summaryrefslogtreecommitdiff
path: root/src
AgeCommit message (Collapse)Author
2012-05-25Implement SP wire protocolMartin Lucina
Implements the SP wire protocol, and infrastructure for legacy wire protocol support. Also added an XS_SERVICE_ID socket option to set the service id and renamed the XS_PROTOCOL option to XS_PATTERN_VERSION. The following pattern versions are supported: PAIR: v3 PUBSUB: v1 (legacy), v4 REQREP: v2 PIPELINE: v3 SURVEY: v2 Note that all existing pattern versions have been bumped by 1 to allow for use of legacy protocols (otherwise there would be no way to distinguish between e.g. PUBSUB v3 and PUBSUB v3 using SP). Signed-off-by: Martin Lucina <martin@lucina.net>
2012-05-21Fixes a conflict with XenStore (seeMartin Lucina
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=668550) Signed-off-by: Martin Lucina <martin@lucina.net>
2012-05-21Expired survey returns ETIMEDOUT instead of EAGAINMartin Sustrik
Up to now, when survey in surveyor pattern expired, xs_recv() returned EAGAIN. That made it impossible to distinguish epired survey from expired receive timeout (XS_RCVTIMEO). This patch make errors different in both cases. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-21Improve error handling on WindowsSteven McCoy
2012-05-21Use global event to synchronise signaler creation on WindowsMartin Sustrik
So far, the event used for signaler creation was local to the session, so, in theory, processes in two different sessions could accidentally interconnect their signalers. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-20Install plugins under $(libdir) to simplify multi-arch packagingGabriele Svelto
Signed-off-by: Gabriele Svelto <gabriele.svelto@gmail.com>
2012-05-18Opening plug-in directory made more robustMartin Sustrik
If plugin directory doesn't exist, is a file or user has no permissions to access it, the plugins are not loaded. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-15Do not crash when multiple peers connect to PAIR socketMartin Hurton
When more then one peer connected to a PAIR socket, an application aborted due to assertion failure. This patch changes the PAIR socket behaviour so that it rejects any further connection requests.
2012-05-13Move inclusion of select() and poll() headers to a single placeGabriele Svelto
Signed-off-by: Gabriele Svelto <gabriele.svelto@gmail.com>
2012-05-12Enabled plug-in loading on all hosts with support for dlopen()Gabriele Svelto
Previously the code used for loading plug-ins would be built only on Linux hosts. Now the required functions are checked for and plug-in loading code is enabled if those are present on all platforms. Using lstat() to check for file types also increases portability to hosts that do not have the dirent.d_type field. If shared libraries support is disabled plug-ins will also be automatically disabled. Signed-off-by: Gabriele Svelto <gabriele.svelto@gmail.com>
2012-05-12Inheritance on signaler's socket on Windows switched offMartin Sustrik
Accidentally, one of the sockets in Windows signaler_t was opened usign WSASocket instead of xs::open_socket. That skipped the standard XS socket initialisation and caused the socket to be inheritable to the child processes. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-12Fix decoder to properly handle large messagesMartin Hurton
The decoder did not properly decode large messages on systems where sizeof (size_t) < sizeof (uint64_t).
2012-05-11Command throttling breaking HWM algorithms -- fixed.Martin Sustrik
When HWM was set to small value it may have happened that command indicating that pipe is ready for writing wasn't processed because of command throttling. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-11Improve command processing efficiencyMartin Sustrik
When processing commands with throttling switched off, RDTSC instruction was executed, but the result was never used. This patch eliminates the unneeded instruction. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-11Fix issue #268Martin Hurton
This patch fixes a bug in the message encoder which was responsible for computing incorrect message offset. The bug affected PGM receiver making it unable to decode inital messages.
2012-05-09Build on Windows and OpenVMS fixedMartin Sustrik
On Windows and OpenVMS there is no IPC transport. This patch fixes the problem introduced by previous patches, where the fact is not taken into account. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-09Trivial coding style fix.Martin Sustrik
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-05-09Whitespace fixes in pgm_socket.cppDouglas Young
2012-05-09Return an error on invalid (e)pgm:// connectionDouglas Young
This prevents the long-standing issue of an invalid pgm or epgm connection string from causing an assert when the connection is attempted and fails.
