As PGM is not capable of passing subscriptions upstream,
subscriptions are ignored at sub side and engine subscribes
for all messages on pub side.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The filtering is now done depending on the socket type. SUB socket
filters the messages (end-to-end filtering) while XSUB relies
on upstream nodes to do (imprefect) filtering.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Till now the VSM buffer was aligned to 16 bit boundary
which could possibly cause problems on RISC architectures
when accessing the message data in unsafe manner.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Storing commands in OS socket buffers caused whole lot of
problems when free space in the buffer ran out. This patch
stores commands in ypipes instead and uses socketpair just
to signal the other thread, ie. at most one byte is stored
in the socketpair at any single instant.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
pthread_* functions return the error number rather than
setting errno. This was not accounter for till now.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
So far the requests in req/rep pattern were delivered to and processed
by worker even though the original requester was dead. Thus,
the worker processing replies with noone to deliver results to.
This optimisation drops requests in two situations:
1. Queued inbound requests in XREP socket when peer disconnects.
2. Queued outbound requests in XREQ when socket is closed.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
This patch introduces two changes:
1. 32-bit ID is used to identify the peer instead of UUID
2. REQ socket seeds the label stack with unique 32-bit request ID
It also drops any replies with non-matching request ID
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
So far there was no distinction between message parts used by 0MQ
and message parts used by user. Now, the message parts used by 0MQ
are marked as 'LABEL'.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
For the platforms that don't support MSG_DONTWAIT the reader
socket in mailbox_t was kept in non-blocking state and flipped
to blocking state and back when blocking read was requested.
Now, the state is preserved between calls and flipped only
if different type of operation (block vs. non-block) is
requested.
Signed-off-by: Fabien Ninoles <fabien@tzone.org>
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
When session is already terminating and reconnection happens at
that point, the session should not create new pipe to its socket.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Introduction of subscription forwarding have broken throughput tests.
This patch changes the pattern used in the tests to PUSH/PULL.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
This option is a performance tweak. In devices XSUB socket filters
the messages just to send them to XPUB socket which filters them
once more. Setting ZMQ_FILTER option to 0 allows to switch the
filtering in XSUB socket off.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
This patch will prevent duplicate matching in devices in the future.
Instead of matching in both XPUB and XSUB, it'll happen only
in XPUB. Receiver endpoint will still filter messages via SUB
socket.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Till now the code was spread over mutliple locations.
Additionally, the code was made more formally correct,
with explicit pipe state machine etc.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
So far, the pipe termination code was spread among socket type
classes, fair queuer, load balancer, etc. This patch moves
all the associated logic to a single place.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
So far, there was a pair of unidirectional pipes between a socket
and a session (or an inproc peer). This resulted in complex
problems with half-closed states and tracking which inpipe
corresponds to which outpipe.
This patch doesn't add any functionality in itself, but is
essential for further work on features like subscription
forwarding.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
ICC doesn't recognise that assert(false) terminates the program
and thus complains that certain functions have no return values.
This patch supplies dummy return values to keep the compiler happy.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
When an inpipe terminated within XREP, it was erased from the array
and thus current_in (which is an index) pointed to a different
element in the array. This caused problems when we were in the
middle of reading a multipart message.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Reaching the HWM caused breaking message atomicity when the
flow was reestablished - initial parts of multipart messages
may have been lost.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
pgm_socket used textural form of UUID to generate GSIs.
The recent patch that removed support for textual UUIDs
broke the functionality. This patch fixes the problem.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The string format of UUID is not used in 0MQ. Further on,
it turns out that UUIDs have fixed microarchitecture-agnostic
binary layout (see RFC4122). Thus, the conversion to string
and back to binary can be avoided.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Fair queueing algorithm was checking whether the current pipe
is not closed in the middle of reading a multipart message.
However, this is OK when the socket is closing down.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
This patch addresses serveral issues:
1. It gathers message related functionality scattered over whole
codebase into a single class.
2. It makes zmq_msg_t an opaque datatype. Internals of the class
don't pollute zmq.h header file.
3. zmq_msg_t size decreases from 48 to 32 bytes. That saves ~33%
of memory in scenarios with large amount of small messages.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
