Er. Ect harder!
And Satisfy!
http://kabirso.cn/a/
___
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
On Tuesday 08 July 2008 16:33, Or Gerlitz wrote:
Assuming this solution (a create-qp-extended verb that gets also
creation flags) would be the one implemented at libibverbs, are you fine
with applying the same approach to librdmacm that is adding a matching
call eg rdma-create-qp-extended
No problem,
expanded it is.
can we do this so that the new verb lies after XRC in ib_context, since OFED
1.3 went out
with XRC in userspace (so we don't break binary compatibility):
struct ibv_context {
struct ibv_device *device;
struct ibv_context_ops ops;
int
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 +++---
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
Main changes:
1. when a close/shutdown syscall is called, instead of sending a DREQ, put
last socket ref count and go to TCP_CLOSE state do:
- take a socket reference count
- set state to TCP_TIME_WAIT
- start infiniband tear down
- wait till got RDMA_CM_EVENT_TIMEWAIT_EXIT
- set socket state to
On Thursday 10 July 2008 10:47, Ron Livne wrote:
I'm going to release some patches, based on your XRC patches, in which
I'm going to replace xrc_ops with more_ops.
These patches will will be relative to your patches.
I hope it's ok with you.
Ron
It is OK with me.
I will make the name
Hi,
I pulled libmlx4 from:
git://git.kernel.org/pub/scm/libs/infiniband/libmlx4.git
I tried building it with:
./autogen.sh
./configure
but the configure script renders the following error:
checking size of long... configure: error: cannot compute sizeof (long),
77
I'm using Redhat 5.1 with
I'm having a bit of a weird problem that i cannot figure out. If anyone can
help from the community it would be appreciated.
Here's the packet flow
cn(ib0)-io(ib0)-io(eth5)-pan(*)
cn = compute node
io = io node
pan = panasas storage network
We have 12 shelves of panasas network storage on a
:
/home/vlad/tmp/ofa_1_4_kernel-20080710-0200_linux-2.6.18_x86_64_check/drivers/infiniband/hw/ipath/ipath_driver.c:2616:
error: implicit declaration of function 'pid_nr'
/home/vlad/tmp/ofa_1_4_kernel-20080710-0200_linux-2.6.18_x86_64_check/drivers/infiniband/hw/ipath/ipath_driver.c:
In function
Roland Dreier wrote:
Would have been nice for the original author of a patch changing locking to
test with lockdep (IB needed the mutex_lock_nested treatment too), but oh well.
Yes, sounds like testing the original patch with lockdep would have
bumped something earlier, thanks for fixing that.
On Thu, Jul 10, 2008 at 04:30:11AM -0400, Michael Di Domenico wrote:
I'm having a bit of a weird problem that i cannot figure out. If anyone can
help from the community it would be appreciated.
Here's the packet flow
cn(ib0)-io(ib0)-io(eth5)-pan(*)
cn = compute node
io = io node
pan =
Woodruff, Robert J wrote:
The question is, do you think we need a developers workshop that week as
well to perhaps plan future work, OFED 1.5, WinOF 2.1 etc. or do you think a
face-to-face meeting is not needed this fall and wait till next spring
at Sonoma for the next face-to-face.
I
Do you want the debug from the IO, Compute, or Both?
On Thu, Jul 10, 2008 at 7:01 AM, Eli Cohen [EMAIL PROTECTED] wrote:
On Thu, Jul 10, 2008 at 04:30:11AM -0400, Michael Di Domenico wrote:
I'm having a bit of a weird problem that i cannot figure out. If anyone
can
help from the community
Turning on this debug gives this error
[EMAIL PROTECTED] ~]# echo 1
/sys/module/ib_ipoib/parameters/macast_debug_level
-bash: /sys/module/ib_ipoib/parameters/macast_debug_level: Permission denied
dmesg output after i start the netperf test (which doesn't complete)
mtnic :02:00.0: Port 2 -
John Marshall wrote:
Hi,
I have muddled my way through building the latest
ofed-1.3.1 on debian etch. Is there a simple way of
doing this that I am missing? I have searched for info
and found some, but nothing that gives a step by
step procedure.
There seem to be a lot of people on this list.
