This email was generated automatically, please do not reply
Common build parameters: --with-ipoib-mod --with-sdp-mod --with-srp-mod
--with-user_mad-mod --with-user_access-mod --with-mthca-mod --with-core-mod
--with-addr_trans-mod --with-rds-mod --with-cxgb3-mod
Passed:
Passed on i686 with
Roland Dreier wrote:
Yes, you can't DMA to VDSO VMA I don't think.
Why not? It's just RAM...
Well ... isn't it read-only?
True... you shouldn't be able to DMA to it. But I assume Dotan is
trying to register the memory with read-only permission and DMA from
it. Dotan, can
Roland Dreier wrote:
If its indeed the case, does removing this line provides a solution to
the problem or just a work around?
Obviously it's just a work around, since it disables this performance
enhancement in the firmware.
OK, commenting these two lines in the init hca code makes the
Hey Roland,
This patch is needed for iw_cxgb3 to handle a change in the cxgb3 driver
posted by Divy that Jeff recently applied. If the cxgb3 change is
destined for 2.6.21, then this change to iw_cxgb3 also needs to go in
(otherwise we get an error log entry for every rdma connection).
It was
Quoting Hal Rosenstock [EMAIL PROTECTED]:
Subject: Re: multicast join failed for...
On Wed, 2007-04-11 at 23:38, Michael S. Tsirkin wrote:
Quoting Hal Rosenstock [EMAIL PROTECTED]:
Subject: Re: multicast join failed for...
On Wed, 2007-04-11 at 15:47, Michael S. Tsirkin wrote:
Quoting G.O. [EMAIL PROTECTED]:
Subject: Re: [ofa-general] does RHEL5 Xen work with OFED?
On 4/10/07, Michael S. Tsirkin [EMAIL PROTECTED] wrote:
Quoting G.O. [EMAIL PROTECTED]:
Subject: Re: [ofa-general] does RHEL5 Xen work with OFED?
On 4/5/07, Scott Weitzenkamp (sweitzen) [EMAIL
Roland:
Thanks for the suggestion. What is the minimum safe value of timeout for
typically IB network with 2-3 level of switch ?
--CQ
-Original Message-
From: Roland Dreier [mailto:[EMAIL PROTECTED]
Sent: Wednesday, April 11, 2007 10:48 PM
To: Tang, Changqing
Cc: Sean Hefty;
Thanks for the suggestion. What is the minimum safe value of timeout for
typically IB network with 2-3 level of switch ?
It depends, since congestion may delay messages for quite a while.
Probably a timeout of 100 milliseconds or so works pretty well. Of
course there is a tradeoff here
I think it's too late for 2.6.21, since this is really not an obvious
change and we don't know how it will interact with all the different
HCAs and FW versions in use.
- R.
___
general mailing list
[EMAIL PROTECTED]
ibv_reg_mr fails for me.
When i added some debug prints i noticed the failure in file:
uverbs_mem.c function: get_page_shift,
find_vma returned NULL.
get_page_shift() doesn't appear in the upstream kernel, so this is
some patch from OFED breaking things I guess. Does the test work with
On Thu, 2007-04-12 at 08:23 -0700, Roland Dreier wrote:
So is the cxgb3 net driver change in question already in Linus's tree?
What is the exact patch that this change goes with?
The patch is the 3rd of 3:
http://marc.info/?l=linux-kernelm=117617444622279w=2
Jeff applied it into his upstream
On Thu, 12 Apr 2007 07:21:55 +0300
Michael S. Tsirkin [EMAIL PROTECTED] wrote:
Quoting Ira Weiny [EMAIL PROTECTED]:
Subject: Re: [ofa-general] Re: multicast join failed for...
On 11 Apr 2007 17:45:54 -0400
Hal Rosenstock [EMAIL PROTECTED] wrote:
On Wed, 2007-04-11 at 15:47,
Roland Dreier wrote:
I think it's too late for 2.6.21, since this is really not an obvious
change and we don't know how it will interact with all the different
HCAs and FW versions in use.
We test it here with all our HCAs (results are good).
In any case we will put it into OFED 1.2
Tziporet, can you open a bug please?
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of
Tziporet Koren
Sent: Thursday, April 12, 2007 8:57 AM
To: Roland Dreier (rdreier)
Cc: [EMAIL PROTECTED]
Subject: Re: [ofa-general] Re: [PATCH] IB/mthca: work
The patch is the 3rd of 3:
http://marc.info/?l=linux-kernelm=117617444622279w=2
Jeff applied it into his upstream tree here:
http://marc.info/?l=linux-netdevm=117630664627997w=2
OK, so it's not in Linus's tree yet. Jeff, how do you want to handle
this? (That last patch breaks
We test it here with all our HCAs (results are good).
In any case we will put it into OFED 1.2
Seems like a good plan. I have no real objection to this patch, but
the 2.6.21 kernel tree is at the stage where we really only want to
merge very urgent fixes. This is something that can go into
I tried adding/removing new storage on sles10. It took few minutes to find
the new target devices (the new target message was showed on
/var/log/messages) then took few minutes to add the path. I did not run
multipath again. The srp_daemon.sh scanned the new target and added path
automatically.
Quoting Ira Weiny [EMAIL PROTECTED]:
Subject: Re: [ofa-general] Re: multicast join failed for...
On Thu, 12 Apr 2007 07:21:55 +0300
Michael S. Tsirkin [EMAIL PROTECTED] wrote:
Quoting Ira Weiny [EMAIL PROTECTED]:
Subject: Re: [ofa-general] Re: multicast join failed for...
On
Hi,
I was partly succesful in inserting the ib_gmthca module into the guest domain.
