5255 | @item advertises the URL of this peer's hostlist address to other peers | ||
5256 | via gossip | ||
5257 | @item automatically learn about hostlist servers from the gossip of other | ||
5258 | peers | ||
5259 | @end itemize | ||
5260 | |||
5261 | @node HOSTLIST - Limitations | ||
5262 | @subsubsection HOSTLIST - Limitations | ||
5263 | |||
5264 | @c %**end of header | ||
5265 | |||
5266 | The HOSTLIST daemon does not: | ||
5267 | |||
5268 | @itemize @bullet | ||
5269 | @item verify the cryptographic information in the HELLO messages | ||
5270 | @item verify the address information in the HELLO messages | ||
5271 | @end itemize | ||
5272 | |||
5273 | @node Interacting with the HOSTLIST daemon | ||
5274 | @subsection Interacting with the HOSTLIST daemon | ||
5275 | |||
5276 | @c %**end of header | ||
5277 | |||
5278 | The HOSTLIST subsystem is currently implemented as a daemon, so there is | ||
5279 | no need for the user to interact with it and therefore there is no | ||
5280 | command line tool and no API to communicate with the daemon. In the | ||
5281 | future, we can envision changing this to allow users to manually trigger | ||
5282 | the download of a hostlist. | ||
5283 | |||
5284 | Since there is no command line interface to interact with HOSTLIST, the | ||
5285 | only way to interact with the hostlist is to use STATISTICS to obtain or | ||
5286 | modify information about the status of HOSTLIST: | ||
5287 | |||
5288 | @example | ||
5289 | $ gnunet-statistics -s hostlist | ||
5290 | @end example | ||
5291 | |||
5292 | @noindent | ||
5293 | In particular, HOSTLIST includes a @strong{persistent} value in statistics | ||
5294 | that specifies when the hostlist server might be queried next. As this | ||
5295 | value is exponentially increasing during runtime, developers may want to | ||
5296 | reset or manually adjust it. Note that HOSTLIST (but not STATISTICS) needs | ||
5297 | to be shutdown if changes to this value are to have any effect on the | ||
5298 | daemon (as HOSTLIST does not monitor STATISTICS for changes to the | ||
5299 | download frequency). | ||
5300 | |||
5301 | @node Hostlist security address validation | ||
5302 | @subsection Hostlist security address validation | ||
5303 | |||
5304 | @c %**end of header | ||
5305 | |||
5306 | Since information obtained from other parties cannot be trusted without | ||
5307 | validation, we have to distinguish between @emph{validated} and | ||
5308 | @emph{not validated} addresses. Before using (and so trusting) | ||
5309 | information from other parties, this information has to be double-checked | ||
5310 | (validated). Address validation is not done by HOSTLIST but by the | ||
5311 | TRANSPORT service. | ||
5312 | |||
5313 | The HOSTLIST component is functionally located between the PEERINFO and | ||
5314 | the TRANSPORT subsystem. When acting as a server, the daemon obtains valid | ||
5315 | (@emph{validated}) peer information (HELLO messages) from the PEERINFO | ||
5316 | service and provides it to other peers. When acting as a client, it | ||
5317 | contacts the HOSTLIST servers specified in the configuration, downloads | ||
5318 | the (unvalidated) list of HELLO messages and forwards these information | ||
5319 | to the TRANSPORT server to validate the addresses. | ||
5320 | |||
5321 | @cindex HOSTLIST daemon | ||
5322 | @node The HOSTLIST daemon | ||
5323 | @subsection The HOSTLIST daemon | ||
5324 | |||
5325 | @c %**end of header | ||
5326 | |||
5327 | The hostlist daemon is the main component of the HOSTLIST subsystem. It is | ||
5328 | started by the ARM service and (if configured) starts the HOSTLIST client | ||
5329 | and server components. | ||
5330 | |||
5331 | If the daemon provides a hostlist itself it can advertise it's own | ||
5332 | hostlist to other peers. To do so it sends a | ||
5333 | @code{GNUNET_MESSAGE_TYPE_HOSTLIST_ADVERTISEMENT} message to other peers | ||
5334 | when they connect to this peer on the CORE level. This hostlist | ||
5335 | advertisement message contains the URL to access the HOSTLIST HTTP | ||
5336 | server of the sender. The daemon may also subscribe to this type of | ||
5337 | message from CORE service, and then forward these kind of message to the | ||
5338 | HOSTLIST client. The client then uses all available URLs to download peer | ||
5339 | information when necessary. | ||
5340 | |||
5341 | When starting, the HOSTLIST daemon first connects to the CORE subsystem | ||
5342 | and if hostlist learning is enabled, registers a CORE handler to receive | ||
5343 | this kind of messages. Next it starts (if configured) the client and | ||
5344 | server. It passes pointers to CORE connect and disconnect and receive | ||
5345 | handlers where the client and server store their functions, so the daemon | ||
5346 | can notify them about CORE events. | ||
5347 | |||
5348 | To clean up on shutdown, the daemon has a cleaning task, shutting down all | ||
5349 | subsystems and disconnecting from CORE. | ||
5350 | |||
5351 | @cindex HOSTLIST server | ||
5352 | @node The HOSTLIST server | ||
5353 | @subsection The HOSTLIST server | ||
5354 | |||
5355 | @c %**end of header | ||
5356 | |||
5357 | The server provides a way for other peers to obtain HELLOs. Basically it | ||
5358 | is a small web server other peers can connect to and download a list of | ||
5359 | HELLOs using standard HTTP; it may also advertise the URL of the hostlist | ||
5360 | to other peers connecting on CORE level. | ||
5361 | |||
5362 | |||
5363 | @menu | ||
5364 | * The HTTP Server:: | ||
5365 | * Advertising the URL:: | ||
5366 | @end menu | ||
5367 | |||
5368 | @node The HTTP Server | ||
5369 | @subsubsection The HTTP Server | ||
5370 | |||
5371 | @c %**end of header | ||
5372 | |||
5373 | During startup, the server starts a web server listening on the port | ||
5374 | specified with the HTTPPORT value (default 8080). In addition it connects | ||
5375 | to the PEERINFO service to obtain peer information. The HOSTLIST server | ||
5376 | uses the GNUNET_PEERINFO_iterate function to request HELLO information for | ||
5377 | all peers and adds their information to a new hostlist if they are | ||
5378 | suitable (expired addresses and HELLOs without addresses are both not | ||
5379 | suitable) and the maximum size for a hostlist is not exceeded | ||
5380 | (MAX_BYTES_PER_HOSTLISTS = 500000). | ||
5381 | When PEERINFO finishes (with a last NULL callback), the server destroys | ||
5382 | the previous hostlist response available for download on the web server | ||
5383 | and replaces it with the updated hostlist. The hostlist format is | ||
5384 | basically a sequence of HELLO messages (as obtained from PEERINFO) without | ||
5385 | any special tokenization. Since each HELLO message contains a size field, | ||
5386 | the response can easily be split into separate HELLO messages by the | ||
5387 | client. | ||
5388 | |||
5389 | A HOSTLIST client connecting to the HOSTLIST server will receive the | ||
5390 | hostlist as a HTTP response and the the server will terminate the | ||
5391 | connection with the result code @code{HTTP 200 OK}. | ||
5392 | The connection will be closed immediately if no hostlist is available. | ||
5393 | |||
5394 | @node Advertising the URL | ||
5395 | @subsubsection Advertising the URL | ||
5396 | |||
5397 | @c %**end of header | ||
5398 | |||
5399 | The server also advertises the URL to download the hostlist to other peers | ||
5400 | if hostlist advertisement is enabled. | ||
5401 | When a new peer connects and has hostlist learning enabled, the server | ||
5402 | sends a @code{GNUNET_MESSAGE_TYPE_HOSTLIST_ADVERTISEMENT} message to this | ||
5403 | peer using the CORE service. | ||
5404 | |||
5405 | @cindex HOSTLIST client | ||
5406 | @node The HOSTLIST client | ||
5407 | @subsection The HOSTLIST client | ||
5408 | |||
5409 | @c %**end of header | ||
5410 | |||
5411 | The client provides the functionality to download the list of HELLOs from | ||
5412 | a set of URLs. | ||
5413 | It performs a standard HTTP request to the URLs configured and learned | ||
5414 | from advertisement messages received from other peers. When a HELLO is | ||
5415 | downloaded, the HOSTLIST client forwards the HELLO to the TRANSPORT | ||
5416 | service for validation. | ||
5417 | |||
5418 | The client supports two modes of operation: | ||
5419 | |||
5420 | @itemize @bullet | ||
5421 | @item download of HELLOs (bootstrapping) | ||
5422 | @item learning of URLs | ||
5423 | @end itemize | ||
5424 | |||
5425 | @menu | ||
5426 | * Bootstrapping:: | ||
5427 | * Learning:: | ||
5428 | @end menu | ||
5429 | |||
5430 | @node Bootstrapping | ||
5431 | @subsubsection Bootstrapping | ||
5432 | |||
5433 | @c %**end of header | ||
5434 | |||
5435 | For bootstrapping, it schedules a task to download the hostlist from the | ||
5436 | set of known URLs. | ||
5437 | The downloads are only performed if the number of current | ||
5438 | connections is smaller than a minimum number of connections | ||
5439 | (at the moment 4). | ||
5440 | The interval between downloads increases exponentially; however, the | ||
5441 | exponential growth is limited if it becomes longer than an hour. | ||
5442 | At that point, the frequency growth is capped at | ||
5443 | (#number of connections * 1h). | ||
5444 | |||
5445 | Once the decision has been taken to download HELLOs, the daemon chooses a | ||
5446 | random URL from the list of known URLs. URLs can be configured in the | ||
5447 | configuration or be learned from advertisement messages. | ||
5448 | The client uses a HTTP client library (libcurl) to initiate the download | ||
5449 | using the libcurl multi interface. | ||
5450 | Libcurl passes the data to the callback_download function which | ||
5451 | stores the data in a buffer if space is available and the maximum size for | ||
5452 | a hostlist download is not exceeded (MAX_BYTES_PER_HOSTLISTS = 500000). | ||
5453 | When a full HELLO was downloaded, the HOSTLIST client offers this | ||
5454 | HELLO message to the TRANSPORT service for validation. | ||
5455 | When the download is finished or failed, statistical information about the | ||
5456 | quality of this URL is updated. | ||
5457 | |||
5458 | @cindex HOSTLIST learning | ||
5459 | @node Learning | ||
5460 | @subsubsection Learning | ||
5461 | |||
5462 | @c %**end of header | ||
5463 | |||
5464 | The client also manages hostlist advertisements from other peers. The | ||
5465 | HOSTLIST daemon forwards @code{GNUNET_MESSAGE_TYPE_HOSTLIST_ADVERTISEMENT} | ||
5466 | messages to the client subsystem, which extracts the URL from the message. | ||
5467 | Next, a test of the newly obtained URL is performed by triggering a | ||
5468 | download from the new URL. If the URL works correctly, it is added to the | ||
5469 | list of working URLs. | ||
5470 | |||
5471 | The size of the list of URLs is restricted, so if an additional server is | ||
5472 | added and the list is full, the URL with the worst quality ranking | ||
5473 | (determined through successful downloads and number of HELLOs e.g.) is | ||
5474 | discarded. During shutdown the list of URLs is saved to a file for | ||
5475 | persistance and loaded on startup. URLs from the configuration file are | ||
5476 | never discarded. | ||
5477 | |||
5478 | @node Usage | ||
5479 | @subsection Usage | ||
5480 | |||
5481 | @c %**end of header | ||
5482 | |||
5483 | To start HOSTLIST by default, it has to be added to the DEFAULTSERVICES | ||
5484 | section for the ARM services. This is done in the default configuration. | ||
5485 | |||
5486 | For more information on how to configure the HOSTLIST subsystem see the | ||
5487 | installation handbook:@ | ||
5488 | Configuring the hostlist to bootstrap@ | ||
5489 | Configuring your peer to provide a hostlist | ||
5490 | |||
5491 | @cindex IDENTITY | ||
5492 | @cindex identity subsystem | ||
5493 | @node GNUnet's IDENTITY subsystem | ||
5494 | @section GNUnet's IDENTITY subsystem | ||
5495 | |||
5496 | @c %**end of header | ||
5497 | |||
5498 | Identities of "users" in GNUnet are called egos. | ||
5499 | Egos can be used as pseudonyms ("fake names") or be tied to an | ||
5500 | organization (for example, "GNU") or even the actual identity of a human. | ||
5501 | GNUnet users are expected to have many egos. They might have one tied to | ||
5502 | their real identity, some for organizations they manage, and more for | ||
5503 | different domains where they want to operate under a pseudonym. | ||
5504 | |||
5505 | The IDENTITY service allows users to manage their egos. The identity | ||
5506 | service manages the private keys egos of the local user; it does not | ||
5507 | manage identities of other users (public keys). Public keys for other | ||
5508 | users need names to become manageable. GNUnet uses the | ||
5509 | @dfn{GNU Name System} (GNS) to give names to other users and manage their | ||
5510 | public keys securely. This chapter is about the IDENTITY service, | ||
5511 | which is about the management of private keys. | ||
5512 | |||
5513 | On the network, an ego corresponds to an ECDSA key (over Curve25519, | ||
5514 | using RFC 6979, as required by GNS). Thus, users can perform actions | ||
5515 | under a particular ego by using (signing with) a particular private key. | ||
5516 | Other users can then confirm that the action was really performed by that | ||
5517 | ego by checking the signature against the respective public key. | ||
5518 | |||
5519 | The IDENTITY service allows users to associate a human-readable name with | ||
5520 | each ego. This way, users can use names that will remind them of the | ||
5521 | purpose of a particular ego. | ||
5522 | The IDENTITY service will store the respective private keys and | ||
5523 | allows applications to access key information by name. | ||
5524 | Users can change the name that is locally (!) associated with an ego. | ||
5525 | Egos can also be deleted, which means that the private key will be removed | ||
5526 | and it thus will not be possible to perform actions with that ego in the | ||
5527 | future. | ||
5528 | |||
5529 | Additionally, the IDENTITY subsystem can associate service functions with | ||
5530 | egos. | ||
5531 | For example, GNS requires the ego that should be used for the shorten | ||
5532 | zone. GNS will ask IDENTITY for an ego for the "gns-short" service. | ||
5533 | The IDENTITY service has a mapping of such service strings to the name of | ||
5534 | the ego that the user wants to use for this service, for example | ||
5535 | "my-short-zone-ego". | ||
5536 | |||
5537 | Finally, the IDENTITY API provides access to a special ego, the | ||
5538 | anonymous ego. The anonymous ego is special in that its private key is not | ||
5539 | really private, but fixed and known to everyone. | ||
5540 | Thus, anyone can perform actions as anonymous. This can be useful as with | ||
5541 | this trick, code does not have to contain a special case to distinguish | ||
5542 | between anonymous and pseudonymous egos. | ||
5543 | |||
5544 | @menu | ||
5545 | * libgnunetidentity:: | ||
5546 | * The IDENTITY Client-Service Protocol:: | ||
5547 | @end menu | ||
5548 | |||
5549 | @cindex libgnunetidentity | ||
5550 | @node libgnunetidentity | ||
5551 | @subsection libgnunetidentity | ||
5552 | @c %**end of header | ||
5553 | |||
5554 | |||
5555 | @menu | ||
5556 | * Connecting to the service:: | ||
5557 | * Operations on Egos:: | ||
5558 | * The anonymous Ego:: | ||
5559 | * Convenience API to lookup a single ego:: | ||
5560 | * Associating egos with service functions:: | ||
5561 | @end menu | ||
5562 | |||
5563 | @node Connecting to the service | ||
5564 | @subsubsection Connecting to the service | ||
5565 | |||
5566 | @c %**end of header | ||
5567 | |||
5568 | First, typical clients connect to the identity service using | ||
5569 | @code{GNUNET_IDENTITY_connect}. This function takes a callback as a | ||
5570 | parameter. | ||
5571 | If the given callback parameter is non-null, it will be invoked to notify | ||
5572 | the application about the current state of the identities in the system. | ||
5573 | |||
5574 | @itemize @bullet | ||
5575 | @item First, it will be invoked on all known egos at the time of the | ||
5576 | connection. For each ego, a handle to the ego and the user's name for the | ||
5577 | ego will be passed to the callback. Furthermore, a @code{void **} context | ||
5578 | argument will be provided which gives the client the opportunity to | ||
5579 | associate some state with the ego. | ||
5580 | @item Second, the callback will be invoked with NULL for the ego, the name | ||
5581 | and the context. This signals that the (initial) iteration over all egos | ||
5582 | has completed. | ||
5583 | @item Then, the callback will be invoked whenever something changes about | ||
5584 | an ego. | ||
5585 | If an ego is renamed, the callback is invoked with the ego handle of the | ||
5586 | ego that was renamed, and the new name. If an ego is deleted, the callback | ||
5587 | is invoked with the ego handle and a name of NULL. In the deletion case, | ||
5588 | the application should also release resources stored in the context. | ||
5589 | @item When the application destroys the connection to the identity service | ||
5590 | using @code{GNUNET_IDENTITY_disconnect}, the callback is again invoked | ||
5591 | with the ego and a name of NULL (equivalent to deletion of the egos). | ||
5592 | This should again be used to clean up the per-ego context. | ||
5593 | @end itemize | ||
5594 | |||
5595 | The ego handle passed to the callback remains valid until the callback is | ||
5596 | invoked with a name of NULL, so it is safe to store a reference to the | ||
5597 | ego's handle. | ||
5598 | |||
5599 | @node Operations on Egos | ||
5600 | @subsubsection Operations on Egos | ||
5601 | |||
5602 | @c %**end of header | ||
5603 | |||
5604 | Given an ego handle, the main operations are to get its associated private | ||
5605 | key using @code{GNUNET_IDENTITY_ego_get_private_key} or its associated | ||
5606 | public key using @code{GNUNET_IDENTITY_ego_get_public_key}. | ||
5607 | |||
5608 | The other operations on egos are pretty straightforward. | ||
5609 | Using @code{GNUNET_IDENTITY_create}, an application can request the | ||
5610 | creation of an ego by specifying the desired name. | ||
5611 | The operation will fail if that name is | ||
5612 | already in use. Using @code{GNUNET_IDENTITY_rename} the name of an | ||
5613 | existing ego can be changed. Finally, egos can be deleted using | ||
5614 | @code{GNUNET_IDENTITY_delete}. All of these operations will trigger | ||
5615 | updates to the callback given to the @code{GNUNET_IDENTITY_connect} | ||
5616 | function of all applications that are connected with the identity service | ||
5617 | at the time. @code{GNUNET_IDENTITY_cancel} can be used to cancel the | ||
5618 | operations before the respective continuations would be called. | ||
5619 | It is not guaranteed that the operation will not be completed anyway, | ||
5620 | only the continuation will no longer be called. | ||
5621 | |||
5622 | @node The anonymous Ego | ||
5623 | @subsubsection The anonymous Ego | ||
5624 | |||
5625 | @c %**end of header | ||
5626 | |||
5627 | A special way to obtain an ego handle is to call | ||
5628 | @code{GNUNET_IDENTITY_ego_get_anonymous}, which returns an ego for the | ||
5629 | "anonymous" user --- anyone knows and can get the private key for this | ||
5630 | user, so it is suitable for operations that are supposed to be anonymous | ||
5631 | but require signatures (for example, to avoid a special path in the code). | ||
5632 | The anonymous ego is always valid and accessing it does not require a | ||
5633 | connection to the identity service. | ||
5634 | |||
5635 | @node Convenience API to lookup a single ego | ||
