On Tuesday 09 May 2006 19:25, Sean Hefty wrote:
>
> This is fine for route lookup, but not rdma_get_option.
>
I missed adding rdma_ib.c to my tag file, so missed the rdma_get_option usage
of local_sa.c -- sorry about that.
My preference is to take the most recent rdma_cm, since there are several
Quoting r. Michael S. Tsirkin <[EMAIL PROTECTED]>:
> Subject: Re: [PATCH] RE: compliancy issue?
>
> Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> > Subject: [PATCH] RE: compliancy issue?
> >
> > >CA4-24.2.3: The connecting peer shall terminate the connection attempt
> > >if ExtMaxAdverts of the HA
Quoting r. Shirley Ma <[EMAIL PROTECTED]>:
> Subject: Re: [openib-general] Re: [PATCH 07/16] ehca: interrupt
> handling?routines
>
>
> "Michael S. Tsirkin" <[EMAIL PROTECTED]> wrote on 05/09/2006 01:20:41 PM:
>
> > Quoting r. Shirley Ma <[EMAIL PROTECTED]>:
> > > My understanding is NAPI handle
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: [PATCH] RE: compliancy issue?
>
> >CA4-24.2.3: The connecting peer shall terminate the connection attempt
> >if ExtMaxAdverts of the HAH is set to zero.
> >
> >This means that SDP must examine the HAH before RTU is sent.
> >But, CMA currently s
On Mon, 2006-05-08 at 16:58 +0300, Michael S. Tsirkin wrote:
> Two nits wrt rdma_cm.h:
>
> /**
> * * rdma_reject - Called on the passive side to reject a connection request.
> */
>
>
> Its OK to call rdma_reject on active side as well, isn't it?
You'll get -EINVAL on iWARP if you do this
"Talpey, Thomas" <[EMAIL PROTECTED]>
wrote on 05/09/2006 06:58:09 PM:
> Of course, the NFS/RDMA copies do not need to compute the checksum,
> so they are more efficient than the socket atop IPoIB. But I am not
> sure that the payload per WQE is important. We are nowhere near the
> op rate of the
http://openib.org/bugzilla/show_bug.cgi?id=68
[EMAIL PROTECTED] changed:
What|Removed |Added
Severity|normal |critical
--- Additional Comments Fro
At 05:47 PM 5/9/2006, Shirley Ma wrote:
>Thanks for sharing these test results.
>
>The netperf/netserver IPoIB over UD mode test spent most of time on copying
>data from user to kernel + checksum(csum_partial_copy_generic), and it only
>can send no more than mtu=2044 ib_post_send() per wiki, wh
There is no need to limit the MTU to 1K on Pathscale HCAs.
This patch fixes the problem.
Signed-off-by: Ralph Campbell <[EMAIL PROTECTED]>
Index: src/userspace/mpi/mvapich-gen2/mpid/ch_gen2/ibverbs_const.h
===
--- src/userspace/mpi/m
http://openib.org/bugzilla/show_bug.cgi?id=78
Summary: OFED 1.0 RC 4 iser install fails if patches already
applied
Product: OpenFabrics Linux
Version: gen2
Platform: All
OS/Version: All
Status: NEW
Severit
Heiko> Yes, I agree. It would not be an optimal solution, because
Heiko> other upper level protocols (e.g. SDP, SRP, etc.) or
Heiko> userspace verbs would not be affected by this
Heiko> changes. Nevertheless, how can an improved "scaling" or
Heiko> "SMP" version of IPoIB look l
Ira wrote,
># odev1 /root > simple_rdma -S
>librdmacm: kernel ABI version 0 doesn't match library version 1.
>Failed to create rdma_cm_id
Sean is correct, the backport puts the abi_version in a different
place to be consistent with how the other kernel modules do it.
Try appying this patch to your
Sean,
Thanks that helped. Changing the /sys/class/misc/rdma_cm/abi_version to
/sys/class/infiniband_ucma/abi_version in the user lib fixed the problem.
I wonder though if this is the correct solution?
Ira
On Tue, 09 May 2006 15:39:39 -0700
Sean Hefty <[EMAIL PROTECTED]> wrote:
> Ira Weiny wr
Ira Weiny wrote:
I have been struggling with getting svn6829 to work and this is one of the
latest issues.
# odev1 /root > simple_rdma -S
librdmacm: kernel ABI version 0 doesn't match library version 1.
