Hi Sasha.
Looks good.
One remark though: all the static functions
should have the __osm_ prefix in their names.
--
Yevgeny
Sasha Khapyorsky wrote:
This improves switch's mcast tables dumping and eliminates multiple file
open/seek/close sequences. In one word - cleanup.
Signed-off-by:
Hello,I have been trying to build OFED source package (1.1 rev 9820) as was described in the HOWTO.build_ofed wiki. The package was built successfully, however i had trouble compilng it. The error I get is a missing library inside OFED's temporary build tree. The error is during the compilation of
Hi Hal
Leaving OSM_LOG_DIR or OSM_CACHE_DIR environment variables
empty will cause OSM to wright log or cache files to /
since OSM runs as root process.
Although one might say that this is just a question of point
of view, I really think that to prevent root directory trashing
(as I did by
On Thu, 2006-10-12 at 20:35, Sasha Khapyorsky wrote:
This improves switch's mcast tables dumping and eliminates multiple file
open/seek/close sequences. In one word - cleanup.
Signed-off-by: Sasha Khapyorsky [EMAIL PROTECTED]
Thanks. Applied.
-- Hal
On Sun, 2006-10-15 at 10:31, Yevgeny Kliteynik wrote:
Hi Hal
This patch fixes a few data type problems with OSM on
64-bit Windows machines.
The changes are done in the following files:
opensm/osm_prtn_config.c
opensm/osm_pkey.c
opensm/osm_qos.c
Note that the casting is done on
Arlin Davis wrote:
Here is a patch to remove uDAPL scm provider from the build since it is
no longer needed nor
supported. This provider was merely a stop gap until uCMA was pushed
into kernel.
Tziporet, can you get this change into OFED 1.1?
Signed-off by: Arlin Davis [EMAIL
Tziporet Koren wrote:
Arlin Davis wrote:
Here is a patch to remove uDAPL scm provider from the build since it
is no longer needed nor
supported. This provider was merely a stop gap until uCMA was pushed
into kernel.
Tziporet, can you get this change into OFED 1.1?
Does this fix a
libibmad: Add support for DrSLID
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
---
Index: ../libibmad/include/infiniband/mad.h
===
--- ../libibmad/include/infiniband/mad.h(revision 9746)
+++
Diags/smpquery: Add support for DrSLID
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
---
Index: src/smpquery.c
===
--- src/smpquery.c (revision 9776)
+++ src/smpquery.c (working copy)
@@ -376,7 +376,8 @@ usage(void)
On Mon, 2006-10-16 at 05:47, Yevgeny Kliteynik wrote:
Hi Hal
Leaving OSM_LOG_DIR or OSM_CACHE_DIR environment variables
empty will cause OSM to wright log or cache files to /
since OSM runs as root process.
Although one might say that this is just a question of point
of view, I really
Or Gerlitz wrote:
Tziporet Koren wrote:
To clarify, without this patch, OFED provides an untested unsupported
udapl provider library (libdaplscm.so). The way i realized that this
happens was when I got bug reports and questions from a customer who
attempted to use it... So it should
Quoting r. Maestas, Christopher Daniel [EMAIL PROTECTED]:
Subject: Re: [openib-general] RHEL5 and OFED ...
Now for userspace - does RHEL5 include at least libibverbs-1.0?
This has been released a while back, and Roland makes regular bugfix
releases.
Here's what I see on a rhel4 u4
Quoting r. Roland Dreier [EMAIL PROTECTED]:
Subject: Re: RHEL5 and OFED ...
Michael 1. The first thing would be to list fixes between 2.6.18
Michael and 2.6.19-rc1 and backport these. Some of them are in
Michael OFED.
If you want to do this, I think it would be great to also
http://openib.org/bugzilla/show_bug.cgi?id=229
[EMAIL PROTECTED] changed:
What|Removed |Added
CC||[EMAIL PROTECTED]
--- You are
Hi.
I started to write man pages for the ib verbs API in the user level.
Those pages are based on the the verbs description (in verbs.h), on the
IB spec and on my experience.
