]> git.openfabrics.org - ~ardavis/dapl.git/log
~ardavis/dapl.git
11 years agoGive error message when fail to open the lock file.
Jianxin Xiong [Thu, 27 Sep 2012 16:49:05 +0000 (09:49 -0700)]
Give error message when fail to open the lock file.

11 years agoAdd a configuration option for the maximum message size.
Jianxin Xiong [Thu, 27 Sep 2012 16:27:28 +0000 (09:27 -0700)]
Add a configuration option for the maximum message size.

11 years agoDefault to /var/run for .pid files
Jianxin Xiong [Wed, 26 Sep 2012 21:09:58 +0000 (14:09 -0700)]
Default to /var/run for .pid files

From: pmmccorm <patrick.m.mccormick@intel.com>

11 years agoOrchestrate the handling of pid and lock files
Jianxin Xiong [Wed, 26 Sep 2012 21:03:58 +0000 (14:03 -0700)]
Orchestrate the handling of pid and lock files

From: pmmccorm <patrick.m.mccormick@intel.com>

11 years agoIncrease the proxy send queue size according to the segmentation setting.
Jianxin Xiong [Tue, 25 Sep 2012 22:36:43 +0000 (15:36 -0700)]
Increase the proxy send queue size according to the segmentation setting.

11 years agoAdd option to allow the proxy buffer be shared among connections from
Jianxin Xiong [Fri, 21 Sep 2012 04:13:16 +0000 (21:13 -0700)]
Add option to allow the proxy buffer be shared among connections from
the same client. The can help greatly reduce the memory consumption of
the porxy server when the number of connections is large.

11 years agoRevamp the instructions for the MCM provider.
Jianxin Xiong [Thu, 20 Sep 2012 20:45:21 +0000 (13:45 -0700)]
Revamp the instructions for the MCM provider.

11 years agoChange the default per connection proxy RDMA buffer from 64MB to 32MB for
Jianxin Xiong [Thu, 20 Sep 2012 02:53:40 +0000 (19:53 -0700)]
Change the default per connection proxy RDMA buffer from 64MB to 32MB for
better scalability. It can still be changed in mpxyd.conf.

11 years agoBug fix: lock/log file path is truncated to 32 characters.
Jianxin Xiong [Wed, 19 Sep 2012 17:07:00 +0000 (10:07 -0700)]
Bug fix: lock/log file path is truncated to 32 characters.

11 years agoIncrease the default queue length for scif_listen() call to 64, and
Jianxin Xiong [Tue, 18 Sep 2012 01:11:44 +0000 (18:11 -0700)]
Increase the default queue length for scif_listen() call to 64, and
add an option to the configuration file to override this setting.

11 years agoUpdate the non-root setup instructions.
Jianxin Xiong [Sat, 15 Sep 2012 00:50:42 +0000 (17:50 -0700)]
Update the non-root setup instructions.

11 years agoChange "MIC Indirect" to "CCL-proxy" in the readme files.
Jianxin Xiong [Fri, 14 Sep 2012 21:03:22 +0000 (14:03 -0700)]
Change "MIC Indirect" to "CCL-proxy" in the readme files.

11 years agoBug fix: MPI connection failure with 64+ ranks due to REJ packet not being delivered...
Jianxin Xiong [Fri, 14 Sep 2012 06:22:58 +0000 (23:22 -0700)]
Bug fix: MPI connection failure with 64+ ranks due to REJ packet not being delivered to the client.

11 years agoCatch proxy server segfault when scaling up.
Jianxin Xiong [Wed, 12 Sep 2012 22:18:23 +0000 (15:18 -0700)]
Catch proxy server segfault when scaling up.

11 years agoImproved error messages with SCIF related calls.
Jianxin Xiong [Wed, 12 Sep 2012 20:53:31 +0000 (13:53 -0700)]
Improved error messages with SCIF related calls.

11 years agoAdd instructions for installation without root privilege.
Jianxin Xiong [Mon, 10 Sep 2012 18:30:52 +0000 (11:30 -0700)]
Add instructions for installation without root privilege.

11 years agoAdd instructions for cluster deployment.
Jianxin Xiong [Thu, 6 Sep 2012 17:21:47 +0000 (10:21 -0700)]
Add instructions for cluster deployment.