5636 | @subsubsection Convenience API to lookup a single ego | ||
5637 | |||
5638 | |||
5639 | As applications commonly simply have to lookup a single ego, there is a | ||
5640 | convenience API to do just that. Use @code{GNUNET_IDENTITY_ego_lookup} to | ||
5641 | lookup a single ego by name. Note that this is the user's name for the | ||
5642 | ego, not the service function. The resulting ego will be returned via a | ||
5643 | callback and will only be valid during that callback. The operation can | ||
5644 | be cancelled via @code{GNUNET_IDENTITY_ego_lookup_cancel} | ||
5645 | (cancellation is only legal before the callback is invoked). | ||
5646 | |||
5647 | @node Associating egos with service functions | ||
5648 | @subsubsection Associating egos with service functions | ||
5649 | |||
5650 | |||
5651 | The @code{GNUNET_IDENTITY_set} function is used to associate a particular | ||
5652 | ego with a service function. The name used by the service and the ego are | ||
5653 | given as arguments. | ||
5654 | Afterwards, the service can use its name to lookup the associated ego | ||
5655 | using @code{GNUNET_IDENTITY_get}. | ||
5656 | |||
5657 | @node The IDENTITY Client-Service Protocol | ||
5658 | @subsection The IDENTITY Client-Service Protocol | ||
5659 | |||
5660 | @c %**end of header | ||
5661 | |||
5662 | A client connecting to the identity service first sends a message with | ||
5663 | type | ||
5664 | @code{GNUNET_MESSAGE_TYPE_IDENTITY_START} to the service. After that, the | ||
5665 | client will receive information about changes to the egos by receiving | ||
5666 | messages of type @code{GNUNET_MESSAGE_TYPE_IDENTITY_UPDATE}. | ||
5667 | Those messages contain the private key of the ego and the user's name of | ||
5668 | the ego (or zero bytes for the name to indicate that the ego was deleted). | ||
5669 | A special bit @code{end_of_list} is used to indicate the end of the | ||
5670 | initial iteration over the identity service's egos. | ||
5671 | |||
5672 | The client can trigger changes to the egos by sending @code{CREATE}, | ||
5673 | @code{RENAME} or @code{DELETE} messages. | ||
5674 | The CREATE message contains the private key and the desired name.@ | ||
5675 | The RENAME message contains the old name and the new name.@ | ||
5676 | The DELETE message only needs to include the name of the ego to delete.@ | ||
5677 | The service responds to each of these messages with a @code{RESULT_CODE} | ||
5678 | message which indicates success or error of the operation, and possibly | ||
5679 | a human-readable error message. | ||
5680 | |||
5681 | Finally, the client can bind the name of a service function to an ego by | ||
5682 | sending a @code{SET_DEFAULT} message with the name of the service function | ||
5683 | and the private key of the ego. | ||
5684 | Such bindings can then be resolved using a @code{GET_DEFAULT} message, | ||
5685 | which includes the name of the service function. The identity service | ||
5686 | will respond to a GET_DEFAULT request with a SET_DEFAULT message | ||
5687 | containing the respective information, or with a RESULT_CODE to | ||
5688 | indicate an error. | ||
5689 | |||
5690 | @cindex NAMESTORE | ||
5691 | @cindex namestore subsystem | ||
5692 | @node GNUnet's NAMESTORE Subsystem | ||
5693 | @section GNUnet's NAMESTORE Subsystem | ||
5694 | |||
5695 | The NAMESTORE subsystem provides persistent storage for local GNS zone | ||
5696 | information. All local GNS zone information are managed by NAMESTORE. It | ||
5697 | provides both the functionality to administer local GNS information (e.g. | ||
5698 | delete and add records) as well as to retrieve GNS information (e.g to | ||
5699 | list name information in a client). | ||
5700 | NAMESTORE does only manage the persistent storage of zone information | ||
5701 | belonging to the user running the service: GNS information from other | ||
5702 | users obtained from the DHT are stored by the NAMECACHE subsystem. | ||
5703 | |||
5704 | NAMESTORE uses a plugin-based database backend to store GNS information | ||
5705 | with good performance. Here sqlite, MySQL and PostgreSQL are supported | ||
5706 | database backends. | ||
5707 | NAMESTORE clients interact with the IDENTITY subsystem to obtain | ||
5708 | cryptographic information about zones based on egos as described with the | ||
5709 | IDENTITY subsystem, but internally NAMESTORE refers to zones using the | ||
5710 | ECDSA private key. | ||
5711 | In addition, it collaborates with the NAMECACHE subsystem and | ||
5712 | stores zone information when local information are modified in the | ||
5713 | GNS cache to increase look-up performance for local information. | ||
5714 | |||
5715 | NAMESTORE provides functionality to look-up and store records, to iterate | ||
5716 | over a specific or all zones and to monitor zones for changes. NAMESTORE | ||
5717 | functionality can be accessed using the NAMESTORE api or the NAMESTORE | ||
5718 | command line tool. | ||
5719 | |||
5720 | @menu | ||
5721 | * libgnunetnamestore:: | ||
5722 | @end menu | ||
5723 | |||
5724 | @cindex libgnunetnamestore | ||
5725 | @node libgnunetnamestore | ||
5726 | @subsection libgnunetnamestore | ||
5727 | |||
5728 | To interact with NAMESTORE clients first connect to the NAMESTORE service | ||
5729 | using the @code{GNUNET_NAMESTORE_connect} passing a configuration handle. | ||
5730 | As a result they obtain a NAMESTORE handle, they can use for operations, | ||
5731 | or NULL is returned if the connection failed. | ||
5732 | |||
5733 | To disconnect from NAMESTORE, clients use | ||
5734 | @code{GNUNET_NAMESTORE_disconnect} and specify the handle to disconnect. | ||
5735 | |||
5736 | NAMESTORE internally uses the ECDSA private key to refer to zones. These | ||
5737 | private keys can be obtained from the IDENTITY subsytem. | ||
5738 | Here @emph{egos} @emph{can be used to refer to zones or the default ego | ||
5739 | assigned to the GNS subsystem can be used to obtained the master zone's | ||
5740 | private key.} | ||
5741 | |||
5742 | |||
5743 | @menu | ||
5744 | * Editing Zone Information:: | ||
5745 | * Iterating Zone Information:: | ||
5746 | * Monitoring Zone Information:: | ||
5747 | @end menu | ||
5748 | |||
5749 | @node Editing Zone Information | ||
5750 | @subsubsection Editing Zone Information | ||
5751 | |||
5752 | @c %**end of header | ||
5753 | |||
5754 | NAMESTORE provides functions to lookup records stored under a label in a | ||
5755 | zone and to store records under a label in a zone. | ||
5756 | |||
5757 | To store (and delete) records, the client uses the | ||
5758 | @code{GNUNET_NAMESTORE_records_store} function and has to provide | ||
5759 | namestore handle to use, the private key of the zone, the label to store | ||
5760 | the records under, the records and number of records plus an callback | ||
5761 | function. | ||
5762 | After the operation is performed NAMESTORE will call the provided | ||
5763 | callback function with the result GNUNET_SYSERR on failure | ||
5764 | (including timeout/queue drop/failure to validate), GNUNET_NO if content | ||
5765 | was already there or not found GNUNET_YES (or other positive value) on | ||
5766 | success plus an additional error message. | ||
5767 | |||
5768 | Records are deleted by using the store command with 0 records to store. | ||
5769 | It is important to note, that records are not merged when records exist | ||
5770 | with the label. | ||
5771 | So a client has first to retrieve records, merge with existing records | ||
5772 | and then store the result. | ||
5773 | |||
5774 | To perform a lookup operation, the client uses the | ||
5775 | @code{GNUNET_NAMESTORE_records_store} function. Here he has to pass the | ||
5776 | namestore handle, the private key of the zone and the label. He also has | ||
5777 | to provide a callback function which will be called with the result of | ||
5778 | the lookup operation: | ||
5779 | the zone for the records, the label, and the records including the | ||
5780 | number of records included. | ||
5781 | |||
5782 | A special operation is used to set the preferred nickname for a zone. | ||
5783 | This nickname is stored with the zone and is automatically merged with | ||
5784 | all labels and records stored in a zone. Here the client uses the | ||
5785 | @code{GNUNET_NAMESTORE_set_nick} function and passes the private key of | ||
5786 | the zone, the nickname as string plus a the callback with the result of | ||
5787 | the operation. | ||
5788 | |||
5789 | @node Iterating Zone Information | ||
5790 | @subsubsection Iterating Zone Information | ||
5791 | |||
5792 | @c %**end of header | ||
5793 | |||
5794 | A client can iterate over all information in a zone or all zones managed | ||
5795 | by NAMESTORE. | ||
5796 | Here a client uses the @code{GNUNET_NAMESTORE_zone_iteration_start} | ||
5797 | function and passes the namestore handle, the zone to iterate over and a | ||
5798 | callback function to call with the result. | ||
5799 | If the client wants to iterate over all the, he passes NULL for the zone. | ||
5800 | A @code{GNUNET_NAMESTORE_ZoneIterator} handle is returned to be used to | ||
5801 | continue iteration. | ||
5802 | |||
5803 | NAMESTORE calls the callback for every result and expects the client to | ||
5804 | call @code{GNUNET_NAMESTORE_zone_iterator_next} to continue to iterate or | ||
5805 | @code{GNUNET_NAMESTORE_zone_iterator_stop} to interrupt the iteration. | ||
5806 | When NAMESTORE reached the last item it will call the callback with a | ||
5807 | NULL value to indicate. | ||
5808 | |||
5809 | @node Monitoring Zone Information | ||
5810 | @subsubsection Monitoring Zone Information | ||
5811 | |||
5812 | @c %**end of header | ||
5813 | |||
5814 | Clients can also monitor zones to be notified about changes. Here the | ||
5815 | clients uses the @code{GNUNET_NAMESTORE_zone_monitor_start} function and | ||
5816 | passes the private key of the zone and and a callback function to call | ||
5817 | with updates for a zone. | ||
5818 | The client can specify to obtain zone information first by iterating over | ||
5819 | the zone and specify a synchronization callback to be called when the | ||
5820 | client and the namestore are synced. | ||
5821 | |||
5822 | On an update, NAMESTORE will call the callback with the private key of the | ||
5823 | zone, the label and the records and their number. | ||
5824 | |||
5825 | To stop monitoring, the client calls | ||
5826 | @code{GNUNET_NAMESTORE_zone_monitor_stop} and passes the handle obtained | ||
5827 | from the function to start the monitoring. | ||
5828 | |||
5829 | @cindex PEERINFO | ||
5830 | @cindex peerinfo subsystem | ||
5831 | @node GNUnet's PEERINFO subsystem | ||
5832 | @section GNUnet's PEERINFO subsystem | ||
5833 | |||
5834 | @c %**end of header | ||
5835 | |||
5836 | The PEERINFO subsystem is used to store verified (validated) information | ||
5837 | about known peers in a persistent way. It obtains these addresses for | ||
5838 | example from TRANSPORT service which is in charge of address validation. | ||
5839 | Validation means that the information in the HELLO message are checked by | ||
5840 | connecting to the addresses and performing a cryptographic handshake to | ||
5841 | authenticate the peer instance stating to be reachable with these | ||
5842 | addresses. | ||
5843 | Peerinfo does not validate the HELLO messages itself but only stores them | ||
5844 | and gives them to interested clients. | ||
5845 | |||
5846 | As future work, we think about moving from storing just HELLO messages to | ||
5847 | providing a generic persistent per-peer information store. | ||
5848 | More and more subsystems tend to need to store per-peer information in | ||
5849 | persistent way. | ||
5850 | To not duplicate this functionality we plan to provide a PEERSTORE | ||
5851 | service providing this functionality. | ||
5852 | |||
5853 | @menu | ||
5854 | * PEERINFO - Features:: | ||
5855 | * PEERINFO - Limitations:: | ||
5856 | * DeveloperPeer Information:: | ||
5857 | * Startup:: | ||
5858 | * Managing Information:: | ||
5859 | * Obtaining Information:: | ||
5860 | * The PEERINFO Client-Service Protocol:: | ||
5861 | * libgnunetpeerinfo:: | ||
5862 | @end menu | ||
5863 | |||
5864 | @node PEERINFO - Features | ||
5865 | @subsection PEERINFO - Features | ||
5866 | |||
5867 | @c %**end of header | ||
5868 | |||
5869 | @itemize @bullet | ||
5870 | @item Persistent storage | ||
5871 | @item Client notification mechanism on update | ||
5872 | @item Periodic clean up for expired information | ||
5873 | @item Differentiation between public and friend-only HELLO | ||
5874 | @end itemize | ||
5875 | |||
5876 | @node PEERINFO - Limitations | ||
5877 | @subsection PEERINFO - Limitations | ||
5878 | |||
5879 | |||
5880 | @itemize @bullet | ||
5881 | @item Does not perform HELLO validation | ||
5882 | @end itemize | ||
5883 | |||
5884 | @node DeveloperPeer Information | ||
5885 | @subsection DeveloperPeer Information | ||
5886 | |||
5887 | @c %**end of header | ||
5888 | |||
5889 | The PEERINFO subsystem stores these information in the form of HELLO | ||
5890 | messages you can think of as business cards. | ||
5891 | These HELLO messages contain the public key of a peer and the addresses | ||
5892 | a peer can be reached under. | ||
5893 | The addresses include an expiration date describing how long they are | ||
5894 | valid. This information is updated regularly by the TRANSPORT service by | ||
5895 | revalidating the address. | ||
5896 | If an address is expired and not renewed, it can be removed from the | ||
5897 | HELLO message. | ||
5898 | |||
5899 | Some peer do not want to have their HELLO messages distributed to other | ||
5900 | peers, especially when GNUnet's friend-to-friend modus is enabled. | ||
5901 | To prevent this undesired distribution. PEERINFO distinguishes between | ||
5902 | @emph{public} and @emph{friend-only} HELLO messages. | ||
5903 | Public HELLO messages can be freely distributed to other (possibly | ||
5904 | unknown) peers (for example using the hostlist, gossiping, broadcasting), | ||
5905 | whereas friend-only HELLO messages may not be distributed to other peers. | ||
5906 | Friend-only HELLO messages have an additional flag @code{friend_only} set | ||
5907 | internally. For public HELLO message this flag is not set. | ||
5908 | PEERINFO does and cannot not check if a client is allowed to obtain a | ||
5909 | specific HELLO type. | ||
5910 | |||
5911 | The HELLO messages can be managed using the GNUnet HELLO library. | ||
5912 | Other GNUnet systems can obtain these information from PEERINFO and use | ||
5913 | it for their purposes. | ||
5914 | Clients are for example the HOSTLIST component providing these | ||
5915 | information to other peers in form of a hostlist or the TRANSPORT | ||
5916 | subsystem using these information to maintain connections to other peers. | ||
5917 | |||
5918 | @node Startup | ||
5919 | @subsection Startup | ||
5920 | |||
5921 | @c %**end of header | ||
5922 | |||
5923 | During startup the PEERINFO services loads persistent HELLOs from disk. | ||
5924 | First PEERINFO parses the directory configured in the HOSTS value of the | ||
5925 | @code{PEERINFO} configuration section to store PEERINFO information. | ||
5926 | For all files found in this directory valid HELLO messages are extracted. | ||
5927 | In addition it loads HELLO messages shipped with the GNUnet distribution. | ||
5928 | These HELLOs are used to simplify network bootstrapping by providing | ||
5929 | valid peer information with the distribution. | ||
5930 | The use of these HELLOs can be prevented by setting the | ||
5931 | @code{USE_INCLUDED_HELLOS} in the @code{PEERINFO} configuration section to | ||
5932 | @code{NO}. Files containing invalid information are removed. | ||
5933 | |||
5934 | @node Managing Information | ||
5935 | @subsection Managing Information | ||
5936 | |||
5937 | @c %**end of header | ||
5938 | |||
5939 | The PEERINFO services stores information about known PEERS and a single | ||
5940 | HELLO message for every peer. | ||
5941 | A peer does not need to have a HELLO if no information are available. | ||
5942 | HELLO information from different sources, for example a HELLO obtained | ||
5943 | from a remote HOSTLIST and a second HELLO stored on disk, are combined | ||
5944 | and merged into one single HELLO message per peer which will be given to | ||
5945 | clients. During this merge process the HELLO is immediately written to | ||
5946 | disk to ensure persistence. | ||
5947 | |||
5948 | PEERINFO in addition periodically scans the directory where information | ||
5949 | are stored for empty HELLO messages with expired TRANSPORT addresses. | ||
5950 | This periodic task scans all files in the directory and recreates the | ||
5951 | HELLO messages it finds. | ||
5952 | Expired TRANSPORT addresses are removed from the HELLO and if the | ||
5953 | HELLO does not contain any valid addresses, it is discarded and removed | ||
5954 | from the disk. | ||
5955 | |||
5956 | @node Obtaining Information | ||
5957 | @subsection Obtaining Information | ||
5958 | |||
5959 | @c %**end of header | ||
5960 | |||
5961 | When a client requests information from PEERINFO, PEERINFO performs a | ||
5962 | lookup for the respective peer or all peers if desired and transmits this | ||
5963 | information to the client. | ||
5964 | The client can specify if friend-only HELLOs have to be included or not | ||
5965 | and PEERINFO filters the respective HELLO messages before transmitting | ||
5966 | information. | ||
5967 | |||
5968 | To notify clients about changes to PEERINFO information, PEERINFO | ||
5969 | maintains a list of clients interested in this notifications. | ||
5970 | Such a notification occurs if a HELLO for a peer was updated (due to a | ||
5971 | merge for example) or a new peer was added. | ||
5972 | |||
5973 | @node The PEERINFO Client-Service Protocol | ||
5974 | @subsection The PEERINFO Client-Service Protocol | ||
5975 | |||
5976 | @c %**end of header | ||
5977 | |||
5978 | To connect and disconnect to and from the PEERINFO Service PEERINFO | ||
5979 | utilizes the util client/server infrastructure, so no special messages | ||
5980 | types are used here. | ||
5981 | |||
5982 | To add information for a peer, the plain HELLO message is transmitted to | ||
5983 | the service without any wrapping. All pieces of information required are | ||
5984 | stored within the HELLO message. | ||
5985 | The PEERINFO service provides a message handler accepting and processing | ||
5986 | these HELLO messages. | ||
5987 | |||
5988 | When obtaining PEERINFO information using the iterate functionality | ||
5989 | specific messages are used. To obtain information for all peers, a | ||
5990 | @code{struct ListAllPeersMessage} with message type | ||
5991 | @code{GNUNET_MESSAGE_TYPE_PEERINFO_GET_ALL} and a flag | ||
5992 | include_friend_only to indicate if friend-only HELLO messages should be | ||
5993 | included are transmitted. If information for a specific peer is required | ||
5994 | a @code{struct ListAllPeersMessage} with | ||
5995 | @code{GNUNET_MESSAGE_TYPE_PEERINFO_GET} containing the peer identity is | ||
5996 | used. | ||
5997 | |||
5998 | For both variants the PEERINFO service replies for each HELLO message it | ||
5999 | wants to transmit with a @code{struct ListAllPeersMessage} with type | ||
6000 | @code{GNUNET_MESSAGE_TYPE_PEERINFO_INFO} containing the plain HELLO. | ||
6001 | The final message is @code{struct GNUNET_MessageHeader} with type | ||
6002 | @code{GNUNET_MESSAGE_TYPE_PEERINFO_INFO}. If the client receives this | ||
6003 | message, it can proceed with the next request if any is pending. | ||
6004 | |||
6005 | @node libgnunetpeerinfo | ||
6006 | @subsection libgnunetpeerinfo | ||
