Hello,
We are just starting out with OFED and I am not sure how to troubleshoot
the following problem.
The setup has two machines, each has a RNIC and they communicate over
TCP/IP. From my understanding, just preloading libsdp.so in a TCP/IP
sockets based app should work fine. Is that correct? I
On Jul 15, 2008, at 6:52 AM, Tziporet Koren wrote:
- What is the proposed release schedule for OFED 1.4?
Alpha - next week and Beta by end of July.
Do you really mean alpha one week and beta the next?
Is there a target final release date for OFED v1.4?
--
Jeff Squyres
Cisco Systems
SWISS ON-LINE LOTTERY
P O Box 1010 Liverpool,
L70 1NL UNITED KINGDOM
REF NO: SL/74/368/05
BATCH NO:SL-121-LT-11-12-05
You won the sum of £1,500 000.00 GBP from our monthly PROGRAM, held on the 17th
July 2008.You are adviced to
Contact Mr.Michael Anderson to claim your prize.
Provide him with
SWISS ON-LINE LOTTERY
P O Box 1010 Liverpool,
L70 1NL UNITED KINGDOM
REF NO: SL/74/368/05
BATCH NO:SL-121-LT-11-12-05
You won the sum of £1,500 000.00 GBP from our monthly PROGRAM, held on the 17th
July 2008.You are adviced to
Contact Mr.Michael Anderson to claim your prize.
Provide him with
Nathan Dauchy wrote:
> Greetings,
>
> We have recently expanded our Infiniband tree and are running into
> problems when all hosts are booted. Details are below. Please let me
> know if there is a more appropriate forum for this issue. Thanks!
>
>
> With less than 600 hosts, everything seems
> If an HCA does not get a SM LID, ib_sa_remove_one() can cause a NULL
> pointer bug when it calls kref_put(). This is easy to see since
> ib_sa_add_one() sets sm_ah to NULL.
Have you seen this in practice? Because ib_sa_add_one() calls
update_sm_ah() before returning, so ib_sa_remove_one() s
I forgot to add that other than this
SystemGUID=0x issue, the HCA appears
to work perfectly.
Thanks,
Craig
Craig Prescott wrote:
Hi;
When we run 'ibdiagnet -r' on our OFED 1.2 cluster,
it bombs with a complaint about a system guid that is
zero on our only PCI-X HCA in the fab
Hi;
When we run 'ibdiagnet -r' on our OFED 1.2 cluster,
it bombs with a complaint about a system guid that is
zero on our only PCI-X HCA in the fabric (see appended).
ibdiagnet seems to be trying to saw off the leading zeroes
from the system guid, and to have nothing left afterwards
seems unexpe
Greetings,
We have recently expanded our Infiniband tree and are running into
problems when all hosts are booted. Details are below. Please let me
know if there is a more appropriate forum for this issue. Thanks!
With less than 600 hosts, everything seems to be working fine. With
more than 6
If an HCA does not get a SM LID, ib_sa_remove_one() can cause a NULL
pointer bug when it calls kref_put(). This is easy to see since
ib_sa_add_one() sets sm_ah to NULL.
Signed-off-by: Ralph Campbell <[EMAIL PROTECTED]>
---
drivers/infiniband/core/sa_query.c |3 ++-
1 files changed, 2 insert
The mpivars.sh and mpivars.csh scripts do not set the LD_LIBRARY_PATH,
which means that mvapich programs may not run unless the path is
explicitly set. In addition, mpivars.csh has a bug where it does not set
the path, either. I will be submitting a similar patch for mvapich2
shortly.
--- mvapich
You need to run set_nodedesc.sh right after the hardware driver is loaded. We
don't use openibd here so I don't know exactly what it does. Our modprobe.conf
runs set_nodedesc.sh as part of the "install" command for the hardware so the
node descriptor is set before the node starts responding to MA
I have one very specific question:
> --- a/drivers/infiniband/hw/mlx4/cq.c
> +++ b/drivers/infiniband/hw/mlx4/cq.c
> @@ -174,7 +174,7 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev,
> int entries, int vector
> if (entries < 1 || entries > dev->dev->caps.max_cqes)
>
It looks like this issue is important to a few of us. Can anyone
comment on these concerns?