11 years agoAllow changing the SCIF port id via the envar DAPL_MCM_PORT_ID so that
Jianxin Xiong [Wed, 5 Sep 2012 18:15:57 +0000 (11:15 -0700)]
Allow changing the SCIF port id via the envar DAPL_MCM_PORT_ID so that
the proxy service can be started by a regular user.

11 years agoMake the default config file consistent with the initialization code.
Jianxin Xiong [Wed, 5 Sep 2012 18:10:49 +0000 (11:10 -0700)]
Make the default config file consistent with the initialization code.

11 years agoPrevent dereferencing a NULL pointer.
Jianxin Xiong [Wed, 5 Sep 2012 17:54:15 +0000 (10:54 -0700)]
Prevent dereferencing a NULL pointer.

11 years agoAdd a readme file for the MCM provider.
Jianxin Xiong [Mon, 20 Aug 2012 18:59:29 +0000 (11:59 -0700)]
Add a readme file for the MCM provider.

11 years agoAdd "buffer_segment_size" to the configuration file.
Jianxin Xiong [Mon, 20 Aug 2012 18:52:57 +0000 (11:52 -0700)]
Add "buffer_segment_size" to the configuration file.

11 years agoBug fix: scif_unregister returns -ENODEV becasue the length is not page aligned.
Jianxin Xiong [Sat, 18 Aug 2012 20:43:30 +0000 (13:43 -0700)]
Bug fix: scif_unregister returns -ENODEV becasue the length is not page aligned.

11 years agoThere is a race condition between dapl_rbuf_add() dapl_rbuf_remove() when
Jianxin Xiong [Sat, 18 Aug 2012 05:30:06 +0000 (22:30 -0700)]
There is a race condition between dapl_rbuf_add() dapl_rbuf_remove() when
the ring buffer is empty or full. This could lead to incorrect item be
dequeued and correct item be discarded.

A temporary workaround is provided to ensure the pending event queue work
correctly. Without this fix dapl_evd_dequeue() could return the wrong event
and causes various application errors (hanging, assertion failures. etc).

Ultimately the ring buffer mechanism needs to the fixed.

11 years agoAdd environment variable DAPL_MCM_ALWAYS_PROXY to allow proxying non-MIC
Jianxin Xiong [Sat, 18 Aug 2012 05:13:37 +0000 (22:13 -0700)]
Add environment variable DAPL_MCM_ALWAYS_PROXY to allow proxying non-MIC
initiated connections. Good for debugging.

11 years agoHandle the (unlikely) corner case that the last sge has length 0.
Jianxin Xiong [Fri, 17 Aug 2012 18:44:27 +0000 (11:44 -0700)]
Handle the (unlikely) corner case that the last sge has length 0.

11 years agoMinor code cleanup.
Jianxin Xiong [Fri, 17 Aug 2012 18:35:25 +0000 (11:35 -0700)]
Minor code cleanup.

11 years agoRecount the actual pending WRs from time to time.
Jianxin Xiong [Fri, 17 Aug 2012 06:27:34 +0000 (23:27 -0700)]
Recount the actual pending WRs from time to time.

11 years agoDon't waste the 64 byte in the proxy buffer between segments.
Jianxin Xiong [Fri, 17 Aug 2012 05:59:20 +0000 (22:59 -0700)]
Don't waste the 64 byte in the proxy buffer between segments.

11 years agoBug fix: the calculation of 'total_offset' could be wrong if a segmented sge is not...
Jianxin Xiong [Thu, 16 Aug 2012 19:34:30 +0000 (12:34 -0700)]
Bug fix: the calculation of 'total_offset' could be wrong if a segmented sge is not the last one.

11 years agoPrevent poll_cnt from becoming negative, which could lead to huge delay.
Jianxin Xiong [Thu, 16 Aug 2012 18:31:22 +0000 (11:31 -0700)]
Prevent poll_cnt from becoming negative, which could lead to huge delay.

11 years agoDon't flood the log with messages about deferred WR posting.
Jianxin Xiong [Thu, 16 Aug 2012 18:28:42 +0000 (11:28 -0700)]
Don't flood the log with messages about deferred WR posting.

11 years agoProxy large writes in segments.
Jianxin Xiong [Wed, 15 Aug 2012 05:32:09 +0000 (22:32 -0700)]
Proxy large writes in segments.