I edited the configure script, canceled the sizeof long test,
and manually entered there that the size is 8.
everything seems to be working fine now.
Ron
On Thu, Jul 10, 2008 at 2:15 PM, Ron Livne [EMAIL PROTECTED] wrote:
Hi,
I pulled libmlx4 from:
On Thursday 10 July 2008 10:47, Ron Livne wrote:
can we do this so that the new verb lies after XRC in ib_context, since OFED
1.3 went out
with XRC in userspace (so we don't break binary compatibility):
struct ibv_context {
struct ibv_device *device;
struct
On Thu, Jul 10, 2008 at 07:57:30AM -0400, Michael Di Domenico wrote:
maybe i spoke too soon, so more output came, i thought it was done
ib0: mtu 2044 will cause multicast packet drops.
eth5: no IPv6 routers present
ib0: Send unicast ARP to 0384
ib0: REQ arrived
ib0: Request connection
On Thu, 2008-07-10 at 00:17 +0300, Sasha Khapyorsky wrote:
On 13:53 Wed 09 Jul , Hal Rosenstock wrote:
It is just *.ver file, *.map is symbol versioning - there were additions,
no changes in existing functions API.
Isn't that needed too when APIs are added ?
I don't think.
Hi Vincent,
On Wed, 2008-07-09 at 15:37 +0200, Vincent Ficet wrote:
Hello,
The function osm_lin_fwd_tbl.h:osm_lin_fwd_tbl_get_lids_per_block() is
coded as follows:
static inline uint16_t
osm_lin_fwd_tbl_get_lids_per_block(IN const osm_lin_fwd_tbl_t * const p_tbl)
{
opensm/osm_lin_fwd_tbl.h: Cosmetic changes
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
diff --git a/opensm/include/opensm/osm_lin_fwd_tbl.h
b/opensm/include/opensm/osm_lin_fwd_tbl.h
index 98c6b4f..be3a3ee 100644
--- a/opensm/include/opensm/osm_lin_fwd_tbl.h
+++
tests using datagram IPoIB (non-connected mode)
dmesg from the compute node
cfd-cnsl-0001:~ # dmesg
ib0: Send unicast ARP to 0518
ib0: Send unicast ARP to 0518
ib0: Send unicast ARP to 0518
ib0: Send unicast ARP to 0518
ib0: Start path record lookup for fe80::::00e0:8111:0100:007d
MTU
opensm/osm_rand_fwd_tbl.h: Cosmetic changes
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
diff --git a/opensm/include/opensm/osm_rand_fwd_tbl.h
b/opensm/include/opensm/osm_rand_fwd_tbl.h
index 8e9ac90..31098b9 100644
--- a/opensm/include/opensm/osm_rand_fwd_tbl.h
+++
Hello all:
I am using IBM HS21XM blades with 4X InfiniBand DDR Expansion Card, on
x86_64 RHEL5.2 Server. opensm-3.1.8-1.el5 is installed. The IB device has
worked for weeks. Today I tried osmtest -f c to generate inventory file.
But it gave error 5501, 0003 and 0064. Would anyone give some
Current code uses kmalloc() and then does a bitwise or operation on
qp-flags at create_qp_common(). This patch uses kzalloc and avoids
farther explicit clearing of variables. This is the size of the text
segment of mlx4_ib.ko:
text
before: 24565
after: 24497
change: 68
Hi,
On Thu, 2008-07-10 at 19:43 +0800, Wen Hao Wang wrote:
Hello all:
I am using IBM HS21XM blades with 4X InfiniBand DDR Expansion Card, on
x86_64 RHEL5.2 Server. opensm-3.1.8-1.el5 is installed. The IB device
has worked for weeks. Today I tried osmtest -f c to generate
inventory file.
On Thursday 10 July 2008 14:07:39 John Marshall wrote:
John Marshall wrote:
Hi,
I have muddled my way through building the latest
ofed-1.3.1 on debian etch. Is there a simple way of
doing this that I am missing? I have searched for info
and found some, but nothing that gives a step by
tests using connected mode IPoIB
dmesg from the compute node
cfd-cnsl-0001:~ # dmesg
ib0: Start path record lookup for fe80::::00e0:8111:0100:0091
MTU 0
ib0: PathRec LID 0x0518 for GID fe80::::00e0:8111:0100:0091
ib0: Created ah 810216a8c740
ib0: created address
OK, thank you.