-Original Message-
From: Chien Tung [mailto:[EMAIL PROTECTED]
Sent: Tuesday, July 15, 2008 4:32 PM
To: John Russo; [EMAIL PROTECTED]
Cc: general@lists.openfabrics.org
Subject: RE: [ofa-general] Qu
Roland Dreier wrote:
> > +int mlx4_qp_get_region(struct mlx4_dev *dev,
> > + enum qp_region region,
> > + int *base_qpn, int *cnt)
>
> I don't seem to see any call to this in the mlx4_en driver you posted to
> netdev, nor to MLX4_QP_REGION_ETH_ADDR. Can you ex
On Tuesday 15 July 2008 22:16, Or Gerlitz wrote:
> I'd be happy to get educated first, why this patch/param is needed at all?
Without this patch, QoS will not work at all.
> and second, why ofed 1.3 is shipped with this patch/param but an opposite
> default value...?
>
We noticed (after ofed 1.3
On Wed, Jul 16, 2008 at 02:12:32PM +0300, Or Gerlitz wrote:
> So the performance improvement was seen with both the multiple EQs and RCA
> patches applied?
Yes.
> I think that it would be better to know the improvement
> achieved by applying each one of them.
>
But the when I said "without RCA"
On 7/16/08, Olga Shern (Voltaire) <[EMAIL PROTECTED]> wrote:
> On 7/16/08, Keshetti Mahesh <[EMAIL PROTECTED]> wrote:
> > > ib-bonding for IPoIB is working only in ACTIVE - PASSIVE mode.
> > > No link aggregation mode is supported.
> >
> >
> > Thanks for the reply. I remember there was a IPOIB-HA t
Dear Webmaster,
My name is James and I represent online-poker-inside.com.
I have a business proposal for you regarding lists.openfabrics.org. Please get
back to me if you are interested and I will explain further so we can make some
business together.
Kind Regards,
James.
Eli Cohen wrote:
RCA stands for a a number of QPs, each connected to its private CQ, and each
such CQ reports to different EQ. The differnet EQs generate different vector
interrupts.
In tests I made on my machine I got the following results with netperf TCP, 4
streams.
Without RCA: 730 MB/s
Wi
Sean Hefty wrote:
Roland, this obviously isn't critical, but it would be nice to get into 2.6.27
if possible, to include it in the next release of the librdmacm.
--- a/include/rdma/rdma_cm.h
+++ b/include/rdma/rdma_cm.h
@@ -57,7 +57,8 @@ enum rdma_cm_event_type {
RDMA_CM_EVENT_DISCONNECT
with linux-2.6.26-rc6
Passed on ppc64 with linux-2.6.16
Passed on ppc64 with linux-2.6.24
Failed:
Build failed on x86_64 with linux-2.6.18
Log:
/home/vlad/tmp/ofa_1_4_kernel-20080716-0200_linux-2.6.18_x86_64_check/drivers/infiniband/hw/ipath/ipath_driver.c:2616:
error: implicit declaration of
On 7/16/08, Keshetti Mahesh <[EMAIL PROTECTED]> wrote:
> > ib-bonding for IPoIB is working only in ACTIVE - PASSIVE mode.
> > No link aggregation mode is supported.
>
>
> Thanks for the reply. I remember there was a IPOIB-HA tool
> included in OFED before which does the same. What happened to that
> ib-bonding for IPoIB is working only in ACTIVE - PASSIVE mode.
> No link aggregation mode is supported.
Thanks for the reply. I remember there was a IPOIB-HA tool
included in OFED before which does the same. What happened to that ?
Is there any work going on to support all 'mode's of ib-bondin
Hi Mahesh,
ib-bonding for IPoIB is working only in ACTIVE - PASSIVE mode.
No link aggregation mode is supported.
Olga
On 7/16/08, Keshetti Mahesh <[EMAIL PROTECTED]> wrote:
> 'ib-bonding.txt' file in the "docs" folder of OFED-1.3 release mentions
> ib-bonding as a HA solution for IPoIB interface
25 matches
Mail list logo
