[EMAIL PROTECTED] wrote on 10/16/2006 01:50:49 PM:
On Mon, 2006-10-16 at 15:25 +0200, Michael S. Tsirkin wrote:
Quoting r. Maestas, Christopher Daniel [EMAIL PROTECTED]:
Subject: Re: [openib-general] RHEL5 and OFED ...
Now for userspace - does RHEL5 include at least libibverbs-1.0?
Hi Troy.
Troy Benjegerdes wrote:
If ibv_reg_mr fails, can an application (or library, such as pvfs)
assume that this is just a temporary error, and try to deregister
some memory, then try again?
I believe that the answer is not always. They may be several reasons for
a memory
Hi
Yossi,
It
should be fixed in OFED-1.1-pre1.
Regards,
Vladimir
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of yosef
etiginSent: Monday, October 16, 2006 11:35 AMTo:
openib-general@openib.orgSubject: [openib-general] building OFED
package from git and svn
Hello,I
(I am taking this back to the openib list because I think the list
needs to hear about real applications that are hitting memory
registration limits)
What are the limits on the ehca memory registrations?
Is there a limit to the number of regions that can be registered?
The numbe rof
Hello Troy and Kyle!
Kyle wrote:
Our app writes out a file once, then reads it in many times through
the pvfs2 system. In the pvfs2 layers, there is memory caching
done at the network level, so memory is registered by the app, and
attempts are made to re-register and/or re-use these
Hi,
What testing did these companies do with rc7?
Still testing rc7. Will post our results probably today evening.
Regards!
Nam
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To
Shirley Ma wrote:
[EMAIL PROTECTED] wrote on 10/16/2006 01:50:49 PM:
On Mon, 2006-10-16 at 15:25 +0200, Michael S. Tsirkin wrote:
Quoting r. Maestas, Christopher Daniel [EMAIL PROTECTED]:
So I would think rhel5 would have at least that or greater. When I
compiled rpms for 1.1rc7
Hi Nam,
Can this patch be saved for the next OFED-1.2 release?
Note: OFED installation script checks that sysfsutils package installed.
Regards,
Vladimir
-Original Message-
From: Hoang-Nam Nguyen [mailto:[EMAIL PROTECTED]
Sent: Wednesday, October 18, 2006 7:02 PM
To: Vladimir
Hi Vlad!
Can this patch be saved for the next OFED-1.2 release?
Note: OFED installation script checks that sysfsutils package installed.
As Michael indicated in previous email configure is a generated file
from autogen.sh. And I'm not sure if your packaging script does generate
it automatically
Thanks for your patch.
I agree with some of the changes you suggest and disagree with others. It
will be useful to post a different patch for each logical change.
1. Fixes bug in srp_daemon for the case where if it is invoked with the
'-e' option, it fails to connect to the SRP targets because
Compilation warning on 64 bit machines was fixed.
Signed-off-by: Dotan Barak [EMAIL PROTECTED]
---
Index: last_stable/src/userspace/librdmacm/examples/rping.c
===
--- last_stable.orig/src/userspace/librdmacm/examples/rping.c
Looks good.
--
Yevgeny
Hal Rosenstock wrote:
OpenSM/osm_port_info_rcv.c: Remove duplicate dump of received PortInfo
in osm_pi_rcv_process
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
Index: opensm/osm_port_info_rcv.c
===
Committed revision 9898.
Thanks,
Steve.
On Thu, 2006-10-19 at 14:58 +0200, Dotan Barak wrote:
Compilation warning on 64 bit machines was fixed.
Signed-off-by: Dotan Barak [EMAIL PROTECTED]
___
openib-general mailing list
Diags/ibportstate: For query op, add peer port checking of link width
and speed active
This is requiresthe combined route support in libibmad
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
---
Index: src/ibportstate.c
===
---
From: Ishai Rabinovitz [mailto:[EMAIL PROTECTED]
Subject: Re: [openib-general] [PATCH] IB/SRP Userspace:
srptools/srp_daemon - Fix connect bug and add support for
user specified initiator extension
Thanks for your patch.
I agree with some of the changes you suggest and disagree
with
Hoang-Nam Nguyen wrote:
As Michael indicated in previous email configure is a generated file
from autogen.sh. And I'm not sure if your packaging script does generate
it automatically when you build the whole OFED tgz-file.
