Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
tag: curl-7_19_3
Fetching contributors…

Cannot retrieve contributors at this time

file 73 lines (58 sloc) 3.886 kb
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73
Implementation of the curl_multi_socket API

  The main ideas of the new API are simply:

   1 - The application can use whatever event system it likes as it gets info
       from libcurl about what file descriptors libcurl waits for what action
       on. (The previous API returns fd_sets which is very select()-centric).

   2 - When the application discovers action on a single socket, it calls
       libcurl and informs that there was action on this particular socket and
       libcurl can then act on that socket/transfer only and not care about
       any other transfers. (The previous API always had to scan through all
       the existing transfers.)

  The idea is that curl_multi_socket() calls a given callback with information
  about what socket to wait for what action on, and the callback only gets
  called if the status of that socket has changed.

  In the API draft from before, we have a timeout argument on a per socket
  basis and we also allowed curl_multi_socket() to pass in an 'easy handle'
  instead of socket to allow libcurl to shortcut a lookup and work on the
  affected easy handle right away. Both these turned out to be bad ideas.

  The timeout argument was removed from the socket callback since after much
  thinking I came to the conclusion that we really don't want to handle
  timeouts on a per socket basis. We need it on a per transfer (easy handle)
  basis and thus we can't provide it in the callbacks in a nice way. Instead,
  we have to offer a curl_multi_timeout() that returns the largest amount of
  time we should wait before we call the "timeout action" of libcurl, to
  trigger the proper internal timeout action on the affected transfer. To get
  this to work, I added a struct to each easy handle in which we store an
  "expire time" (if any). The structs are then "splay sorted" so that we can
  add and remove times from the linked list and yet somewhat swiftly figure
  out 1 - how long time there is until the next timer expires and 2 - which
  timer (handle) should we take care of now. Of course, the upside of all this
  is that we get a curl_multi_timeout() that should also work with old-style
  applications that use curl_multi_perform().

  We also added a timer callback that makes libcurl call the application when
  the timeout value changes, and you set that with curl_multi_setopt().

  We created an internal "socket to easy handles" hash table that given
  a socket (file descriptor) return the easy handle that waits for action on
  that socket. This hash is made using the already existing hash code
  (previously only used for the DNS cache).

  To make libcurl able to report plain sockets in the socket callback, we had
  to re-organize the internals of the curl_multi_fdset() etc so that the
  conversion from sockets to fd_sets for that function is only done in the
  last step before the data is returned. I also had to extend c-ares to get a
  function that can return plain sockets, as that library too returned only
  fd_sets and that is no longer good enough. The changes done to c-ares have
  been committed and are available in the c-ares CVS repository destined to be
  included in the c-ares 1.3.1 release.

  We have done a test runs with up to 9000 connections (with a single active
  one). The curl_multi_socket() invoke then takes less than 10 microseconds in
  average (using the read-only-1-byte-at-a-time hack). We are now below the
  60 microseconds "per socket action" goal (the extra 50 is the time libevent
  needs).

Status Right Now

  The curl_multi_socket() API is implemented according to how it is
  documented. We deem it ready to use.

    http://curl.haxx.se/libcurl/c/curl_multi_socket.html
    http://curl.haxx.se/libcurl/c/curl_multi_timeout.html
    http://curl.haxx.se/libcurl/c/curl_multi_setopt.html

What is Left for the curl_multi_socket API

  Real world usage!
Something went wrong with that request. Please try again.