This is the debug output I get while inserting the module :
modprobe ib_gmthca mc=1 domain=2
host ip 192.168.0.3
domain 2
[drivers/infiniband/hw/gmthca/../../utils/kernel_socket.c:62],1Connection ok
Never mind, I see that cxgb3 - missing CPL hanler and register
setting. has appeared in Linus's tree. Steve, I'll ask Linus to pull
this fix today. Jeff, never mind my question since it's too late now.
- R.
___
general mailing list
[EMAIL PROTECTED]
Linus, please pull from
master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband.git for-linus
This tree is also available from kernel.org mirrors at:
git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband.git
for-linus
This will fix up some problems with
Quoting Shirley Ma [EMAIL PROTECTED]:
Subject: Re: [PATCH] IB/mthca: work around kernel QP starvation
Hello Michael,
We saw the same problem. Is a userspace patch needed?
Thanks
Shirley Ma
No, we are protecting kernel QPs from being starved by userspace,
and we can't trust userspace
On Thu, 2007-04-12 at 13:36, Roland Dreier wrote:
Definitely a big improvement to readability. However, I don't like
the smi_type name, since the enum is not really a type but rather an
action:
+enum smi_type {
+ IB_SMI_DISCARD,
+ IB_SMI_HANDLE
+};
+
+enum
BTW, Roland, why aren't we using txqueuelen ifconfig/ethtool options here?
The ifconfig option is about the TX queue outside the driver's
hardware queue. Not sure what ethtool is setting. I think the main
reasons why we're not using ethtool are:
- the patches I got didn't do it
- it would
Hello Michael,
Could you please create a patch against OFED-1.1? And in the future
what's the process to apply this kind of patch for prev OFED release?
Thanks
Shirley Ma
IBM Linux Technology Center
15300 SW Koll Parkway
Beaverton, OR 97006-6063
Phone(Fax): (503)
Hello James,
We are trying to find a way for the OpenIB-cma uDAPL layer to coexist
with SGI's xpmem uDAPL on a single system.
Obviously, the installation scriptlets for xpmem uDAPL need to
add lines into /etc/dat.conf when xpmem UDAPL is installed.
Since a static version of /etc/dat.conf
I am running into a number of mthca issues listed below and need help with
them.
1. I am using linux-2.6.21-rc5 and I see this Oops when I modprobe
ib_mthca (on ppc64)
Apr 12 14:11:19 elm3b37 kernel: ib_mthca 0002:d9:00.0: HCA FW version
3.3.3 is old (3.4.0 is current).
Apr 12 14:11:19
Rick,
SDP is easy to get with netperf by running LD_PRELOAD=libsdp.so
netperf/neterver, so in my opinion SDP is already covered.
Here's a list of things I'd like to see from netperf, in priority order:
1) IP multicast
2) Test that uses multiple concurrent sockets at the same time
3) RDS
I
Scott Weitzenkamp (sweitzen) wrote:
Rick,
SDP is easy to get with netperf by running LD_PRELOAD=libsdp.so
netperf/neterver, so in my opinion SDP is already covered.
The one and IMO very big worry I have about using LD_PRELOAD is that it
does not change the netperf test banner. So, it leaves
I just tagged the 1.1 release of libibverbs and pushed it out to
my git tree on kernel.org:
git://git.kernel.org/pub/scm/libs/infiniband/libibverbs.git
(the name of the tag is libibverbs-1.1).
I've also copied a tarball into my home directory on openfabrics.org,
with sha1sum:
On Thu, 12 Apr 2007 20:16:32 +0300
Michael S. Tsirkin [EMAIL PROTECTED] wrote:
Quoting Ira Weiny [EMAIL PROTECTED]:
Subject: Re: [ofa-general] Re: multicast join failed for...
On Thu, 12 Apr 2007 07:21:55 +0300
Michael S. Tsirkin [EMAIL PROTECTED] wrote:
Quoting Ira Weiny [EMAIL
I was using default netperf params, throughput is stable now that I use
-- -s 349520 -S 349520 -m 65536 to force socket buffer and message
sizes.
Scott
-Original Message-
From: Or Gerlitz [mailto:[EMAIL PROTECTED]
Sent: Thursday, April 12, 2007 4:54 AM
To: Moni Shoua
Cc: Scott
Quoting Rick Jones [EMAIL PROTECTED]:
Subject: Re: desired netperf mods?
Scott Weitzenkamp (sweitzen) wrote:
Rick,
SDP is easy to get with netperf by running LD_PRELOAD=libsdp.so
netperf/neterver, so in my opinion SDP is already covered.
The one and IMO very big worry I have about
The job will continue running though, and when you diagnose the problem
and disconnect the bad node, rate will be back to high.
So what's the problem?
What would bring the rate back up?
Halting all multicast traffic across the subnet to handle a flaky node wanting
to join some multicast
Caderousse was choking; he sent embarrassed looked wed taken around for somehat
You say must see wait delight for what? asked Caderousse. copy island What must
mine successfully then be done?Yes, injure glass for I can follow you short
quality no farther, and I only
Yes.sadly jump That is
Quoting Sean Hefty [EMAIL PROTECTED]:
Subject: RE: [ofa-general] Re: multicast join failed for...
The job will continue running though, and when you diagnose the problem
and disconnect the bad node, rate will be back to high.
So what's the problem?
What would bring the rate back
When the node is diagnosed and disconnected, SM will bring the rate
back up.
But how? Doesn't it require re-registration of all multicast groups and
clients registered for SA events?
As I said, there are tens of ways a bad node can hurt performance,
and we don't/can't handle them. Why focus on
37 matches
Mail list logo