The actual patches in configure.in and config.h.in are needed because
Hi Or,
Please see below,
Regards,
Vladimir
2) OPENIB_PARAMS documentation
Doing some probing, i undersrand that to set this or that
option to the
build i need to set this or that --with or --without
directive to the
SOURCES/openib-1.1/ofed_scripts/configure script and this is
Initial value of LFT block variable was double calculated (first time
as 'startblock' and then block = startblock/BLOCK_SIZE).
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
---
diags/src/ibroute.c |2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/diags/src/ibroute.c
Trivial comments fixing - incorrect function name in one place and
incorrect function description in another.
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
---
osm/include/opensm/osm_port.h |6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)
diff --git
On Thu, 2006-10-19 at 09:50, Sasha Khapyorsky wrote:
Trivial comments fixing - incorrect function name in one place and
incorrect function description in another.
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
Thanks. Applied.
-- Hal
___
On Thu, 2006-10-19 at 09:49, Sasha Khapyorsky wrote:
Initial value of LFT block variable was double calculated (first time
as 'startblock' and then block = startblock/BLOCK_SIZE).
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
Good catch. Thanks. Applied.
-- Hal
Hi,
The release is closed. We only updating the documents now (will be
closed in the coming few hours only).
Since ehca is in technology preview state these issues are not blockers.
Please document all issues in ehca release_notes (or send me parts you
want to include).
Even though ehca is in
On Thu, 2006-10-19 at 07:09 +0200, Michael S. Tsirkin wrote:
Quoting r. Doug Ledford [EMAIL PROTECTED]:
Subject: Re: [openfabrics-ewg] RHEL5 and OFED ...
On Wed, 2006-10-18 at 09:29 +0200, Michael S. Tsirkin wrote:
Quoting r. Doug Ledford [EMAIL PROTECTED]:
From our dicussion, it
Hi Nam,
The code that fixes lib/lib64 issue on ppc64 and libehca.so to be in
libehca RPM is in OFED-1.1.
Regards,
Vladimir
-Original Message-
From: Hoang-Nam Nguyen [mailto:[EMAIL PROTECTED]
Sent: Thursday, October 19, 2006 4:41 PM
To: Tziporet Koren
Cc: Michael S. Tsirkin;
Shirley I am not sure whether this already has an answer. The
Shirley justification is madvise(..., MADV_DONTFORK) is used to
Shirley make fork() work for verbs consumers in the recent
Shirley packages. I hope same patch will be in libehca.
That's not in any released version of
Hi Vlad,
The code that fixes lib/lib64 issue on ppc64 and libehca.so to be in
libehca RPM is in OFED-1.1.
Thanks! Can you pls confirm if configure.in and config.h.in are in
OFED-1.1 as well?
Regards
Nam
___
openib-general mailing list
Doug Ledford wrote:
... and reviewing arpingib
(which I'm going to remove from the ipoibtools and fix the native arping
in RHEL5 to work properly over IB without needing a new flag, the -A or
-U flags should be sufficient assuming those modes worked at all over IB
which they don't in either
No, the updated configure.in and config.h.in are not in OFED-1.1.
In any case, I believe that most of the checks you have added to
configure scripts are provided by OFED installation scripts.
So, in case OFED installation fails, ehca configure would fail as well.
Regards,
Vladimir
so this means this problem we try to fix is already covered by your new
setup?
Do you have a version we could try?
Gruss / Regards . . . Christoph Raisch
christoph raisch, HCAD teamlead, IODF2 (d/3627), ibm boeblingen lab,
phone: (+49/0)7031-16 4584, fax: -16 2042, loc: 71032-05-003, internet:
Krishna Kumar2 wrote:
I felt that since the time_after() check matched (in all likelyhood) due
to the
processing of the cancellation, ECANCELLED is more appropriate to return.
