Hi,
You can't connect to localhost/127.0.0.1/0.0.0.0/lo interface with SDP.
These ip's are not of IB interfaces so SDP (actually CM) is having
trouble connecting to it.
If you would have used libsdp instead of changing your source code - you
could configure it to use TCP to these addresses
Hi,
I noticed this mail after the private mail...
As I said before - you should run epmd using LD_PRELOAD and make sure that in
libsdp.conf calls to 127.0.0.1 are done using TCP and not SDP.
--
Amir Vadai
Software Eng.
Mellanox Technologies
mailto: am...@mellanox.co.il
Tel +972-3-6259539
are changed those days.
Bye,
Amir.
--
Amir Vadai
Software Eng.
Mellanox Technologies
mailto: am...@mellanox.co.il
Tel +972-3-6259539
On 07/17/2009 06:25 PM, Lars Ellenberg wrote:
Lets say my typical usage pattern is:
two nodes, alice and bob.
two tcp connections via SDP, long lifetime
-
Van: general-boun...@lists.openfabrics.org
[mailto:general-boun...@lists.openfabrics.org] Namens Amir Vadai
Verzonden: woensdag 15 juli 2009 12:28
Aan: general@lists.openfabrics.org
CC: Amir Vadai
Onderwerp: [ofa-general] [PATCH] sdp: fix some warning and bugs in porting
to ofed 1.5
Signed
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp.h |9 +++--
drivers/infiniband/ulp/sdp/sdp_cma.c |2 +-
drivers/infiniband/ulp/sdp/sdp_main.c | 17 -
drivers/infiniband/ulp/sdp/sdp_rx.c | 14 --
4 files changed
-boun...@lists.openfabrics.org
[mailto:general-boun...@lists.openfabrics.org] Namens Amir Vadai
Verzonden: woensdag 15 juli 2009 12:28
Aan: general@lists.openfabrics.org
CC: Amir Vadai
Onderwerp: [ofa-general] [PATCH] sdp: fix some warning and bugs in porting
to ofed 1.5
Signed-off-by: Amir
Hi,
I will post a fix soon.
--
Amir Vadai
Software Eng.
Mellanox Technologies
mailto: am...@mellanox.co.il
Tel +972-3-6259539
On 07/14/2009 03:01 PM, Jason Gunthorpe wrote:
Hey all,
I'm trying to use SDP with stock 2.6.30.1 plus the
'drivers/infiniband/ulp/sdp' directory from
ofa_kernel
sizes do you use? what is the setup in general?
Let me know if you find something interesting.
- Amir
On 07/14/2009 05:31 PM, Jason Gunthorpe wrote:
On Tue, Jul 14, 2009 at 04:14:59PM +0300, Amir Vadai wrote:
Hi,
I will post a fix soon.
Thanks Amir!
BTW - we are testing SDP here
this helps,
Maarten
-Oorspronkelijk bericht-
Van: Amir Vadai [mailto:am...@mellanox.co.il]
Verzonden: dinsdag 14 juli 2009 16:37
Aan: Jason Gunthorpe
CC: general-list
Onderwerp: Re: [ofa-general] SDP and stock kernel gets BUG?
OFED-1.5 SDP has many changes in the data path
Lars Hi,
I'v found the bug.
a fix patch is attached.
I will commit it after the weekend (In Israel weekend is Friday-Saturday).
- Amir
On 07/08/2009 01:17 PM, Lars Ellenberg wrote:
On Wed, Jul 08, 2009 at 11:12:15AM +0300, Amir Vadai wrote:
Lars Hi,
I opened a bug in our bugzilla
e3 e2
r...@kugel:/home/lars/DRBD/IB_SDP# LD_PRELOAD=libsdp.so perl my_client.pl
rum-ib0
2009-07-07 19:07:14 my_client.pl 4306: recv payload [12]: expected 22530, but
received 22529 byte; last bytes received: 55 e4 e3 e2
any suggestions how to proceed from here?
--
Amir Vadai
see below
On 07/08/2009 01:17 PM, Lars Ellenberg wrote:
On Wed, Jul 08, 2009 at 11:12:15AM +0300, Amir Vadai wrote:
Lars Hi,
I opened a bug in our bugzilla
(https://bugs.openfabrics.org/show_bug.cgi?id=1672).
I couldn't reproduce it on my setup: SLES 10SP2, stock kernel, same ofed git
ok - bug is reproduced now on your kernel+config
will check it now.
On 07/08/2009 01:33 PM, Amir Vadai wrote:
see below
On 07/08/2009 01:17 PM, Lars Ellenberg wrote:
On Wed, Jul 08, 2009 at 11:12:15AM +0300, Amir Vadai wrote:
Lars Hi,
I opened a bug in our bugzilla
(https
list in the CC and my direct mail in the TO...