Failed to create rdma_cm_id
This is a little rdma app I wrote which uses the rdma_cm userspa
I have been struggling with getting svn6829 to work and this is one of the
latest issues.
# odev1 /root > simple_rdma -S
librdmacm: kernel ABI version 0 doesn't match library version 1.
Failed to create rdma_cm_id
This is a little rdma app I wrote which uses the rdma_cm userspace lib.
The code w
>CA4-24.2.3: The connecting peer shall terminate the connection attempt
>if ExtMaxAdverts of the HAH is set to zero.
>
>This means that SDP must examine the HAH before RTU is sent.
>But, CMA currently sends RTU from cma_rep_recv, before notifying
>the user.
Can you try this simple patch and see if
Thanks for sharing these test results.
The netperf/netserver IPoIB over UD
mode test spent most of time on copying data from user to kernel + checksum(csum_partial_copy_generic),
and it only can send no more than mtu=2044 ib_post_send() per wiki, which
definitely limits its performance compared
Michael S. Tsirkin wrote:
Until the underlying IB stack supports loopback connections on a non-active
port
How do you mean? You can already create loopback connections as per IB spec - it
works already.
It works if the port is active. I don't believe that there's any code to
support connect
"Michael S. Tsirkin" <[EMAIL PROTECTED]>
wrote on 05/09/2006 01:20:41 PM:
> Quoting r. Shirley Ma <[EMAIL PROTECTED]>:
> > My understanding is NAPI handle interrutps CQ callbacks on the
same CPU.
>
> My understanding is NAPI disables interrupts under high RX load. No?
>
> --
> MST
Yes, NAPI d
Shirley, Hassan - I am *very* interested in these results, and I
want to at least mention that I'm doing similar NFS/RDMA testing,
and getting some contrasting results.
> 699040 699040 1638460.003668.07 (458MB/s)
> cpu utilization was around 95%.
On my dual-2.4GHz Xeon, with the relati
On Tue, May 09, 2006 at 12:41:30PM -0700, Shirley Ma wrote:
> Grant Grundler <[EMAIL PROTECTED]> wrote on 05/09/2006 12:05:38 PM:
>
> > On Tue, May 09, 2006 at 11:36:45AM -0700, Shirley Ma wrote:
> > > What throughput did you get on two CPUs?
> >
> > With one CPU, I get ~2.5-2.8 Gb/s. With two CP
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> my thinking is to have the RDMA CM select the first active port when
> connecting in loopback.
OK. The way to do this I guess is to scan all devices along the lines of
what xxx_ip_dev_find does, and check which one is up.
--
MST
_
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: RE: [PATCH] cm refcount race fix
>
> Here's a patch for all of the files that you listed.
>
> I did do some basic testing and didn't see any issues.
Looks fine, changes look trivial which is good. I'll test a bit tomorrow.
--
MST
__
Quoting r. Shirley Ma <[EMAIL PROTECTED]>:
> My understanding is NAPI handle interrutps CQ callbacks on the same CPU.
My understanding is NAPI disables interrupts under high RX load. No?
--
MST
___
openib-general mailing list
openib-general@openib.org
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Until the underlying IB stack supports loopback connections on a non-active
> port
How do you mean? You can already create loopback connections as per IB spec - it
works already.
--
MST
___
openib-general m
[EMAIL PROTECTED] wrote on 05/09/2006
11:57:01 AM:
> On 09.05.2006, at 18:49, Michael S. Tsirkin wrote:
>
> >> The trivial way to do it would be to use the same idea as
the current
> >> ehca driver: just create a thread for receive CQ events and
a thread
> >> for send CQ events, and defer CQ pol
Grant Grundler <[EMAIL PROTECTED]> wrote on 05/09/2006
12:05:38 PM:
> On Tue, May 09, 2006 at 11:36:45AM -0700, Shirley Ma wrote:
> > What throughput did you get on two CPUs?
>
> With one CPU, I get ~2.5-2.8 Gb/s. With two CPUs: 3.5-3.6 Gb/s.
> The last SVN version I tested was 2.6.15 + r4929 (s
Michael S. Tsirkin wrote:
I thought about this too. People actually do expect loopback to work when link
is down. I guess we could create "loopback" path record, with parameters such
as SL editable from sysfs.
Until the underlying IB stack supports loopback connections on a non-active
port, m
I finally looked this over.
First, this should be two patches: making srp_sg_tablesize tunable
should be a separate change from making it possible to specify
max_cmd_per_lun for a target.
