On 2/7/2010 6:39 PM, Steve Wise wrote:
If ofed-1.5.1 is based on 2.6.33 then it will get this patch
automatically (assuming it goes upstream and makes 2.6.33). Or we can
pull it in as a kernel_patches/fixes/ patch.
OFED 1.5.1 is not based on 2.6.33, but on 2.6.30, so we need the patch
On 2/8/2010 8:02 AM, Sean Hefty wrote:
Since iWarp devices are not guaranteed to support loopback connections,
prevent rdma_bind_addr from associating the loopback address with
an iWarp device.
Signed-off-by: Sean Heftysean.he...@intel.com
Steve
Have you tested this patch?
When accepted
Currently, the recv buffer posting logic is based on the transactional
nature of iser which allows for posting a buffer before sending a PDU.
Change this to post only when the number of outstanding recv buffers is
below a water mark and in a batched manner, thus simplifying and
optimizing the data
With both the posting and reaping of recv buffers being in the
completion path, their outstanding number counter need not be atomic.
Signed-off-by: Or Gerlitz ogerl...@voltaire.com
---
drivers/infiniband/ulp/iser/iscsi_iser.h |2 +-
drivers/infiniband/ulp/iser/iser_initiator.c |6
Use a different CQ for send completions, where send completions are
being polled by the interrupt driven recv completion handler.
As such, interrupts aren't used for the send CQ.
Signed-off-by: Or Gerlitz ogerl...@voltaire.com
---
drivers/infiniband/ulp/iser/iscsi_iser.h |3
Simplify and shrink the logic/code used for the send descriptors.
Changes include removal of struct iser_dto which is unnecessary
abstraction, use struct iser_regd_buf only for handling SCSI commands,
use dma_sync instead of dma_map/unmap, etc.
Signed-off-by: Or Gerlitz ogerl...@voltaire.com
---
Two minor flows in iser's data path still use allocations,
move them to be atomic as a preperation step towards moving
to use libiscsi passthrough mode.
Signed-off-by: Or Gerlitz ogerl...@voltaire.com
---
drivers/infiniband/ulp/iser/iser_initiator.c |2 +-
currently iser recv completion flow takes the session lock twice.
optimize it to avoid the first one by letting iser_task_rdma_finalize()
be called only from the cleanup_task callback invoked by iscsi_free_task,
thus reducing the contention on the session lock between the scsi
command submission
currently iser recv completion flow takes the session lock twice.
optimize it to avoid the first one by letting iser_task_rdma_finalize()
be called only from the cleanup_task callback invoked by iscsi_free_task,
thus reducing the contention on the session lock between the scsi
command submission
On Sun, Feb 7, 2010 at 4:47 AM, Eli Dorfman (Voltaire)
dorfman@gmail.com wrote:
Hal Rosenstock wrote:
On Fri, Feb 5, 2010 at 9:18 AM, Eli Dorfman dorfman@gmail.com wrote:
On Thu, Feb 4, 2010 at 10:52 PM, Hal Rosenstock
hal.rosenst...@gmail.com wrote:
On Thu, Feb 4, 2010 at 12:43 PM,
Signed-off-by: Vladimir Sokolovsky v...@mellanox.co.il
---
src/mlx4.c |2 ++
1 files changed, 2 insertions(+), 0 deletions(-)
diff --git a/src/mlx4.c b/src/mlx4.c
index 1295c53..973df68 100644
--- a/src/mlx4.c
+++ b/src/mlx4.c
@@ -66,6 +66,8 @@ struct {
HCA(MELLANOX, 0x6354), /*
Tziporet Koren wrote:
On 2/7/2010 6:39 PM, Steve Wise wrote:
If ofed-1.5.1 is based on 2.6.33 then it will get this patch
automatically (assuming it goes upstream and makes 2.6.33). Or we can
pull it in as a kernel_patches/fixes/ patch.
OFED 1.5.1 is not based on 2.6.33, but on 2.6.30, so
This patch doesn't solve the openmpi/IB regression. So for OFED, IMO,
we need a different patch...
Tziporet Koren wrote:
On 2/8/2010 8:02 AM, Sean Hefty wrote:
Since iWarp devices are not guaranteed to support loopback connections,
prevent rdma_bind_addr from associating the loopback
Hal Rosenstock wrote:
On Sun, Feb 7, 2010 at 4:47 AM, Eli Dorfman (Voltaire)
dorfman@gmail.com wrote:
Hal Rosenstock wrote:
On Fri, Feb 5, 2010 at 9:18 AM, Eli Dorfman dorfman@gmail.com wrote:
On Thu, Feb 4, 2010 at 10:52 PM, Hal Rosenstock
hal.rosenst...@gmail.com wrote:
On Thu,
This patch doesn't solve the openmpi/IB regression. So for OFED,
IMO, we need a different patch...
If this doesn't solve the regression the we should have a different
patch for upstream too. The goal for 2.6.33 should be to keep open mpi
working, even if that requires us to go back to old
On Monday 08 February 2010 05:21:50 you wrote:
Octavian Purdila wrote:
On Friday 05 February 2010 06:45:38 you wrote:
Again, using bitmap algorithm is not a problem and it's better, the
problem is sysctl interface, how would you plan to interact with users
via sysctl/proc if you use
[PATCH 0/2] fix SRQ WQE buffer initialization in liblmx4 and in mlx4_ib
http://www.spinics.net/lists/linux-rdma/msg01911.html
This isn't a patch AFAICT.