- Amir
On 07/02/2009 11:14 AM, Lars Ellenberg wrote:
On Wed, Jul 01, 2009 at 07:53:16PM +0300, Amir Vadai wrote:
I understand that you're using ofed-1.5.
We should check if those bugs exist in 1.4.1 too - since there were many
changes in SDP
Hi Lars,
This is the right place for posting patches.
I will commit it ASAP into both branches.
--
Amir Vadai
Software Eng.
Mellanox Technologies
mailto: am...@mellanox.co.il
Tel +972-3-6259539
On 07/01/2009 03:52 PM, Lars Ellenberg wrote:
Hi there.
Please direct me to the right place
Thanks to Lars Ellenberg for finding the leak
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp_main.c |2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/drivers/infiniband/ulp/sdp/sdp_main.c
b/drivers/infiniband/ulp/sdp/sdp_main.c
index
:02:17PM +0300, Amir Vadai wrote:
Subject: Re: [patch] fix SDP page leak in sdp_bz_cleanup
In-Reply-To: 4a4b5e59.2030...@mellanox.co.il
Hi Lars,
This is the right place for posting patches.
I will commit it ASAP into both branches.
Thanks for that one.
now, let me summarize some other
Following is a series of patches to improve SDP small-mid sized packets
bandwidth performance.
Patches were splitted from 1 big patch for readability.
The full patch is at the end of the series.
--
Amir Vadai
Software Eng.
Mellanox Technologies
mailto: am...@mellanox.co.il
Tel +972-3-6259539
* use rcvbuf_initial - fix server to use rcvbuf_initial_size as initial
* hw interrupt moderation
* fix interrupts statistics
* fix max frags bug
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp.h |7 +--
drivers/infiniband/ulp/sdp/sdp_bcopy.c
Send first packet immediately - even when nagle is on
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp.h |5 +++
drivers/infiniband/ulp/sdp/sdp_bcopy.c | 48 ++--
drivers/infiniband/ulp/sdp/sdp_rx.c|6 +++-
drivers
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp_cma.c |7 ---
drivers/infiniband/ulp/sdp/sdp_rx.c |3 +--
drivers/infiniband/ulp/sdp/sdp_tx.c |2 ++
3 files changed, 7 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/ulp/sdp
Make HW RC interrupt moderation adaptive.
For latency sensitive traffic - no moderation.
In BW mode - maximal moderation.
Use heuristics to detect mode.:w
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp.h | 42
drivers/infiniband/ulp/sdp
Queue RX packets in interrupt context - reduce unnecessary context switches.
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp.h | 140 +++
drivers/infiniband/ulp/sdp/sdp_bcopy.c | 66 ++
drivers/infiniband/ulp/sdp/sdp_cma.c | 114
poll tx cq with timer instead of interrupts
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/sdp.h | 34 --
drivers/infiniband/ulp/sdp/sdp_bcopy.c | 222 ---
drivers/infiniband/ulp/sdp/sdp_cma.c | 72 +++
drivers
* statistics infrastructure for sdp
* RX/TX packets dumps
Signed-off-by: Amir Vadai am...@mellanox.co.il
---
drivers/infiniband/ulp/sdp/Makefile|2 +-
drivers/infiniband/ulp/sdp/sdp.h | 108 +-
drivers/infiniband/ulp/sdp/sdp_bcopy.c | 80 +++-
drivers/infiniband/ulp
on the cluster I am working on. Where can I
get test cases (such as pt2pt communications) to carry out these
tests? What approach did you use in your benchmark test that you sent
me (your sent message attached)?
Thanks in advance!
On Thu, Jun 11, 2009 at 11:22 AM, Amir Vadai am...@mellanox.co.il
.
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
--
Amir Vadai
Software Eng
Sean Hi,
I need to know how is the LocalAckTimeout for a new QP is calculated
(for an SDP QP).
Is there a way to change it through a module parameter?
If not, what is the right way to change it?
Thanks,
Amir
___
general mailing list
Hal Hi,
Assuming I am using OpenSM - how can I tell it to do it?
Thanks,
Amir
From: Hal Rosenstock [mailto:hal.rosenst...@gmail.com]
Sent: Thu 21-May-09 4:20 PM
To: Amir Vadai
Cc: Sean Hefty; Nimrod Gindi; OpenIB
Subject: Re: [ofa-general] Default value
; Amir Vadai
Cc: Nimrod Gindi; OpenIB
Subject: RE: [ofa-general] Default value of LocalAckTimeout for a new QP
Sean Hi,
I need to know how is the LocalAckTimeout for a new QP is
calculated (for an
SDP QP).