Attached is a file with the man pages (binary + pod source files +
Makefile) of the following verbs:
Mr Sten Williams,
We have won a price off 850.000,-- US dollars but
we have to pay 500 pound delivery cost in frond.
Can u ashore me that we won 850.000,-- US dollars
en will you send it on paper to me.
I only want to know because i don't want to pay
500,-- pound and recieve nothing.
Bill,
2 small changes to the diagram on slide
6.
SRP box should be yellow since it is IB
specific.
Drop the word "R-NIC" from the User APIs
box.
I think we can improve this diagram
message.
Both kernel and user API boxes for "verbs/API" should be
non-colored "common".
Thanks,
Arkady
But we can document this in the release notes.
Arlin - can you update uDAPL release notes with this or do you want me
to do it.
Go ahead and update the release notes. If you could also apply the small
dat.conf patch that removes SCM from the configuration it would reduce
the changes of
Go ahead and update the release notes. If you could also apply the small
dat.conf patch that removes SCM from the configuration it would reduce
the changes of customer picking up SCM by mistake. Most customers refer
to dat.conf for supported configurations.
-arlin
OK I will
Tziporet
This patch is already in.
We will publish latest pre-release version tomorrow so everybody can do
latest checks.
Is this OK?
Tziporet
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Scott
Weitzenkamp (sweitzen)
Sent: Sunday, October 15, 2006 10:16 PM
To:
Hi, Roland,
If we only support GSO enablement in ethtool, there is no problem. What I meant is anything related to MAC address in ethtool utility needs to be updated for IB device.
Do you like the idea to add ethtool support in IPoIB? Do you want me to work on this?
Thanks
Shirley
On Sun, 15 Oct 2006, Roland Dreier wrote:
However, this is a somewhat weird symptom, although I can imagine that
out-of-order doorbells cause extra completions or something like that,
which causes IPoIB to overrun the send queue.
Adding the mmiowb()s definitely fixes things?
At least
Shirley If we only support GSO enablement in ethtool, there is no
Shirley problem. What I meant is anything related to MAC address
Shirley in ethtool utility needs to be updated for IB device.
Shirley Do you like the idea to add ethtool support in IPoIB? Do
Shirley you want me
On Sun, 2006-10-15 at 09:39 -0700, Roland Dreier wrote:
I've been meaning to mention this... I have a preliminary version in
git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband.git
ipoib-napi
There are further changes I would like to add on top of that, but
comments on
Yes, that would be great.
Scott Weitzenkamp
SQA and Release Manager
Server Virtualization Business Unit
Cisco Systems
-Original Message-
From: Tziporet Koren [mailto:[EMAIL PROTECTED]
Sent: Monday, October 16, 2006 9:26 AM
To: Scott Weitzenkamp (sweitzen); Tziporet Koren;
[EMAIL
Eli Please diff to see my comments. Generaly it looks like the
Eli condition on netif_rx_reschedule() should be inverted.
Why? A return value of 0 means that the reschedule failed (probably
because the poll routine is already running somewhere else) and the
poll routine should just
akepner At least with the workload that we used to reproduce this
akepner bug, yes. (The workload was simply 2 ttcp processes, each
akepner placed on a different node of an Altix.) Without the
akepner mmiowb()s things would hang very reliably and very
akepner quickly (within a
On Mon, 2006-10-16 at 09:48 -0700, Roland Dreier wrote:
Eli Please diff to see my comments. Generaly it looks like the
Eli condition on netif_rx_reschedule() should be inverted.
Why? A return value of 0 means that the reschedule failed (probably
because the poll routine is already
Actually, one further question:
@@ -314,6 +316,9 @@ void mthca_cq_clean(struct mthca_dev *de
wmb();
cq-cons_index += nfreed;
update_cons_index(dev, cq, nfreed);
+/* use mmiowb to ensure update is ordered properly
+ *
Good. Then after enabling GSO, we can chain multiple packets together in IPoIB for one doorbell to send large packet.
Thanks
Shirley Ma
IBM Linux Technology Center
15300 SW Koll Parkway
Beaverton, OR 97006-6063
Phone(Fax): (503) 578-7638___
Shirley Good. Then after enabling GSO, we can chain multiple
Shirley packets together in IPoIB for one doorbell to send large
Shirley packet.