6007 | |||
6008 | @c %**end of header | ||
6009 | |||
6010 | The PEERINFO API consists mainly of three different functionalities: | ||
6011 | |||
6012 | @itemize @bullet | ||
6013 | @item maintaining a connection to the service | ||
6014 | @item adding new information to the PEERINFO service | ||
6015 | @item retrieving information from the PEERINFO service | ||
6016 | @end itemize | ||
6017 | |||
6018 | @menu | ||
6019 | * Connecting to the PEERINFO Service:: | ||
6020 | * Adding Information to the PEERINFO Service:: | ||
6021 | * Obtaining Information from the PEERINFO Service:: | ||
6022 | @end menu | ||
6023 | |||
6024 | @node Connecting to the PEERINFO Service | ||
6025 | @subsubsection Connecting to the PEERINFO Service | ||
6026 | |||
6027 | @c %**end of header | ||
6028 | |||
6029 | To connect to the PEERINFO service the function | ||
6030 | @code{GNUNET_PEERINFO_connect} is used, taking a configuration handle as | ||
6031 | an argument, and to disconnect from PEERINFO the function | ||
6032 | @code{GNUNET_PEERINFO_disconnect}, taking the PEERINFO | ||
6033 | handle returned from the connect function has to be called. | ||
6034 | |||
6035 | @node Adding Information to the PEERINFO Service | ||
6036 | @subsubsection Adding Information to the PEERINFO Service | ||
6037 | |||
6038 | @c %**end of header | ||
6039 | |||
6040 | @code{GNUNET_PEERINFO_add_peer} adds a new peer to the PEERINFO subsystem | ||
6041 | storage. This function takes the PEERINFO handle as an argument, the HELLO | ||
6042 | message to store and a continuation with a closure to be called with the | ||
6043 | result of the operation. | ||
6044 | The @code{GNUNET_PEERINFO_add_peer} returns a handle to this operation | ||
6045 | allowing to cancel the operation with the respective cancel function | ||
6046 | @code{GNUNET_PEERINFO_add_peer_cancel}. To retrieve information from | ||
6047 | PEERINFO you can iterate over all information stored with PEERINFO or you | ||
6048 | can tell PEERINFO to notify if new peer information are available. | ||
6049 | |||
6050 | @node Obtaining Information from the PEERINFO Service | ||
6051 | @subsubsection Obtaining Information from the PEERINFO Service | ||
6052 | |||
6053 | @c %**end of header | ||
6054 | |||
6055 | To iterate over information in PEERINFO you use | ||
6056 | @code{GNUNET_PEERINFO_iterate}. | ||
6057 | This function expects the PEERINFO handle, a flag if HELLO messages | ||
6058 | intended for friend only mode should be included, a timeout how long the | ||
6059 | operation should take and a callback with a callback closure to be called | ||
6060 | for the results. | ||
6061 | If you want to obtain information for a specific peer, you can specify | ||
6062 | the peer identity, if this identity is NULL, information for all peers are | ||
6063 | returned. The function returns a handle to allow to cancel the operation | ||
6064 | using @code{GNUNET_PEERINFO_iterate_cancel}. | ||
6065 | |||
6066 | To get notified when peer information changes, you can use | ||
6067 | @code{GNUNET_PEERINFO_notify}. | ||
6068 | This function expects a configuration handle and a flag if friend-only | ||
6069 | HELLO messages should be included. The PEERINFO service will notify you | ||
6070 | about every change and the callback function will be called to notify you | ||
6071 | about changes. The function returns a handle to cancel notifications | ||
6072 | with @code{GNUNET_PEERINFO_notify_cancel}. | ||
6073 | |||
6074 | @cindex PEERSTORE subsystem | ||
6075 | @node GNUnet's PEERSTORE subsystem | ||
6076 | @section GNUnet's PEERSTORE subsystem | ||
6077 | |||
6078 | @c %**end of header | ||
6079 | |||
6080 | GNUnet's PEERSTORE subsystem offers persistent per-peer storage for other | ||
6081 | GNUnet subsystems. GNUnet subsystems can use PEERSTORE to persistently | ||
6082 | store and retrieve arbitrary data. | ||
6083 | Each data record stored with PEERSTORE contains the following fields: | ||
6084 | |||
6085 | @itemize @bullet | ||
6086 | @item subsystem: Name of the subsystem responsible for the record. | ||
6087 | @item peerid: Identity of the peer this record is related to. | ||
6088 | @item key: a key string identifying the record. | ||
6089 | @item value: binary record value. | ||
6090 | @item expiry: record expiry date. | ||
6091 | @end itemize | ||
6092 | |||
6093 | @menu | ||
6094 | * Functionality:: | ||
6095 | * Architecture:: | ||
6096 | * libgnunetpeerstore:: | ||
6097 | @end menu | ||
6098 | |||
6099 | @node Functionality | ||
6100 | @subsection Functionality | ||
6101 | |||
6102 | @c %**end of header | ||
6103 | |||
6104 | Subsystems can store any type of value under a (subsystem, peerid, key) | ||
6105 | combination. A "replace" flag set during store operations forces the | ||
6106 | PEERSTORE to replace any old values stored under the same | ||
6107 | (subsystem, peerid, key) combination with the new value. | ||
6108 | Additionally, an expiry date is set after which the record is *possibly* | ||
6109 | deleted by PEERSTORE. | ||
6110 | |||
6111 | Subsystems can iterate over all values stored under any of the following | ||
6112 | combination of fields: | ||
6113 | |||
6114 | @itemize @bullet | ||
6115 | @item (subsystem) | ||
6116 | @item (subsystem, peerid) | ||
6117 | @item (subsystem, key) | ||
6118 | @item (subsystem, peerid, key) | ||
6119 | @end itemize | ||
6120 | |||
6121 | Subsystems can also request to be notified about any new values stored | ||
6122 | under a (subsystem, peerid, key) combination by sending a "watch" | ||
6123 | request to PEERSTORE. | ||
6124 | |||
6125 | @node Architecture | ||
6126 | @subsection Architecture | ||
6127 | |||
6128 | @c %**end of header | ||
6129 | |||
6130 | PEERSTORE implements the following components: | ||
6131 | |||
6132 | @itemize @bullet | ||
6133 | @item PEERSTORE service: Handles store, iterate and watch operations. | ||
6134 | @item PEERSTORE API: API to be used by other subsystems to communicate and | ||
6135 | issue commands to the PEERSTORE service. | ||
6136 | @item PEERSTORE plugins: Handles the persistent storage. At the moment, | ||
6137 | only an "sqlite" plugin is implemented. | ||
6138 | @end itemize | ||
6139 | |||
6140 | @cindex libgnunetpeerstore | ||
6141 | @node libgnunetpeerstore | ||
6142 | @subsection libgnunetpeerstore | ||
6143 | |||
6144 | @c %**end of header | ||
6145 | |||
6146 | libgnunetpeerstore is the library containing the PEERSTORE API. Subsystems | ||
6147 | wishing to communicate with the PEERSTORE service use this API to open a | ||
6148 | connection to PEERSTORE. This is done by calling | ||
6149 | @code{GNUNET_PEERSTORE_connect} which returns a handle to the newly | ||
6150 | created connection. | ||
6151 | This handle has to be used with any further calls to the API. | ||
6152 | |||
6153 | To store a new record, the function @code{GNUNET_PEERSTORE_store} is to | ||
6154 | be used which requires the record fields and a continuation function that | ||
6155 | will be called by the API after the STORE request is sent to the | ||
6156 | PEERSTORE service. | ||
6157 | Note that calling the continuation function does not mean that the record | ||
6158 | is successfully stored, only that the STORE request has been successfully | ||
6159 | sent to the PEERSTORE service. | ||
6160 | @code{GNUNET_PEERSTORE_store_cancel} can be called to cancel the STORE | ||
6161 | request only before the continuation function has been called. | ||
6162 | |||
6163 | To iterate over stored records, the function | ||
6164 | @code{GNUNET_PEERSTORE_iterate} is | ||
6165 | to be used. @emph{peerid} and @emph{key} can be set to NULL. An iterator | ||
6166 | callback function will be called with each matching record found and a | ||
6167 | NULL record at the end to signal the end of result set. | ||
6168 | @code{GNUNET_PEERSTORE_iterate_cancel} can be used to cancel the ITERATE | ||
6169 | request before the iterator callback is called with a NULL record. | ||
6170 | |||
6171 | To be notified with new values stored under a (subsystem, peerid, key) | ||
6172 | combination, the function @code{GNUNET_PEERSTORE_watch} is to be used. | ||
6173 | This will register the watcher with the PEERSTORE service, any new | ||
6174 | records matching the given combination will trigger the callback | ||
6175 | function passed to @code{GNUNET_PEERSTORE_watch}. This continues until | ||
6176 | @code{GNUNET_PEERSTORE_watch_cancel} is called or the connection to the | ||
6177 | service is destroyed. | ||
6178 | |||
6179 | After the connection is no longer needed, the function | ||
6180 | @code{GNUNET_PEERSTORE_disconnect} can be called to disconnect from the | ||
6181 | PEERSTORE service. | ||
6182 | Any pending ITERATE or WATCH requests will be destroyed. | ||
6183 | If the @code{sync_first} flag is set to @code{GNUNET_YES}, the API will | ||
6184 | delay the disconnection until all pending STORE requests are sent to | ||
6185 | the PEERSTORE service, otherwise, the pending STORE requests will be | ||
6186 | destroyed as well. | ||
6187 | |||
6188 | @cindex SET Subsystem | ||
6189 | @node GNUnet's SET Subsystem | ||
6190 | @section GNUnet's SET Subsystem | ||
6191 | |||
6192 | @c %**end of header | ||
6193 | |||
6194 | The SET service implements efficient set operations between two peers | ||
6195 | over a mesh tunnel. | ||
6196 | Currently, set union and set intersection are the only supported | ||
6197 | operations. Elements of a set consist of an @emph{element type} and | ||
6198 | arbitrary binary @emph{data}. | ||
6199 | The size of an element's data is limited to around 62 KB. | ||
6200 | |||
6201 | @menu | ||
6202 | * Local Sets:: | ||
6203 | * Set Modifications:: | ||
6204 | * Set Operations:: | ||
6205 | * Result Elements:: | ||
6206 | * libgnunetset:: | ||
6207 | * The SET Client-Service Protocol:: | ||
6208 | * The SET Intersection Peer-to-Peer Protocol:: | ||
6209 | * The SET Union Peer-to-Peer Protocol:: | ||
6210 | @end menu | ||
6211 | |||
6212 | @node Local Sets | ||
6213 | @subsection Local Sets | ||
6214 | |||
6215 | @c %**end of header | ||
6216 | |||
6217 | Sets created by a local client can be modified and reused for multiple | ||
6218 | operations. As each set operation requires potentially expensive special | ||
6219 | auxilliary data to be computed for each element of a set, a set can only | ||
6220 | participate in one type of set operation (i.e. union or intersection). | ||
6221 | The type of a set is determined upon its creation. | ||
6222 | If a the elements of a set are needed for an operation of a different | ||
6223 | type, all of the set's element must be copied to a new set of appropriate | ||
6224 | type. | ||
6225 | |||
6226 | @node Set Modifications | ||
6227 | @subsection Set Modifications | ||
6228 | |||
6229 | @c %**end of header | ||
6230 | |||
6231 | Even when set operations are active, one can add to and remove elements | ||
6232 | from a set. | ||
6233 | However, these changes will only be visible to operations that have been | ||
6234 | created after the changes have taken place. That is, every set operation | ||
6235 | only sees a snapshot of the set from the time the operation was started. | ||
6236 | This mechanism is @emph{not} implemented by copying the whole set, but by | ||
6237 | attaching @emph{generation information} to each element and operation. | ||
6238 | |||
6239 | @node Set Operations | ||
6240 | @subsection Set Operations | ||
6241 | |||
6242 | @c %**end of header | ||
6243 | |||
6244 | Set operations can be started in two ways: Either by accepting an | ||
6245 | operation request from a remote peer, or by requesting a set operation | ||
6246 | from a remote peer. | ||
6247 | Set operations are uniquely identified by the involved @emph{peers}, an | ||
6248 | @emph{application id} and the @emph{operation type}. | ||
6249 | |||
6250 | The client is notified of incoming set operations by @emph{set listeners}. | ||
6251 | A set listener listens for incoming operations of a specific operation | ||
6252 | type and application id. | ||
6253 | Once notified of an incoming set request, the client can accept the set | ||
6254 | request (providing a local set for the operation) or reject it. | ||
6255 | |||
6256 | @node Result Elements | ||
6257 | @subsection Result Elements | ||
6258 | |||
6259 | @c %**end of header | ||
6260 | |||
6261 | The SET service has three @emph{result modes} that determine how an | ||
6262 | operation's result set is delivered to the client: | ||
6263 | |||
6264 | @itemize @bullet | ||
6265 | @item @strong{Full Result Set.} All elements of set resulting from the set | ||
6266 | operation are returned to the client. | ||
6267 | @item @strong{Added Elements.} Only elements that result from the | ||
6268 | operation and are not already in the local peer's set are returned. | ||
6269 | Note that for some operations (like set intersection) this result mode | ||
6270 | will never return any elements. | ||
6271 | This can be useful if only the remove peer is actually interested in | ||
6272 | the result of the set operation. | ||
6273 | @item @strong{Removed Elements.} Only elements that are in the local | ||
6274 | peer's initial set but not in the operation's result set are returned. | ||
6275 | Note that for some operations (like set union) this result mode will | ||
6276 | never return any elements. This can be useful if only the remove peer is | ||
6277 | actually interested in the result of the set operation. | ||
6278 | @end itemize | ||
6279 | |||
6280 | @cindex libgnunetset | ||
6281 | @node libgnunetset | ||
6282 | @subsection libgnunetset | ||
6283 | |||
6284 | @c %**end of header | ||
6285 | |||
6286 | @menu | ||
6287 | * Sets:: | ||
6288 | * Listeners:: | ||
6289 | * Operations:: | ||
6290 | * Supplying a Set:: | ||
6291 | * The Result Callback:: | ||
6292 | @end menu | ||
6293 | |||
6294 | @node Sets | ||
6295 | @subsubsection Sets | ||
6296 | |||
6297 | @c %**end of header | ||
6298 | |||
6299 | New sets are created with @code{GNUNET_SET_create}. Both the local peer's | ||
6300 | configuration (as each set has its own client connection) and the | ||
6301 | operation type must be specified. | ||
6302 | The set exists until either the client calls @code{GNUNET_SET_destroy} or | ||
6303 | the client's connection to the service is disrupted. | ||
6304 | In the latter case, the client is notified by the return value of | ||
6305 | functions dealing with sets. This return value must always be checked. | ||
6306 | |||
6307 | Elements are added and removed with @code{GNUNET_SET_add_element} and | ||
6308 | @code{GNUNET_SET_remove_element}. | ||
6309 | |||
6310 | @node Listeners | ||
6311 | @subsubsection Listeners | ||
6312 | |||
6313 | @c %**end of header | ||
6314 | |||
6315 | Listeners are created with @code{GNUNET_SET_listen}. Each time time a | ||
6316 | remote peer suggests a set operation with an application id and operation | ||
6317 | type matching a listener, the listener's callback is invoked. | ||
6318 | The client then must synchronously call either @code{GNUNET_SET_accept} | ||
6319 | or @code{GNUNET_SET_reject}. Note that the operation will not be started | ||
6320 | until the client calls @code{GNUNET_SET_commit} | ||
6321 | (see Section "Supplying a Set"). | ||
6322 | |||
6323 | @node Operations | ||
6324 | @subsubsection Operations | ||
6325 | |||
6326 | @c %**end of header | ||
6327 | |||
6328 | Operations to be initiated by the local peer are created with | ||
6329 | @code{GNUNET_SET_prepare}. Note that the operation will not be started | ||
6330 | until the client calls @code{GNUNET_SET_commit} | ||
6331 | (see Section "Supplying a Set"). | ||
6332 | |||
6333 | @node Supplying a Set | ||
6334 | @subsubsection Supplying a Set | ||
6335 | |||
6336 | @c %**end of header | ||
6337 | |||
6338 | To create symmetry between the two ways of starting a set operation | ||
6339 | (accepting and nitiating it), the operation handles returned by | ||
6340 | @code{GNUNET_SET_accept} and @code{GNUNET_SET_prepare} do not yet have a | ||
6341 | set to operate on, thus they can not do any work yet. | ||
6342 | |||
6343 | The client must call @code{GNUNET_SET_commit} to specify a set to use for | ||
6344 | an operation. @code{GNUNET_SET_commit} may only be called once per set | ||
6345 | operation. | ||
6346 | |||
6347 | @node The Result Callback | ||
6348 | @subsubsection The Result Callback | ||
6349 | |||
6350 | @c %**end of header | ||
6351 | |||
6352 | Clients must specify both a result mode and a result callback with | ||
6353 | @code{GNUNET_SET_accept} and @code{GNUNET_SET_prepare}. The result | ||
6354 | callback with a status indicating either that an element was received, or | ||
6355 | the operation failed or succeeded. | ||
6356 | The interpretation of the received element depends on the result mode. | ||
6357 | The callback needs to know which result mode it is used in, as the | ||
6358 | arguments do not indicate if an element is part of the full result set, | ||
6359 | or if it is in the difference between the original set and the final set. | ||
6360 | |||
6361 | @node The SET Client-Service Protocol | ||
6362 | @subsection The SET Client-Service Protocol | ||
6363 | |||
6364 | @c %**end of header | ||
6365 | |||
6366 | @menu | ||
6367 | * Creating Sets:: | ||
6368 | * Listeners2:: | ||
6369 | * Initiating Operations:: | ||
6370 | * Modifying Sets:: | ||
6371 | * Results and Operation Status:: | ||
6372 | * Iterating Sets:: | ||
6373 | @end menu | ||
6374 | |||
6375 | @node Creating Sets | ||
6376 | @subsubsection Creating Sets | ||
6377 | |||
6378 | @c %**end of header | ||
6379 | |||
6380 | For each set of a client, there exists a client connection to the service. | ||
6381 | Sets are created by sending the @code{GNUNET_SERVICE_SET_CREATE} message | ||
6382 | over a new client connection. Multiple operations for one set are | ||
6383 | multiplexed over one client connection, using a request id supplied by | ||
6384 | the client. | ||
6385 | |||
6386 | @node Listeners2 | ||
6387 | @subsubsection Listeners2 | ||
6388 | |||
6389 | @c %**end of header | ||
6390 | |||
6391 | Each listener also requires a seperate client connection. By sending the | ||
6392 | @code{GNUNET_SERVICE_SET_LISTEN} message, the client notifies the service | ||
6393 | of the application id and operation type it is interested in. A client | ||
6394 | rejects an incoming request by sending @code{GNUNET_SERVICE_SET_REJECT} | ||
6395 | on the listener's client connection. | ||
6396 | In contrast, when accepting an incoming request, a | ||
6397 | @code{GNUNET_SERVICE_SET_ACCEPT} message must be sent over the@ set that | ||
6398 | is supplied for the set operation. | ||
6399 | |||
6400 | @node Initiating Operations | ||
6401 | @subsubsection Initiating Operations | ||
6402 | |||
6403 | @c %**end of header | ||
6404 | |||
6405 | Operations with remote peers are initiated by sending a | ||
6406 | @code{GNUNET_SERVICE_SET_EVALUATE} message to the service. The@ client | ||
6407 | connection that this message is sent by determines the set to use. | ||
6408 | |||
6409 | @node Modifying Sets | ||
6410 | @subsubsection Modifying Sets | ||
6411 | |||
6412 | @c %**end of header | ||
6413 | |||
6414 | Sets are modified with the @code{GNUNET_SERVICE_SET_ADD} and | ||
6415 | @code{GNUNET_SERVICE_SET_REMOVE} messages. | ||
6416 | |||
6417 | |||
6418 | @c %@menu | ||
6419 | @c %* Results and Operation Status:: | ||
6420 | @c %* Iterating Sets:: | ||
6421 | @c %@end menu | ||
6422 | |||
6423 | @node Results and Operation Status | ||
6424 | @subsubsection Results and Operation Status | ||
6425 | @c %**end of header | ||
6426 | |||
6427 | The service notifies the client of result elements and success/failure of | ||
6428 | a set operation with the @code{GNUNET_SERVICE_SET_RESULT} message. | ||
