]> git.openfabrics.org - ~shefty/librdmacm.git/log
~shefty/librdmacm.git
11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:33:01 +0000 (11:33 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:33:01 +0000 (11:33 -0700)]
pop

11 years agorsockets: Use wr_id to determine completion type
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Use wr_id to determine completion type

If a work request has completed in error, the completion type
field is undefined.  Use the wr_id to determine if the failed
completion was a send or receive.

This fixes an issue where MPI can hang during finalize.  With
both sides of a connection shutting down simultaneously, one
side may complete quicker and delete its QP before the other
side receives an acknowledgement to their disconnect message.
Eventually, the disconnect message will time out, but because
the completion type field is undefined, it may be processed
as a failed receive, rather than a failed send.  The end
result is that the second side hangs waiting for the send to
complete.

This problem showed up more easily after commit
2e5b0fc95964f74ea59dd725e849027faa0cd526, but existed beforehand.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
11 years agorefresh
Sean Hefty [Mon, 30 Jul 2012 18:29:19 +0000 (11:29 -0700)]
refresh

11 years agorefresh
Sean Hefty [Mon, 30 Jul 2012 18:29:19 +0000 (11:29 -0700)]
refresh

11 years agorefresh (create temporary patch)
Sean Hefty [Mon, 30 Jul 2012 18:28:50 +0000 (11:28 -0700)]
refresh (create temporary patch)

11 years agorefresh (create temporary patch)
Sean Hefty [Mon, 30 Jul 2012 18:28:50 +0000 (11:28 -0700)]
refresh (create temporary patch)

11 years agoRefresh of rs-ftp
Sean Hefty [Mon, 30 Jul 2012 18:28:50 +0000 (11:28 -0700)]
Refresh of rs-ftp

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:28:47 +0000 (11:28 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:28:47 +0000 (11:28 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:28:46 +0000 (11:28 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:28:46 +0000 (11:28 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:23:25 +0000 (11:23 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:23:25 +0000 (11:23 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:02:15 +0000 (11:02 -0700)]
pop

11 years agopop
Sean Hefty [Mon, 30 Jul 2012 18:02:15 +0000 (11:02 -0700)]
pop

11 years agocommit
Sean Hefty [Mon, 30 Jul 2012 18:02:02 +0000 (11:02 -0700)]
commit

11 years agocommit
Sean Hefty [Mon, 30 Jul 2012 18:02:02 +0000 (11:02 -0700)]
commit

12 years agorefresh
Sean Hefty [Sat, 28 Jul 2012 00:57:23 +0000 (17:57 -0700)]
refresh

12 years agorefresh
Sean Hefty [Sat, 28 Jul 2012 00:57:23 +0000 (17:57 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Sat, 28 Jul 2012 00:57:23 +0000 (17:57 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Sat, 28 Jul 2012 00:57:23 +0000 (17:57 -0700)]
refresh (create temporary patch)

12 years agoRefresh of fast-disc
Sean Hefty [Sat, 28 Jul 2012 00:57:23 +0000 (17:57 -0700)]
Refresh of fast-disc

12 years agorefresh
Sean Hefty [Sat, 28 Jul 2012 00:35:36 +0000 (17:35 -0700)]
refresh

12 years agorefresh
Sean Hefty [Sat, 28 Jul 2012 00:35:36 +0000 (17:35 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Sat, 28 Jul 2012 00:35:36 +0000 (17:35 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Sat, 28 Jul 2012 00:35:36 +0000 (17:35 -0700)]
refresh (create temporary patch)

12 years agoRefresh of fast-disc
Sean Hefty [Sat, 28 Jul 2012 00:35:36 +0000 (17:35 -0700)]
Refresh of fast-disc

12 years agorsocket: Improve disconnect time under normal conditions
Sean Hefty [Fri, 27 Jul 2012 17:46:42 +0000 (10:46 -0700)]
rsocket: Improve disconnect time under normal conditions

When both sides of a connection attempt to close at the same
time, one of the two sides can easily get an error when sending
a disconnect message.  This results in that side hanging
during close until the send times out.  (The time out is caused
by the remote side destroying its QP.)