How does that work? GSO doesn't change the hard_start_xmit()
interface, does it?
- R.
___
Diags/ibnetdiscover: More changes for IB routers
Signed-off-by: Hal Rosenstock [EMAIL PROTECTED]
Index: src/ibnetdiscover.c
===
--- src/ibnetdiscover.c (revision 9827)
+++ src/ibnetdiscover.c (working copy)
@@ -461,14 +461,30 @@
Roland Dreier [EMAIL PROTECTED] wrote on 10/16/2006 10:37:12 AM:
Shirley Good. Then after enabling GSO, we can chain multiple
Shirley packets together in IPoIB for one doorbell to send large
Shirley packet.
How does that work? GSO doesn't change the hard_start_xmit()
interface, does
Shirley No, it doesn't. I am thinking to add enqueue/dequeue
Shirley multiple packets in qdisc. It would benifit other
Shirley networking device.
So am I understanding correctly -- this is other work that is
independent of GSO? Is the plan to add a new optional driver method
that
Roland Dreier [EMAIL PROTECTED] wrote on 10/16/2006 10:49:32 AM:
Shirley No, it doesn't. I am thinking to add enqueue/dequeue
Shirley multiple packets in qdisc. It would benifit other
Shirley networking device.
So am I understanding correctly -- this is other work that is
Hi,
We will publish 1.1-pre1 package tomorrow (Tue. 17-Oct)
Only blocker issues from RC7 will be updated:
1. SRP fix for Cisco FC gateway
2. Small updates for the install
currently we're working on the one install issue as I mentioned in another
thread.
We found out that the 64- and 32-bit
Tziporet Koren wrote:
This is the plan to do the 1.1 release this week:
We will publish 1.1-pre1 package tomorrow (Tue. 17-Oct)
Only blocker issues from RC7 will be updated:
1. SRP fix for Cisco FC gateway
2. Small updates for the install
3. Fix in diagnet to support SM on a
Eitan Zahavi wrote:
I agree that layering on top is easier. But does it really solve the
bug? I think not. If you would REPLACE the API and not provide both options
(above and below refcount enforcement ) it would make sense to me.
We disagree on the philosophy here. I view ib_mad as
This plan is OK with Cisco.
Scott
Weitzenkamp
SQA and Release
Manager
Server Virtualization
Business Unit
Cisco Systems
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Tziporet
KorenSent: Monday, October 16, 2006 10:04 AMTo: Open
FabricsCc: openibSubject:
On 11:47 Mon 16 Oct , Yevgeny Kliteynik wrote:
Hi Hal
Leaving OSM_LOG_DIR or OSM_CACHE_DIR environment variables
empty will cause OSM to wright log or cache files to /
since OSM runs as root process.
Although one might say that this is just a question of point
of view, I really
Roland,
Don't know why I have trouble to get this patch from your git tree. Do you mind to post this patch here so I can test the performance over ehca?
Thanks
Shirley Ma___
openib-general mailing list
openib-general@openib.org
Sean Hefty wrote:
pri_path = alt_path = path 1
works
no, I haven't tested that. I can try that too, if u think that can
provide useful info..
I misunderstood one of your earlier e-mails then. I threw together a test case
to try this, and it worked for me. Can you see if the same
I'm beginning work on adding InformInfo/Notice event reporting to the IB stack,
and I'd like any input on potential implementations, as well as intended usage.
Clients use InformInfo to register for events, with registration tracked on a
per source QP basis. Given this, possible approaches
On Mon, 16 Oct 2006, Roland Dreier wrote:
OK, cool. Sounds convincing to me. BTW -- are there Altix systems
with PCIe? Have you tested the mthca_arbel_xxx (mem-free PCIe HCA)
changes, or just the mthca_tavor_xxx (PCI-X HCA) parts?
Yes, there are PCIe Altices, but I was unable to reproduce
now I am able to create connections with pri_path = path1, alt_path = path2
with the following change in the code.