11 years agoReplace scif_writeto with scif_readfrom, preparing for segmentation support.
Jianxin Xiong [Tue, 14 Aug 2012 20:00:58 +0000 (13:00 -0700)]
Replace scif_writeto with scif_readfrom, preparing for segmentation support.

11 years agoMake both address and size 64 bytes aligned when calling scif_writeto().
Jianxin Xiong [Tue, 14 Aug 2012 05:23:41 +0000 (22:23 -0700)]
Make both address and size 64 bytes aligned when calling scif_writeto().

11 years agoPort numbers need to be unique within each physical IB device (md),
Jianxin Xiong [Fri, 10 Aug 2012 00:55:04 +0000 (17:55 -0700)]
Port numbers need to be unique within each physical IB device (md),
otherwise connection requests between multiple clients of the same
proxy server can be matched incorrectly.

11 years agoCorrect many mismatches between the log messages and the actual operations.
Jianxin Xiong [Tue, 7 Aug 2012 15:53:18 +0000 (08:53 -0700)]
Correct many mismatches between the log messages and the actual operations.

11 years agoCorrect the MR size calculation when registering with SCIF.
Jianxin Xiong [Tue, 7 Aug 2012 05:53:19 +0000 (22:53 -0700)]
Correct the MR size calculation when registering with SCIF.

11 years agoMPXYD: only post an incoming write WR when there is no pending WR on the QP,
Jianxin Xiong [Sat, 4 Aug 2012 00:32:31 +0000 (17:32 -0700)]
MPXYD: only post an incoming write WR when there is no pending WR on the QP,
otherwise there might be ordering issues.

11 years agoFix the (nil) "wr_id" output of posted pending work request.
Jianxin Xiong [Sat, 4 Aug 2012 00:25:44 +0000 (17:25 -0700)]
Fix the (nil) "wr_id" output of posted pending work request.

11 years agoBug fix: connection error with "unexpected DAPL event 0x4003" when more then one...
Jianxin Xiong [Fri, 3 Aug 2012 03:30:34 +0000 (20:30 -0700)]
Bug fix: connection error with "unexpected DAPL event 0x4003" when more then one client exists.

11 years agoCorrect the range of valid ops in mix_op_str().
Jianxin Xiong [Thu, 2 Aug 2012 21:38:43 +0000 (14:38 -0700)]
Correct the range of valid ops in mix_op_str().

11 years agoFix conenction setup deadlock with more than one client connecting to MPXYD.
Jianxin Xiong [Thu, 2 Aug 2012 18:45:50 +0000 (11:45 -0700)]
Fix conenction setup deadlock with more than one client connecting to MPXYD.

11 years agoAdd wr_end field for entry count and hd/tl adjustments
Arlin Davis [Wed, 25 Jul 2012 06:24:59 +0000 (23:24 -0700)]
Add wr_end field for entry count and hd/tl adjustments

MPI now works pingpong. Still not stress testing well.
Some hang conditions need debugged

11 years agodtest update for unaligned/aligned data option, unaligned by default
sean.hefty@intel.com [Mon, 23 Jul 2012 22:20:08 +0000 (15:20 -0700)]
dtest update for unaligned/aligned data option, unaligned by default

set -a for page alignment.

return EINVAL for anything other than sends or writes.

11 years agordma write working both directions. MIC to XEON, XEON to MIC.
Arlin Davis [Mon, 23 Jul 2012 06:42:16 +0000 (23:42 -0700)]
rdma write working both directions. MIC to XEON, XEON to MIC.

updated memory registration to adjust for SCIF page alignment

12 years agoadd progress thread processing of delayed WR's
Arlin Davis [Tue, 17 Jul 2012 17:44:19 +0000 (10:44 -0700)]
add progress thread processing of delayed WR's

12 years agordma writes working
Arlin Davis [Tue, 17 Jul 2012 05:04:18 +0000 (22:04 -0700)]
rdma writes working

12 years agordma write support coded, compiled not tested
sean.hefty@intel.com [Sun, 15 Jul 2012 07:11:56 +0000 (00:11 -0700)]
rdma write support coded, compiled not tested

12 years agoSends/receives working, Xeon to MIC
Arlin Davis [Wed, 11 Jul 2012 16:33:19 +0000 (09:33 -0700)]
Sends/receives working, Xeon to MIC

12 years agoupdate debug message
Arlin Davis [Thu, 5 Jul 2012 16:05:36 +0000 (09:05 -0700)]
update debug message