It is most likely that if both conditions are true, that a cancelled
operation led to
the time_after() match
Krishna Kumar2 wrote:
This will mean that a deref is wrongly done if a loopback or zero address
is
passed to this function, without it having done a ref inc. I do think this
case
requires a variable to indicate whether a ref was got or not. Assuming
that is
true, I will submit a patch
Vladimir Sokolovsky wrote:
OPENIB_PARAMS supports only the following options:
--with-memtrack
--without-modprobe
--with-madeye-mod
--without-ipoibconf
CONFIG_INFINIBAND_IPOIB_DEBUG option is set by default if IPoIB selected
You can see all configuration options in
Hi Ira.
Ira Weiny wrote:
I got the following error running with OFED 1.1 on a modified 2.6.9 RHEL4
kernel. Hal mentioned that there might be a catastrophic error recovery patch
submitted since then? I can't find a mention of that in the mailing list. If
possible I would like to try such a
In case osmtest failed to remove some MC group,
dumping all the MC groups that still remain,
and their member.
--
Yevgeny
Signed-off-by: Yevgeny Kliteynik [EMAIL PROTECTED]
Index: osmtest/osmt_multicast.c
===
---
Vlad,
I am trying to build OFED1.1-rc7 on a system where i have built the RH4 U3
kernel manually from its source rpm (that prepared kernel-2.6.9-34.EL.src.rpm
for being built and then did make, make modules_install, make install and
reboot)
and get the errors below.
Do you have any idea what i
On Thu, 2006-10-19 at 08:15 -0700, Shirley Ma wrote:
Roland Dreier [EMAIL PROTECTED] wrote on 10/19/2006 07:51:31 AM:
Shirley I am not sure whether this already has an answer. The
Shirley justification is madvise(..., MADV_DONTFORK) is used to
Shirley make fork() work for verbs
On Thu, 2006-10-19 at 17:02 +0200, Or Gerlitz wrote:
Doug Ledford wrote:
... and reviewing arpingib
(which I'm going to remove from the ipoibtools and fix the native arping
in RHEL5 to work properly over IB without needing a new flag, the -A or
-U flags should be sufficient assuming those
On Oct 18, 2006, at 8:10 AM, Jeff Squyres wrote:
One feature that bugzilla has (and that seems to be disabled in
openib bugzilla
:() is mail integration, where I can Cc bugzilla and mail contents
will get
attached to bug report. I was hoping that new server will have this
capability. Does
Hi Or,
I think that required for 2.6.9-34.EL kernel backport patches from
kernel_patches/backport/2.6.9_U3 directory are not applied by configure
script.
You should change kernel name to be 2.6.9-34*.
Regards,
Vladimir
-Original Message-
From: Or Gerlitz [mailto:[EMAIL PROTECTED]
Shirley How can RHEL5 pick up this particular patch? Applications
Shirley with fork() depend on this patch.
It can't really, since it breaks the libibverbs ABI and therefore has
to be part of a major release.
___
openib-general mailing list
Roland Dreier [EMAIL PROTECTED] wrote on 10/19/2006 09:19:14 AM:
Shirley How can RHEL5 pick up this particular patch? Applications
Shirley with fork() depend on this patch.
It can't really, since it breaks the libibverbs ABI and therefore has
to be part of a major release.
Then we
On Thu, 2006-10-19 at 07:09 +0200, Michael S. Tsirkin wrote:
Yeah, I'll look over the diff today.
How does it look?
Not too far in yet, but the srp_topspin patch in the
kernel_patches/fixes directory appears to have munged whitespace.
--
Doug Ledford [EMAIL PROTECTED]
GPG
Doug, what udev does RHEL5 beta have? Any plans to
upgrade udev for RHEL4 U5?
Scott
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Ishai
RabinovitzSent: Tuesday, October 17, 2006 5:36 AMTo:
Sharma, Karun Cc: [EMAIL PROTECTED];
OpenSM/osm_lid_mgr.c: Fix base LID if needed to eliminate potential
infinite loop
If SMA responds with base LID of 0x in PortInfo, the for loops
following the call to osm_port_get_lid_range_ho would cause an infinite
loop.
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
Index:
On Thu, 2006-10-19 at 11:24 -0700, Scott Weitzenkamp (sweitzen) wrote:
Doug, what udev does RHEL5 beta have? Any plans to upgrade udev for
RHEL4 U5?