2012-05-06Return error if an invalid connection string is usedDouglas Young
This patch reintroduces the behaviour that if a tcp:// or ipc:// connection string which is invalid is passed to xs_connect, then an error is reported, instead of asserting at connection time
2012-05-04Correct handling of connect() errors in tcp, ipc_connecter_tMartin Lucina
EAGAIN was being used as a translation value for EINPROGRESS, thus shadowing a real EAGAIN return value from the OS. This caused later assertions of "Invalid argument" in stream_engine.cpp when it attempted to use a socket which was not connected. I also add EINTR to mean EINPROGRESS, as per the POSIX and FreeBSD documentation which specifies that a connect() call interrupted due to a signal will complete asynchronously. Signed-off-by: Martin Lucina <martin@lucina.net>
2012-05-02Conditionally include header files required by select() and poll()Gabriele Svelto
Feature checks are introduced to check for all the headers required by the select() and poll() calls. Include files are then included conditionally without the use of any OS-specific directive. The change also fixes a couple of problems: - Fixed compilation under FreeBSD, NetBSD and OpenBSD when forcing the use of select() in the poller - Quieted a warning mixed-sign comparison warning on FreeBSD caused by FD_SETSIZE being declared as an unsigned constant on that OS - Removed the obsolescent AC_HEADER_TIME macro from the configure script Signed-off-by: Gabriele Svelto <gabriele.svelto@gmail.com>
2012-05-01Revert "Implement protocol versioning (except PGM)"Martin Sustrik
This reverts commit 512f3a604924fec9d89e2b4bfd6f73aa66309fa7.
2012-04-30'own' command replaced by synchronous callMartin Sustrik
The commands was a vestige of the removed 'durable sockets' feature Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-30Fix premature poll reset_polloutPaul Colomiets
Signed-off-by: Paul Colomiets <paul@colomiets.name>
2012-04-29Implement protocol versioning (except PGM)Martin Lucina
Implements SP protocol versioning, legacy protocol support, and the following pattern protocol versions: PAIR: v2 PUBSUB: v1 (legacy), v3 REQREP: v1 PIPELINE: v2 SURVEY: v1 Engine support is only for stream_engine_t at this stage. Signed-off-by: Martin Lucina <martin@lucina.net>
2012-04-29Fix TCP/IPC socket shutdownMartin Sustrik
When socket in the process of asynchronous connect is being closed and the fact that there is no peer is found out at the same time close() may return ECONNRESET. This patch handles this situation decently. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-29atomic: prever inline asm over GCC builtins for armv7aAndrew W. Nosenko
As per Patrick Trantham comment http://groups.crossroads.io/r/topic/510xaadGgWVcXFdATnhAuN the GCC __sync_*() builtins are more generic but slower than specialized hand-written assembly for armv7a case. Therefore, the code is reordered for prefer inline assembly for armv7a over GCC builtins. * src/atomic_counter.hpp: (atomic_counter_t::add): (atomic_counter_t::sub): * src/atomic_ptr.hpp: (atomic_ptr_t::xchg): (atomic_ptr_t::cas): Prefer armv7a specific inline asm over GCC __sync_*() builtins.
2012-04-29atomic: revisit the atomic operation checks and their usageAndrew W. Nosenko
* configure.ac: Check for working Solaris/NetBSD-style atomic.h independly on OS. Add check for GCC-style __sync_*() builtins. New defines: XS_ATOMIC_GCC_SYNC, XS_ATOMIC_SOLARIS. Removed define: XS_FORCE_MUTEXES. * src/atomic_counter.hpp: (atomic_counter_t::add): (atomic_counter_t::sub): * src/atomic_ptr.hpp: (atomic_ptr_t::xchg): (atomic_ptr_t::cas): Use result of these checks. Preference order: 1. GCC-style __sync_*() builtins 2. Inline asm (x86, x86-64, armv7a) 3. Solaris/NetBSD-style atomic.h, Windows-specific API 4. Fallback to mutex-based implementation
2012-04-29atomic: prefer GCC __sync_*() builtins over inline asmAndrew W. Nosenko
Use GCC __sync_*() builtins when compiler claims to be GCC (GCC itself, Clang...) It can be disabled explicitly by using XS_DISABLE_GCC_SYNC_BUILTINS define. Just for any case. * src/atomic_counter.hpp [__GNUC__ && !XS_DISABLE_GCC_SYNC_BUILTINS]: (atomic_counter_t::add): (atomic_counter_t::sub): * src/atomic_ptr.hpp [__GNUC__ && !XS_DISABLE_GCC_SYNC_BUILTINS]: (atomic_ptr_t::xchg): (atomic_ptr_t::cas): Prefer GCC __sync_*() builtins over inline asm.