The srp_sg_tablesize change makes the default number of SG entries
quite a bit larger than it is now, which m
Here's a patch for all of the files that you listed.
I did do some basic testing and didn't see any issues.
Signed-off-by: Sean Hefty <[EMAIL PROTECTED]>
---
Index: mad_rmpp.c
===
--- mad_rmpp.c (revision 6884)
+++ mad_rmpp.c (work
BTW, does Mellanox (or anyone else) have any numbers showing that
using FMRs makes any difference in performance on a semi-realistic benchmark?
- R.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-gene
Tziporet> Roland, What is the status of the FMR patch? When do
Tziporet> you expect it to be on the trunk?
It seems good. I should check it into svn today. I still need to
figure out how I want to put it on the for-2.6.18 git branch, since it
depends on some earlier fixes that Linus has
On Tue, May 09, 2006 at 11:36:45AM -0700, Shirley Ma wrote:
> What throughput did you get on two CPUs?
With one CPU, I get ~2.5-2.8 Gb/s. With two CPUs: 3.5-3.6 Gb/s.
The last SVN version I tested was 2.6.15 + r4929 (several monthes ago).
To be clear, "one CPU" means the netperf process is bound
Roland,
What is the status of the FMR patch?
When do you expect it to be on the trunk?
Thanks
Tziporet
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Vu Pham
Sent: Monday, May 08, 2006 10:12 PM
To: Roland Dreier
Cc: openib-general@openib.org
Subject: Re:
On 09.05.2006, at 18:49, Michael S. Tsirkin wrote:
The trivial way to do it would be to use the same idea as the current
ehca driver: just create a thread for receive CQ events and a thread
for send CQ events, and defer CQ polling into those two threads.
For RX, isn't this basically what NAPI
Quoting r. Shirley Ma <[EMAIL PROTECTED]>:
> No, CPU utilization wasn't reduced. When you use single CQ, NAPI polls on
> both RX/TX.
I think NAPI's point is to reduce the interrupt rate.
Wouldn't this reduce CPU load?
> netperf, iperf, mpstat, netpipe, oprofiling, what's your suggestion?
netper
"Michael S. Tsirkin" <[EMAIL PROTECTED]>
wrote on 05/09/2006 11:44:52 AM:
> Quoting r. Shirley Ma <[EMAIL PROTECTED]>:
> > According to some results from different resouces, NAPI only
gives
> 3%-10% performance improvement on single CQ.
>
> When you say performance you mean bandwidth.
> But I th
Quoting r. Shirley Ma <[EMAIL PROTECTED]>:
> According to some results from different resouces, NAPI only gives 3%-10%
> performance improvement on single CQ.
When you say performance you mean bandwidth.
But I think it should improve the CPU utilization on RX side significantly.
If it does, that
Roland Dreier <[EMAIL PROTECTED]> wrote on 05/09/2006
11:36:07 AM:
> But there's no reason why the two threads can't be pinned to different
> CPUs or given exclusive CPU masks, exactly the same way that ehca
> implements it.
>
> - R.
I could try this. Let's see how much latency increase
there.
Quoting r. [EMAIL PROTECTED] <[EMAIL PROTECTED]>:
> Subject: Re: NOP problem in ib_mthca on OFED RC4
>
>
> Michael,
>
> > > > Which FW revision do you have?
> > > >
> > > The "ibstat" command shows:
> > >
> > > CA type: MT25204
> > > Number of ports: 1
> > > Firmware vers
Shirley> I have done some patch like that on top of splitting
Shirley> CQ. The problem I found that hardware interrupt favors
Shirley> one CPU. Most of the time these two threads are running
Shirley> on the same cpu according to my debug output. You can
Shirley> easily find out
Michael,
> > > Which FW revision do you have?
> > >
> > The "ibstat" command shows:
> >
> > CA type: MT25204
> > Number of ports: 1
> > Firmware version: 1.0.800
> > Hardware version: a0
> > Node GUID: 0x0002c90200216dc4
> > System image GUID: 0x00
Grant,
> You might also try the netperf "-T"
option to bind the netserver
> process to a different CPU. If two CPUs are available, IPoIB
throughput
> is better with the interrupts handled on a different CPU than the
one
> handling the data.
>
> grant
What throughput did you get on two CPUs?
