HACKING 7.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201
  1. 0. The buildbot.
  2. http://tor-buildbot.freehaven.net:8010/
  3. 0.1. Useful command-lines that are non-trivial to reproduce but can
  4. help with tracking bugs or leaks.
  5. dmalloc -l ~/dmalloc.log
  6. (run the commands it tells you)
  7. ./configure --with-dmalloc
  8. valgrind --leak-check=yes --error-limit=no --show-reachable=yes src/or/tor
  9. 1. Coding conventions
  10. 1.0. Whitespace and C conformance
  11. Invoke "make check-spaces" from time to time, so it can tell you about
  12. deviations from our C whitespace style. Generally, we use:
  13. - Unix-style line endings
  14. - K&R-style indentation
  15. - No space before newlines
  16. - A blank line at the end of each file
  17. - Never more than one blank line in a row
  18. - Always spaces, never tabs
  19. - No more than 79-columns per line.
  20. - Two spaces per indent.
  21. - A space between control keywords and their corresponding paren
  22. "if (x)", "while (x)", and "switch (x)", never "if(x)", "while(x)", or
  23. "switch(x)".
  24. - A space between anything and an open brace.
  25. - No space between a function name and an opening paren. "puts(x)", not
  26. "puts (x)".
  27. - Function declarations at the start of the line.
  28. We try hard to build without warnings everywhere. In particular, if you're
  29. using gcc, you should invoke the configure script with the option
  30. "--enable-gcc-warnings". This will give a bunch of extra warning flags to
  31. the compiler, and help us find divergences from our preferred C style.
  32. 1.1. Details
  33. Use tor_malloc, tor_free, tor_strdup, and tor_gettimeofday instead of their
  34. generic equivalents. (They always succeed or exit.)
  35. You can get a full list of the compatibility functions that Tor provides
  36. by looking through src/common/util.h and src/common/compat.h.
  37. Use 'INLINE' instead of 'inline', so that we work properly on Windows.
  38. 1.2. Calling and naming conventions
  39. Whenever possible, functions should return -1 on error and 0 on success.
  40. For multi-word identifiers, use lowercase words combined with
  41. underscores. (e.g., "multi_word_identifier"). Use ALL_CAPS for macros and
  42. constants.
  43. Typenames should end with "_t".
  44. Function names should be prefixed with a module name or object name. (In
  45. general, code to manipulate an object should be a module with the same
  46. name as the object, so it's hard to tell which convention is used.)
  47. Functions that do things should have imperative-verb names
  48. (e.g. buffer_clear, buffer_resize); functions that return booleans should
  49. have predicate names (e.g. buffer_is_empty, buffer_needs_resizing).
  50. 1.3. What To Optimize
  51. Don't optimize anything if it's not in the critical path. Right now,
  52. the critical path seems to be AES, logging, and the network itself.
  53. Feel free to do your own profiling to determine otherwise.
  54. 1.4. Log conventions
  55. http://wiki.noreply.org/noreply/TheOnionRouter/TorFAQ#LogLevels
  56. No error or warning messages should be expected during normal OR or OP
  57. operation.
  58. If a library function is currently called such that failure always
  59. means ERR, then the library function should log WARN and let the caller
  60. log ERR.
  61. [XXX Proposed convention: every message of severity INFO or higher should
  62. either (A) be intelligible to end-users who don't know the Tor source; or
  63. (B) somehow inform the end-users that they aren't expected to understand
  64. the message (perhaps with a string like "internal error"). Option (A) is
  65. to be preferred to option (B). -NM]
  66. 1.5. Doxygen
  67. We use the 'doxygen' utility to generate documentation from our
  68. source code. Here's how to use it:
  69. 1. Begin every file that should be documented with
  70. /**
  71. * \file filename.c
  72. * \brief Short desccription of the file.
  73. **/
  74. (Doxygen will recognize any comment beginning with /** as special.)
  75. 2. Before any function, structure, #define, or variable you want to
  76. document, add a comment of the form:
  77. /** Describe the function's actions in imperative sentences.
  78. *
  79. * Use blank lines for paragraph breaks
  80. * - and
  81. * - hyphens
  82. * - for
  83. * - lists.
  84. *
  85. * Write <b>argument_names</b> in boldface.
  86. *
  87. * \code
  88. * place_example_code();
  89. * between_code_and_endcode_commands();
  90. * \endcode
  91. */
  92. 3. Make sure to escape the characters "<", ">", "\", "%" and "#" as "\<",
  93. "\>", "\\", "\%", and "\#".
  94. 4. To document structure members, you can use two forms:
  95. struct foo {
  96. /** You can put the comment before an element; */
  97. int a;
  98. int b; /**< Or use the less-than symbol to put the comment
  99. * after the element. */
  100. };
  101. 5. To generate documentation from the Tor source code, type:
  102. $ doxygen -g
  103. To generate a file called 'Doxyfile'. Edit that file and run
  104. 'doxygen' to generate the API documentation.
  105. 6. See the Doxygen manual for more information; this summary just
  106. scratches the surface.
  107. 2. Code notes
  108. 2.1. Dataflows
  109. 2.1.1. How Incoming data is handled
  110. There are two paths for data arriving at Tor over the network: regular
  111. TCP data, and DNS.
  112. 2.1.1.1. TCP.
  113. When Tor takes information over the network, it uses the functions
  114. read_to_buf() and read_to_buf_tls() in buffers.c. These read from a
  115. socket or an SSL* into a buffer_t, which is an mbuf-style linkedlist
  116. of memory chunks.
  117. read_to_buf() and read_to_buf_tls() are called only from
  118. connection_read_to_buf() in connection.c. It takes a connection_t
  119. pointer, and reads data into it over the network, up to the
  120. connection's current bandwidth limits. It places that data into the
  121. "inbuf" field of the connection, and then:
  122. - Adjusts the connection's want-to-read/want-to-write status as
  123. appropriate.
  124. - Increments the read and written counts for the connection as
  125. appropriate.
  126. - Adjusts bandwidth buckets as appropriate.
  127. connection_read_to_buf() is called only from connection_handle_read().
  128. The connection_handle_read() function is called whenever libevent
  129. decides (based on select, poll, epoll, kqueue, etc) that there is data
  130. to read from a connection. If any data is read,
  131. connection_handle_read() calls connection_process_inbuf() to see if
  132. any of the data can be processed. If the connection was closed,
  133. connection_handle_read() calls connection_reached_eof().
  134. Connection_process_inbuf() and connection_reached_eof() both dispatch
  135. based on the connection type to determine what to do with the data
  136. that's just arrived on the connection's inbuf field. Each type of
  137. connection has its own version of these functions. For example,
  138. directory connections process incoming data in
  139. connection_dir_process_inbuf(), while OR connections process incoming
  140. data in connection_or_process_inbuf(). These
  141. connection_*_process_inbuf() functions extract data from the
  142. connection's inbuf field (a buffer_t), using functions from buffers.c.
  143. Some of these accessor functions are straightforward data extractors
  144. (like fetch_from_buf()); others do protocol-specific parsing.
  145. 2.1.1.2. DNS
  146. Tor launches (and optionally accepts) DNS requests using the code in
  147. eventdns.c, which is a copy of libevent's evdns.c. (We don't use
  148. libevent's version because it is not yet in the versions of libevent
  149. all our users have.) DNS replies are read in nameserver_read();
  150. DNS queries are read in server_port_read().