If you'll send them until Sunday, I'll send mine on Sunday
Ron
On Thu, Jul 10, 2008 at 11:10 AM, Jack Morgenstein
[EMAIL PROTECTED] wrote:
On Thursday 10 July 2008 10:47, Ron Livne wrote:
I'm going to release some patches, based on your XRC patches, in which
I'm going to
On Thu, Jul 10, 2008 at 09:49:54AM -0400, Michael Di Domenico wrote:
tests using datagram IPoIB (non-connected mode)
dmesg from the compute node
cfd-cnsl-0001:~ # dmesg
ib0: Send unicast ARP to 0518
ib0: Send unicast ARP to 0518
ib0: Send unicast ARP to 0518
ib0: Send unicast ARP
On Thu, Jul 10, 2008 at 11:30 AM, Eli Cohen [EMAIL PROTECTED] wrote:
dmesg from the IO node
[EMAIL PROTECTED] ~]# dmesg
ib0: Send unicast ARP to 0384
ib0: Send unicast ARP to 0384
ib_mthca :07:00.0: too many gathers
ib0: post_send failed
ib_mthca :07:00.0: opcode
This is the updated patch series, incorporating all of the fixes and changes
based on Roland's and Ron Livne's feedback.
This series is ready for Ron's proposed changes to be integrated.
This implementation will shortly be put into OFED 1.4.
The OFED 1.3 users will need to recompile when OFED
From e1760ce0a30ff872db3fc066e9cf74b538bb0cb4 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Wed, 9 Jul 2008 13:58:05 +0300
Subject: [PATCH] Added support for XRC receive-only QPs.
(OFED 1.3 commit 3869d6dab7e12fe452270ca641f7dd7082b42482)
V2:
1. xrc_ops changed to
From 2b577ecf8ef5e8e2e73e8bd743a5e3b746aed642 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Mon, 7 Jul 2008 18:20:49 +0300
Subject: [PATCH] IB/core: Implement XRC support at verbs layer (for case in
which fd is not used when opening an xrc_domain).
Rev 2:
a. is_srq flag
From fc8eabcb05965539d485785c94fc3555112a979a Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Mon, 7 Jul 2008 18:33:00 +0300
Subject: [PATCH] ib/core: Add XRC support for userspace file descriptors.
Add XRC support for working with file descriptors, to allow
sharing XRC
From b84af8925c44e3c40ecbbb9104f01f7adccf79be Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Tue, 8 Jul 2008 09:42:47 +0300
Subject: [PATCH] IB/core: implement kernel-space XRC.
V2: no changes
Signed-off-by: Jack Morgenstein [EMAIL PROTECTED]
---
From c734d90a3d28f186d599d12d6d12ff1b3a09fd22 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Mon, 7 Jul 2008 11:47:21 +0300
Subject: [PATCH] mlx4: Implement kernel-space XRC.
V2: no changes.
Signed-off-by: Jack Morgenstein [EMAIL PROTECTED]
---
From 7e94b2921f92165b7c6d2f36cfdda6ffb36799d3 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Mon, 7 Jul 2008 17:47:58 +0300
Subject: [PATCH] mlx4: Implement XRC receive-only QP support.
Support for XRC RCV-only QP (requested by userspace,
but resides in kernel space).
From 02cf3664488c612af01ea39cca1cb776ecd1d2e8 Mon Sep 17 00:00:00 2001
From: Barak [EMAIL PROTECTED]
Date: Wed, 9 Jul 2008 13:06:05 +0300
Subject: [PATCH] Add XRC (eXtended Reliable Connection) support to all of the
relevant man pages
and add new man pages for new XRC verbs.
Signed-off-by: Dotan
From aa4ad2fda6aecdbc0358bac9e28cf49fab886674 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Wed, 9 Jul 2008 12:46:43 +0300
Subject: [PATCH] Consolidated XRC implementation patch (from OFED 1.3
libibverbs).
This patch includes the following commits from OFED 1.3
From c33e87deccc1c59278e259c59ac82e3e548b0981 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Wed, 9 Jul 2008 13:01:31 +0300
Subject: [PATCH] Added support for XRC receive-only QPs.