T
On Mon, May 08, 2006 at 05:52:17PM -0500, Hassan M. Jafri wrote:
> I cant crank out more than 150 MB/sec with my 2.0 GHz xeons. verbs level
> benchmarks, however give decent numbers for bandwidth. With netperf, the
> server side CPU usage is 99% which is much higher than other posted
> bandwidth
[EMAIL PROTECTED] wrote on 05/09/2006
09:49:19 AM:
> Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> > The trivial way to do it would be to use the same idea as the
current
> > ehca driver: just create a thread for receive CQ events and a
thread
> > for send CQ events, and defer CQ polling into t
Sorry, subject should be "opensm: low-level QoS implementation (take 2)"
On 21:00 Tue 09 May , Sasha Khapyorsky wrote:
> Hello,
>
> There is "take 2" of support for low level Quality of Service (QoS)
> parameters configuration and setup in OpenSM.
>
> The changes against previous version are
Dotan Barak wrote:
Compilation of the gen2 driver generate compilation warnings when compiling
the libibcm.
I've committed changes that should eliminate these warnings. I've also updated
librdmacm to replace deprecated function/structure names with the latest names.
Thanks.
- Sean
__
>> Sync up with Sean on commits.
>
>I'm watching for Sean's commit. Did I miss it?
It looks like my commit failed for some reason, and I missed it. I've just
re-committed the changes, which should be in revision 7019.
- Sean
___
openib-general mailing
This new option '--no_qos' (or '-O') will disable QoS setup globally in
OpenSM.
Signed-off-by: Sasha Khapyorsky <[EMAIL PROTECTED]>
---
osm/include/opensm/osm_subnet.h |1 +
osm/opensm/main.c | 10 +-
osm/opensm/osm_qos.c|3 +++
osm/opensm/osm_subnet.
Basic low-level QoS implementation. The main procedure (osm_qos_setup())
will be called from resweeper (after configuration refreshing). And
then this will setup low level QoS related ports' attributes
(PortInfo:VLHighLimit, VL*Arbitration and SL2VLMapping tables).
Different port categories (HCA,
Trivial low-level QoS configuration parameters description, definition
and processing.
Signed-off-by: Sasha Khapyorsky <[EMAIL PROTECTED]>
---
osm/doc/qos-config.txt | 44 +
osm/include/opensm/osm_subnet.h | 81
osm/opensm/osm_subnet.c |
Hello,
There is "take 2" of support for low level Quality of Service (QoS)
parameters configuration and setup in OpenSM.
The changes against previous version are:
- consistent default values for QoS configuration parameters (as
suggested by Eitan Zahavi)
- global '--no_qos' option which disabl
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
The changes and patch are:
Ralph Campbell:
IB
osmtest: Add rudimentary SA MultiPathRecord tests
Signed-off-by: Hal Rosenstock <[EMAIL PROTECTED]>
Index: osmtest/osmtest.c
===
--- osmtest/osmtest.c (revision 7007)
+++ osmtest/osmtest.c (working copy)
@@ -1,4 +1,5 @@
/*
+ * Co
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: RE: [PATCH] cm refcount race fix
>
> Here's a patch that should fix both the IB CM and RDMA CM using
> completions rather than spinlock / wait objects.
>
> Michael, can you test that this version works for you?
Looks sane, I'll test tomorrow.
Here's a patch that should fix both the IB CM and RDMA CM using
completions rather than spinlock / wait objects.
Michael, can you test that this version works for you?
Signed-off-by: Sean Hefty <[EMAIL PROTECTED]>
---
Index: cm.c
===
OpenSM: Add SA client API for MultiPathRecord queries
Signed-off-by: Hal Rosenstock <[EMAIL PROTECTED]>
Index: include/vendor/osm_vendor_sa_api.h
===
--- include/vendor/osm_vendor_sa_api.h (revision 7007)
+++ include/vendor/osm_vend
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> The trivial way to do it would be to use the same idea as the current
> ehca driver: just create a thread for receive CQ events and a thread
> for send CQ events, and defer CQ polling into those two threads.
For RX, isn't this basically what NAPI is
On Tue, 2006-05-09 at 12:29, Roland Dreier wrote:
> Leonid> we need a possibility to reset the port counters in
> Leonid> /sys/class/infiniband/mthca0/ports/1/counters/.
>
> Why do you need this possibility? Having counters reset locally is
> going to confuse any performance manager
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: Re: CMA: compliancy issue?
>
> Michael S. Tsirkin wrote:
> >>>From iSER point of view, this approach is fine, and it would allow for
> >>some future flexibility to reject the REP. We prefer to implement it
> >>only for 2.6.19, that is when 2.6.