It's a function of the packet life time on the path from source to
destination
Thanks,
Will try it.
- Amir
From: Hal Rosenstock [mailto:hal.rosenst...@gmail.com]
Sent: Thu 21-May-09 11:01 PM
To: Amir Vadai
Cc: Sean Hefty; Nimrod Gindi; OpenIB
Subject: Re: [ofa-general] Default value of LocalAckTimeout for a new QP
On Thu, May 21
visit http://openib.org/mailman/listinfo/openib-general
--
Amir Vadai
Software Eng.
Mellanox Technologies
mailto: am...@mellanox.co.il
Tel +972-3-6259539
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman
Hi,
I don't know of such a tool - except of course IB analyzer.
I hope to add in the next release statistics in /proc for sdp, and you
could use it than.
- Amir
On 02/28/2009 01:36 AM, Sumeet Lahorani wrote:
Hi,
Is there a tool to observe the SDP throughput while a workload is in
Sorry for the late answer.
If you run iperf as root - sdp could use zero copy which should boost
the performance.
- Amir
Jan Ruffing wrote:
Hello,
I'm new to Infiniband and still trying to get a grasp on what performance it
can realistically deliver.
The two directly connected test
To be more specific - you need to have CAP_IPC_LOCK capability (will
enable sdp to pin pages while they are zero copied).
super use have it - and regular user could be configured to have it.
- Amir
Amir Vadai wrote:
Sorry for the late answer.
If you run iperf as root - sdp could use zero
Both bugs on SDP are fixed (BUG1348, BUG1349) - Currently there are no
major bugs relevant to this release.
- Amir
Tziporet Koren wrote:
Hi All,
I have Just reviewed bugs status with Vlad.
We have 11 major and critical bugs, and we will not be able to fix all
of them in one week
Thus
BUG1279 and BUG1331 are very minor bugs and won't be fixed for the release.
Tziporet Koren wrote:
Hi,
This is the bugs status
Bug owners - please update bugs status (I think I saw some commits so
maybe some of them are already fixed) and see if they are really
critical for the release
to send
between
ia64 hosts:
When doing SDP transfers from an IA64 to any other host (IA64, x86,
x86_64) through ttcp, I got this message:
- Amir
Amir Vadai wrote:
Hi,
Please open a bug in https://bugs.openfabrics.org/ (make sure it is not
a duplicate)
I guess you have some endianess
I opened a bug in bugzilla with your research:
https://bugs.openfabrics.org/show_bug.cgi?id=1311
Nicolas Morey Chaisemartin wrote:
Amir Vadai a écrit :
I asked our IB expert Jack for hints and he told me this:
From Section 11.6.2 (COMPLETION RETURN STATUS0 of the IB Spec volume
1, revision
---
drivers/infiniband/ulp/sdp/sdp_bcopy.c |4 ++--
drivers/infiniband/ulp/sdp/sdp_cma.c |8
drivers/infiniband/ulp/sdp/sdp_main.c | 11 +--
3 files changed, 11 insertions(+), 12 deletions(-)
diff --git a/drivers/infiniband/ulp/sdp/sdp_bcopy.c
Hi,
Please open a bug in https://bugs.openfabrics.org/ (make sure it is not
a duplicate)
I guess you have some endianess problem since ia64 is big endian and x86 is
little endian.
Try running the test on a stock Redhat/SLES kernel.
- Amir
Nicolas Morey Chaisemartin wrote:
Hi,
I am
as it segment size.
This fix the bug here:
https://bugs.openfabrics.org/show_bug.cgi?id=1300
Signed-off-by: David Wilder [EMAIL PROTECTED]
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp.h |1 +
drivers/infiniband/ulp/sdp/sdp_bcopy.c |8
drivers
Fixes bug at:
https://bugs.openfabrics.org/show_bug.cgi?id=1302
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp_bcopy.c |5 +
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/drivers/infiniband/ulp/sdp/sdp_bcopy.c
b/drivers/infiniband/ulp/sdp
Set initial recv buffer according to incoming hha header.
Fixed bugzilla 1086: SDP Linux and SDP windows don't work togeather
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp.h |2 ++
drivers/infiniband/ulp/sdp/sdp_bcopy.c | 15 +++
drivers
wrote:
Hi,
I'm attempting to do this with IO signals - install a signal handler
that
will be called when the connect fails, and it will do the fallback.
--Yossi
Amir Vadai wrote:
Yossi Hi,
I'm on vacation till Monday.
I'll check when can we have the full fix
Yossi Hi,
I'm on vacation till Monday.
I'll check when can we have the full fix - and if it is not in the near future
we'll put your patch till the full fix be prepared.
