HACKING 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309
  1. 0. Useful tools.
  2. 0.0 The buildbot.
  3. https://buildbot.vidalia-project.net/one_line_per_build
  4. 0.1. Useful command-lines that are non-trivial to reproduce but can
  5. help with tracking bugs or leaks.
  6. 0.1.1. Dmalloc
  7. dmalloc -l ~/dmalloc.log
  8. (run the commands it tells you)
  9. ./configure --with-dmalloc
  10. 0.2.2. Valgrind
  11. valgrind --leak-check=yes --error-limit=no --show-reachable=yes src/or/tor
  12. (Note that if you get a zillion openssl warnings, you will also need to
  13. pass --undef-value-errors=no to valgrind, or rebuild your openssl
  14. with -DPURIFY.)
  15. 0.2. Running gcov for unit test coverage
  16. make clean
  17. make CFLAGS='-g -fprofile-arcs -ftest-coverage'
  18. ./src/test/test
  19. cd src/common; gcov *.[ch]
  20. cd ../or; gcov *.[ch]
  21. Then, look at the .gcov files. '-' before a line means that the
  22. compiler generated no code for that line. '######' means that the
  23. line was never reached. Lines with numbers were called that number
  24. of times.
  25. 1. Coding conventions
  26. 1.0. Whitespace and C conformance
  27. Invoke "make check-spaces" from time to time, so it can tell you about
  28. deviations from our C whitespace style. Generally, we use:
  29. - Unix-style line endings
  30. - K&R-style indentation
  31. - No space before newlines
  32. - A blank line at the end of each file
  33. - Never more than one blank line in a row
  34. - Always spaces, never tabs
  35. - No more than 79-columns per line.
  36. - Two spaces per indent.
  37. - A space between control keywords and their corresponding paren
  38. "if (x)", "while (x)", and "switch (x)", never "if(x)", "while(x)", or
  39. "switch(x)".
  40. - A space between anything and an open brace.
  41. - No space between a function name and an opening paren. "puts(x)", not
  42. "puts (x)".
  43. - Function declarations at the start of the line.
  44. We try hard to build without warnings everywhere. In particular, if you're
  45. using gcc, you should invoke the configure script with the option
  46. "--enable-gcc-warnings". This will give a bunch of extra warning flags to
  47. the compiler, and help us find divergences from our preferred C style.
  48. 1.0.1. Getting emacs to edit Tor source properly.
  49. Hi, folks! Nick here. I like to put the following snippet in my .emacs
  50. file:
  51. (add-hook 'c-mode-hook
  52. (lambda ()
  53. (font-lock-mode 1)
  54. (set-variable 'show-trailing-whitespace t)
  55. (let ((fname (expand-file-name (buffer-file-name))))
  56. (cond
  57. ((string-match "^/home/nickm/src/libevent" fname)
  58. (set-variable 'indent-tabs-mode t)
  59. (set-variable 'c-basic-offset 4)
  60. (set-variable 'tab-width 4))
  61. ((string-match "^/home/nickm/src/tor" fname)
  62. (set-variable 'indent-tabs-mode nil)
  63. (set-variable 'c-basic-offset 2))
  64. ((string-match "^/home/nickm/src/openssl" fname)
  65. (set-variable 'indent-tabs-mode t)
  66. (set-variable 'c-basic-offset 8)
  67. (set-variable 'tab-width 8))
  68. ))))
  69. You'll note that it defaults to showing all trailing whitespace. The
  70. "cond" test detects whether the file is one of a few C free software
  71. projects that I often edit, and sets up the indentation level and tab
  72. preferences to match what they want.
  73. If you want to try this out, you'll need to change the filename regex
  74. patterns to match where you keep your Tor files.