Leonid> we need a possibility to reset the port counters in
Leonid> /sys/class/infiniband/mthca0/ports/1/counters/.
Why do you need this possibility? Having counters reset locally is
going to confuse any performance manager running on the fabric that
might be reading the counters rem
Hi Sasha
Thanks for clearing the issues.
I'm OK with the RFC.
Eitan Zahavi
Design Technology Director
Mellanox Technologies LTD
Tel:+972-4-9097208
Fax:+972-4-9593245
P.O. Box 586 Yokneam 20692 ISRAEL
> -Original Message-
> From: Sasha Khapyorsky [mailto:[EMAIL PROTECTED]
> Sent: Tuesday
Michael S. Tsirkin wrote:
From iSER point of view, this approach is fine, and it would allow for
some future flexibility to reject the REP. We prefer to implement it
only for 2.6.19, that is when 2.6.18-rc1 is out.
Let us start by implementing this in SVN trunk. Sean, if you agree too, can yo
Jack Morgenstein wrote:
Use of local_sa in the rdma_cm kernel module is already patched out for OFED.
local_sa is used ONLY in kernel cma.c ( static function
cma_resolve_ib_route(), which calls ib_get_path_rec()).
Icall to ib_get_path_rec() is eliminated, and we call cma_query_ib_route()
direct
Heiko> Yes, I agree. It would not be an optimal solution, because
Heiko> other upper level protocols (e.g. SDP, SRP, etc.) or
Heiko> userspace verbs would not be affected by this
Heiko> changes. Nevertheless, how can an improved "scaling" or
Heiko> "SMP" version of IPoIB look li
On 18:04 Tue 09 May , Eitan Zahavi wrote:
>
> > > [EZ] Please note that algorithm to validate the applicability of the
> > > above on the
> > > particular fabric is still required as not all devices
> support
> > > the 16 VLs
> >
> > VL numbers are translated according to port's capa
Quoting r. [EMAIL PROTECTED] <[EMAIL PROTECTED]>:
> Subject: Re: NOP problem in ib_mthca on OFED RC4
>
>
> Michael,
>
> "Michael S. Tsirkin" <[EMAIL PROTECTED]> wrote on 05/08/2006 10:32:01 PM:
>
> > Quoting r. [EMAIL PROTECTED] <[EMAIL PROTECTED]>:
> > > When I had the problem previously, Rola
Michael,
"Michael S. Tsirkin" <[EMAIL PROTECTED]>
wrote on 05/08/2006 10:32:01 PM:
> Quoting r. [EMAIL PROTECTED] <[EMAIL PROTECTED]>:
> > When I had the problem previously, Roland Drier suggested trying
> to load the ib_mthca module with "fw_cmd_doorbell=0", which
did
> avoid the error then,
Hi Sasha,
I have cut and paste only the sections I would like to further comment
on.
Eitan
> > [EZ] Please note that algorithm to validate the applicability of the
> > above on the
> > particular fabric is still required as not all devices
support
> > the 16 VLs
>
> VL numbers are tra
Quoting r. keshetti mahesh <[EMAIL PROTECTED]>:
> Subject: problem with applications like telnet over SDP
>
> hello
>
> can anybody tell me ..do i need to configure anything
> to make applications like TELNET,FTP,...work over SDP
>
> The normal socket applications are working fine over
> SDP wit
Hi Eitan,
On 14:20 Tue 09 May , Eitan Zahavi wrote:
>
> It is great that you work on QoS implementation. In general I see this
> simple extension of the SM capabilities as very useful one. But I think
> it would have been better if you first send out the RFC for the proposed
> functionality a
On Sun, 2006-05-07 at 10:07 -0700, Ranjit Pandit wrote:
> Please mark me as the default owner of RDS bugs.
As it turns out, I "fired and forgot", and didn't check the result of
creating the component. It failed because you don't have a Bugzilla
account. Please create one.
http://openib
hello
can anybody tell me ..do i need to configure anything
to make applications like TELNET,FTP,...work over SDP
The normal socket applications are working fine over
SDP with the setup i am using.
i have configured /etc/libsdp.conf like below
>match listen *:*
>match destination *:*
>match pro
Driver Version: openib_gen2-20060509-0800 (REV=7002)
Here are the compilation warnings:
make[2]: Entering directory `/tmp/openib_gen2/last_stable/src/userspace/libibcm'
if /bin/sh ./libtool --tag=CC --mode=compile gcc -DHAVE_CONFIG_H -I. -I. -I.