6429 | |||
6430 | @node Iterating Sets | ||
6431 | @subsubsection Iterating Sets | ||
6432 | |||
6433 | @c %**end of header | ||
6434 | |||
6435 | All elements of a set can be requested by sending | ||
6436 | @code{GNUNET_SERVICE_SET_ITER_REQUEST}. The server responds with | ||
6437 | @code{GNUNET_SERVICE_SET_ITER_ELEMENT} and eventually terminates the | ||
6438 | iteration with @code{GNUNET_SERVICE_SET_ITER_DONE}. | ||
6439 | After each received element, the client | ||
6440 | must send @code{GNUNET_SERVICE_SET_ITER_ACK}. Note that only one set | ||
6441 | iteration may be active for a set at any given time. | ||
6442 | |||
6443 | @node The SET Intersection Peer-to-Peer Protocol | ||
6444 | @subsection The SET Intersection Peer-to-Peer Protocol | ||
6445 | |||
6446 | @c %**end of header | ||
6447 | |||
6448 | The intersection protocol operates over CADET and starts with a | ||
6449 | GNUNET_MESSAGE_TYPE_SET_P2P_OPERATION_REQUEST being sent by the peer | ||
6450 | initiating the operation to the peer listening for inbound requests. | ||
6451 | It includes the number of elements of the initiating peer, which is used | ||
6452 | to decide which side will send a Bloom filter first. | ||
6453 | |||
6454 | The listening peer checks if the operation type and application | ||
6455 | identifier are acceptable for its current state. | ||
6456 | If not, it responds with a GNUNET_MESSAGE_TYPE_SET_RESULT and a status of | ||
6457 | GNUNET_SET_STATUS_FAILURE (and terminates the CADET channel). | ||
6458 | |||
6459 | If the application accepts the request, the listener sends back a | ||
6460 | @code{GNUNET_MESSAGE_TYPE_SET_INTERSECTION_P2P_ELEMENT_INFO} if it has | ||
6461 | more elements in the set than the client. | ||
6462 | Otherwise, it immediately starts with the Bloom filter exchange. | ||
6463 | If the initiator receives a | ||
6464 | @code{GNUNET_MESSAGE_TYPE_SET_INTERSECTION_P2P_ELEMENT_INFO} response, | ||
6465 | it beings the Bloom filter exchange, unless the set size is indicated to | ||
6466 | be zero, in which case the intersection is considered finished after | ||
6467 | just the initial handshake. | ||
6468 | |||
6469 | |||
6470 | @menu | ||
6471 | * The Bloom filter exchange:: | ||
6472 | * Salt:: | ||
6473 | @end menu | ||
6474 | |||
6475 | @node The Bloom filter exchange | ||
6476 | @subsubsection The Bloom filter exchange | ||
6477 | |||
6478 | @c %**end of header | ||
6479 | |||
6480 | In this phase, each peer transmits a Bloom filter over the remaining | ||
6481 | keys of the local set to the other peer using a | ||
6482 | @code{GNUNET_MESSAGE_TYPE_SET_INTERSECTION_P2P_BF} message. This | ||
6483 | message additionally includes the number of elements left in the sender's | ||
6484 | set, as well as the XOR over all of the keys in that set. | ||
6485 | |||
6486 | The number of bits 'k' set per element in the Bloom filter is calculated | ||
6487 | based on the relative size of the two sets. | ||
6488 | Furthermore, the size of the Bloom filter is calculated based on 'k' and | ||
6489 | the number of elements in the set to maximize the amount of data filtered | ||
6490 | per byte transmitted on the wire (while avoiding an excessively high | ||
6491 | number of iterations). | ||
6492 | |||
6493 | The receiver of the message removes all elements from its local set that | ||
6494 | do not pass the Bloom filter test. | ||
6495 | It then checks if the set size of the sender and the XOR over the keys | ||
6496 | match what is left of his own set. If they do, he sends a | ||
6497 | @code{GNUNET_MESSAGE_TYPE_SET_INTERSECTION_P2P_DONE} back to indicate | ||
6498 | that the latest set is the final result. | ||
6499 | Otherwise, the receiver starts another Bloom fitler exchange, except | ||
6500 | this time as the sender. | ||
6501 | |||
6502 | @node Salt | ||
6503 | @subsubsection Salt | ||
6504 | |||
6505 | @c %**end of header | ||
6506 | |||
6507 | Bloomfilter operations are probablistic: With some non-zero probability | ||
6508 | the test may incorrectly say an element is in the set, even though it is | ||
6509 | not. | ||
6510 | |||
6511 | To mitigate this problem, the intersection protocol iterates exchanging | ||
6512 | Bloom filters using a different random 32-bit salt in each iteration (the | ||
6513 | salt is also included in the message). | ||
6514 | With different salts, set operations may fail for different elements. | ||
6515 | Merging the results from the executions, the probability of failure drops | ||
6516 | to zero. | ||
6517 | |||
6518 | The iterations terminate once both peers have established that they have | ||
6519 | sets of the same size, and where the XOR over all keys computes the same | ||
6520 | 512-bit value (leaving a failure probability of 2-511). | ||
6521 | |||
6522 | @node The SET Union Peer-to-Peer Protocol | ||
6523 | @subsection The SET Union Peer-to-Peer Protocol | ||
6524 | |||
6525 | @c %**end of header | ||
6526 | |||
6527 | The SET union protocol is based on Eppstein's efficient set reconciliation | ||
6528 | without prior context. You should read this paper first if you want to | ||
6529 | understand the protocol. | ||
6530 | |||
6531 | The union protocol operates over CADET and starts with a | ||
6532 | GNUNET_MESSAGE_TYPE_SET_P2P_OPERATION_REQUEST being sent by the peer | ||
6533 | initiating the operation to the peer listening for inbound requests. | ||
6534 | It includes the number of elements of the initiating peer, which is | ||
6535 | currently not used. | ||
6536 | |||
6537 | The listening peer checks if the operation type and application | ||
6538 | identifier are acceptable for its current state. If not, it responds with | ||
6539 | a @code{GNUNET_MESSAGE_TYPE_SET_RESULT} and a status of | ||
6540 | @code{GNUNET_SET_STATUS_FAILURE} (and terminates the CADET channel). | ||
6541 | |||
6542 | If the application accepts the request, it sends back a strata estimator | ||
6543 | using a message of type GNUNET_MESSAGE_TYPE_SET_UNION_P2P_SE. The | ||
6544 | initiator evaluates the strata estimator and initiates the exchange of | ||
6545 | invertible Bloom filters, sending a GNUNET_MESSAGE_TYPE_SET_UNION_P2P_IBF. | ||
6546 | |||
6547 | During the IBF exchange, if the receiver cannot invert the Bloom filter or | ||
6548 | detects a cycle, it sends a larger IBF in response (up to a defined | ||
6549 | maximum limit; if that limit is reached, the operation fails). | ||
6550 | Elements decoded while processing the IBF are transmitted to the other | ||
6551 | peer using GNUNET_MESSAGE_TYPE_SET_P2P_ELEMENTS, or requested from the | ||
6552 | other peer using GNUNET_MESSAGE_TYPE_SET_P2P_ELEMENT_REQUESTS messages, | ||
6553 | depending on the sign observed during decoding of the IBF. | ||
6554 | Peers respond to a GNUNET_MESSAGE_TYPE_SET_P2P_ELEMENT_REQUESTS message | ||
6555 | with the respective element in a GNUNET_MESSAGE_TYPE_SET_P2P_ELEMENTS | ||
6556 | message. If the IBF fully decodes, the peer responds with a | ||
6557 | GNUNET_MESSAGE_TYPE_SET_UNION_P2P_DONE message instead of another | ||
6558 | GNUNET_MESSAGE_TYPE_SET_UNION_P2P_IBF. | ||
6559 | |||
6560 | All Bloom filter operations use a salt to mingle keys before hasing them | ||
6561 | into buckets, such that future iterations have a fresh chance of | ||
6562 | succeeding if they failed due to collisions before. | ||
6563 | |||
6564 | @cindex STATISTICS subsystem | ||
6565 | @node GNUnet's STATISTICS subsystem | ||
6566 | @section GNUnet's STATISTICS subsystem | ||
6567 | |||
6568 | @c %**end of header | ||
6569 | |||
6570 | In GNUnet, the STATISTICS subsystem offers a central place for all | ||
6571 | subsystems to publish unsigned 64-bit integer run-time statistics. | ||
6572 | Keeping this information centrally means that there is a unified way for | ||
6573 | the user to obtain data on all subsystems, and individual subsystems do | ||
6574 | not have to always include a custom data export method for performance | ||
6575 | metrics and other statistics. For example, the TRANSPORT system uses | ||
6576 | STATISTICS to update information about the number of directly connected | ||
6577 | peers and the bandwidth that has been consumed by the various plugins. | ||
6578 | This information is valuable for diagnosing connectivity and performance | ||
6579 | issues. | ||
6580 | |||
6581 | Following the GNUnet service architecture, the STATISTICS subsystem is | ||
6582 | divided into an API which is exposed through the header | ||
6583 | @strong{gnunet_statistics_service.h} and the STATISTICS service | ||
6584 | @strong{gnunet-service-statistics}. The @strong{gnunet-statistics} | ||
6585 | command-line tool can be used to obtain (and change) information about | ||
6586 | the values stored by the STATISTICS service. The STATISTICS service does | ||
6587 | not communicate with other peers. | ||
6588 | |||
6589 | Data is stored in the STATISTICS service in the form of tuples | ||
6590 | @strong{(subsystem, name, value, persistence)}. The subsystem determines | ||
6591 | to which other GNUnet's subsystem the data belongs. name is the name | ||
6592 | through which value is associated. It uniquely identifies the record | ||
6593 | from among other records belonging to the same subsystem. | ||
6594 | In some parts of the code, the pair @strong{(subsystem, name)} is called | ||
6595 | a @strong{statistic} as it identifies the values stored in the STATISTCS | ||
6596 | service.The persistence flag determines if the record has to be preserved | ||
6597 | across service restarts. A record is said to be persistent if this flag | ||
6598 | is set for it; if not, the record is treated as a non-persistent record | ||
6599 | and it is lost after service restart. Persistent records are written to | ||
6600 | and read from the file @strong{statistics.data} before shutdown | ||
6601 | and upon startup. The file is located in the HOME directory of the peer. | ||
6602 | |||
6603 | An anomaly of the STATISTICS service is that it does not terminate | ||
6604 | immediately upon receiving a shutdown signal if it has any clients | ||
6605 | connected to it. It waits for all the clients that are not monitors to | ||
6606 | close their connections before terminating itself. | ||
6607 | This is to prevent the loss of data during peer shutdown --- delaying the | ||
6608 | STATISTICS service shutdown helps other services to store important data | ||
6609 | to STATISTICS during shutdown. | ||
6610 | |||
6611 | @menu | ||
6612 | * libgnunetstatistics:: | ||
6613 | * The STATISTICS Client-Service Protocol:: | ||
6614 | @end menu | ||
6615 | |||
6616 | @cindex libgnunetstatistics | ||
6617 | @node libgnunetstatistics | ||
6618 | @subsection libgnunetstatistics | ||
6619 | |||
6620 | @c %**end of header | ||
6621 | |||
6622 | @strong{libgnunetstatistics} is the library containing the API for the | ||
6623 | STATISTICS subsystem. Any process requiring to use STATISTICS should use | ||
6624 | this API by to open a connection to the STATISTICS service. | ||
6625 | This is done by calling the function @code{GNUNET_STATISTICS_create()}. | ||
6626 | This function takes the subsystem's name which is trying to use STATISTICS | ||
6627 | and a configuration. | ||
6628 | All values written to STATISTICS with this connection will be placed in | ||
6629 | the section corresponding to the given subsystem's name. | ||
6630 | The connection to STATISTICS can be destroyed with the function | ||
6631 | @code{GNUNET_STATISTICS_destroy()}. This function allows for the | ||
6632 | connection to be destroyed immediately or upon transferring all | ||
6633 | pending write requests to the service. | ||
6634 | |||
6635 | Note: STATISTICS subsystem can be disabled by setting @code{DISABLE = YES} | ||
6636 | under the @code{[STATISTICS]} section in the configuration. With such a | ||
6637 | configuration all calls to @code{GNUNET_STATISTICS_create()} return | ||
6638 | @code{NULL} as the STATISTICS subsystem is unavailable and no other | ||
6639 | functions from the API can be used. | ||
6640 | |||
6641 | |||
6642 | @menu | ||
6643 | * Statistics retrieval:: | ||
6644 | * Setting statistics and updating them:: | ||
6645 | * Watches:: | ||
6646 | @end menu | ||
6647 | |||
6648 | @node Statistics retrieval | ||
6649 | @subsubsection Statistics retrieval | ||
6650 | |||
6651 | @c %**end of header | ||
6652 | |||
6653 | Once a connection to the statistics service is obtained, information | ||
6654 | about any other system which uses statistics can be retrieved with the | ||
6655 | function GNUNET_STATISTICS_get(). | ||
6656 | This function takes the connection handle, the name of the subsystem | ||
6657 | whose information we are interested in (a @code{NULL} value will | ||
6658 | retrieve information of all available subsystems using STATISTICS), the | ||
6659 | name of the statistic we are interested in (a @code{NULL} value will | ||
6660 | retrieve all available statistics), a continuation callback which is | ||
6661 | called when all of requested information is retrieved, an iterator | ||
6662 | callback which is called for each parameter in the retrieved information | ||
6663 | and a closure for the aforementioned callbacks. The library then invokes | ||
6664 | the iterator callback for each value matching the request. | ||
6665 | |||
6666 | Call to @code{GNUNET_STATISTICS_get()} is asynchronous and can be | ||
6667 | canceled with the function @code{GNUNET_STATISTICS_get_cancel()}. | ||
6668 | This is helpful when retrieving statistics takes too long and especially | ||
6669 | when we want to shutdown and cleanup everything. | ||
6670 | |||
6671 | @node Setting statistics and updating them | ||
6672 | @subsubsection Setting statistics and updating them | ||
6673 | |||
6674 | @c %**end of header | ||
6675 | |||
6676 | So far we have seen how to retrieve statistics, here we will learn how we | ||
6677 | can set statistics and update them so that other subsystems can retrieve | ||
6678 | them. | ||
6679 | |||
6680 | A new statistic can be set using the function | ||
6681 | @code{GNUNET_STATISTICS_set()}. | ||
6682 | This function takes the name of the statistic and its value and a flag to | ||
6683 | make the statistic persistent. | ||
6684 | The value of the statistic should be of the type @code{uint64_t}. | ||
6685 | The function does not take the name of the subsystem; it is determined | ||
6686 | from the previous @code{GNUNET_STATISTICS_create()} invocation. If | ||
6687 | the given statistic is already present, its value is overwritten. | ||
6688 | |||
6689 | An existing statistics can be updated, i.e its value can be increased or | ||
6690 | decreased by an amount with the function | ||
6691 | @code{GNUNET_STATISTICS_update()}. | ||
6692 | The parameters to this function are similar to | ||
6693 | @code{GNUNET_STATISTICS_set()}, except that it takes the amount to be | ||
6694 | changed as a type @code{int64_t} instead of the value. | ||
6695 | |||
6696 | The library will combine multiple set or update operations into one | ||
6697 | message if the client performs requests at a rate that is faster than the | ||
6698 | available IPC with the STATISTICS service. Thus, the client does not have | ||
6699 | to worry about sending requests too quickly. | ||
6700 | |||
6701 | @node Watches | ||
6702 | @subsubsection Watches | ||
6703 | |||
6704 | @c %**end of header | ||
6705 | |||
6706 | As interesting feature of STATISTICS lies in serving notifications | ||
6707 | whenever a statistic of our interest is modified. | ||
6708 | This is achieved by registering a watch through the function | ||
6709 | @code{GNUNET_STATISTICS_watch()}. | ||
6710 | The parameters of this function are similar to those of | ||
6711 | @code{GNUNET_STATISTICS_get()}. | ||
6712 | Changes to the respective statistic's value will then cause the given | ||
6713 | iterator callback to be called. | ||
6714 | Note: A watch can only be registered for a specific statistic. Hence | ||
6715 | the subsystem name and the parameter name cannot be @code{NULL} in a | ||
6716 | call to @code{GNUNET_STATISTICS_watch()}. | ||
6717 | |||
6718 | A registered watch will keep notifying any value changes until | ||
6719 | @code{GNUNET_STATISTICS_watch_cancel()} is called with the same | ||
6720 | parameters that are used for registering the watch. | ||
6721 | |||
6722 | @node The STATISTICS Client-Service Protocol | ||
6723 | @subsection The STATISTICS Client-Service Protocol | ||
6724 | @c %**end of header | ||
6725 | |||
6726 | |||
6727 | @menu | ||
6728 | * Statistics retrieval2:: | ||
6729 | * Setting and updating statistics:: | ||
6730 | * Watching for updates:: | ||
6731 | @end menu | ||
6732 | |||
6733 | @node Statistics retrieval2 | ||
6734 | @subsubsection Statistics retrieval2 | ||
6735 | |||
6736 | @c %**end of header | ||
6737 | |||
6738 | To retrieve statistics, the client transmits a message of type | ||
6739 | @code{GNUNET_MESSAGE_TYPE_STATISTICS_GET} containing the given subsystem | ||
6740 | name and statistic parameter to the STATISTICS service. | ||
6741 | The service responds with a message of type | ||
6742 | @code{GNUNET_MESSAGE_TYPE_STATISTICS_VALUE} for each of the statistics | ||
6743 | parameters that match the client request for the client. The end of | ||
6744 | information retrieved is signaled by the service by sending a message of | ||
6745 | type @code{GNUNET_MESSAGE_TYPE_STATISTICS_END}. | ||
6746 | |||
6747 | @node Setting and updating statistics | ||
6748 | @subsubsection Setting and updating statistics | ||
6749 | |||
6750 | @c %**end of header | ||
6751 | |||
6752 | The subsystem name, parameter name, its value and the persistence flag are | ||
6753 | communicated to the service through the message | ||
6754 | @code{GNUNET_MESSAGE_TYPE_STATISTICS_SET}. | ||
6755 | |||
6756 | When the service receives a message of type | ||
6757 | @code{GNUNET_MESSAGE_TYPE_STATISTICS_SET}, it retrieves the subsystem | ||
6758 | name and checks for a statistic parameter with matching the name given in | ||
6759 | the message. | ||
6760 | If a statistic parameter is found, the value is overwritten by the new | ||
6761 | value from the message; if not found then a new statistic parameter is | ||
6762 | created with the given name and value. | ||
6763 | |||
6764 | In addition to just setting an absolute value, it is possible to perform a | ||
6765 | relative update by sending a message of type | ||
6766 | @code{GNUNET_MESSAGE_TYPE_STATISTICS_SET} with an update flag | ||
6767 | (@code{GNUNET_STATISTICS_SETFLAG_RELATIVE}) signifying that the value in | ||
6768 | the message should be treated as an update value. | ||
6769 | |||
6770 | @node Watching for updates | ||
6771 | @subsubsection Watching for updates | ||
6772 | |||
6773 | @c %**end of header | ||
6774 | |||
6775 | The function registers the watch at the service by sending a message of | ||
6776 | type @code{GNUNET_MESSAGE_TYPE_STATISTICS_WATCH}. The service then sends | ||
6777 | notifications through messages of type | ||
6778 | @code{GNUNET_MESSAGE_TYPE_STATISTICS_WATCH_VALUE} whenever the statistic | ||
6779 | parameter's value is changed. | ||
6780 | |||
6781 | @cindex DHT | ||
6782 | @cindex Distributed Hash Table | ||
6783 | @node GNUnet's Distributed Hash Table (DHT) | ||
6784 | @section GNUnet's Distributed Hash Table (DHT) | ||
6785 | |||
6786 | @c %**end of header | ||
6787 | |||
6788 | GNUnet includes a generic distributed hash table that can be used by | ||
6789 | developers building P2P applications in the framework. | ||
6790 | This section documents high-level features and how developers are | ||
6791 | expected to use the DHT. | ||
6792 | We have a research paper detailing how the DHT works. | ||
6793 | Also, Nate's thesis includes a detailed description and performance | ||