We can reduce the chance of this occurring by immediately
assuming that the disconnect has been successful once we've
received the remote side's disconnect message.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsocket: Improve disconnect time under normal conditions
Sean Hefty [Fri, 27 Jul 2012 17:46:42 +0000 (10:46 -0700)]
rsocket: Improve disconnect time under normal conditions

When both sides of a connection attempt to close at the same
time, one of the two sides can easily get an error when sending
a disconnect message.  This results in that side hanging
during close until the send times out.  (The time out is caused
by the remote side destroying its QP.)

We can reduce the chance of this occurring by immediately
assuming that the disconnect has been successful once we've
received the remote side's disconnect message.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agonew
Sean Hefty [Fri, 27 Jul 2012 17:46:42 +0000 (10:46 -0700)]
new

12 years agonew
Sean Hefty [Fri, 27 Jul 2012 17:46:42 +0000 (10:46 -0700)]
new

12 years agorsocket: Improve disconnect time under normal conditions
Sean Hefty [Fri, 27 Jul 2012 17:46:42 +0000 (10:46 -0700)]
rsocket: Improve disconnect time under normal conditions

When both sides of a connection attempt to close at the same
time, one of the two sides can easily get an error when sending
a disconnect message.  This results in that side hanging
during close until the send times out.  (The time out is caused
by the remote side destroying its QP.)

We can reduce the chance of this occurring by immediately
assuming that the disconnect has been successful once we've
received the remote side's disconnect message.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorename
Sean Hefty [Fri, 27 Jul 2012 17:43:12 +0000 (10:43 -0700)]
rename

12 years agorename
Sean Hefty [Fri, 27 Jul 2012 17:43:12 +0000 (10:43 -0700)]
rename

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:32:19 +0000 (23:32 -0700)]
refresh

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:32:19 +0000 (23:32 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:32:19 +0000 (23:32 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:32:19 +0000 (23:32 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Fri, 27 Jul 2012 06:32:18 +0000 (23:32 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:31:37 +0000 (23:31 -0700)]
refresh

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:31:37 +0000 (23:31 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:31:37 +0000 (23:31 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:31:37 +0000 (23:31 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Fri, 27 Jul 2012 06:31:37 +0000 (23:31 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:30:08 +0000 (23:30 -0700)]
refresh

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:30:08 +0000 (23:30 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:29:34 +0000 (23:29 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:29:34 +0000 (23:29 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Fri, 27 Jul 2012 06:29:33 +0000 (23:29 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:29:06 +0000 (23:29 -0700)]
refresh

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:29:06 +0000 (23:29 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:25:04 +0000 (23:25 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:25:04 +0000 (23:25 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Fri, 27 Jul 2012 06:25:04 +0000 (23:25 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:11:20 +0000 (23:11 -0700)]
refresh

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 06:11:20 +0000 (23:11 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:11:20 +0000 (23:11 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 06:11:20 +0000 (23:11 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Fri, 27 Jul 2012 06:11:19 +0000 (23:11 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 05:59:56 +0000 (22:59 -0700)]
refresh

12 years agorefresh
Sean Hefty [Fri, 27 Jul 2012 05:59:56 +0000 (22:59 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 05:59:56 +0000 (22:59 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Fri, 27 Jul 2012 05:59:56 +0000 (22:59 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Fri, 27 Jul 2012 05:59:56 +0000 (22:59 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 23:18:08 +0000 (16:18 -0700)]
refresh

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 23:18:08 +0000 (16:18 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 23:18:08 +0000 (16:18 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 23:18:08 +0000 (16:18 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Thu, 26 Jul 2012 23:18:08 +0000 (16:18 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 23:17:30 +0000 (16:17 -0700)]
refresh

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 23:17:30 +0000 (16:17 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 23:17:30 +0000 (16:17 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 23:17:30 +0000 (16:17 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Thu, 26 Jul 2012 23:17:30 +0000 (16:17 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 22:43:37 +0000 (15:43 -0700)]
refresh

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 22:43:37 +0000 (15:43 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 22:43:37 +0000 (15:43 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 22:43:37 +0000 (15:43 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Thu, 26 Jul 2012 22:43:36 +0000 (15:43 -0700)]
Refresh of mpi-finalize

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 22:37:14 +0000 (15:37 -0700)]
refresh