RHEL5 currently has something like 0.97 I think. For RHEL4.5, I don't
currently have any plans to update udev so drastically as it would
require
Roland Dreier wrote:
chas i would guess the read to the mmio region is flushing the
chas writes to the config register but the read happens too
chas soon after those writes. on a more mundance computer, the
chas write/write/read probably wouldnt be batched together.
config
Quoting r. Doug Ledford [EMAIL PROTECTED]:
Subject: Re: [openfabrics-ewg] RHEL5 and OFED ...
On Thu, 2006-10-19 at 07:09 +0200, Michael S. Tsirkin wrote:
Yeah, I'll look over the diff today.
How does it look?
Not too far in yet, but the srp_topspin patch in the
I'm not sure the standard OpenIB NetPIPE runs can reproduce this
type of workload. However, we have developed a working PVFS2-
NetPIPE module which can reproduce this problem on occassion, if
there is interest in further testing this on your end, I can make
it available.
Yes. Please send
From: Ishai Rabinovitz [EMAIL PROTECTED]
This makes SRP more robust in presence of hardware errors
and is closer to behaviour suggested by IB spec,
reducing chance of stale packets.
Signed-off-by: Ishai Rabinovitz [EMAIL PROTECTED]
Signed-off-by: Michael S. Tsirkin [EMAIL PROTECTED]
---
Hello,
On Thu, 2006-10-19 at 21:54 +0200, Michael S. Tsirkin wrote:
Quoting r. Doug Ledford [EMAIL PROTECTED]:
Subject: Re: [openfabrics-ewg] RHEL5 and OFED ...
On Thu, 2006-10-19 at 07:09 +0200, Michael S. Tsirkin wrote:
Yeah, I'll look over the diff today.
How does it look?
Hi,
On executing the command 'rpm -ivh *.rpm ' the following error
message came up:-
error: Failed dependencies:
kernel(kernel) = 07562a5eb4f39f26 is needed by
kernel-ib-1.0-2.6.16.21_0.8_smp.x86_64
kernel(drivers) = 2b3023c350dc4c0d is needed by
Quoting r. Shirley Ma [EMAIL PROTECTED]:
Subject: Re: [PATCH/RFC 1/2] IB: Return maybe_missed_event hint from
ib_req_notify_cq()
Roland Dreier [EMAIL PROTECTED] wrote on 10/18/2006 01:55:13 PM:
I would like to understand why there's a throughput difference with
scaling turned off, since
On Thu, Oct 19, 2006 at 11:37:55AM -0400, Doug Ledford wrote:
and ISTR that it
isn't even required by the MPI spec since that leaves behavior of an MPI
app undefined after a fork() call and hence any application written to
depend on undefined behavior is broken by design,
Doug,
There are
This is extension to existing OpenSM 'file' routing engine. This adds
ability to dump switch lid matrices (aka min hops tables) to file and
later to load it as is (currently).
The usage is similar to unicast forwarding tables loading from dump
file (introduced by 'file' routing engine), but new
This adds new method named build_lid_matrices() to OpenSM routing engine
structure. When defined this method will be used by ucast_mgr_process()
for switch min hop tables (aka lid matrices) preparation. In case of
failure default lid matrix creation algorithm will be used.
Signed-off-by: Sasha
This adds dumping switches lid matrices to the file
'opensm-lid-matrix.dump'. Like other routing related dumps this code
will be activated when OSM_LOG_ROUTING logging flag is set.
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
---
osm/opensm/osm_ucast_mgr.c | 32
This unifies ucsat_mgr dumper. Main goal is to provide infrastructure
for different dump file generation using the same routines.
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
---
osm/opensm/osm_ucast_mgr.c | 104 +++-
1 files changed, 55
This is bonus - switch forwarding tables dump compatible with output
produced by dump_lfts.sh and which can be used as input for unicast
forwarding tables loader (with -R 'file' -U file-name). The dump file
name is 'opensm-lfts.sh' and will be generate if OSM_LOG_ROUTING
logging flag is set.