6794 | analysis (in chapter 6). | ||
6795 | |||
6796 | Key features of GNUnet's DHT include: | ||
6797 | |||
6798 | @itemize @bullet | ||
6799 | @item stores key-value pairs with values up to (approximately) 63k in size | ||
6800 | @item works with many underlay network topologies (small-world, random | ||
6801 | graph), underlay does not need to be a full mesh / clique | ||
6802 | @item support for extended queries (more than just a simple 'key'), | ||
6803 | filtering duplicate replies within the network (bloomfilter) and content | ||
6804 | validation (for details, please read the subsection on the block library) | ||
6805 | @item can (optionally) return paths taken by the PUT and GET operations | ||
6806 | to the application | ||
6807 | @item provides content replication to handle churn | ||
6808 | @end itemize | ||
6809 | |||
6810 | GNUnet's DHT is randomized and unreliable. Unreliable means that there is | ||
6811 | no strict guarantee that a value stored in the DHT is always | ||
6812 | found --- values are only found with high probability. | ||
6813 | While this is somewhat true in all P2P DHTs, GNUnet developers should be | ||
6814 | particularly wary of this fact (this will help you write secure, | ||
6815 | fault-tolerant code). Thus, when writing any application using the DHT, | ||
6816 | you should always consider the possibility that a value stored in the | ||
6817 | DHT by you or some other peer might simply not be returned, or returned | ||
6818 | with a significant delay. | ||
6819 | Your application logic must be written to tolerate this (naturally, some | ||
6820 | loss of performance or quality of service is expected in this case). | ||
6821 | |||
6822 | @menu | ||
6823 | * Block library and plugins:: | ||
6824 | * libgnunetdht:: | ||
6825 | * The DHT Client-Service Protocol:: | ||
6826 | * The DHT Peer-to-Peer Protocol:: | ||
6827 | @end menu | ||
6828 | |||
6829 | @node Block library and plugins | ||
6830 | @subsection Block library and plugins | ||
6831 | |||
6832 | @c %**end of header | ||
6833 | |||
6834 | @menu | ||
6835 | * What is a Block?:: | ||
6836 | * The API of libgnunetblock:: | ||
6837 | * Queries:: | ||
6838 | * Sample Code:: | ||
6839 | * Conclusion2:: | ||
6840 | @end menu | ||
6841 | |||
6842 | @node What is a Block? | ||
6843 | @subsubsection What is a Block? | ||
6844 | |||
6845 | @c %**end of header | ||
6846 | |||
6847 | Blocks are small (< 63k) pieces of data stored under a key (struct | ||
6848 | GNUNET_HashCode). Blocks have a type (enum GNUNET_BlockType) which defines | ||
6849 | their data format. Blocks are used in GNUnet as units of static data | ||
6850 | exchanged between peers and stored (or cached) locally. | ||
6851 | Uses of blocks include file-sharing (the files are broken up into blocks), | ||
6852 | the VPN (DNS information is stored in blocks) and the DHT (all | ||
6853 | information in the DHT and meta-information for the maintenance of the | ||
6854 | DHT are both stored using blocks). | ||
6855 | The block subsystem provides a few common functions that must be | ||
6856 | available for any type of block. | ||
6857 | |||
6858 | @cindex libgnunetblock API | ||
6859 | @node The API of libgnunetblock | ||
6860 | @subsubsection The API of libgnunetblock | ||
6861 | |||
6862 | @c %**end of header | ||
6863 | |||
6864 | The block library requires for each (family of) block type(s) a block | ||
6865 | plugin (implementing @file{gnunet_block_plugin.h}) that provides basic | ||
6866 | functions that are needed by the DHT (and possibly other subsystems) to | ||
6867 | manage the block. | ||
6868 | These block plugins are typically implemented within their respective | ||
6869 | subsystems. | ||
6870 | The main block library is then used to locate, load and query the | ||
6871 | appropriate block plugin. | ||
6872 | Which plugin is appropriate is determined by the block type (which is | ||
6873 | just a 32-bit integer). Block plugins contain code that specifies which | ||
6874 | block types are supported by a given plugin. The block library loads all | ||
6875 | block plugins that are installed at the local peer and forwards the | ||
6876 | application request to the respective plugin. | ||
6877 | |||
6878 | The central functions of the block APIs (plugin and main library) are to | ||
6879 | allow the mapping of blocks to their respective key (if possible) and the | ||
6880 | ability to check that a block is well-formed and matches a given | ||
6881 | request (again, if possible). | ||
6882 | This way, GNUnet can avoid storing invalid blocks, storing blocks under | ||
6883 | the wrong key and forwarding blocks in response to a query that they do | ||
6884 | not answer. | ||
6885 | |||
6886 | One key function of block plugins is that it allows GNUnet to detect | ||
6887 | duplicate replies (via the Bloom filter). All plugins MUST support | ||
6888 | detecting duplicate replies (by adding the current response to the | ||
6889 | Bloom filter and rejecting it if it is encountered again). | ||
6890 | If a plugin fails to do this, responses may loop in the network. | ||
6891 | |||
6892 | @node Queries | ||
6893 | @subsubsection Queries | ||
6894 | @c %**end of header | ||
6895 | |||
6896 | The query format for any block in GNUnet consists of four main components. | ||
6897 | First, the type of the desired block must be specified. Second, the query | ||
6898 | must contain a hash code. The hash code is used for lookups in hash | ||
6899 | tables and databases and must not be unique for the block (however, if | ||
6900 | possible a unique hash should be used as this would be best for | ||
6901 | performance). | ||
6902 | Third, an optional Bloom filter can be specified to exclude known results; | ||
6903 | replies that hash to the bits set in the Bloom filter are considered | ||
6904 | invalid. False-positives can be eliminated by sending the same query | ||
6905 | again with a different Bloom filter mutator value, which parameterizes | ||
6906 | the hash function that is used. | ||
6907 | Finally, an optional application-specific "eXtended query" (xquery) can | ||
6908 | be specified to further constrain the results. It is entirely up to | ||
6909 | the type-specific plugin to determine whether or not a given block | ||
6910 | matches a query (type, hash, Bloom filter, and xquery). | ||
6911 | Naturally, not all xquery's are valid and some types of blocks may not | ||
6912 | support Bloom filters either, so the plugin also needs to check if the | ||
6913 | query is valid in the first place. | ||
6914 | |||
6915 | Depending on the results from the plugin, the DHT will then discard the | ||
6916 | (invalid) query, forward the query, discard the (invalid) reply, cache the | ||
6917 | (valid) reply, and/or forward the (valid and non-duplicate) reply. | ||
6918 | |||
6919 | @node Sample Code | ||
6920 | @subsubsection Sample Code | ||
6921 | |||
6922 | @c %**end of header | ||
6923 | |||
6924 | The source code in @strong{plugin_block_test.c} is a good starting point | ||
6925 | for new block plugins --- it does the minimal work by implementing a | ||
6926 | plugin that performs no validation at all. | ||
6927 | The respective @strong{Makefile.am} shows how to build and install a | ||
6928 | block plugin. | ||
6929 | |||
6930 | @node Conclusion2 | ||
6931 | @subsubsection Conclusion2 | ||
6932 | |||
6933 | @c %**end of header | ||
6934 | |||
6935 | In conclusion, GNUnet subsystems that want to use the DHT need to define a | ||
6936 | block format and write a plugin to match queries and replies. For testing, | ||
6937 | the "@code{GNUNET_BLOCK_TYPE_TEST}" block type can be used; it accepts | ||
6938 | any query as valid and any reply as matching any query. | ||
6939 | This type is also used for the DHT command line tools. | ||
6940 | However, it should NOT be used for normal applications due to the lack | ||
6941 | of error checking that results from this primitive implementation. | ||
6942 | |||
6943 | @cindex libgnunetdht | ||
6944 | @node libgnunetdht | ||
6945 | @subsection libgnunetdht | ||
6946 | |||
6947 | @c %**end of header | ||
6948 | |||
6949 | The DHT API itself is pretty simple and offers the usual GET and PUT | ||
6950 | functions that work as expected. The specified block type refers to the | ||
6951 | block library which allows the DHT to run application-specific logic for | ||
6952 | data stored in the network. | ||
6953 | |||
6954 | |||
6955 | @menu | ||
6956 | * GET:: | ||
6957 | * PUT:: | ||
6958 | * MONITOR:: | ||
6959 | * DHT Routing Options:: | ||
6960 | @end menu | ||
6961 | |||
6962 | @node GET | ||
6963 | @subsubsection GET | ||
6964 | |||
6965 | @c %**end of header | ||
6966 | |||
6967 | When using GET, the main consideration for developers (other than the | ||
6968 | block library) should be that after issuing a GET, the DHT will | ||
6969 | continuously cause (small amounts of) network traffic until the operation | ||
6970 | is explicitly canceled. | ||
6971 | So GET does not simply send out a single network request once; instead, | ||
6972 | the DHT will continue to search for data. This is needed to achieve good | ||
6973 | success rates and also handles the case where the respective PUT | ||
6974 | operation happens after the GET operation was started. | ||
6975 | Developers should not cancel an existing GET operation and then | ||
6976 | explicitly re-start it to trigger a new round of network requests; | ||
6977 | this is simply inefficient, especially as the internal automated version | ||
6978 | can be more efficient, for example by filtering results in the network | ||
6979 | that have already been returned. | ||
6980 | |||
6981 | If an application that performs a GET request has a set of replies that it | ||
6982 | already knows and would like to filter, it can call@ | ||
6983 | @code{GNUNET_DHT_get_filter_known_results} with an array of hashes over | ||
6984 | the respective blocks to tell the DHT that these results are not | ||
6985 | desired (any more). | ||
6986 | This way, the DHT will filter the respective blocks using the block | ||
6987 | library in the network, which may result in a significant reduction in | ||
6988 | bandwidth consumption. | ||
6989 | |||
6990 | @node PUT | ||
6991 | @subsubsection PUT | ||
6992 | |||
6993 | @c %**end of header | ||
6994 | |||
6995 | In contrast to GET operations, developers @strong{must} manually re-run | ||
6996 | PUT operations periodically (if they intend the content to continue to be | ||
6997 | available). Content stored in the DHT expires or might be lost due to | ||
6998 | churn. | ||
6999 | Furthermore, GNUnet's DHT typically requires multiple rounds of PUT | ||
7000 | operations before a key-value pair is consistently available to all | ||
7001 | peers (the DHT randomizes paths and thus storage locations, and only | ||
7002 | after multiple rounds of PUTs there will be a sufficient number of | ||
7003 | replicas in large DHTs). An explicit PUT operation using the DHT API will | ||
7004 | only cause network traffic once, so in order to ensure basic availability | ||
7005 | and resistance to churn (and adversaries), PUTs must be repeated. | ||
7006 | While the exact frequency depends on the application, a rule of thumb is | ||
7007 | that there should be at least a dozen PUT operations within the content | ||
7008 | lifetime. Content in the DHT typically expires after one day, so | ||
7009 | DHT PUT operations should be repeated at least every 1-2 hours. | ||
7010 | |||
7011 | @node MONITOR | ||
7012 | @subsubsection MONITOR | ||
7013 | |||
7014 | @c %**end of header | ||
7015 | |||
7016 | The DHT API also allows applications to monitor messages crossing the | ||
7017 | local DHT service. | ||
7018 | The types of messages used by the DHT are GET, PUT and RESULT messages. | ||
7019 | Using the monitoring API, applications can choose to monitor these | ||
7020 | requests, possibly limiting themselves to requests for a particular block | ||
7021 | type. | ||
7022 | |||
7023 | The monitoring API is not only usefu only for diagnostics, it can also be | ||
7024 | used to trigger application operations based on PUT operations. | ||
7025 | For example, an application may use PUTs to distribute work requests to | ||
7026 | other peers. | ||
7027 | The workers would then monitor for PUTs that give them work, instead of | ||
7028 | looking for work using GET operations. | ||
7029 | This can be beneficial, especially if the workers have no good way to | ||
7030 | guess the keys under which work would be stored. | ||
7031 | Naturally, additional protocols might be needed to ensure that the desired | ||
7032 | number of workers will process the distributed workload. | ||
7033 | |||
7034 | @node DHT Routing Options | ||
7035 | @subsubsection DHT Routing Options | ||
7036 | |||
7037 | @c %**end of header | ||
7038 | |||
7039 | There are two important options for GET and PUT requests: | ||
7040 | |||
7041 | @table @asis | ||
7042 | @item GNUNET_DHT_RO_DEMULITPLEX_EVERYWHERE This option means that all | ||
7043 | peers should process the request, even if their peer ID is not closest to | ||
7044 | the key. For a PUT request, this means that all peers that a request | ||
7045 | traverses may make a copy of the data. | ||
7046 | Similarly for a GET request, all peers will check their local database | ||
7047 | for a result. Setting this option can thus significantly improve caching | ||
7048 | and reduce bandwidth consumption --- at the expense of a larger DHT | ||
7049 | database. If in doubt, we recommend that this option should be used. | ||
7050 | @item GNUNET_DHT_RO_RECORD_ROUTE This option instructs the DHT to record | ||
7051 | the path that a GET or a PUT request is taking through the overlay | ||
7052 | network. The resulting paths are then returned to the application with | ||
7053 | the respective result. This allows the receiver of a result to construct | ||
7054 | a path to the originator of the data, which might then be used for | ||
7055 | routing. Naturally, setting this option requires additional bandwidth | ||
7056 | and disk space, so applications should only set this if the paths are | ||
7057 | needed by the application logic. | ||
7058 | @item GNUNET_DHT_RO_FIND_PEER This option is an internal option used by | ||
7059 | the DHT's peer discovery mechanism and should not be used by applications. | ||
7060 | @item GNUNET_DHT_RO_BART This option is currently not implemented. It may | ||
7061 | in the future offer performance improvements for clique topologies. | ||
7062 | @end table | ||
7063 | |||
7064 | @node The DHT Client-Service Protocol | ||
7065 | @subsection The DHT Client-Service Protocol | ||
7066 | |||
7067 | @c %**end of header | ||
7068 | |||
7069 | @menu | ||
7070 | * PUTting data into the DHT:: | ||
7071 | * GETting data from the DHT:: | ||
7072 | * Monitoring the DHT:: | ||
7073 | @end menu | ||
7074 | |||
7075 | @node PUTting data into the DHT | ||
7076 | @subsubsection PUTting data into the DHT | ||
7077 | |||
7078 | @c %**end of header | ||
7079 | |||
7080 | To store (PUT) data into the DHT, the client sends a | ||
7081 | @code{struct GNUNET_DHT_ClientPutMessage} to the service. | ||
7082 | This message specifies the block type, routing options, the desired | ||
7083 | replication level, the expiration time, key, | ||
7084 | value and a 64-bit unique ID for the operation. The service responds with | ||
7085 | a @code{struct GNUNET_DHT_ClientPutConfirmationMessage} with the same | ||
7086 | 64-bit unique ID. Note that the service sends the confirmation as soon as | ||
7087 | it has locally processed the PUT request. The PUT may still be | ||
7088 | propagating through the network at this time. | ||
7089 | |||
7090 | In the future, we may want to change this to provide (limited) feedback | ||
7091 | to the client, for example if we detect that the PUT operation had no | ||
7092 | effect because the same key-value pair was already stored in the DHT. | ||
7093 | However, changing this would also require additional state and messages | ||
7094 | in the P2P interaction. | ||
7095 | |||
7096 | @node GETting data from the DHT | ||
7097 | @subsubsection GETting data from the DHT | ||
7098 | |||
7099 | @c %**end of header | ||
7100 | |||
7101 | To retrieve (GET) data from the DHT, the client sends a | ||
7102 | @code{struct GNUNET_DHT_ClientGetMessage} to the service. The message | ||
7103 | specifies routing options, a replication level (for replicating the GET, | ||
7104 | not the content), the desired block type, the key, the (optional) | ||
7105 | extended query and unique 64-bit request ID. | ||
7106 | |||
7107 | Additionally, the client may send any number of | ||
7108 | @code{struct GNUNET_DHT_ClientGetResultSeenMessage}s to notify the | ||
7109 | service about results that the client is already aware of. | ||
7110 | These messages consist of the key, the unique 64-bit ID of the request, | ||
7111 | and an arbitrary number of hash codes over the blocks that the client is | ||
7112 | already aware of. As messages are restricted to 64k, a client that | ||
7113 | already knows more than about a thousand blocks may need to send | ||
7114 | several of these messages. Naturally, the client should transmit these | ||
7115 | messages as quickly as possible after the original GET request such that | ||
7116 | the DHT can filter those results in the network early on. Naturally, as | ||
7117 | these messages are send after the original request, it is conceivalbe | ||
7118 | that the DHT service may return blocks that match those already known | ||
7119 | to the client anyway. | ||
7120 | |||
7121 | In response to a GET request, the service will send @code{struct | ||
7122 | GNUNET_DHT_ClientResultMessage}s to the client. These messages contain the | ||
7123 | block type, expiration, key, unique ID of the request and of course the | ||
7124 | value (a block). Depending on the options set for the respective | ||
7125 | operations, the replies may also contain the path the GET and/or the PUT | ||
7126 | took through the network. | ||
7127 | |||
7128 | A client can stop receiving replies either by disconnecting or by sending | ||
7129 | a @code{struct GNUNET_DHT_ClientGetStopMessage} which must contain the | ||
7130 | key and the 64-bit unique ID of the original request. Using an | ||
7131 | explicit "stop" message is more common as this allows a client to run | ||
7132 | many concurrent GET operations over the same connection with the DHT | ||
7133 | service --- and to stop them individually. | ||
7134 | |||
7135 | @node Monitoring the DHT | ||
7136 | @subsubsection Monitoring the DHT | ||
7137 | |||
7138 | @c %**end of header | ||
7139 | |||
7140 | To begin monitoring, the client sends a | ||
7141 | @code{struct GNUNET_DHT_MonitorStartStop} message to the DHT service. | ||
7142 | In this message, flags can be set to enable (or disable) monitoring of | ||
7143 | GET, PUT and RESULT messages that pass through a peer. The message can | ||
7144 | also restrict monitoring to a particular block type or a particular key. | ||
7145 | Once monitoring is enabled, the DHT service will notify the client about | ||
7146 | any matching event using @code{struct GNUNET_DHT_MonitorGetMessage}s for | ||
7147 | GET events, @code{struct GNUNET_DHT_MonitorPutMessage} for PUT events | ||
7148 | and @code{struct GNUNET_DHT_MonitorGetRespMessage} for RESULTs. Each of | ||
7149 | these messages contains all of the information about the event. | ||
7150 | |||
7151 | @node The DHT Peer-to-Peer Protocol | ||
7152 | @subsection The DHT Peer-to-Peer Protocol | ||
7153 | @c %**end of header | ||
7154 | |||
7155 | |||
7156 | @menu | ||
7157 | * Routing GETs or PUTs:: | ||
7158 | * PUTting data into the DHT2:: | ||
7159 | * GETting data from the DHT2:: | ||
