980 Commits

Author SHA1 Message Date
Sergey KHripchenko
057fab09a8 fixes for zmq_unbind() / zmq_disconnect() usage corner cases
1. when we call zmq_bind()/zmq_connect() to create endpoint
we send ourselfs(through launch_child()) command to process_own(endpoint)
(and add it to own_t::owned)
in the application thread we could call zmq_unbind() / zmq_disconnect() _BEFORE_
we run process_own() in ZMQ thread and in this situation we will be unable to find it in
own_t::owned. in other words own_t::owned.find(endpoint) will not be deleted but it will be deleted from
socket_base_t::endpoints.

2. when you zmq_unbind() the lisnening TCP/IPC socket was terminated only in destructor...
so the whole ZMQ_LINGER time listening TCP/IPC socket was able to accept() new connections
but unable to handle them.

this all geting even worse since unfortunately zmq has a bug and '*_listener_t' object not terminated
untill the socket's zmq_close().
AT LEAST FOR PUSH SOCKETS.
Everything is ok for SUB sockets.

Easy to reproduce without my fix:

zmq_socket(PUSH)
zmq_bind(tcp);
// connect to  it from PULL socket
zmq_unbind(tcp);

sleep(forever)

// netstat -anp | grep 'tcp listening socket'

With my fix you could see that after zmq_unbind(tcp) all previously connected tcp sessions
will not be finished untill the zmq_close(socket) regardless of ZMQ_LINGER value.

(*_listener_t terminates all owned session_base_t(connect=false) and they call pipe_t::terminate()
which in turn should call session_base_t::terminated() but this never happens)
2012-04-21 18:56:10 +04:00
Sergey KHripchenko
74ae19ac1f spaces deleted 2012-04-21 18:36:20 +04:00
Sergey KHripchenko
06b2eae87d small wording change 2012-04-21 08:12:59 +04:00
Sergey KHripchenko
2faa4c487f stupid bug. I feel ashamed ^) 2012-04-21 07:55:54 +04:00
Sergey KHripchenko
2eb6b32ef7 fixes for
make[2]: Entering directory `/home/laotse/src/abs/zeromq-git/src/libzmq-build/src'
  CXX    libzmq_la-address.lo
address.cpp: In destructor 'zmq::address_t::~address_t()':
address.cpp:41:29: error: deleting object of polymorphic class type 'zmq::tcp_address_t' which has non-virtual destructor might cause undefined behaviour [-Werror=delete-non-virtual-dtor]
cc1plus: all warnings being treated as errors
2012-04-20 19:16:26 +04:00
Sergey KHripchenko
489481857a 2nd try wuth sock->unbind() and sock->disconnect(). now with blackjack and const char*'s 2012-04-20 18:59:08 +04:00
Sergey KHripchenko
7b8e728e43 implement zmq_unbind(),zmq_disconnect(), zmq->sock->getsockopt(ZMQ_LAST_ENDPOINT_ID) 2012-04-18 23:42:11 +04:00
Sergey KHripchenko
b89a53ee7a fix bug in zmq::tcp_address_t::resolve_interface() where all resolved interface ip's overwrited by 0.0.0.0 2012-04-18 14:06:39 +04:00
Sergey KHripchenko
c77dc98b5c some spaces cleanups + delete unused anymore zmq::max_sockets + some minor code chages 2012-04-13 13:26:57 +04:00
Sergey KHripchenko
acba6bdd6c Implement ZMQ_TCP_ACCEPT_FILTER setsockopt() for listening TCP sockets.
Assign arbitrary number of filters that will be applied for each new TCP transport
connection on a listening socket.
If no filters applied, then TCP transport allows connections from any ip.
If at least one filter is applied then new connection source ip should be matched.
To clear all filters call zmq_setsockopt(socket, ZMQ_TCP_ACCEPT_FILTER, NULL, 0).
Filter is a null-terminated string with ipv6 or ipv4 CIDR.

