Ralph Campbell wrote:
I guess what I'm objecting to is hard coding mlx4. I was trying to think of a
way that would allow other HCAs to support the block loopback option in the
future. It looks like ipoib sets IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK for
kernel QPs but this isn't defined in
Hefty, Sean wrote:
One could argue that this change is reasonable regardless of the OFED kernel
patch. It avoids sending multicast traffic when the destination is local. The
main drawback beyond the extra code is that a node can't send a multicast
message to itself, with the intent that
On Tue, Nov 16, 2010 at 02:43:46PM -0800, Roland Dreier wrote:
Where is the change in handling in the members_count field? In the firmware?
The firmware now distinguishes between differnet multicast addresses
according to a protocol field located in the high MS byte which
previously was
Hi Ido,
We came into a situation where running rdma_lat with vs with out the -c flag,
which means w. or w.o using the rdma-cm introduces a notable ~1us difference in
latency for 1k messages, that is ~3us w.o using rdma-cm and 3.9us when using
the rdma-cm.
I have reproduced that now with the
Or Gerlitz wrote:
Now, Jack, using this patch,
I could get rdma_lat which doesn't use the rdma-cm, which means setting all
the low-level QP params
by the hand to produce the SAME result of 3.9us as with the rdma-cm, as you
can see its one liner
patch which uses higher MTU of 2048 vs the
We must update the control segment before marking it as valid.
Signed-off-by: Eli Cohen e...@mellanox.co.il
---
drivers/infiniband/hw/mlx4/qp.c |6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index
To be precise, the bit avoids recieving multicast packets by the QP that
--sent-- it, not by other QPs subscribed to that group on the same
node/hca, the patch change-log even states that Inter QP multicast
packets on the relevant HCA will still be delivered. You can test that
with two mckey
I guess what I'm objecting to is hard coding mlx4.
I was trying to think of a way that would allow other HCAs
to support the block loopback option in the future.
It looks like ipoib sets IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK
for kernel QPs but this isn't defined in libibverbs yet.
If there's
Where is the change in handling in the members_count field? In the
firmware?
The firmware now distinguishes between differnet multicast addresses
according to a protocol field located in the high MS byte which
previously was reserved. The patches that Alekseys will will probably
OFED 1.5.2 introduced a kernel patch that disables looping
back multicast messages to the sending QP. This capability
was enabled by default. The current ACM address resolution
protocol depends on this feature of multicast. To handle
the case where multicast loopback has been disabled, add an
I was under the impression that a different module is appropriate here
based on the structure of the IB modules and similar modules like
netfilter_netlink, and other architectures in the kernel. However, I'm
naturally open to other views. If the other way around is the consensus
On Wed, Nov 17, 2010 at 12:43 PM, Steve Wise
sw...@opengridcomputing.com wrote:
I think NFSRDMA server will close the connection after 5 minutes of
inactivity...
Thanks Steve. I mistook the messages as being a problem that may have
been related to nfs speed issues. I think it is something else
Nov 15 09:39:00 node4 kernel: rpcrdma: connection to
192.168.0.100:20049 on mlx4_0, memreg 5 slots 32 ird 16
and then 5 minutes later:
Nov 15 09:44:00 node4 kernel: rpcrdma: connection to
192.168.0.100:20049 closed (-103)
I think NFSRDMA server will close the
I have what is probably a silly question
If I compile the rdma_bw example from perftests with g++, it doesn't
work... granted I have to make a few changes wrt structure
initialization, but I would think it should behave as when built with
gcc... I am getting an error message in
On 11/16/10 1:39 PM, Or Gerlitz wrote:
Tom Tuckert...@ogc.us wrote:
This patch changes the bus mapping logic to avoid page_address() where necessary
Hi Tom,
Does when necessary comes to say that invocations of page_address
which remained in the code after this patch was applied are safe
On Wed, Nov 17, 2010 at 09:38:47AM -0800, Roland Dreier wrote:
Where is the change in handling in the members_count field? In the
firmware?
The firmware now distinguishes between differnet multicast addresses
according to a protocol field located in the high MS byte which
On Wed, Nov 17, 2010 at 01:34:29PM -0800, Roland Dreier wrote:
In fact I'm not sure even for embedded it's worth making this something
that can be configured out. We probably have too many config options
that no one ever uses as it is. Are there any embedded systems that are
both so small
17 matches
Mail list logo