-I./include -I../libibverbs/include -g
make[2]: Entering directory `/tmp/openib_gen2/last_stable/src/userspace/libibcm'
if /bin/sh ./libtool --tag=CC --mode=compile gcc -DHAVE_CONFIG_H -I. -I. -I.
-I./include -I../libibverbs/include -g -Wall -D_GNU_SOURCE -g
-O2 -MT cm.lo -MD -MP -MF ".deps/cm.Tpo" -c -o cm.lo `test -f 'src/cm.c' ||
I think the capability to reset counters is useful.
Some comments on the patch:
Quoting r. Leonid Arsh <[EMAIL PROTECTED]>:
> Subject: [openib-general][RFC][PATCH] core/sysfs.c: ability to reset port
> counters
>
>
> Hello,
>
> we need a possibility to reset the port counters in
> /sys/c
> +#include
> +#include
> +#include
Please don't use directly ever. Always include
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/
Roland Dreier wrote:
Heiko> Originaly, we had the same idea as you mentioned, that it
Heiko> would be better to do this in the higher levels. The point
Heiko> is that we can't see so far any simple posibility how this
Heiko> can done in the OpenIB stack, the TCP/IP network layer o
Quoting r. zhu shi song <[EMAIL PROTECTED]>:
> Subject: Re: sdp can't support many connections (>2000)
>
> ab send the request to squid cache server running on
> Machine B. Then squid send the real request to google
> website.
> So how can I upgrade my version to solve the
> problem?
>
> zh
Hello,
we need a possibility to reset the port counters in
/sys/class/infiniband/mthca0/ports/1/counters/.
The attached patch implements the ability to reset the counters by writing
to the sysfs counter files.
The patch uses the same process_mad() mechanism as in the counter re
ab send the request to squid cache server running on
Machine B. Then squid send the real request to google
website.
So how can I upgrade my version to solve the
problem?
zhu
--- "Michael S. Tsirkin" <[EMAIL PROTECTED]> wrote:
> Quoting r. zhu shi song <[EMAIL PROTECTED]>:
> > Subject: sdp
Quoting r. zhu shi song <[EMAIL PROTECTED]>:
> Subject: sdp can't support many connections (>2000)
>
>after I have upgraded firmware to MT25204 1.0.8
> from 1.0.1, sdp doesn't cause the kernel panic again.
> But after executing LD_PRELOAD=libsdp.so ab -c 1000 -n
> 1000 -X 193.12.10.14:3129 ht
Hi Sasha, Hal,
It is great that you work on QoS implementation. In general I see this
simple extension of the SM capabilities as very useful one. But I think
it would have been better if you first send out the RFC for the proposed
functionality and only later implement it (as was done on the parti
after I have upgraded firmware to MT25204 1.0.8
from 1.0.1, sdp doesn't cause the kernel panic again.
But after executing LD_PRELOAD=libsdp.so ab -c 1000 -n
1000 -X 193.12.10.14:3129 http://www.google.com about
three times continually, I can't connect Machine B
from Machine A. I must reboot th
On Tuesday 09 May 2006 02:46, you wrote:
> The end-users should probably decide that. These changes use the local_sa,
> which has not been queued to be merged upstream yet.
Use of local_sa in the rdma_cm kernel module is already patched out for OFED.
local_sa is used ONLY in kernel cma.c ( static
Quoting r. Or Gerlitz <[EMAIL PROTECTED]>:
> >From iSER point of view, this approach is fine, and it would allow for
> some future flexibility to reject the REP. We prefer to implement it
> only for 2.6.19, that is when 2.6.18-rc1 is out.
Let us start by implementing this in SVN trunk. Sean, if yo
OS: Redhat AMD64/EMT64 64-bit Enterprise Linux v3
update 4
IB: IBG2.0.1 downloaded from mellanox website
HCA: MT 25204
APP: squid-2.5-stable13, ab(apache benchmark
distributed with httpd 2.2.0)
I use two HCA cards to connect two computers directly.
on Machine B run command:
LD_PRELOAD=libsdp.so sq
OS: Redhat AMD64/EMT64 64-bit Enterprise Linux v3
update 4
IB: IBG2.0.1 downloaded from mellanox website
HCA: MT 25204
APP: squid-2.5-stable13, ab(apache benchmark
distributed with httpd 2.2.0)
I use two HCA cards to connect two computers directly.
on Machine B run command:
LD_PRELOAD=libsdp.so sq
88 matches
Mail list logo