For example:
localhost
127.0.0.1
mail.ru/24
::1
::1/128
3ffe:1::
3ffe:1::/56

Returns -1 if the filter couldn't be assigned(format error or ipv6 filter with ZMQ_IPV4ONLY set)

P.S.
The only thing that worries me is that I had to re-enable 'default assign by reference constructor/operator'
for 'tcp_address_t' (and for my inherited class tcp_address_mask_t) to store it in std::vector in 'options_t'...
2012-04-12 18:37:14 +04:00
Martin Hurton
cfa6f4bf51 Fix issue #335
The CreateEvent function requests EVENT_ALL_ACCESS access rights
when the event object already exists. This causes problems
when the event object is created from a service.
The solution is to call OpenEvent function when the CreateEvent
failed due to access control.
The proper solution would be to use CreateEventEx function, but
this one is not available on Windows XP.
2012-04-11 11:41:10 +02:00
Mikko Koppanen
70cf2d41ba ETIMEDOUT is a valid error code here. Fixes intermittent assertion failure on laggy networks 2012-04-10 09:55:10 +09:00
Pieter Hintjens
3cd0b1b647 Merge pull request #306 from shripchenko/master
Implement ZMQ_TCP_KEEPALIVE* family of options to adjust TCP keepalives
2012-04-09 10:14:35 -07:00
Sergey KHripchenko
4b303402a7 more flexible keepalive options detection + awful options documentation 2012-04-06 20:04:35 +04:00
Sergey KHripchenko
0c3d917926 initial implementation of tuning TCP keep-alives for TCP socket
currently not fully cross-platform
work on linux
possibly work in *bsd
and could be enhanced to work on windows
2012-04-05 19:39:53 +04:00
Martin Hurton
12c0c6aa74 Fix identity exchange for inproc transport
The socket connecting using the inproc transport never
received the identity of the remote peer.
2012-04-05 16:00:41 +02:00
Martin Hurton
b41036aa2a Flush identity message for inproc transport
The scoket implementation for inproc transfer failed to flush
identity message. The result was that the identity message
was not delivered until after the user sent the first message.

The identity message was never delivered if the user
used the socket only to receive messages.
2012-04-05 01:19:13 +02:00
Pieter Hintjens
a969028977 Merge pull request #302 from hurtonm/router_bugfix
Fix error in router socket introduced in the previous commit
2012-04-03 00:34:21 -07:00
Martin Hurton
9a60b3a2a1 Fix error in router socket introduced in the previous commit 2012-04-03 09:29:33 +02:00
Ian Barber
2faedd9e08 Merge pull request #300 from hurtonm/router_send_fixes
router: always respect message boundaries
2012-04-02 13:23:52 -07:00
Martin Hurton
600488fa9c router: always respect message boundaries
The current implementaion of router socket does not
handle the full pipe and unroutable messages properly.
Namely, in those cases, the socket could route some
message parts into a wrong connection.
2012-04-02 21:07:32 +02:00
Martin Hurton
f037290dc3 router: reimplement peer identification
The new implementation allows one to send messages through
a router socket to a peer even before receiving
messages from this peer.

Fixes issue #304
2012-04-02 20:45:10 +02:00
Pieter Hintjens
9728706b10 Added optval checking in zmq_ctx_set 2012-03-29 07:46:46 -05:00
Pieter Hintjens
eb8217bb28 Fixed zmq_init(-1) issue 2012-03-29 00:47:17 -05:00
Ian Barber
cecc790ca3 Merge pull request #297 from hurtonm/code_cleanup
Code cleanup
2012-03-28 03:51:13 -07:00
Martin Hurton
1e2564a55f Fix return value of send for router socket 2012-03-28 07:45:57 +02:00
Martin Hurton
77d93d70f0 Simplify use of posix_assert in mutex.hpp
It is the job of the posix_assert macro to check
the value. No need to do it twice.