[PATCH 1/2] libmlx4: initialize SRQ scatter entries when creating an SRQ
http://www.spinics.net/lists/linux-rdma/msg01912.html
Just
On Mon, Feb 8, 2010 at 11:05 AM, Eli Dorfman (Voltaire)
dorfman@gmail.com wrote:
Hal Rosenstock wrote:
On Sun, Feb 7, 2010 at 4:47 AM, Eli Dorfman (Voltaire)
dorfman@gmail.com wrote:
Hal Rosenstock wrote:
On Fri, Feb 5, 2010 at 9:18 AM, Eli Dorfman dorfman@gmail.com wrote:
On
Jason Gunthorpe wrote:
On Mon, Feb 08, 2010 at 08:52:10AM -0800, Roland Dreier wrote:
This patch doesn't solve the openmpi/IB regression. So for OFED,
IMO, we need a different patch...
If this doesn't solve the regression the we should have a different
patch for upstream too. The
to 127.0.0.1. Sean will this work?
If we agree to do this for 2.6.33, then I'll build/test this and resubmit.
rdma/cma: Disallow binding rdma endpoints to 127.0.0.1.
Currently this functionality breaks openmpi. Once openmpi is fixed to
correctly ignore 127.0.0.1 as a valid external rdma
IMO 127.0.0.1 should be for SW loopback, not HW RDMA loopback.
I disagree, but what does it matter? So, we add a 'software' loopback that uses
127.0.0.1. Openmpi still wouldn't work.
I will commit to get the fix in openmpi asap.
If we don't care if the fix is in the kernel or user space, then
Sean Hefty wrote:
IMO 127.0.0.1 should be for SW loopback, not HW RDMA loopback.
I disagree, but what does it matter? So, we add a 'software' loopback that uses
127.0.0.1. Openmpi still wouldn't work.
I guess that's true.
I will commit to get the fix in openmpi asap.
If
Hi Roland,
Any thoughts on this patch series?
Thanks,
/ac
* Alex Chiang achi...@hp.com:
This is v2 of a patch series that increases the maximum number of
IB HCAs supported per system.
The original mail thread is here:
http://lkml.org/lkml/2010/1/29/346
One note, I decided to
Sorry -- I missed many of these mails today due to mail filtering (don't ask).
FWIW:
- I'm not opposed to adding LOOPBACK checks into OMPI to avoid this problem
(I'm waiting for a patch, actually). I'm just saying that we're not going to
get a release out immediately with this fix. Our next
Here is a not tested patch that I think removes support for binding
to 127.0.0.1. Sean will this work?
If we agree to do this for 2.6.33, then I'll build/test this and resubmit.
rdma/cma: Disallow binding rdma endpoints to 127.0.0.1.
Currently this functionality breaks openmpi. Once
On Feb 8, 2010, at 5:09 PM, Jason Gunthorpe wrote:
DESCRIPTION
Associates a source address with an rdma_cm_id. The address may be
wildcarded. If binding to a specific local address, the rdma_cm_id
will also be bound to a local RDMA device.
This statement is trying
On Feb 8, 2010, at 5:13 PM, Sean Hefty wrote:
Are you certain that rdma_bind_addr does NOT work with 127.0.0.1, and that
this
is now the problem?
It does appear to work on OFED 1.4 and on 2.6.26 based on ucmatose. Is the
problem really with rdma_bind_addr succeeding, or with
On my OFED 1.4.1 RHEL4u6 systems, rdma_bind_addr() fails when attempting to
bind to 127.0.0.1 per the email I sent Friday:
http://www.spinics.net/lists/linux-rdma/msg02568.html
This is what I see over IB on 2.6.26, with a couple extra prints added to
cmatose:
Sean, can you try openmpi? It fails for me, and yet ucmatose succeeds.
I don't understand the difference yet...
Sean Hefty wrote:
On my OFED 1.4.1 RHEL4u6 systems, rdma_bind_addr() fails when attempting to
bind to 127.0.0.1 per the email I sent Friday:
Sean, can you try openmpi? It fails for me, and yet ucmatose succeeds.
I don't understand the difference yet...
I believe the issue is that rdma_bind_addr succeeds (returns 0), but no device
is assigned to the rdma_cm_id (verbs field is NULL).
This was a change from commit
On Feb 8, 2010, at 6:48 PM, Sean Hefty wrote:
rc = rdma_bind_addr(cm_id, ipaddr);
if (rc || !cm_id-verbs) {
rc = OMPI_SUCCESS;
goto out3;
}
Ah, yes! Per the OMPI code you cited, I amended my printf's and see:
[svbu-mpi.cisco.com:19315] FAILED to bind to 127.0.0.1:
Jeff Squyres wrote:
On Feb 8, 2010, at 7:30 PM, Pradeep Satyanarayana wrote:
elm3b199:/usr/lib # /usr/mpi/gcc/openmpi-1.4.1/bin/mpirun -np 2 --bynode
--mca btl_openib_cpc_include rdmacm ring
--
mpirun was unable to
On Feb 8, 2010, at 7:50 PM, Pradeep Satyanarayana wrote:
No, there is none. I got this command from one of the mails in the thread.
What should I use instead?
You need to compile and run an MPI program. ring is a typical test program
that sends a message around in a ring. I think that OFED
Hi Sasha,
I have incorporated your feedback and sent you the modified patch with a new
subject line [PATCH V2] opensm: Add option to specify prefix to syslog
messages
Regards,
Benjamin
Sasha Khapyorsky wrote:
Hi Benjamin,
On 14:16 Thu 21 Jan , Arputham Benjamin wrote:
Added a text
34 matches
Mail list logo