7160 | @end menu | ||
7161 | |||
7162 | @node Routing GETs or PUTs | ||
7163 | @subsubsection Routing GETs or PUTs | ||
7164 | |||
7165 | @c %**end of header | ||
7166 | |||
7167 | When routing GETs or PUTs, the DHT service selects a suitable subset of | ||
7168 | neighbours for forwarding. The exact number of neighbours can be zero or | ||
7169 | more and depends on the hop counter of the query (initially zero) in | ||
7170 | relation to the (log of) the network size estimate, the desired | ||
7171 | replication level and the peer's connectivity. | ||
7172 | Depending on the hop counter and our network size estimate, the selection | ||
7173 | of the peers maybe randomized or by proximity to the key. | ||
7174 | Furthermore, requests include a set of peers that a request has already | ||
7175 | traversed; those peers are also excluded from the selection. | ||
7176 | |||
7177 | @node PUTting data into the DHT2 | ||
7178 | @subsubsection PUTting data into the DHT2 | ||
7179 | |||
7180 | @c %**end of header | ||
7181 | |||
7182 | To PUT data into the DHT, the service sends a @code{struct PeerPutMessage} | ||
7183 | of type @code{GNUNET_MESSAGE_TYPE_DHT_P2P_PUT} to the respective | ||
7184 | neighbour. | ||
7185 | In addition to the usual information about the content (type, routing | ||
7186 | options, desired replication level for the content, expiration time, key | ||
7187 | and value), the message contains a fixed-size Bloom filter with | ||
7188 | information about which peers (may) have already seen this request. | ||
7189 | This Bloom filter is used to ensure that DHT messages never loop back to | ||
7190 | a peer that has already processed the request. | ||
7191 | Additionally, the message includes the current hop counter and, depending | ||
7192 | on the routing options, the message may include the full path that the | ||
7193 | message has taken so far. | ||
7194 | The Bloom filter should already contain the identity of the previous hop; | ||
7195 | however, the path should not include the identity of the previous hop and | ||
7196 | the receiver should append the identity of the sender to the path, not | ||
7197 | its own identity (this is done to reduce bandwidth). | ||
7198 | |||
7199 | @node GETting data from the DHT2 | ||
7200 | @subsubsection GETting data from the DHT2 | ||
7201 | |||
7202 | @c %**end of header | ||
7203 | |||
7204 | A peer can search the DHT by sending @code{struct PeerGetMessage}s of type | ||
7205 | @code{GNUNET_MESSAGE_TYPE_DHT_P2P_GET} to other peers. In addition to the | ||
7206 | usual information about the request (type, routing options, desired | ||
7207 | replication level for the request, the key and the extended query), a GET | ||
7208 | request also again contains a hop counter, a Bloom filter over the peers | ||
7209 | that have processed the request already and depending on the routing | ||
7210 | options the full path traversed by the GET. | ||
7211 | Finally, a GET request includes a variable-size second Bloom filter and a | ||
7212 | so-called Bloom filter mutator value which together indicate which | ||
7213 | replies the sender has already seen. During the lookup, each block that | ||
7214 | matches they block type, key and extended query is additionally subjected | ||
7215 | to a test against this Bloom filter. | ||
7216 | The block plugin is expected to take the hash of the block and combine it | ||
7217 | with the mutator value and check if the result is not yet in the Bloom | ||
7218 | filter. The originator of the query will from time to time modify the | ||
7219 | mutator to (eventually) allow false-positives filtered by the Bloom filter | ||
7220 | to be returned. | ||
7221 | |||
7222 | Peers that receive a GET request perform a local lookup (depending on | ||
7223 | their proximity to the key and the query options) and forward the request | ||
7224 | to other peers. | ||
7225 | They then remember the request (including the Bloom filter for blocking | ||
7226 | duplicate results) and when they obtain a matching, non-filtered response | ||
7227 | a @code{struct PeerResultMessage} of type | ||
7228 | @code{GNUNET_MESSAGE_TYPE_DHT_P2P_RESULT} is forwarded to the previous | ||
7229 | hop. | ||
7230 | Whenver a result is forwarded, the block plugin is used to update the | ||
7231 | Bloom filter accordingly, to ensure that the same result is never | ||
7232 | forwarded more than once. | ||
7233 | The DHT service may also cache forwarded results locally if the | ||
7234 | "CACHE_RESULTS" option is set to "YES" in the configuration. | ||
7235 | |||
7236 | @node The GNU Name System (GNS) | ||
7237 | @section The GNU Name System (GNS) | ||
7238 | |||
7239 | @c %**end of header | ||
7240 | |||
7241 | The GNU Name System (GNS) is a decentralized database that enables users | ||
7242 | to securely resolve names to values. | ||
7243 | Names can be used to identify other users (for example, in social | ||
7244 | networking), or network services (for example, VPN services running at a | ||
7245 | peer in GNUnet, or purely IP-based services on the Internet). | ||
7246 | Users interact with GNS by typing in a hostname that ends in ".gnu" | ||
7247 | or ".zkey". | ||
7248 | |||
7249 | Videos giving an overview of most of the GNS and the motivations behind | ||
7250 | it is available here and here. | ||
7251 | The remainder of this chapter targets developers that are familiar with | ||
7252 | high level concepts of GNS as presented in these talks. | ||
7253 | @c TODO: Add links to here and here and to these. | ||
7254 | |||
7255 | GNS-aware applications should use the GNS resolver to obtain the | ||
7256 | respective records that are stored under that name in GNS. | ||
7257 | Each record consists of a type, value, expiration time and flags. | ||
7258 | |||
7259 | The type specifies the format of the value. Types below 65536 correspond | ||
7260 | to DNS record types, larger values are used for GNS-specific records. | ||
7261 | Applications can define new GNS record types by reserving a number and | ||
7262 | implementing a plugin (which mostly needs to convert the binary value | ||
7263 | representation to a human-readable text format and vice-versa). | ||
7264 | The expiration time specifies how long the record is to be valid. | ||
7265 | The GNS API ensures that applications are only given non-expired values. | ||
7266 | The flags are typically irrelevant for applications, as GNS uses them | ||
7267 | internally to control visibility and validity of records. | ||
7268 | |||
7269 | Records are stored along with a signature. | ||
7270 | The signature is generated using the private key of the authoritative | ||
7271 | zone. This allows any GNS resolver to verify the correctness of a | ||
7272 | name-value mapping. | ||
7273 | |||
7274 | Internally, GNS uses the NAMECACHE to cache information obtained from | ||
7275 | other users, the NAMESTORE to store information specific to the local | ||
7276 | users, and the DHT to exchange data between users. | ||
7277 | A plugin API is used to enable applications to define new GNS | ||
7278 | record types. | ||
7279 | |||
7280 | @menu | ||
7281 | * libgnunetgns:: | ||
7282 | * libgnunetgnsrecord:: | ||
7283 | * GNS plugins:: | ||
7284 | * The GNS Client-Service Protocol:: | ||
7285 | * Hijacking the DNS-Traffic using gnunet-service-dns:: | ||
7286 | * Serving DNS lookups via GNS on W32:: | ||
7287 | @end menu | ||
7288 | |||
7289 | @node libgnunetgns | ||
7290 | @subsection libgnunetgns | ||
7291 | |||
7292 | @c %**end of header | ||
7293 | |||
7294 | The GNS API itself is extremely simple. Clients first connec to the | ||
7295 | GNS service using @code{GNUNET_GNS_connect}. | ||
7296 | They can then perform lookups using @code{GNUNET_GNS_lookup} or cancel | ||
7297 | pending lookups using @code{GNUNET_GNS_lookup_cancel}. | ||
7298 | Once finished, clients disconnect using @code{GNUNET_GNS_disconnect}. | ||
7299 | |||
7300 | @menu | ||
7301 | * Looking up records:: | ||
7302 | * Accessing the records:: | ||
7303 | * Creating records:: | ||
7304 | * Future work:: | ||
7305 | @end menu | ||
7306 | |||
7307 | @node Looking up records | ||
7308 | @subsubsection Looking up records | ||
7309 | |||
7310 | @c %**end of header | ||
7311 | |||
7312 | @code{GNUNET_GNS_lookup} takes a number of arguments: | ||
7313 | |||
7314 | @table @asis | ||
7315 | @item handle This is simply the GNS connection handle from | ||
7316 | @code{GNUNET_GNS_connect}. | ||
7317 | @item name The client needs to specify the name to | ||
7318 | be resolved. This can be any valid DNS or GNS hostname. | ||
7319 | @item zone The client | ||
7320 | needs to specify the public key of the GNS zone against which the | ||
7321 | resolution should be done (the ".gnu" zone). | ||
7322 | Note that a key must be provided, even if the name ends in ".zkey". | ||
7323 | This should typically be the public key of the master-zone of the user. | ||
7324 | @item type This is the desired GNS or DNS record type | ||
7325 | to look for. While all records for the given name will be returned, this | ||
7326 | can be important if the client wants to resolve record types that | ||
7327 | themselves delegate resolution, such as CNAME, PKEY or GNS2DNS. | ||
7328 | Resolving a record of any of these types will only work if the respective | ||
7329 | record type is specified in the request, as the GNS resolver will | ||
7330 | otherwise follow the delegation and return the records from the | ||
7331 | respective destination, instead of the delegating record. | ||
7332 | @item only_cached This argument should typically be set to | ||
7333 | @code{GNUNET_NO}. Setting it to @code{GNUNET_YES} disables resolution via | ||
7334 | the overlay network. | ||
7335 | @item shorten_zone_key If GNS encounters new names during resolution, | ||
7336 | their respective zones can automatically be learned and added to the | ||
7337 | "shorten zone". If this is desired, clients must pass the private key of | ||
7338 | the shorten zone. If NULL is passed, shortening is disabled. | ||
7339 | @item proc This argument identifies | ||
7340 | the function to call with the result. It is given proc_cls, the number of | ||
7341 | records found (possilby zero) and the array of the records as arguments. | ||
7342 | proc will only be called once. After proc,> has been called, the lookup | ||
7343 | must no longer be cancelled. | ||
7344 | @item proc_cls The closure for proc. | ||
7345 | @end table | ||
7346 | |||
7347 | @node Accessing the records | ||
7348 | @subsubsection Accessing the records | ||
7349 | |||
7350 | @c %**end of header | ||
7351 | |||
7352 | The @code{libgnunetgnsrecord} library provides an API to manipulate the | ||
7353 | GNS record array that is given to proc. In particular, it offers | ||
7354 | functions such as converting record values to human-readable | ||
7355 | strings (and back). However, most @code{libgnunetgnsrecord} functions are | ||
7356 | not interesting to GNS client applications. | ||
7357 | |||
7358 | For DNS records, the @code{libgnunetdnsparser} library provides | ||
7359 | functions for parsing (and serializing) common types of DNS records. | ||
7360 | |||
7361 | @node Creating records | ||
7362 | @subsubsection Creating records | ||
7363 | |||
7364 | @c %**end of header | ||
7365 | |||
7366 | Creating GNS records is typically done by building the respective record | ||
7367 | information (possibly with the help of @code{libgnunetgnsrecord} and | ||
7368 | @code{libgnunetdnsparser}) and then using the @code{libgnunetnamestore} to | ||
7369 | publish the information. The GNS API is not involved in this | ||
7370 | operation. | ||
7371 | |||
7372 | @node Future work | ||
7373 | @subsubsection Future work | ||
7374 | |||
7375 | @c %**end of header | ||
7376 | |||
7377 | In the future, we want to expand @code{libgnunetgns} to allow | ||
7378 | applications to observe shortening operations performed during GNS | ||
7379 | resolution, for example so that users can receive visual feedback when | ||
7380 | this happens. | ||
7381 | |||
7382 | @node libgnunetgnsrecord | ||
7383 | @subsection libgnunetgnsrecord | ||
7384 | |||
7385 | @c %**end of header | ||
7386 | |||
7387 | The @code{libgnunetgnsrecord} library is used to manipulate GNS | ||
7388 | records (in plaintext or in their encrypted format). | ||
7389 | Applications mostly interact with @code{libgnunetgnsrecord} by using the | ||
7390 | functions to convert GNS record values to strings or vice-versa, or to | ||
7391 | lookup a GNS record type number by name (or vice-versa). | ||
7392 | The library also provides various other functions that are mostly | ||
7393 | used internally within GNS, such as converting keys to names, checking for | ||
7394 | expiration, encrypting GNS records to GNS blocks, verifying GNS block | ||
7395 | signatures and decrypting GNS records from GNS blocks. | ||
7396 | |||
7397 | We will now discuss the four commonly used functions of the API.@ | ||
7398 | @code{libgnunetgnsrecord} does not perform these operations itself, | ||
7399 | but instead uses plugins to perform the operation. | ||
7400 | GNUnet includes plugins to support common DNS record types as well as | ||
7401 | standard GNS record types. | ||
7402 | |||
7403 | @menu | ||
7404 | * Value handling:: | ||
7405 | * Type handling:: | ||
7406 | @end menu | ||
7407 | |||
7408 | @node Value handling | ||
7409 | @subsubsection Value handling | ||
7410 | |||
7411 | @c %**end of header | ||
7412 | |||
7413 | @code{GNUNET_GNSRECORD_value_to_string} can be used to convert | ||
7414 | the (binary) representation of a GNS record value to a human readable, | ||
7415 | 0-terminated UTF-8 string. | ||
7416 | NULL is returned if the specified record type is not supported by any | ||
7417 | available plugin. | ||
7418 | |||
7419 | @code{GNUNET_GNSRECORD_string_to_value} can be used to try to convert a | ||
7420 | human readable string to the respective (binary) representation of | ||
7421 | a GNS record value. | ||
7422 | |||
7423 | @node Type handling | ||
7424 | @subsubsection Type handling | ||
7425 | |||
7426 | @c %**end of header | ||
7427 | |||
7428 | @code{GNUNET_GNSRECORD_typename_to_number} can be used to obtain the | ||
7429 | numeric value associated with a given typename. For example, given the | ||
7430 | typename "A" (for DNS A reocrds), the function will return the number 1. | ||
7431 | A list of common DNS record types is | ||
7432 | @uref{http://en.wikipedia.org/wiki/List_of_DNS_record_types, here}. | ||
7433 | Note that not all DNS record types are supported by GNUnet GNSRECORD | ||
7434 | plugins at this time. | ||
7435 | |||
7436 | @code{GNUNET_GNSRECORD_number_to_typename} can be used to obtain the | ||
7437 | typename associated with a given numeric value. | ||
7438 | For example, given the type number 1, the function will return the | ||
7439 | typename "A". | ||
7440 | |||
7441 | @node GNS plugins | ||
7442 | @subsection GNS plugins | ||
7443 | |||
7444 | @c %**end of header | ||
7445 | |||
7446 | Adding a new GNS record type typically involves writing (or extending) a | ||
7447 | GNSRECORD plugin. The plugin needs to implement the | ||
7448 | @code{gnunet_gnsrecord_plugin.h} API which provides basic functions that | ||
7449 | are needed by GNSRECORD to convert typenames and values of the respective | ||
7450 | record type to strings (and back). | ||
7451 | These gnsrecord plugins are typically implemented within their respective | ||
7452 | subsystems. | ||
7453 | Examples for such plugins can be found in the GNSRECORD, GNS and | ||
7454 | CONVERSATION subsystems. | ||
7455 | |||
7456 | The @code{libgnunetgnsrecord} library is then used to locate, load and | ||
7457 | query the appropriate gnsrecord plugin. | ||
7458 | Which plugin is appropriate is determined by the record type (which is | ||
7459 | just a 32-bit integer). The @code{libgnunetgnsrecord} library loads all | ||
7460 | block plugins that are installed at the local peer and forwards the | ||
7461 | application request to the plugins. If the record type is not | ||
7462 | supported by the plugin, it should simply return an error code. | ||
7463 | |||
7464 | The central functions of the block APIs (plugin and main library) are the | ||
7465 | same four functions for converting between values and strings, and | ||
7466 | typenames and numbers documented in the previous subsection. | ||
7467 | |||
7468 | @node The GNS Client-Service Protocol | ||
7469 | @subsection The GNS Client-Service Protocol | ||
7470 | @c %**end of header | ||
7471 | |||
7472 | The GNS client-service protocol consists of two simple messages, the | ||
7473 | @code{LOOKUP} message and the @code{LOOKUP_RESULT}. Each @code{LOOKUP} | ||
7474 | message contains a unique 32-bit identifier, which will be included in the | ||
7475 | corresponding response. Thus, clients can send many lookup requests in | ||
7476 | parallel and receive responses out-of-order. | ||
7477 | A @code{LOOKUP} request also includes the public key of the GNS zone, | ||
7478 | the desired record type and fields specifying whether shortening is | ||
7479 | enabled or networking is disabled. Finally, the @code{LOOKUP} message | ||
7480 | includes the name to be resolved. | ||
7481 | |||
7482 | The response includes the number of records and the records themselves | ||
7483 | in the format created by @code{GNUNET_GNSRECORD_records_serialize}. | ||
7484 | They can thus be deserialized using | ||
7485 | @code{GNUNET_GNSRECORD_records_deserialize}. | ||
7486 | |||
7487 | @node Hijacking the DNS-Traffic using gnunet-service-dns | ||
7488 | @subsection Hijacking the DNS-Traffic using gnunet-service-dns | ||
7489 | |||
7490 | @c %**end of header | ||
7491 | |||
7492 | This section documents how the gnunet-service-dns (and the | ||
7493 | gnunet-helper-dns) intercepts DNS queries from the local system. | ||
7494 | This is merely one method for how we can obtain GNS queries. | ||
7495 | It is also possible to change @code{resolv.conf} to point to a machine | ||
7496 | running @code{gnunet-dns2gns} or to modify libc's name system switch | ||
7497 | (NSS) configuration to include a GNS resolution plugin. | ||
7498 | The method described in this chaper is more of a last-ditch catch-all | ||
7499 | approach. | ||
7500 | |||
7501 | @code{gnunet-service-dns} enables intercepting DNS traffic using policy | ||
7502 | based routing. | ||
7503 | We MARK every outgoing DNS-packet if it was not sent by our application. | ||
7504 | Using a second routing table in the Linux kernel these marked packets are | ||
7505 | then routed through our virtual network interface and can thus be | ||
7506 | captured unchanged. | ||
7507 | |||
7508 | Our application then reads the query and decides how to handle it: A | ||
7509 | query to an address ending in ".gnu" or ".zkey" is hijacked by | ||
7510 | @code{gnunet-service-gns} and resolved internally using GNS. | ||
7511 | In the future, a reverse query for an address of the configured virtual | ||
7512 | network could be answered with records kept about previous forward | ||
7513 | queries. | ||
7514 | Queries that are not hijacked by some application using the DNS service | ||
7515 | will be sent to the original recipient. | ||
7516 | The answer to the query will always be sent back through the virtual | ||
7517 | interface with the original nameserver as source address. | ||
7518 | |||
7519 | |||
7520 | @menu | ||
7521 | * Network Setup Details:: | ||
7522 | @end menu | ||
7523 | |||
7524 | @node Network Setup Details | ||
7525 | @subsubsection Network Setup Details | ||
7526 | |||
7527 | @c %**end of header | ||
7528 | |||
7529 | The DNS interceptor adds the following rules to the Linux kernel: | ||
7530 | @example | ||
7531 | iptables -t mangle -I OUTPUT 1 -p udp --sport $LOCALPORT --dport 53 \ | ||