  75. If you *only* use emacs to edit Tor, you could always just say:
  76. (add-hook 'c-mode-hook
  77. (lambda ()
  78. (font-lock-mode 1)
  79. (set-variable 'show-trailing-whitespace t)
  80. (set-variable 'indent-tabs-mode nil)
  81. (set-variable 'c-basic-offset 2)))
  82. There is probably a better way to do this. No, we are probably not going
  83. to clutter the files with emacs stuff.
  84. 1.1. Details
  85. Use tor_malloc, tor_free, tor_strdup, and tor_gettimeofday instead of their
  86. generic equivalents. (They always succeed or exit.)
  87. You can get a full list of the compatibility functions that Tor provides by
  88. looking through src/common/util.h and src/common/compat.h. You can see the
  89. available containers in src/common/containers.h. You should probably
  90. familiarize yourself with these modules before you write too much code,
  91. or else you'll wind up reinventing the wheel.
  92. Use 'INLINE' instead of 'inline', so that we work properly on Windows.
  93. 1.2. Calling and naming conventions
  94. Whenever possible, functions should return -1 on error and 0 on success.
  95. For multi-word identifiers, use lowercase words combined with
  96. underscores. (e.g., "multi_word_identifier"). Use ALL_CAPS for macros and
  97. constants.
  98. Typenames should end with "_t".
  99. Function names should be prefixed with a module name or object name. (In
  100. general, code to manipulate an object should be a module with the same
  101. name as the object, so it's hard to tell which convention is used.)
  102. Functions that do things should have imperative-verb names
  103. (e.g. buffer_clear, buffer_resize); functions that return booleans should
  104. have predicate names (e.g. buffer_is_empty, buffer_needs_resizing).
  105. If you find that you have four or more possible return code values, it's
  106. probably time to create an enum. If you find that you are passing three or
  107. more flags to a function, it's probably time to create a flags argument
  108. that takes a bitfield.
  109. 1.3. What To Optimize
  110. Don't optimize anything if it's not in the critical path. Right now,
  111. the critical path seems to be AES, logging, and the network itself.
  112. Feel free to do your own profiling to determine otherwise.
  113. 1.4. Log conventions
  114. https://wiki.torproject.org/noreply/TheOnionRouter/TorFAQ#LogLevels
  115. No error or warning messages should be expected during normal OR or OP
  116. operation.
  117. If a library function is currently called such that failure always
  118. means ERR, then the library function should log WARN and let the caller
  119. log ERR.
  120. [XXX Proposed convention: every message of severity INFO or higher should
  121. either (A) be intelligible to end-users who don't know the Tor source; or
  122. (B) somehow inform the end-users that they aren't expected to understand
  123. the message (perhaps with a string like "internal error"). Option (A) is
  124. to be preferred to option (B). -NM]
  125. 1.5. Doxygen
  126. We use the 'doxygen' utility to generate documentation from our
  127. source code. Here's how to use it:
  128. 1. Begin every file that should be documented with
  129. /**
  130. * \file filename.c
  131. * \brief Short description of the file.
  132. **/
  133. (Doxygen will recognize any comment beginning with /** as special.)
  134. 2. Before any function, structure, #define, or variable you want to
  135. document, add a comment of the form:
  136. /** Describe the function's actions in imperative sentences.
  137. *
  138. * Use blank lines for paragraph breaks
  139. * - and
  140. * - hyphens
  141. * - for
  142. * - lists.
  143. *
  144. * Write <b>argument_names</b> in boldface.
  145. *
  146. * \code
  147. * place_example_code();
  148. * between_code_and_endcode_commands();
  149. * \endcode
  150. */
  151. 3. Make sure to escape the characters "<", ">", "\", "%" and "#" as "\<",
  152. "\>", "\\", "\%", and "\#".
  153. 4. To document structure members, you can use two forms:
  154. struct foo {
  155. /** You can put the comment before an element; */
  156. int a;
  157. int b; /**< Or use the less-than symbol to put the comment
  158. * after the element. */
  159. };
  160. 5. To generate documentation from the Tor source code, type:
  161. $ doxygen -g
  162. To generate a file called 'Doxyfile'. Edit that file and run
  163. 'doxygen' to generate the API documentation.