- Amir
-Original Message-
From: Yossi Etigin [mailto:[EMAIL PROTECTED]
Sent: Mon 8/25/2008 6:18 PM
To: Amir Vadai
Cc
Devesh Hi,
You could use some strategies:
1. check the size column of ib_sdp module when issueing lsmod
2. check in /proc/slabinfo values of slab named 'SDP'
3. check values in /proc/meminfo before and after loading and using ib_sdp
module
- Amir
_
From: [EMAIL PROTECTED]
in error flow, sdp_destroy_work is placed in workqueue and sometimes the
user application call close() that destruct the socket before sdp_destroy_work
is called.
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp_main.c |3 +++
1 files changed, 3 insertions(+), 0
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/core/cma.c |7 ++-
include/rdma/rdma_cm.h|3 ++-
2 files changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index ae11d5c..9fa1371 100644
I sent the patch to the list - and also attached the patch mail here.
Thanks,
- Amir
-Original Message-
From: Sean Hefty [mailto:[EMAIL PROTECTED]
Sent: Monday, July 14, 2008 8:28 PM
To: Amir Vadai
Cc: general@lists.openfabrics.org; Oren Duer; Tziporet Koren
Subject: RE: [ofa
Sean Hi,
Acquiring the handler_mutex explicitly sounds like a good idea.
When will it be commited? Will we have it in 2.6.27?
Thanks,
- Amir
-Original Message-
From: Sean Hefty [mailto:[EMAIL PROTECTED]
Sent: Wednesday, July 02, 2008 9:48 PM
To: Amir Vadai
Cc: general
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp_bcopy.c | 209 ++--
1 files changed, 116 insertions(+), 93 deletions(-)
diff --git a/drivers/infiniband/ulp/sdp/sdp_bcopy.c
b/drivers/infiniband/ulp/sdp/sdp_bcopy.c
index 05825bf..3a8d5ac
Fix all places that change sk_state directly to use sdp_set_state()
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp.h | 12 ++--
drivers/infiniband/ulp/sdp/sdp_bcopy.c |4 ++--
drivers/infiniband/ulp/sdp/sdp_cma.c |6 +++---
drivers
Notify the ULP when a IB_CM_TIMEWAIT_EXIT occurs.
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
.../fixes/cma_0050_timewait_finished_event.patch | 45
1 files changed, 45 insertions(+), 0 deletions(-)
create mode 100644 kernel_patches/fixes
the
socket closing when CM finish.
5. when sdp_post_sends send SDP_MID_DISCONNECT according to socket state
multiple DISCONNECT could be sent - changed it to be triggered by a flag.
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp.h | 20 +++---
drivers/infiniband
Fix all places that change sk_state directly to use sdp_set_state()
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp.h | 12 ++--
drivers/infiniband/ulp/sdp/sdp_bcopy.c |4 ++--
drivers/infiniband/ulp/sdp/sdp_cma.c |6 +++---
drivers
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp_bcopy.c | 209 ++--
1 files changed, 116 insertions(+), 93 deletions(-)
diff --git a/drivers/infiniband/ulp/sdp/sdp_bcopy.c
b/drivers/infiniband/ulp/sdp/sdp_bcopy.c
index 05825bf..3a8d5ac
the
socket closing when CM finish.
5. when sdp_post_sends send SDP_MID_DISCONNECT according to socket state
multiple DISCONNECT could be sent - changed it to be triggered by a flag.
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
---
drivers/infiniband/ulp/sdp/sdp.h | 20 +++---
drivers/infiniband
To: Amir Vadai
Cc: Sean Hefty; Oren Duer; general@lists.openfabrics.org
Subject: Re: [ofa-general] cma: add support for TIMEWAIT_EXIT event (second try)
Amir Vadai wrote:
SDP needs to be notified when CM exits the TIMEWAIT state.
CM does notify the CMA but CMA doesn't pass it to the upper layer
/IB_CM_DREP_RECEIVED as it is done today. Instead
I changed the check at the beginning of cma_ib_handler() as you can
see in the patch.
Signed-off-by: Amir Vadai [EMAIL PROTECTED]
--- include/rdma/rdma_cm.h.orig2008-06-18 15:04:22.560904000 +0300
+++ include/rdma/rdma_cm.h2008-06-11 11:55
I understand - I'll make sure the flow you described will be used.
Thanks a lot,
- Amir.
-Original Message-
From: Sean Hefty [mailto:[EMAIL PROTECTED]
Sent: ה 17 אפריל 2008 21:52
To: Amir Vadai
Cc: general@lists.openfabrics.org; Oren Duer
Subject: RE: CM goes to timewait state without
61 matches
Mail list logo