7532 | -j ACCEPT iptables -t mangle -I OUTPUT 2 -p udp --dport 53 -j MARK \ | ||
7533 | --set-mark 3 ip rule add fwmark 3 table2 ip route add default via \ | ||
7534 | $VIRTUALDNS table2 | ||
7535 | @end example | ||
7536 | |||
7537 | @c FIXME: Rewrite to reflect display which is no longer content by line | ||
7538 | @c FIXME: due to the < 74 characters limit. | ||
7539 | Line 1 makes sure that all packets coming from a port our application | ||
7540 | opened beforehand (@code{$LOCALPORT}) will be routed normally. | ||
7541 | Line 2 marks every other packet to a DNS-Server with mark 3 (chosen | ||
7542 | arbitrarily). The third line adds a routing policy based on this mark | ||
7543 | 3 via the routing table. | ||
7544 | |||
7545 | @node Serving DNS lookups via GNS on W32 | ||
7546 | @subsection Serving DNS lookups via GNS on W32 | ||
7547 | |||
7548 | @c %**end of header | ||
7549 | |||
7550 | This section documents how the libw32nsp (and | ||
7551 | gnunet-gns-helper-service-w32) do DNS resolutions of DNS queries on the | ||
7552 | local system. This only applies to GNUnet running on W32. | ||
7553 | |||
7554 | W32 has a concept of "Namespaces" and "Namespace providers". | ||
7555 | These are used to present various name systems to applications in a | ||
7556 | generic way. | ||
7557 | Namespaces include DNS, mDNS, NLA and others. For each namespace any | ||
7558 | number of providers could be registered, and they are queried in an order | ||
7559 | of priority (which is adjustable). | ||
7560 | |||
7561 | Applications can resolve names by using WSALookupService*() family of | ||
7562 | functions. | ||
7563 | |||
7564 | However, these are WSA-only facilities. Common BSD socket functions for | ||
7565 | namespace resolutions are gethostbyname and getaddrinfo (among others). | ||
7566 | These functions are implemented internally (by default - by mswsock, | ||
7567 | which also implements the default DNS provider) as wrappers around | ||
7568 | WSALookupService*() functions (see "Sample Code for a Service Provider" | ||
7569 | on MSDN). | ||
7570 | |||
7571 | On W32 GNUnet builds a libw32nsp - a namespace provider, which can then be | ||
7572 | installed into the system by using w32nsp-install (and uninstalled by | ||
7573 | w32nsp-uninstall), as described in "Installation Handbook". | ||
7574 | |||
7575 | libw32nsp is very simple and has almost no dependencies. As a response to | ||
7576 | NSPLookupServiceBegin(), it only checks that the provider GUID passed to | ||
7577 | it by the caller matches GNUnet DNS Provider GUID, checks that name being | ||
7578 | resolved ends in ".gnu" or ".zkey", then connects to | ||
7579 | gnunet-gns-helper-service-w32 at 127.0.0.1:5353 (hardcoded) and sends the | ||
7580 | name resolution request there, returning the connected socket to the | ||
7581 | caller. | ||
7582 | |||
7583 | When the caller invokes NSPLookupServiceNext(), libw32nsp reads a | ||
7584 | completely formed reply from that socket, unmarshalls it, then gives | ||
7585 | it back to the caller. | ||
7586 | |||
7587 | At the moment gnunet-gns-helper-service-w32 is implemented to ever give | ||
7588 | only one reply, and subsequent calls to NSPLookupServiceNext() will fail | ||
7589 | with WSA_NODATA (first call to NSPLookupServiceNext() might also fail if | ||
7590 | GNS failed to find the name, or there was an error connecting to it). | ||
7591 | |||
7592 | gnunet-gns-helper-service-w32 does most of the processing: | ||
7593 | |||
7594 | @itemize @bullet | ||
7595 | @item Maintains a connection to GNS. | ||
7596 | @item Reads GNS config and loads appropriate keys. | ||
7597 | @item Checks service GUID and decides on the type of record to look up, | ||
7598 | refusing to make a lookup outright when unsupported service GUID is | ||
7599 | passed. | ||
7600 | @item Launches the lookup | ||
7601 | @end itemize | ||
7602 | |||
7603 | When lookup result arrives, gnunet-gns-helper-service-w32 forms a complete | ||
7604 | reply (including filling a WSAQUERYSETW structure and, possibly, a binary | ||
7605 | blob with a hostent structure for gethostbyname() client), marshalls it, | ||
7606 | and sends it back to libw32nsp. If no records were found, it sends an | ||
7607 | empty header. | ||
7608 | |||
7609 | This works for most normal applications that use gethostbyname() or | ||
7610 | getaddrinfo() to resolve names, but fails to do anything with | ||
7611 | applications that use alternative means of resolving names (such as | ||
7612 | sending queries to a DNS server directly by themselves). | ||
7613 | This includes some of well known utilities, like "ping" and "nslookup". | ||
7614 | |||
7615 | @node The GNS Namecache | ||
7616 | @section The GNS Namecache | ||
7617 | |||
7618 | @c %**end of header | ||
7619 | |||
7620 | The NAMECACHE subsystem is responsible for caching (encrypted) resolution | ||
7621 | results of the GNU Name System (GNS). GNS makes zone information available | ||
7622 | to other users via the DHT. However, as accessing the DHT for every | ||
7623 | lookup is expensive (and as the DHT's local cache is lost whenever the | ||
7624 | peer is restarted), GNS uses the NAMECACHE as a more persistent cache for | ||
7625 | DHT lookups. | ||
7626 | Thus, instead of always looking up every name in the DHT, GNS first | ||
7627 | checks if the result is already available locally in the NAMECACHE. | ||
7628 | Only if there is no result in the NAMECACHE, GNS queries the DHT. | ||
7629 | The NAMECACHE stores data in the same (encrypted) format as the DHT. | ||
7630 | It thus makes no sense to iterate over all items in the | ||
7631 | NAMECACHE --- the NAMECACHE does not have a way to provide the keys | ||
7632 | required to decrypt the entries. | ||
7633 | |||
7634 | Blocks in the NAMECACHE share the same expiration mechanism as blocks in | ||
7635 | the DHT --- the block expires wheneever any of the records in | ||
7636 | the (encrypted) block expires. | ||
7637 | The expiration time of the block is the only information stored in | ||
7638 | plaintext. The NAMECACHE service internally performs all of the required | ||
7639 | work to expire blocks, clients do not have to worry about this. | ||
7640 | Also, given that NAMECACHE stores only GNS blocks that local users | ||
7641 | requested, there is no configuration option to limit the size of the | ||
7642 | NAMECACHE. It is assumed to be always small enough (a few MB) to fit on | ||
7643 | the drive. | ||
7644 | |||
7645 | The NAMECACHE supports the use of different database backends via a | ||
7646 | plugin API. | ||
7647 | |||
7648 | @menu | ||
7649 | * libgnunetnamecache:: | ||
7650 | * The NAMECACHE Client-Service Protocol:: | ||
7651 | * The NAMECACHE Plugin API:: | ||
7652 | @end menu | ||
7653 | |||
7654 | @node libgnunetnamecache | ||
7655 | @subsection libgnunetnamecache | ||
7656 | |||
7657 | @c %**end of header | ||
7658 | |||
7659 | The NAMECACHE API consists of five simple functions. First, there is | ||
7660 | @code{GNUNET_NAMECACHE_connect} to connect to the NAMECACHE service. | ||
7661 | This returns the handle required for all other operations on the | ||
7662 | NAMECACHE. Using @code{GNUNET_NAMECACHE_block_cache} clients can insert a | ||
7663 | block into the cache. | ||
7664 | @code{GNUNET_NAMECACHE_lookup_block} can be used to lookup blocks that | ||
7665 | were stored in the NAMECACHE. Both operations can be cancelled using | ||
7666 | @code{GNUNET_NAMECACHE_cancel}. Note that cancelling a | ||
7667 | @code{GNUNET_NAMECACHE_block_cache} operation can result in the block | ||
7668 | being stored in the NAMECACHE --- or not. Cancellation primarily ensures | ||
7669 | that the continuation function with the result of the operation will no | ||
7670 | longer be invoked. | ||
7671 | Finally, @code{GNUNET_NAMECACHE_disconnect} closes the connection to the | ||
7672 | NAMECACHE. | ||
7673 | |||
7674 | The maximum size of a block that can be stored in the NAMECACHE is | ||
7675 | @code{GNUNET_NAMECACHE_MAX_VALUE_SIZE}, which is defined to be 63 kB. | ||
7676 | |||
7677 | @node The NAMECACHE Client-Service Protocol | ||
7678 | @subsection The NAMECACHE Client-Service Protocol | ||
7679 | |||
7680 | @c %**end of header | ||
7681 | |||
7682 | All messages in the NAMECACHE IPC protocol start with the | ||
7683 | @code{struct GNUNET_NAMECACHE_Header} which adds a request | ||
7684 | ID (32-bit integer) to the standard message header. | ||
7685 | The request ID is used to match requests with the | ||
7686 | respective responses from the NAMECACHE, as they are allowed to happen | ||
7687 | out-of-order. | ||
7688 | |||
7689 | |||
7690 | @menu | ||
7691 | * Lookup:: | ||
7692 | * Store:: | ||
7693 | @end menu | ||
7694 | |||
7695 | @node Lookup | ||
7696 | @subsubsection Lookup | ||
7697 | |||
7698 | @c %**end of header | ||
7699 | |||
7700 | The @code{struct LookupBlockMessage} is used to lookup a block stored in | ||
7701 | the cache. | ||
7702 | It contains the query hash. The NAMECACHE always responds with a | ||
7703 | @code{struct LookupBlockResponseMessage}. If the NAMECACHE has no | ||
7704 | response, it sets the expiration time in the response to zero. | ||
7705 | Otherwise, the response is expected to contain the expiration time, the | ||
7706 | ECDSA signature, the derived key and the (variable-size) encrypted data | ||
7707 | of the block. | ||
7708 | |||
7709 | @node Store | ||
7710 | @subsubsection Store | ||
7711 | |||
7712 | @c %**end of header | ||
7713 | |||
7714 | The @code{struct BlockCacheMessage} is used to cache a block in the | ||
7715 | NAMECACHE. | ||
7716 | It has the same structure as the @code{struct LookupBlockResponseMessage}. | ||
7717 | The service responds with a @code{struct BlockCacheResponseMessage} which | ||
7718 | contains the result of the operation (success or failure). | ||
7719 | In the future, we might want to make it possible to provide an error | ||
7720 | message as well. | ||
7721 | |||
7722 | @node The NAMECACHE Plugin API | ||
7723 | @subsection The NAMECACHE Plugin API | ||
7724 | @c %**end of header | ||
7725 | |||
7726 | The NAMECACHE plugin API consists of two functions, @code{cache_block} to | ||
7727 | store a block in the database, and @code{lookup_block} to lookup a block | ||
7728 | in the database. | ||
7729 | |||
7730 | |||
7731 | @menu | ||
7732 | * Lookup2:: | ||
7733 | * Store2:: | ||
7734 | @end menu | ||
7735 | |||
7736 | @node Lookup2 | ||
7737 | @subsubsection Lookup2 | ||
7738 | |||
7739 | @c %**end of header | ||
7740 | |||
7741 | The @code{lookup_block} function is expected to return at most one block | ||
7742 | to the iterator, and return @code{GNUNET_NO} if there were no non-expired | ||
7743 | results. | ||
7744 | If there are multiple non-expired results in the cache, the lookup is | ||
7745 | supposed to return the result with the largest expiration time. | ||
7746 | |||
7747 | @node Store2 | ||
7748 | @subsubsection Store2 | ||
7749 | |||
7750 | @c %**end of header | ||
7751 | |||
7752 | The @code{cache_block} function is expected to try to store the block in | ||
7753 | the database, and return @code{GNUNET_SYSERR} if this was not possible | ||
7754 | for any reason. | ||
7755 | Furthermore, @code{cache_block} is expected to implicitly perform cache | ||
7756 | maintenance and purge blocks from the cache that have expired. Note that | ||
7757 | @code{cache_block} might encounter the case where the database already has | ||
7758 | another block stored under the same key. In this case, the plugin must | ||
7759 | ensure that the block with the larger expiration time is preserved. | ||
7760 | Obviously, this can done either by simply adding new blocks and selecting | ||
7761 | for the most recent expiration time during lookup, or by checking which | ||
7762 | block is more recent during the store operation. | ||
7763 | |||
7764 | @node The REVOCATION Subsystem | ||
7765 | @section The REVOCATION Subsystem | ||
7766 | @c %**end of header | ||
7767 | |||
7768 | The REVOCATION subsystem is responsible for key revocation of Egos. | ||
7769 | If a user learns that theis private key has been compromised or has lost | ||
7770 | it, they can use the REVOCATION system to inform all of the other users | ||
7771 | that their private key is no longer valid. | ||
7772 | The subsystem thus includes ways to query for the validity of keys and to | ||
7773 | propagate revocation messages. | ||
7774 | |||
7775 | @menu | ||
7776 | * Dissemination:: | ||
7777 | * Revocation Message Design Requirements:: | ||
7778 | * libgnunetrevocation:: | ||
7779 | * The REVOCATION Client-Service Protocol:: | ||
7780 | * The REVOCATION Peer-to-Peer Protocol:: | ||
7781 | @end menu | ||
7782 | |||
7783 | @node Dissemination | ||
7784 | @subsection Dissemination | ||
7785 | |||
7786 | @c %**end of header | ||
7787 | |||
7788 | When a revocation is performed, the revocation is first of all | ||
7789 | disseminated by flooding the overlay network. | ||
7790 | The goal is to reach every peer, so that when a peer needs to check if a | ||
7791 | key has been revoked, this will be purely a local operation where the | ||
7792 | peer looks at his local revocation list. Flooding the network is also the | ||
7793 | most robust form of key revocation --- an adversary would have to control | ||
7794 | a separator of the overlay graph to restrict the propagation of the | ||
7795 | revocation message. Flooding is also very easy to implement --- peers that | ||
7796 | receive a revocation message for a key that they have never seen before | ||
7797 | simply pass the message to all of their neighbours. | ||
7798 | |||
7799 | Flooding can only distribute the revocation message to peers that are | ||
7800 | online. | ||
7801 | In order to notify peers that join the network later, the revocation | ||
7802 | service performs efficient set reconciliation over the sets of known | ||
7803 | revocation messages whenever two peers (that both support REVOCATION | ||
7804 | dissemination) connect. | ||
7805 | The SET service is used to perform this operation efficiently. | ||
7806 | |||
7807 | @node Revocation Message Design Requirements | ||
7808 | @subsection Revocation Message Design Requirements | ||
7809 | |||
7810 | @c %**end of header | ||
7811 | |||
7812 | However, flooding is also quite costly, creating O(|E|) messages on a | ||
7813 | network with |E| edges. | ||
7814 | Thus, revocation messages are required to contain a proof-of-work, the | ||
7815 | result of an expensive computation (which, however, is cheap to verify). | ||
7816 | Only peers that have expended the CPU time necessary to provide | ||
7817 | this proof will be able to flood the network with the revocation message. | ||
7818 | This ensures that an attacker cannot simply flood the network with | ||
7819 | millions of revocation messages. The proof-of-work required by GNUnet is | ||
7820 | set to take days on a typical PC to compute; if the ability to quickly | ||
7821 | revoke a key is needed, users have the option to pre-compute revocation | ||
7822 | messages to store off-line and use instantly after their key has expired. | ||
7823 | |||
7824 | Revocation messages must also be signed by the private key that is being | ||
7825 | revoked. Thus, they can only be created while the private key is in the | ||
7826 | possession of the respective user. This is another reason to create a | ||
7827 | revocation message ahead of time and store it in a secure location. | ||
7828 | |||
7829 | @node libgnunetrevocation | ||
7830 | @subsection libgnunetrevocation | ||
7831 | |||
7832 | @c %**end of header | ||
7833 | |||
7834 | The REVOCATION API consists of two parts, to query and to issue | ||
7835 | revocations. | ||
7836 | |||
7837 | |||
7838 | @menu | ||
7839 | * Querying for revoked keys:: | ||
7840 | * Preparing revocations:: | ||
7841 | * Issuing revocations:: | ||
7842 | @end menu | ||
7843 | |||
7844 | @node Querying for revoked keys | ||
7845 | @subsubsection Querying for revoked keys | ||
7846 | |||
7847 | @c %**end of header | ||
7848 | |||
7849 | @code{GNUNET_REVOCATION_query} is used to check if a given ECDSA public | ||
7850 | key has been revoked. | ||
7851 | The given callback will be invoked with the result of the check. | ||
7852 | The query can be cancelled using @code{GNUNET_REVOCATION_query_cancel} on | ||
7853 | the return value. | ||
7854 | |||
7855 | @node Preparing revocations | ||
7856 | @subsubsection Preparing revocations | ||
7857 | |||
7858 | @c %**end of header | ||
7859 | |||
7860 | It is often desirable to create a revocation record ahead-of-time and | ||
7861 | store it in an off-line location to be used later in an emergency. | ||
7862 | This is particularly true for GNUnet revocations, where performing the | ||
7863 | revocation operation itself is computationally expensive and thus is | ||
7864 | likely to take some time. | ||
7865 | Thus, if users want the ability to perform revocations quickly in an | ||
7866 | emergency, they must pre-compute the revocation message. | ||
7867 | The revocation API enables this with two functions that are used to | ||
7868 | compute the revocation message, but not trigger the actual revocation | ||
7869 | operation. | ||
7870 | |||
7871 | @code{GNUNET_REVOCATION_check_pow} should be used to calculate the | ||
7872 | proof-of-work required in the revocation message. This function takes the | ||
7873 | public key, the required number of bits for the proof of work (which in | ||
7874 | GNUnet is a network-wide constant) and finally a proof-of-work number as | ||
7875 | arguments. | ||
7876 | The function then checks if the given proof-of-work number is a valid | ||
7877 | proof of work for the given public key. Clients preparing a revocation | ||
7878 | are expected to call this function repeatedly (typically with a | ||
7879 | monotonically increasing sequence of numbers of the proof-of-work number) | ||
7880 | until a given number satisfies the check. | ||
7881 | That number should then be saved for later use in the revocation | ||
7882 | operation. | ||
7883 | |||
7884 | @code{GNUNET_REVOCATION_sign_revocation} is used to generate the | ||
7885 | signature that is required in a revocation message. | ||
7886 | It takes the private key that (possibly in the future) is to be revoked | ||
7887 | and returns the signature. | ||
7888 | The signature can again be saved to disk for later use, which will then | ||
7889 | allow performing a revocation even without access to the private key. | ||
7890 | |||
7891 | @node Issuing revocations | ||
7892 | @subsubsection Issuing revocations | ||
7893 | |||
7894 | |||
7895 | Given a ECDSA public key, the signature from @code{GNUNET_REVOCATION_sign} | ||
7896 | and the proof-of-work, | ||
7897 | @code{GNUNET_REVOCATION_revoke} can be used to perform the | ||
7898 | actual revocation. The given callback is called upon completion of the | ||
7899 | operation. @code{GNUNET_REVOCATION_revoke_cancel} can be used to stop the | ||
7900 | library from calling the continuation; however, in that case it is | ||
7901 | undefined whether or not the revocation operation will be executed. | ||
7902 | |||
7903 | @node The REVOCATION Client-Service Protocol | ||
7904 | @subsection The REVOCATION Client-Service Protocol | ||
7905 | |||
7906 | |||
7907 | The REVOCATION protocol consists of four simple messages. | ||
7908 | |||
7909 | A @code{QueryMessage} containing a public ECDSA key is used to check if a | ||
7910 | particular key has been revoked. The service responds with a | ||
7911 | @code{QueryResponseMessage} which simply contains a bit that says if the | ||
7912 | given public key is still valid, or if it has been revoked. | ||
7913 | |||
7914 | The second possible interaction is for a client to revoke a key by | ||