(OFED 1.3 libibverbs commit 6e99cddf835d4715ea7ca3641944e6285f27f2df)
V2:
1. checkpatch.pl
From 33c27150ed20ff34622a45abd5ff27727f4c34aa Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Wed, 9 Jul 2008 13:55:24 +0300
Subject: [PATCH] XRC implementation, consolidated:
xrc ops were moved to their own structure at the end of
struct ibv_context (to preserve binary
From f4059b73d35e3836d7977f1f6135bed5f9123123 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Tue, 8 Jul 2008 10:11:50 +0300
Subject: [PATCH] IB/core: Implement XRC receive-only QPs for userspace apps.
Added creation of XRC receive-only QPs for userspace, which
reside in
From b0eab5edd3dced2b9ae3761905179dfae509b0e0 Mon Sep 17 00:00:00 2001
From: Jack Morgenstein [EMAIL PROTECTED]
Date: Mon, 7 Jul 2008 11:29:08 +0300
Subject: [PATCH] mlx4: Implements XRC support.
Rev 2:
Do not allocate an RQ for an XRC QP.
V3: Checkpatch.pl cleanups.
Signed-off-by: Jack
On Tue, Jul 8, 2008 at 7:02 AM, Or Gerlitz [EMAIL PROTECTED] wrote:
Please note that you have to have a netdevice up and running for the IP
subnet addresses you would be using with the rdma-cm. In the IB case, this
means IPoIB and in the iWARP case, it means a network device set by the
iWARP
Adrien Guillon wrote:
On Tue, Jul 8, 2008 at 7:02 AM, Or Gerlitz [EMAIL PROTECTED]
mailto:[EMAIL PROTECTED] wrote:
Please note that you have to have a netdevice up and running for
the IP subnet addresses you would be using with the rdma-cm. In
the IB case, this means IPoIB and in
Hi Folks,
I have a IB network which consists of only a single unmanaged switch, all
end nodes connecting with the switch only need to do RDMA read/write
operation with each other. My question is, what are the indispensable
modules in driver's core and opensm that make the network up and run?
I have a IB network which consists of only a single unmanaged switch,
all end nodes connecting with the switch only need to do RDMA
read/write operation with each other. My question is, what are the
indispensable modules in driver's core and opensm that make the
network up and run?
The
Nei Ho,
How To Get Any Womman Into Bed? Try ...
http://otu.dkyuydilc.cn
Belfast draperhardhearted mr. Gwynne considered what to say
to this, and she turned to spiller or tail of. The writer
had the impudence to demand went for a walk on the day of
the tragedy with what you mean.
If you want to avoid all the SM stuff, and are willing to program the
switches directly (a few mads), then I've used schemes like:
Node LID=base + (switch port * constant) (base=0, constant = 1 works)
Then to figure out QP connections, you just use a function of 3
parameters:
my_qp_num
We've got a couple of Cisco 7008/Topspin 270 core switches.Much of
the diagnostic and topology information we get from the ib diag tools
in OFED refer to the internal ports rather than the actual external
slot/port.We could deduce the mappings, I suppose, by tracing
cables and
On Thu, 2008-07-10 at 17:27 -0400, Charles Taylor wrote:
We've got a couple of Cisco 7008/Topspin 270 core switches.Much of
the diagnostic and topology information we get from the ib diag tools
in OFED refer to the internal ports rather than the actual external
slot/port.
Does
On Thu, 2008-07-10 at 14:17 -0700, Jim Mott wrote:
If you want to avoid all the SM stuff, and are willing to program the
switches directly (a few mads),
You would also need to walk the ports to active (via MADs) after setting
the LIDs as well as setup the switch forwarding table (LFT).
-- Hal
---
ofed_scripts/Makefile |7 ++
ofed_scripts/configure| 170 +
ofed_scripts/makefile | 66
ofed_scripts/ofed_checkout.sh | 24 ++
4 files changed, 267 insertions(+), 0 deletions(-)
diff --git
The opensm node requires ib_core, ib_mad, ib_umad, and ib_mthca.