  164. 6. See the Doxygen manual for more information; this summary just
  165. scratches the surface.
  166. 1.5.1. Doxygen comment conventions
  167. Say what functions do as a series of one or more imperative sentences, as
  168. though you were telling somebody how to be the function. In other words,
  169. DO NOT say:
  170. /** The strtol function parses a number.
  171. *
  172. * nptr -- the string to parse. It can include whitespace.
  173. * endptr -- a string pointer to hold the first thing that is not part
  174. * of the number, if present.
  175. * base -- the numeric base.
  176. * returns: the resulting number.
  177. */
  178. long strtol(const char *nptr, char **nptr, int base);
  179. Instead, please DO say:
  180. /** Parse a number in radix <b>base</b> from the string <b>nptr</b>,
  181. * and return the result. Skip all leading whitespace. If
  182. * <b>endptr</b> is not NULL, set *<b>endptr</b> to the first character
  183. * after the number parsed.
  184. **/
  185. long strtol(const char *nptr, char **nptr, int base);
  186. Doxygen comments are the contract in our abstraction-by-contract world: if
  187. the functions that call your function rely on it doing something, then your
  188. function should mention that it does that something in the documentation.
  189. If you rely on a function doing something beyond what is in its
  190. documentation, then you should watch out, or it might do something else
  191. later.
  192. 2. Code notes
  193. 2.1. Dataflows
  194. 2.1.1. How Incoming data is handled
  195. There are two paths for data arriving at Tor over the network: regular
  196. TCP data, and DNS.
  197. 2.1.1.1. TCP.
  198. When Tor takes information over the network, it uses the functions
  199. read_to_buf() and read_to_buf_tls() in buffers.c. These read from a
  200. socket or an SSL* into a buffer_t, which is an mbuf-style linkedlist
  201. of memory chunks.
  202. read_to_buf() and read_to_buf_tls() are called only from
  203. connection_read_to_buf() in connection.c. It takes a connection_t
  204. pointer, and reads data into it over the network, up to the
  205. connection's current bandwidth limits. It places that data into the
  206. "inbuf" field of the connection, and then:
  207. - Adjusts the connection's want-to-read/want-to-write status as
  208. appropriate.
  209. - Increments the read and written counts for the connection as
  210. appropriate.
  211. - Adjusts bandwidth buckets as appropriate.
  212. connection_read_to_buf() is called only from connection_handle_read().
  213. The connection_handle_read() function is called whenever libevent
  214. decides (based on select, poll, epoll, kqueue, etc) that there is data
  215. to read from a connection. If any data is read,
  216. connection_handle_read() calls connection_process_inbuf() to see if
  217. any of the data can be processed. If the connection was closed,
  218. connection_handle_read() calls connection_reached_eof().
  219. Connection_process_inbuf() and connection_reached_eof() both dispatch
  220. based on the connection type to determine what to do with the data
  221. that's just arrived on the connection's inbuf field. Each type of
  222. connection has its own version of these functions. For example,
  223. directory connections process incoming data in
  224. connection_dir_process_inbuf(), while OR connections process incoming
  225. data in connection_or_process_inbuf(). These
  226. connection_*_process_inbuf() functions extract data from the
  227. connection's inbuf field (a buffer_t), using functions from buffers.c.
  228. Some of these accessor functions are straightforward data extractors
  229. (like fetch_from_buf()); others do protocol-specific parsing.
  230. 2.1.1.2. DNS
  231. Tor launches (and optionally accepts) DNS requests using the code in
  232. eventdns.c, which is a copy of libevent's evdns.c. (We don't use
  233. libevent's version because it is not yet in the versions of libevent
  234. all our users have.) DNS replies are read in nameserver_read();
  235. DNS queries are read in server_port_read().