12 years agoMCM development
Arlin Davis [Thu, 5 Jul 2012 16:01:47 +0000 (09:01 -0700)]
MCM development

12 years agoupdated for QP, CM services
Arlin Davis [Tue, 26 Jun 2012 18:35:53 +0000 (11:35 -0700)]
updated for  QP, CM services

12 years agoadd mcm
Arlin Davis [Wed, 13 Jun 2012 01:33:45 +0000 (18:33 -0700)]
add mcm

12 years agoupdate makefile, dat.conf
Arlin Davis [Wed, 13 Jun 2012 01:33:14 +0000 (18:33 -0700)]
update makefile, dat.conf

12 years agoadd extension.h update mpxyd.c
Arlin Davis [Wed, 6 Jun 2012 19:19:19 +0000 (12:19 -0700)]
add extension.h update mpxyd.c

12 years agoadd mcm.c and mix.c, MIC indirect CM code and MIC indirect eXchange messager
Arlin Davis [Mon, 4 Jun 2012 23:00:41 +0000 (16:00 -0700)]
add mcm.c and mix.c, MIC indirect CM code and MIC indirect eXchange messager

12 years agoupdate for chkconfig
Arlin Davis [Thu, 31 May 2012 22:56:20 +0000 (15:56 -0700)]
update for chkconfig

12 years agoupdate makefile
Arlin Davis [Thu, 31 May 2012 22:11:07 +0000 (15:11 -0700)]
update makefile

12 years agoupdate spec file
Arlin Davis [Thu, 31 May 2012 21:47:00 +0000 (14:47 -0700)]
update spec file

12 years agomakefile update
Arlin Davis [Thu, 31 May 2012 21:44:16 +0000 (14:44 -0700)]
makefile update

12 years agoupdate Makefile move daplmpxy to mpxyd.init.in
Arlin Davis [Thu, 31 May 2012 21:41:03 +0000 (14:41 -0700)]
update Makefile move daplmpxy to mpxyd.init.in

12 years agoupdate makefile
Arlin Davis [Thu, 31 May 2012 21:38:40 +0000 (14:38 -0700)]
update makefile

12 years agoadd mpxyd.conf
Arlin Davis [Tue, 22 May 2012 18:55:06 +0000 (11:55 -0700)]
add mpxyd.conf

12 years agoupdate makefile for mpdxyd.conf
Arlin Davis [Tue, 22 May 2012 18:31:58 +0000 (11:31 -0700)]
update makefile for mpdxyd.conf

12 years agoupdate makefile
Arlin Davis [Tue, 22 May 2012 17:47:31 +0000 (10:47 -0700)]
update makefile

12 years agoupdate makefile
Arlin Davis [Tue, 22 May 2012 17:44:12 +0000 (10:44 -0700)]
update makefile

12 years agoadd mpxyd.c
Arlin Davis [Mon, 21 May 2012 22:40:30 +0000 (15:40 -0700)]
add mpxyd.c

12 years agofix makefile
Arlin Davis [Mon, 21 May 2012 22:37:12 +0000 (15:37 -0700)]
fix makefile

12 years agofix makefile
Arlin Davis [Mon, 21 May 2012 22:34:19 +0000 (15:34 -0700)]
fix makefile

12 years agofix makefile
Arlin Davis [Mon, 21 May 2012 22:32:27 +0000 (15:32 -0700)]
fix makefile

12 years agoconfigure changes.
Arlin Davis [Mon, 21 May 2012 22:27:53 +0000 (15:27 -0700)]
configure changes.

12 years agoupdate configure.in
Arlin Davis [Mon, 21 May 2012 22:24:35 +0000 (15:24 -0700)]
update configure.in

12 years agoChange names to mpxy
Arlin Davis [Mon, 21 May 2012 22:22:00 +0000 (15:22 -0700)]
Change names to mpxy

12 years agopxd.c
Arlin Davis [Wed, 16 May 2012 23:23:16 +0000 (16:23 -0700)]
pxd.c

12 years agopxd.c
Arlin Davis [Wed, 16 May 2012 23:22:33 +0000 (16:22 -0700)]
pxd.c

12 years agopxd.c
Arlin Davis [Wed, 16 May 2012 23:21:37 +0000 (16:21 -0700)]
pxd.c

12 years agomakefile
Arlin Davis [Wed, 16 May 2012 23:20:00 +0000 (16:20 -0700)]
makefile