This adds lid matrices dump file parser and loader. It is part of 'file'
routing engine, the file name should be specified in OpenSM command line
with -M or --lid_matrix_file option (no default value). Example of valid
usage is:
opensm -R file -M ./opensm-lid-matrix.dump
The file format is
Hi,
I ran into the same problem. What I did to get past this was:
mv /usr/lib/rpm/find-provides.ksyms /usr/lib/rpm/find-provides.ksyms.org
mv /usr/lib/rpm/find-requires.ksyms /usr/lib/rpm/find-requires.ksyms.org
mv /usr/lib/rpm/find-supplements.ksyms
/usr/lib/rpm/find-supplements.ksyms.org
Thanks Michael for all these tips. I have tried several suggestions as you proposed here. I couldn't see performance any better. The TCP_RR is dropped to 472 trans/s from about 18,000 trans/s , and TCP_STREAM BW is dropped to 1/3 as before ( ehca + scaling code) with same TCP configuration, send
OK, as promised I redid the request notify patches according to
Michael's suggestion to add a new flag. I think I like this a lot
better -- I'll send out the new patches as replies to this email for
comments.
- R.
___
openib-general mailing list
The semantics defined by the InfiniBand specification say that
completion events are only generated when a completions is added to a
completion queue (CQ) after completion notification is requested. In
other words, this means that the following race is possible:
while (CQ is not empty)
QLogic have tests OFED 1.1pre1 and are happy with the results. We have
tests UD, UC, RC, IPoIB, SDP and uDAPL.
Regards,
Robert.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To
Do you have any performance data?
Scott Weitzenkamp
SQA and Release Manager
Server Virtualization Business Unit
Cisco Systems
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Robert Walsh
Sent: Thursday, October 19, 2006 4:23 PM
To: [EMAIL
Cisco is happy with OFED 1.1 pre1, we only did light testing because no
C changes were made. The following bugs have been tested and closed.
273 OFED 1.1 rc7 does not work with Cisco FC Gateway
278 OFED 1.1: two copies of openib.spec in openib-1.1.tgz
268 OFED openibd script
This removes obsolete now shared sm-p_report_buf buffer and cleans
up related code.
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
---
osm/include/opensm/osm_base.h |5 --
osm/include/opensm/osm_sm.h|2 -
osm/include/opensm/osm_state_mgr.h |8 ---
Hi,
Where could I find the file openib-1.1.tgz ? Couldn't find it on
www.openib.org ...Thanks!
Vishal
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit
Roland,
I have applied this patch and updated patch 2/2. You will send out an updated patch 2/2, I think.
I did some extra modification in ipoib code, (which has more extra repolls). I do see around 10% or more performance improvement now with this change on both scaling and none scaling code.
Convert the IP-over-InfiniBand network device driver over to using
NAPI to handle all completions (both receive and send).
Signed-off-by: Roland Dreier [EMAIL PROTECTED]
---
drivers/infiniband/ulp/ipoib/ipoib.h |1
drivers/infiniband/ulp/ipoib/ipoib_ib.c | 78
I have applied this patch and updated patch 2/2. You will send out an
updated patch 2/2, I think.
Sorry, messed that up. I just sent out the patch.
I did some extra modification in ipoib code, (which has more extra
repolls). I do see around 10% or more performance improvement now with
Roland Dreier [EMAIL PROTECTED] wrote on 10/19/2006 07:39:25 PM:
I have applied this patch and updated patch 2/2. You will send out an
updated patch 2/2, I think.
Sorry, messed that up. I just sent out the patch.
No problem, I did same change.
You mean you add more calls to
23454:Config Write REG = 01 TYPE = 1BE = Req = (0,0,0)
Tag = 1 Bus = 1 Device = 0 Function = 0 WAIT = 2
23462:Memory Rd DW A = 00280698 BE = Req = (0,0,0) Tag = 0
WAIT = 2
23470:Split compl. Lower A = 00 Req = (0,0,0) Tag = 0 Comp
I run out of ideas why losing 2/3 of the throughput and got 476 trans/s. So
I assumed there was always a missed event, then ipoib would stay in its
napi poll within its scheduled time. That's why it helps. This is really a
hack, doesn't address the problem. It sacrificed cpu utilization
Roland Dreier [EMAIL PROTECTED] wrote on 10/19/2006 09:10:35 PM:
It's entirely possible that my implementation of the missing event
hint in ehca is wrong. I just guessed based on how poll CQ is
implemented -- if the consumer requests a hint about missing events,
then I lock the CQ and check
75 matches
Mail list logo