This patch uses basic or extended counters which can be read by a command
interface, to report counters for all the QPs that work on an rdmaoe port. This
effectively allows to implement performance counter ala IB.
Signed-off-by: Eli Cohen e...@mellanox.co.il
---
drivers/infiniband/hw/mlx4/mad.c
On 11/4/2010 10:45 PM, Craig Prescott wrote:
Hi,
We have a new-ish node whose IPoIB access was interrupted last night;
last message in the syslog was:
ib0: failed send event (status=2, wrid=74 vend_err 67)
The node has a Mellanox ConnectX HCA (MHQH19-XTC).
I understand the status
On 11/9/2010 9:22 PM, Usha Srinivasan wrote:
Hello,
Can someone from Mellanox tell me what the vendor error 0x32 means? I am
getting this error for wc.opcode 128 (IB_WC_RECV) wc.status 4
(IB_WC_LOC_PROT_ERR).
This error means corrupted MPT or PD violation
Tziporet
--
To
Usha Srinivasan wrote:
Can someone from Mellanox tell me what the vendor error 0x32 means? I am
getting this error for wc.opcode 128 (IB_WC_RECV) wc.status 4
(IB_WC_LOC_PROT_ERR). I am running ofed 1.5.2 and am getting it on both rhel5
and sles11
You can't count on the wc.opcode when the
On 11/09/10 20:44, Jason Gunthorpe wrote:
Broadly it looks to me like your actions are in the wrong order.
A poll based RDMA loop should look like this:
- exit poll
- Check poll bit
- call ibv_get_cq_event
- call ibv_req_notify_cq
- repeatedly call ibv_poll_cq (while rc == num requested)
-
On 11/10/10 10:30, Andrea Gozzelino wrote:
Hi Jonathan,
I wrote down a test (latency and transfer speed) with RDMA.
Server and client work with the same code and they change defined size
buffers for n times (loop). In the makefile.txt, you can find an help to
use the code.
I tested Intel
It seems there are several ways to transfer data via RDMA: Send/Send
w/immediate, Receive, RDMA Write/RDMA Write w/immediate, the Atomic
extensions, and then the various transport modes variations with some
of these operations. As an RDMA/IB newbie, it is somewhat confusing
when
Could I get some clarification on where there is no ordering
guarantee? The WC's do not necessarily come back in the order that the
sends were posted?
For a given queue, completions are always returned in the order that
work requests were posted. However there is no ordering between
On Wed, Nov 10, 2010 at 09:43:12AM -0800, Roland Dreier wrote:
Could I get some clarification on where there is no ordering
guarantee? The WC's do not necessarily come back in the order that the
sends were posted?
For a given queue, completions are always returned in the order that
On Wed, Nov 10, 2010 at 02:39:03PM +, Jonathan Rosser wrote:
Continually posting sends and recvs will get you into trouble, you
will run out of recvs and get RNR's. These days the wisdom for
implementing RDMA is that you should have explicit message flow
OK - I appreciate that a real
On one of our test clusters, I'm seeing a situation where multicast traffic is
not received on the QP which sends a message. All nodes that I actually tested
on the cluster have this problem (5 of 15). The devinfo from one of these is
below:
hca_id: mlx4_0
transport:
The cluster is running OFED [...] I do not see this problem on 2.6.36
I think you hit this ofed only patch
http://git.openfabrics.org/git?p=ofed_1_5/linux-
2.6.git;a=blob;f=kernel_patches/fixes/mlx4_0290_mcast_loopback.patch;h=786a
3926529befac2c2d1fa6d8c36bada79d61a7;hb=HEAD
Yes - this
Structure ib_uverbs_qp_attr is copied to userland with allmost all
fields uninitialized (140 bytes on x86). It leads to leaking of
contents of kernel stack memory.
I don't think most of the fields are uninitialized... we have:
memset(qp_attr, 0, sizeof qp_attr);
and then later
On Wed, Nov 10, 2010 at 10:14:16AM -0700, i...@celticblues.com wrote:
It seems there are several ways to transfer data via RDMA: Send/Send
w/immediate, Receive, RDMA Write/RDMA Write w/immediate, the Atomic
extensions, and then the various transport modes variations with
some of these
Sean, what is intended for qp_state handling here? It seems
ib_copy_qp_attr_to_user() should either clear it or set it to something
sensible.
I'm not sure what the original intent was, but both libibcm and librdmacm
provide the qp_state as input to the init_qp_attr calls. It doesn't end up
15 matches
Mail list logo