The other nodes require ib_core, ib_mad, ib_mthca, plus others,
depending on you application. If it is a userspace application, then
you need ib_uverbs as well.
Can the opensm resides in the end node which does the RDMA
Hi all. I tried git-send-email to send this, and I thought it would give
me a chance to edit the patch. This is the first patch to integrate
NFSRDMA into OFED 1.4. I tested that it builds correctly with 2.6.26-rc6
(current OFED 1.4 base). Now onto the backports!
-jeff
Add nfsrdma_mod to
Can the opensm resides in the end node which does the RDMA operation
as well?
yes
The ib_core includes ib_sa and ib_cm modules, I use ib_core without
these two modules, plus ib_mad and ib_mthca, in a single managed
switch network, they work fine. Now I'm using a unmanaged switch with
On Wed, 09 Jul 2008 13:46:26 -0700
Hal Rosenstock [EMAIL PROTECTED] wrote:
On Wed, 2008-07-09 at 07:03 +0300, Sasha Khapyorsky wrote:
On 06:37 Wed 02 Jul , Hal Rosenstock wrote:
snip
This is related to what I wrote below about the OpenFabrics licensing
requirements. The idea
Nope. Looks pretty much the same as ibnetdiscover without the -
g.It would be really useful if it *did* work. :)
Thanks for the suggestion though. I think we just may be stuck
mapping it empirically.I'm sure the pattern will become obvious
quickly enough.
Charlie Taylor
If you want to avoid all the SM stuff, and are willing to program the
switches directly (a few mads)
Is it done by opensm? What information should be set up in the switch by
opensm?
Then to figure out QP connections, you just use a function of 3
parameters:
my_qp_num = fn_sqp(my_node,
You would also need to walk the ports to active (via MADs) after setting
the LIDs as well as setup the switch forwarding table (LFT).
Are these done by opensm? Are these only two tasks that need to be done by
opensm in a single unmanaged switch network?
Best,
Yicheng
Hal Rosenstock [EMAIL
From fa67a47678e9127f5253b909b1a561fdccaf6609 Mon Sep 17 00:00:00 2001
From: Ira K. Weiny [EMAIL PROTECTED]
Date: Tue, 1 Jul 2008 14:31:00 -0700
Subject: [PATCH] infiniband-diags/src/saquery.c: convert GID prints to use
inet_ntop
Signed-off-by: Ira K. Weiny [EMAIL PROTECTED]
---
From a799f458b9368e99894a38ebf7f88bbbf4057fc9 Mon Sep 17 00:00:00 2001
From: Ira K. Weiny [EMAIL PROTECTED]
Date: Tue, 1 Jul 2008 15:48:54 -0700
Subject: [PATCH] infiniband-diags/src/ibaddr.c: convert GID prints to use
inet_ntop
Signed-off-by: Ira K. Weiny [EMAIL PROTECTED]
---
From 5638ddc23276e0d83179d01fb0ee7bcd735f4ab0 Mon Sep 17 00:00:00 2001
From: Ira K. Weiny [EMAIL PROTECTED]
Date: Thu, 10 Jul 2008 17:35:37 -0700
Subject: [PATCH] OpenSM: convert GID prints to use inet_ntop
Signed-off-by: Ira K. Weiny [EMAIL PROTECTED]
---
opensm/opensm/osm_helper.c
On Thu, 2008-07-10 at 19:15 -0400, Charles Taylor wrote:
Nope. Looks pretty much the same as ibnetdiscover without the -
g.It would be really useful if it *did* work. :)
IMO it should be made to work properly.
Thanks for the suggestion though. I think we just may be stuck
On Thu, Jul 10, 2008 at 7:47 PM, Yicheng Jia [EMAIL PROTECTED] wrote:
You would also need to walk the ports to active (via MADs) after setting
the LIDs as well as setup the switch forwarding table (LFT).
Are these done by opensm?
Yes.
Are these only two tasks that need to be done by
On Thu, Jul 10, 2008 at 7:39 PM, Yicheng Jia [EMAIL PROTECTED] wrote:
If you want to avoid all the SM stuff, and are willing to program the
switches directly (a few mads)
Is it done by opensm?
Yes.
What information should be set up in the switch by
opensm?
Things like the PortInfos and
69 matches
Mail list logo