The patch also fixes some whitespace problems.
2012-03-28 07:28:43 +02:00
Martin Hurton
e7674025e5 Remove the launch_sibling method
The method is not used anymore.
2012-03-28 07:28:43 +02:00
Martin Hurton
6f47f1281b Do not pass a message to the check_write method
The check_write method does not use the passed message.
The parameter was needed to implement the swap.
As the swap is not supported anymore, it is safe to remove this parameter.
2012-03-28 07:28:42 +02:00
Martin Hurton
9c46e87588 tcp_address: minor cleanups 2012-03-27 07:44:39 +02:00
Martin Hurton
700e08c3c2 tcp_address: make port number conversion more robust
This still rejects 00 as port number.
2012-03-27 07:44:10 +02:00
Martin Hurton
c428f6aece tcp_address: check address length before manipulating it 2012-03-27 07:37:28 +02:00
Pieter Hintjens
2bcc300a0f Fixed issue #348 2012-03-26 16:31:12 -05:00
Ian Barber
4c93fc2587 Merge pull request #291 from rcxdude/bugfix
Fix for issue #307
2012-03-26 08:12:53 -07:00
Ian Barber
64732aeca3 Merge pull request #293 from hurtonm/fix_getaddrinfo_for_ipv6_on_freebsd
Don't set the AI_V4MAPPED flag on FreeBSD
2012-03-26 08:12:12 -07:00
Martin Hurton
c2e9997a3c Don't set the AI_V4MAPPED flag on FreeBSD
The problem is that even though the AI_V4MAPPED flag is defined
on FreeBSD, the getaddrinfo function does not support it and
returns EAI_BADFLAGS.

The patch also sets the flag on Windows if it is defined there.
This is true for Windows Vista and later.

Fixes issue #331.
2012-03-26 00:14:17 +02:00
Martin Hurton
00b4571bf1 Fixed issue #334 2012-03-25 20:06:06 +02:00
Douglas Young
084e1c2193 Fix for issue #307
dist was skipping over pipes when one failed because the non-working pipe got
swapped with a working pipe but the write was never retried on that pipe
2012-03-25 18:11:14 +01:00
Pieter Hintjens
82c06e4417 Fixed issue #325 2012-03-23 17:29:29 -05:00
Pieter Hintjens
3585ec8aea Fixed Makefile for ROUTER/DEALER rename 2012-03-22 16:48:29 -05:00
Pieter Hintjens
27c28bdc2e Mark sockets as invalid when closed, not when destroyed
Previously, sockets were still "valid" after being closed and only marked
as invalid when destroyed. This meant programs could access closed sockets.
Now the socket is marked "invalid" when closed.
2012-03-22 15:56:30 -05:00
Pieter Hintjens
c12fedc70a Completed internal renaming of XREP/XREQ to ROUTER/DEALER 2012-03-22 11:36:19 -05:00
Pieter Hintjens
75809b27da Fixed issue LIBZMQ-345 - race condition in ctx.socket/term allows segfault 2012-03-22 11:04:01 -05:00
Ian Barber
3b483a8dd7 Merge pull request #285 from hurtonm/ctx_patches
Ctx patches
2012-03-22 03:32:43 -07:00
Martin Hurton
151d0717bb Minor cleanups
* corrects whitespace errors
* fixes spelling errors in comments
* prefers #ifdef to #if defined
2012-03-22 07:47:39 +01:00
Martin Hurton
ae2b938330 register_endpoint: simplify locking 2012-03-22 07:47:39 +01:00
Martin Hurton
e56f698023 choose_io_thread: do not assert when no I/O thread was chosen 2012-03-22 07:47:39 +01:00
Martin Hurton
f944336ee0 Fix race condition in find_endpoint 2012-03-22 07:47:38 +01:00
Pieter Hintjens
860e1d24c0 Fixed issue #344 2012-03-21 17:49:50 -05:00
Ian Barber
8a64d80fba Merge pull request #283 from pieterh/master
Fixed issue LIBZMQ-343 - zmq_msg_get/set are too complex
2012-03-21 13:21:21 -07:00