12 years agoupdate pxd.c
Arlin Davis [Wed, 16 May 2012 23:16:07 +0000 (16:16 -0700)]
update pxd.c

12 years agomakefile
Arlin Davis [Wed, 16 May 2012 23:14:52 +0000 (16:14 -0700)]
makefile

12 years agomakefile.am
Arlin Davis [Wed, 16 May 2012 23:09:56 +0000 (16:09 -0700)]
makefile.am

12 years agoadd svc/pxm.c for deamon
Arlin Davis [Wed, 16 May 2012 23:06:43 +0000 (16:06 -0700)]
add svc/pxm.c for deamon

12 years agomakefile.am
Arlin Davis [Wed, 16 May 2012 23:02:53 +0000 (16:02 -0700)]
makefile.am

12 years agoMakefile.am
Arlin Davis [Wed, 16 May 2012 22:59:39 +0000 (15:59 -0700)]
Makefile.am

12 years agoconfigure.in
Arlin Davis [Wed, 16 May 2012 22:56:45 +0000 (15:56 -0700)]
configure.in

12 years agoupdate configure.in
Arlin Davis [Wed, 16 May 2012 22:54:44 +0000 (15:54 -0700)]
update configure.in

12 years agoadd daplpdx.init.in
Arlin Davis [Wed, 16 May 2012 22:51:17 +0000 (15:51 -0700)]
add daplpdx.init.in

12 years agoupdate for daplpxd daemon
Arlin Davis [Wed, 16 May 2012 22:48:55 +0000 (15:48 -0700)]
update for daplpxd daemon

12 years agoadd libdaplopcm.map
Arlin Davis [Wed, 16 May 2012 19:01:24 +0000 (12:01 -0700)]
add libdaplopcm.map

12 years agofix makefile.am
Arlin Davis [Wed, 16 May 2012 18:54:51 +0000 (11:54 -0700)]
fix makefile.am

12 years agofix makefile.am
Arlin Davis [Wed, 16 May 2012 18:53:14 +0000 (11:53 -0700)]
fix makefile.am

12 years agoadd pcm files
Arlin Davis [Wed, 16 May 2012 18:46:59 +0000 (11:46 -0700)]
add pcm files

12 years agofix configure.in
Arlin Davis [Wed, 16 May 2012 18:42:42 +0000 (11:42 -0700)]
fix configure.in

12 years agoAdd configuration files for new Proxy RDMA/SCIF Provider
Arlin Davis [Wed, 16 May 2012 18:28:09 +0000 (11:28 -0700)]
Add configuration files for new Proxy RDMA/SCIF Provider

This proxy RDMA provider is a interbus service within a platform
to enable offloading of RDMA writes from a small core (MIC) to
large core resource. Data sourced on a MIC that is
destined for remote nodes on the fabric will be transfered
to large core via SCIF and than to remote node.

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>
12 years agocommon: allow qp modify in init state
Arlin Davis [Mon, 14 May 2012 21:51:38 +0000 (14:51 -0700)]
common: allow qp modify in init state

Allow consumer to modify attributes via dat_ep_modify
in init state.

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>
12 years agocommon: check for valid states during ep posting
Arlin Davis [Thu, 10 May 2012 21:57:31 +0000 (14:57 -0700)]
common: check for valid states during ep posting

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>
12 years agodat.conf: keep list of providers in order for backward compatibility
Arlin Davis [Thu, 10 May 2012 20:35:55 +0000 (13:35 -0700)]
dat.conf: keep list of providers in order for backward compatibility

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>
12 years agoucm: record and silently drop a duplicate reject CM message
Arlin Davis [Thu, 10 May 2012 17:49:09 +0000 (10:49 -0700)]
ucm: record and silently drop a duplicate reject CM message

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>
12 years agowindows: new version of getlocalipaddr not portable
Arlin Davis [Wed, 25 Apr 2012 20:37:53 +0000 (13:37 -0700)]
windows: new version of getlocalipaddr not portable

revert to the original getaddrinfo method for windows

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>
12 years agodapltest: DFLT_QLEN is defined in multiple tests
Arlin Davis [Wed, 25 Apr 2012 20:36:52 +0000 (13:36 -0700)]
dapltest: DFLT_QLEN is defined in multiple tests

add #ifdef checking in transaction test.

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>