2012-04-29Implement atomic operations for armv7aPatrick Trantham
This commit implements atomic operations for the armv7a architecture using gcc inline assembly. This offers higher performance compared to pthread mutexes. Tested on an am3517 evm, clocked at 600MHz: ./inproc_thr 200 1000000 ------------------------ 53-60K messages / sec, pthread mutexes 73-90K messages / sec, assembly atomic ops ./inproc_lat 200 1000000 ------------------------ average latency: 42.234 [us], pthread mutexes average latency: 35.496 [us], assembly atomic ops
2012-04-29Fix a bug in prefix filter compaction logicAdam Guthrie
2012-04-27Truncate the message when it is sent to XSUBMartin Sustrik
This patch is based on patch by Martin Hurton from 0MQ project. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-25Solaris build fixedMartin Sustrik
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-25tcp_address_t and ipc_address_t merged into a single classMartin Sustrik
This patch is a preliminary work for mergine TCP and IPC connecters and listeners. The ultimated goal is to avoid code duplication occuring in these classes. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-22Initialise fd_set once for signler_t objectMartin Sustrik
Optimisation. Up to now new fd_set was initialised in each signaler_wait call. Now the fd_set is initialised once when signaler is created. This is useful espacially on Windows where fd_set is list of pointers rather than bitset and thus can be rather large. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-21Select polling mechanisms at one placeMartin Sustrik
This patch provides a single place for selecting polling mechanisms (polling.hpp). Up to now the selection was spread among the build system and several source files. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-21Fix a warning in MSVC buildMartin Sustrik
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-21xs_shutdown implementedMartin Sustrik
This patch allows for partial shutdown of the socket. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-20Clock instance removed from surveyor_tMartin Sustrik
SURVEYOR socket now uses clock instance from socket_base_t for measuring survey timeout. This allows for better performance in tight xs_recv() loops. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-18xs_zmq: initialize size in zmq_getsockoptAdam Guthrie
This fixes a critical bug in zmq_getsockopt.
2012-04-18Improve efficiency of time measurement in tight loopsMartin Sustrik
This patch instantiate a clock_t instance for each XS socket. Thus, it is shared between subsequent calls to xs_recv (and xs_send). That in turn significantly limits the number of invocations of getimeofday (or similar) when timeouts are used and recv/send is called in a tight loop. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-18Discard pending surveys when (X)SURVEYOR socket disappearsMartin Sustrik
This is an optimisation. The idea is that when surveyor disappears it makes no sense to process any surveys it have sent. There are two cases handled: - when (X)SURVEYOR socket is closed any unsent outbound messages are discarded - when (X)RESPONDENT detects that its peer have disconnected it drops any unprocessed messages that have been received from it Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-17XS_SURVEY_TIMEOUT socket option added.Martin Sustrik
It can be used to timeout the survey. Value is in milliseconds and -1 means infinite (default). Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-17"Survey" pattern implementedMartin Sustrik
Survey pattern is "multicast with reply". There are two roles: surveyor and respondent. Surveyor publishes a survey which gets delivered to all connected respondents. Each repondent can send a response to the survey. All the responses are delivered to the original surveyor. Once the surveyor decides that the survey is over (e.g. deadline was reached) it can send initiate survey. Late responses from old surveys are automatically discarded by the surveyor socket. Socket types: SURVEYOR, XSURVEYOR, RESPONDENT, XRESPONDENT Patch also includes a test program with surveoyr, two respondents and an intermediary device. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-16Few trivial fixes to REQ/REP pattern.Martin Sustrik
- couple of comments adjusted - empty 'stack delimiter' message correctly closed - non-reachable function xrep_t::rollback() removed Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-16Automotic loading of plugins (Linux)Martin Sustrik
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-16Define the XS_PREFIX_PATH macro with the installation prefixGabriele Svelto
Signed-off-by: Gabriele Svelto <gabriele.svelto@gmail.com>
2012-04-16Automatic loading of plugins on WindowsMartin Sustrik
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
2012-04-16Honour --with-poller option for synchronous pollingMartin Sustrik
While --with-poller configure option could be used for selecting the polling mechanism for I/O threads (asynchronous), synchronous polling (xs_poll, signaler_t) was not affected by the option. This patch uses selected mechanism is it is either 'select' or 'poll'. Othrwise it chooses the mechanism according to the platform. Signed-off-by: Martin Sustrik <sustrik@250bpm.com>