7915 | passing a @code{RevokeMessage} to the service. The @code{RevokeMessage} | ||
7916 | contains the ECDSA public key to be revoked, a signature by the | ||
7917 | corresponding private key and the proof-of-work, The service responds | ||
7918 | with a @code{RevocationResponseMessage} which can be used to indicate | ||
7919 | that the @code{RevokeMessage} was invalid (i.e. proof of work incorrect), | ||
7920 | or otherwise indicates that the revocation has been processed | ||
7921 | successfully. | ||
7922 | |||
7923 | @node The REVOCATION Peer-to-Peer Protocol | ||
7924 | @subsection The REVOCATION Peer-to-Peer Protocol | ||
7925 | |||
7926 | @c %**end of header | ||
7927 | |||
7928 | Revocation uses two disjoint ways to spread revocation information among | ||
7929 | peers. | ||
7930 | First of all, P2P gossip exchanged via CORE-level neighbours is used to | ||
7931 | quickly spread revocations to all connected peers. | ||
7932 | Second, whenever two peers (that both support revocations) connect, | ||
7933 | the SET service is used to compute the union of the respective revocation | ||
7934 | sets. | ||
7935 | |||
7936 | In both cases, the exchanged messages are @code{RevokeMessage}s which | ||
7937 | contain the public key that is being revoked, a matching ECDSA signature, | ||
7938 | and a proof-of-work. | ||
7939 | Whenever a peer learns about a new revocation this way, it first | ||
7940 | validates the signature and the proof-of-work, then stores it to disk | ||
7941 | (typically to a file $GNUNET_DATA_HOME/revocation.dat) and finally | ||
7942 | spreads the information to all directly connected neighbours. | ||
7943 | |||
7944 | For computing the union using the SET service, the peer with the smaller | ||
7945 | hashed peer identity will connect (as a "client" in the two-party set | ||
7946 | protocol) to the other peer after one second (to reduce traffic spikes | ||
7947 | on connect) and initiate the computation of the set union. | ||
7948 | All revocation services use a common hash to identify the SET operation | ||
7949 | over revocation sets. | ||
7950 | |||
7951 | The current implementation accepts revocation set union operations from | ||
7952 | all peers at any time; however, well-behaved peers should only initiate | ||
7953 | this operation once after establishing a connection to a peer with a | ||
7954 | larger hashed peer identity. | ||
7955 | |||
7956 | @cindex gnunet-fs | ||
7957 | @cindex FS | ||
7958 | @cindex FS subsystem | ||
7959 | @node GNUnet's File-sharing (FS) Subsystem | ||
7960 | @section GNUnet's File-sharing (FS) Subsystem | ||
7961 | |||
7962 | @c %**end of header | ||
7963 | |||
7964 | This chapter describes the details of how the file-sharing service works. | ||
7965 | As with all services, it is split into an API (libgnunetfs), the service | ||
7966 | process (gnunet-service-fs) and user interface(s). | ||
7967 | The file-sharing service uses the datastore service to store blocks and | ||
7968 | the DHT (and indirectly datacache) for lookups for non-anonymous | ||
7969 | file-sharing. | ||
7970 | Furthermore, the file-sharing service uses the block library (and the | ||
7971 | block fs plugin) for validation of DHT operations. | ||
7972 | |||
7973 | In contrast to many other services, libgnunetfs is rather complex since | ||
7974 | the client library includes a large number of high-level abstractions; | ||
7975 | this is necessary since the Fs service itself largely only operates on | ||
7976 | the block level. | ||
7977 | The FS library is responsible for providing a file-based abstraction to | ||
7978 | applications, including directories, meta data, keyword search, | ||
7979 | verification, and so on. | ||
7980 | |||
7981 | The method used by GNUnet to break large files into blocks and to use | ||
7982 | keyword search is called the | ||
7983 | "Encoding for Censorship Resistant Sharing" (ECRS). | ||
7984 | ECRS is largely implemented in the fs library; block validation is also | ||
7985 | reflected in the block FS plugin and the FS service. | ||
7986 | ECRS on-demand encoding is implemented in the FS service. | ||
7987 | |||
7988 | NOTE: The documentation in this chapter is quite incomplete. | ||
7989 | |||
7990 | @menu | ||
7991 | * Encoding for Censorship-Resistant Sharing (ECRS):: | ||
7992 | * File-sharing persistence directory structure:: | ||
7993 | @end menu | ||
7994 | |||
7995 | @cindex ecrs | ||
7996 | @cindex Encoding for Censorship-Resistant Sharing | ||
7997 | @node Encoding for Censorship-Resistant Sharing (ECRS) | ||
7998 | @subsection Encoding for Censorship-Resistant Sharing (ECRS) | ||
7999 | |||
8000 | @c %**end of header | ||
8001 | |||
8002 | When GNUnet shares files, it uses a content encoding that is called ECRS, | ||
8003 | the Encoding for Censorship-Resistant Sharing. | ||
8004 | Most of ECRS is described in the (so far unpublished) research paper | ||
8005 | attached to this page. ECRS obsoletes the previous ESED and ESED II | ||
8006 | encodings which were used in GNUnet before version 0.7.0. | ||
8007 | The rest of this page assumes that the reader is familiar with the | ||
8008 | attached paper. What follows is a description of some minor extensions | ||
8009 | that GNUnet makes over what is described in the paper. | ||
8010 | The reason why these extensions are not in the paper is that we felt | ||
8011 | that they were obvious or trivial extensions to the original scheme and | ||
8012 | thus did not warrant space in the research report. | ||
8013 | |||
8014 | @menu | ||
8015 | * Namespace Advertisements:: | ||
8016 | * KSBlocks:: | ||
8017 | @end menu | ||
8018 | |||
8019 | @node Namespace Advertisements | ||
8020 | @subsubsection Namespace Advertisements | ||
8021 | |||
8022 | @c %**end of header | ||
8023 | @c %**FIXME: all zeroses -> ? | ||
8024 | |||
8025 | An @code{SBlock} with identifier all zeros is a signed | ||
8026 | advertisement for a namespace. This special @code{SBlock} contains | ||
8027 | metadata describing the content of the namespace. | ||
8028 | Instead of the name of the identifier for a potential update, it contains | ||
8029 | the identifier for the root of the namespace. | ||
8030 | The URI should always be empty. The @code{SBlock} is signed with the | ||
8031 | content provder's RSA private key (just like any other SBlock). Peers | ||
8032 | can search for @code{SBlock}s in order to find out more about a namespace. | ||
8033 | |||
8034 | @node KSBlocks | ||
8035 | @subsubsection KSBlocks | ||
8036 | |||
8037 | @c %**end of header | ||
8038 | |||
8039 | GNUnet implements @code{KSBlocks} which are @code{KBlocks} that, instead | ||
8040 | of encrypting a CHK and metadata, encrypt an @code{SBlock} instead. | ||
8041 | In other words, @code{KSBlocks} enable GNUnet to find @code{SBlocks} | ||
8042 | using the global keyword search. | ||
8043 | Usually the encrypted @code{SBlock} is a namespace advertisement. | ||
8044 | The rationale behind @code{KSBlock}s and @code{SBlock}s is to enable | ||
8045 | peers to discover namespaces via keyword searches, and, to associate | ||
8046 | useful information with namespaces. When GNUnet finds @code{KSBlocks} | ||
8047 | during a normal keyword search, it adds the information to an internal | ||
8048 | list of discovered namespaces. Users looking for interesting namespaces | ||
8049 | can then inspect this list, reducing the need for out-of-band discovery | ||
8050 | of namespaces. | ||
8051 | Naturally, namespaces (or more specifically, namespace advertisements) can | ||
8052 | also be referenced from directories, but @code{KSBlock}s should make it | ||
8053 | easier to advertise namespaces for the owner of the pseudonym since they | ||
8054 | eliminate the need to first create a directory. | ||
8055 | |||
8056 | Collections are also advertised using @code{KSBlock}s. | ||
8057 | |||
8058 | @table @asis | ||
8059 | @item Attachment Size | ||
8060 | @item ecrs.pdf 270.68 KB | ||
8061 | @item https://gnunet.org/sites/default/files/ecrs.pdf | ||
8062 | @end table | ||
8063 | |||
8064 | @node File-sharing persistence directory structure | ||
8065 | @subsection File-sharing persistence directory structure | ||
8066 | |||
8067 | @c %**end of header | ||
8068 | |||
8069 | This section documents how the file-sharing library implements | ||
8070 | persistence of file-sharing operations and specifically the resulting | ||
8071 | directory structure. | ||
8072 | This code is only active if the @code{GNUNET_FS_FLAGS_PERSISTENCE} flag | ||
8073 | was set when calling @code{GNUNET_FS_start}. | ||
8074 | In this case, the file-sharing library will try hard to ensure that all | ||
8075 | major operations (searching, downloading, publishing, unindexing) are | ||
8076 | persistent, that is, can live longer than the process itself. | ||
8077 | More specifically, an operation is supposed to live until it is | ||
8078 | explicitly stopped. | ||
8079 | |||
8080 | If @code{GNUNET_FS_stop} is called before an operation has been stopped, a | ||
8081 | @code{SUSPEND} event is generated and then when the process calls | ||
8082 | @code{GNUNET_FS_start} next time, a @code{RESUME} event is generated. | ||
8083 | Additionally, even if an application crashes (segfault, SIGKILL, system | ||
8084 | crash) and hence @code{GNUNET_FS_stop} is never called and no | ||
8085 | @code{SUSPEND} events are generated, operations are still resumed (with | ||
8086 | @code{RESUME} events). | ||
8087 | This is implemented by constantly writing the current state of the | ||
8088 | file-sharing operations to disk. | ||
8089 | Specifically, the current state is always written to disk whenever | ||
8090 | anything significant changes (the exception are block-wise progress in | ||
8091 | publishing and unindexing, since those operations would be slowed down | ||
8092 | significantly and can be resumed cheaply even without detailed | ||
8093 | accounting). | ||
8094 | Note that if the process crashes (or is killed) during a serialization | ||
8095 | operation, FS does not guarantee that this specific operation is | ||
8096 | recoverable (no strict transactional semantics, again for performance | ||
8097 | reasons). However, all other unrelated operations should resume nicely. | ||
8098 | |||
8099 | Since we need to serialize the state continuously and want to recover as | ||
8100 | much as possible even after crashing during a serialization operation, | ||
8101 | we do not use one large file for serialization. | ||
8102 | Instead, several directories are used for the various operations. | ||
8103 | When @code{GNUNET_FS_start} executes, the master directories are scanned | ||
8104 | for files describing operations to resume. | ||
8105 | Sometimes, these operations can refer to related operations in child | ||
8106 | directories which may also be resumed at this point. | ||
8107 | Note that corrupted files are cleaned up automatically. | ||
8108 | However, dangling files in child directories (those that are not | ||
8109 | referenced by files from the master directories) are not automatically | ||
8110 | removed. | ||
8111 | |||
8112 | Persistence data is kept in a directory that begins with the "STATE_DIR" | ||
8113 | prefix from the configuration file | ||
8114 | (by default, "$SERVICEHOME/persistence/") followed by the name of the | ||
8115 | client as given to @code{GNUNET_FS_start} (for example, "gnunet-gtk") | ||
8116 | followed by the actual name of the master or child directory. | ||
8117 | |||
8118 | The names for the master directories follow the names of the operations: | ||
8119 | |||
8120 | @itemize @bullet | ||
8121 | @item "search" | ||
8122 | @item "download" | ||
8123 | @item "publish" | ||
8124 | @item "unindex" | ||
8125 | @end itemize | ||
8126 | |||
8127 | Each of the master directories contains names (chosen at random) for each | ||
8128 | active top-level (master) operation. | ||
8129 | Note that a download that is associated with a search result is not a | ||
8130 | top-level operation. | ||
8131 | |||
8132 | In contrast to the master directories, the child directories are only | ||
8133 | consulted when another operation refers to them. | ||
8134 | For each search, a subdirectory (named after the master search | ||
8135 | synchronization file) contains the search results. | ||
8136 | Search results can have an associated download, which is then stored in | ||
8137 | the general "download-child" directory. | ||
8138 | Downloads can be recursive, in which case children are stored in | ||
8139 | subdirectories mirroring the structure of the recursive download | ||
8140 | (either starting in the master "download" directory or in the | ||
8141 | "download-child" directory depending on how the download was initiated). | ||
8142 | For publishing operations, the "publish-file" directory contains | ||
8143 | information about the individual files and directories that are part of | ||
8144 | the publication. | ||
8145 | However, this directory structure is flat and does not mirror the | ||
8146 | structure of the publishing operation. | ||
8147 | Note that unindex operations cannot have associated child operations. | ||
8148 | |||
8149 | @cindex REGEX subsystem | ||
8150 | @cindex regex subsystem | ||
8151 | @node GNUnet's REGEX Subsystem | ||
8152 | @section GNUnet's REGEX Subsystem | ||
8153 | |||
8154 | @c %**end of header | ||
8155 | |||
8156 | Using the REGEX subsystem, you can discover peers that offer a particular | ||
8157 | service using regular expressions. | ||
8158 | The peers that offer a service specify it using a regular expressions. | ||
8159 | Peers that want to patronize a service search using a string. | ||
8160 | The REGEX subsystem will then use the DHT to return a set of matching | ||
8161 | offerers to the patrons. | ||
8162 | |||
8163 | For the technical details, we have Max's defense talk and Max's Master's | ||
8164 | thesis. | ||
8165 | |||
8166 | @c An additional publication is under preparation and available to | ||
8167 | @c team members (in Git). | ||
8168 | @c FIXME: Where is the file? Point to it. Assuming that it's szengel2012ms | ||
8169 | |||
8170 | @menu | ||
8171 | * How to run the regex profiler:: | ||
8172 | @end menu | ||
8173 | |||
8174 | @node How to run the regex profiler | ||
8175 | @subsection How to run the regex profiler | ||
8176 | |||
8177 | @c %**end of header | ||
8178 | |||
8179 | The gnunet-regex-profiler can be used to profile the usage of mesh/regex | ||
8180 | for a given set of regular expressions and strings. | ||
8181 | Mesh/regex allows you to announce your peer ID under a certain regex and | ||
8182 | search for peers matching a particular regex using a string. | ||
8183 | See @uref{https://gnunet.org/szengel2012ms, szengel2012ms} for a full | ||
8184 | introduction. | ||
8185 | |||
8186 | First of all, the regex profiler uses GNUnet testbed, thus all the | ||
8187 | implications for testbed also apply to the regex profiler | ||
8188 | (for example you need password-less ssh login to the machines listed in | ||
8189 | your hosts file). | ||
8190 | |||
8191 | @strong{Configuration} | ||
8192 | |||
8193 | Moreover, an appropriate configuration file is needed. | ||
8194 | Generally you can refer to the | ||
8195 | @file{contrib/regex_profiler_infiniband.conf} file in the sourcecode | ||
8196 | of GNUnet for an example configuration. | ||
8197 | In the following paragraph the important details are highlighted. | ||
8198 | |||
8199 | Announcing of the regular expressions is done by the | ||
8200 | gnunet-daemon-regexprofiler, therefore you have to make sure it is | ||
8201 | started, by adding it to the AUTOSTART set of ARM: | ||
8202 | |||
8203 | @example | ||
8204 | [regexprofiler] | ||
8205 | AUTOSTART = YES | ||
8206 | @end example | ||
8207 | |||
8208 | @noindent | ||
8209 | Furthermore you have to specify the location of the binary: | ||
8210 | |||
8211 | @example | ||
8212 | [regexprofiler] | ||
8213 | # Location of the gnunet-daemon-regexprofiler binary. | ||
8214 | BINARY = /home/szengel/gnunet/src/mesh/.libs/gnunet-daemon-regexprofiler | ||
8215 | # Regex prefix that will be applied to all regular expressions and | ||
8216 | # search string. | ||
8217 | REGEX_PREFIX = "GNVPN-0001-PAD" | ||
8218 | @end example | ||
8219 | |||
8220 | @noindent | ||
8221 | When running the profiler with a large scale deployment, you probably | ||
8222 | want to reduce the workload of each peer. | ||
8223 | Use the following options to do this. | ||
8224 | |||
8225 | @example | ||
8226 | [dht] | ||
8227 | # Force network size estimation | ||
8228 | FORCE_NSE = 1 | ||
8229 | |||
8230 | [dhtcache] | ||
8231 | DATABASE = heap | ||
8232 | # Disable RC-file for Bloom filter? (for benchmarking with limited IO | ||
8233 | # availability) | ||
8234 | DISABLE_BF_RC = YES | ||
8235 | # Disable Bloom filter entirely | ||
8236 | DISABLE_BF = YES | ||
8237 | |||
8238 | [nse] | ||
8239 | # Minimize proof-of-work CPU consumption by NSE | ||
8240 | WORKBITS = 1 | ||
8241 | @end example | ||
8242 | |||
8243 | @noindent | ||
8244 | @strong{Options} | ||
8245 | |||
8246 | To finally run the profiler some options and the input data need to be | ||
8247 | specified on the command line. | ||
8248 | |||
8249 | @example | ||
8250 | gnunet-regex-profiler -c config-file -d log-file -n num-links \ | ||
8251 | -p path-compression-length -s search-delay -t matching-timeout \ | ||
8252 | -a num-search-strings hosts-file policy-dir search-strings-file | ||
8253 | @end example | ||
8254 | |||
8255 | @noindent | ||
8256 | Where... | ||
8257 | |||
8258 | @itemize @bullet | ||
8259 | @item ... @code{config-file} means the configuration file created earlier. | ||
8260 | @item ... @code{log-file} is the file where to write statistics output. | ||
8261 | @item ... @code{num-links} indicates the number of random links between | ||
8262 | started peers. | ||
8263 | @item ... @code{path-compression-length} is the maximum path compression | ||
8264 | length in the DFA. | ||
8265 | @item ... @code{search-delay} time to wait between peers finished linking | ||
8266 | and starting to match strings. | ||
8267 | @item ... @code{matching-timeout} timeout after which to cancel the | ||
8268 | searching. | ||
8269 | @item ... @code{num-search-strings} number of strings in the | ||
8270 | search-strings-file. | ||
8271 | @item ... the @code{hosts-file} should contain a list of hosts for the | ||
8272 | testbed, one per line in the following format: | ||
8273 | |||
8274 | @itemize @bullet | ||
8275 | @item @code{user@@host_ip:port} | ||
8276 | @end itemize | ||
8277 | @item ... the @code{policy-dir} is a folder containing text files | ||
8278 | containing one or more regular expressions. A peer is started for each | ||
8279 | file in that folder and the regular expressions in the corresponding file | ||
8280 | are announced by this peer. | ||
8281 | @item ... the @code{search-strings-file} is a text file containing search | ||
8282 | strings, one in each line. | ||
8283 | @end itemize | ||
8284 | |||
8285 | @noindent | ||
8286 | You can create regular expressions and search strings for every AS in the | ||
8287 | Internet using the attached scripts. You need one of the | ||
8288 | @uref{http://data.caida.org/datasets/routing/routeviews-prefix2as/, CAIDA routeviews prefix2as} | ||
8289 | data files for this. Run | ||
8290 | |||
8291 | @example | ||
8292 | create_regex.py <filename> <output path> | ||
8293 | @end example | ||
8294 | |||
8295 | @noindent | ||
8296 | to create the regular expressions and | ||
8297 | |||
8298 | @example | ||
8299 | create_strings.py <input path> <outfile> | ||
8300 | @end example | ||
8301 | |||
8302 | @noindent | ||
8303 | to create a search strings file from the previously created | ||
8304 | regular expressions. | ||