12 years agorefresh
Sean Hefty [Thu, 26 Jul 2012 22:37:14 +0000 (15:37 -0700)]
refresh

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 22:37:14 +0000 (15:37 -0700)]
refresh (create temporary patch)

12 years agorefresh (create temporary patch)
Sean Hefty [Thu, 26 Jul 2012 22:37:14 +0000 (15:37 -0700)]
refresh (create temporary patch)

12 years agoRefresh of mpi-finalize
Sean Hefty [Thu, 26 Jul 2012 22:37:14 +0000 (15:37 -0700)]
Refresh of mpi-finalize

12 years agorsockets: Use wr_id to determine completion type
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Use wr_id to determine completion type

If a work request has completed in error, the completion type
field is undefined.  Use the wr_id to determine if the failed
completion was a send or receive.

This fixes an issue where MPI can hang during finalize.  With
both sides of a connection shutting down simultaneously, one
side may complete quicker and delete its QP before the other
side receives an acknowledgement to their disconnect message.
Eventually, the disconnect message will time out, but because
the completion type field is undefined, it may be processed
as a failed receive, rather than a failed send.  The end
result is that the second side hangs waiting for the send to
complete.

This problem showed up more easily after commit
2e5b0fc95964f74ea59dd725e849027faa0cd526, but existed beforehand.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Use wr_id to determine completion type
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Use wr_id to determine completion type

If a work request has completed in error, the completion type
field is undefined.  Use the wr_id to determine if the failed
completion was a send or receive.

This fixes an issue where MPI can hang during finalize.  With
both sides of a connection shutting down simultaneously, one
side may complete quicker and delete its QP before the other
side receives an acknowledgement to their disconnect message.
Eventually, the disconnect message will time out, but because
the completion type field is undefined, it may be processed
as a failed receive, rather than a failed send.  The end
result is that the second side hangs waiting for the send to
complete.

This problem showed up more easily after commit
2e5b0fc95964f74ea59dd725e849027faa0cd526, but existed beforehand.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Use wr_id to determine completion type
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Use wr_id to determine completion type

If a work request has completed in error, the completion type
field is undefined.  Use the wr_id to determine if the failed
completion was a send or receive.

This fixes an issue where MPI can hang during finalize.  With
both sides of a connection shutting down simultaneously, one
side may complete quicker and delete its QP before the other
side receives an acknowledgement to their disconnect message.
Eventually, the disconnect message will time out, but because
the completion type field is undefined, it may be processed
as a failed receive, rather than a failed send.  The end
result is that the second side hangs waiting for the send to
complete.

This problem showed up more easily after commit
2e5b0fc95964f74ea59dd725e849027faa0cd526, but existed beforehand.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Use wr_id to determine completion type
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Use wr_id to determine completion type

If a work request has completed in error, the completion type
field is undefined.  Use the wr_id to determine if the failed
completion was a send or receive.

This fixes an issue where MPI can hang during finalize.  With
both sides of a connection shutting down simultaneously, one
side may complete quicker and delete its QP before the other
side receives an acknowledgement to their disconnect message.
Eventually, the disconnect message will time out, but because
the completion type field is undefined, it may be processed
as a failed receive, rather than a failed send.  The end
result is that the second side hangs waiting for the send to
complete.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Fix shutdown
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Fix shutdown

mpi finalize hangs - fix it

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Fix shutdown
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Fix shutdown

mpi finalize hangs - fix it

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Fix shutdown
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Fix shutdown

mpi finalize hangs - fix it

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Fix shutdown
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Fix shutdown

mpi finalize hangs - fix it

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Fix shutdown
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Fix shutdown

mpi finalize hangs - fix it

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agorsockets: Fix shutdown
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Fix shutdown

mpi finalize hangs - fix it

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agonew
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
new

12 years agonew
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
new

12 years agorsockets: Fix shutdown
Sean Hefty [Thu, 26 Jul 2012 22:35:32 +0000 (15:35 -0700)]
rsockets: Fix shutdown

mpi finalize hangs - fix it

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
12 years agopop
Sean Hefty [Thu, 26 Jul 2012 19:09:32 +0000 (12:09 -0700)]
pop

12 years agopop
Sean Hefty [Thu, 26 Jul 2012 19:09:32 +0000 (12:09 -0700)]
pop