I specify the port_num and al_port_num before calling ib_modify_qp() to
change
state to RTR (earlier I was changing this when modifying state to
IB_QPS_INIT).
Doesn't
On Mon, 16 Oct 2006, Roland Dreier wrote:
Would it be cleaner just to put an mmiowb() inside update_cons_index()
in the case where the doorbell gets rung? That way there's no
unnecessary mmiowb() in the memfree case (which doesn't use a doorbell
or do any writing to PCI at all).
On second
From: Sean Hefty
Sent: Monday, October 16, 2006 3:29 PM
To: openib
Subject: [openib-general] [RFC] Notice/InformInfo event reporting
I'm beginning work on adding InformInfo/Notice event reporting to the
IB
stack,
and I'd like any input on potential implementations, as well as
intended
Quoting r. Doug Ledford [EMAIL PROTECTED]:
1. The first thing would be to list fixes between 2.6.18 and 2.6.19-rc1 and
backport these. Some of them are in OFED.
That would be helpful. Since 2.6.19-rc looks to have integrated the
iWARP merge, the fixes are no doubt mixed in with a bunch
Quoting r. Roland Dreier [EMAIL PROTECTED]:
There are further changes I would like to add on top of that, but
comments on the two patches there would be appreciated.
A small optimization:
if (missed_event netif_rx_reschedule(dev, 0))
should be, I think
if
akepner On second (or eighth, or whatever) look, yeah, that's
akepner better and simpler, too. I'll repost a fixed up version.
No need -- I'll just revise it in my tree (since I already grabbed
your patch).
Thanks,
Roland
___
openib-general
somenath wrote:
Doesn't ib_cm_init_qp_attr() set this for you?
No, it doesn't. it returns me
attr_mask=0x12d181
port=0x0 alt_port=0x0
Okay - there was a fix to the cm.c file (svn rev 8267) that added setting the
alternate port number when initializing the QP attributes. Apparently
Rimmer, Todd wrote:
My recommendation is option 2.
Thanks for the response.
In large fabrics the SA can be a bottleneck. It is best for an end node
to register with the SA only for the events which are of actual interest
to the end node.
Which part of the SA is the bottleneck? Is it the
Sorry, I don't know much about git...
I tried to git the module code for OFED 1.1 rc7 this using the following
command. What am I doing wrong?
14:42:40 git clone git://www.mellanox.co.il/~git/infinibandref
fatal: unexpected EOF
fetch-pack from 'git://www.mellanox.co.il/~git/infinibandref'
Hi,
I have been trying to run dapltest with trunk rev 9717 with linux kernel
2.6.18 and I get an error. The error and configuration is shown below.
Basically, the rdma_cm device is not created under /dev/infiniband. I
am wondering if this is a known problem and how to solve it.
Thanks,
Steve
From: Sean Hefty [mailto:[EMAIL PROTECTED]
Sent: Monday, October 16, 2006 5:33 PM
To: Rimmer, Todd; Matt Leininger
Cc: openib
Subject: Re: [openib-general] [RFC] Notice/InformInfo event reporting
Rimmer, Todd wrote:
My recommendation is option 2.
Thanks for the response.
In large
Steve Smaldone wrote:
Hi,
I have been trying to run dapltest with trunk rev 9717 with linux kernel
2.6.18 and I get an error. The error and configuration is shown below.
Basically, the rdma_cm device is not created under /dev/infiniband. I
am wondering if this is a known problem and how
Hi,
Sorry for replying to myself, but I loaded rdma_ucm and the rdma_cm
device appears. However, it now fails with the following:
$ ./dapltest -T S -D IB1
...
DAT Registry: dat_ia_openv (IB1,1:2,0) called
DAT Registry: IA IB1, trying to load library /usr/local/lib/libdapl.so
DAT Registry:
Hi Hal.
[snip]
So will you supply another patch with this approach ?
-- Hal
Here it is.
--
Yevgeny
Signed-off-by: Yevgeny Kliteynik [EMAIL PROTECTED]
Index: opensm/osm_db_files.c
===
--- opensm/osm_db_files.c
Rimmer, Todd wrote:
In a functioning fabric, events will be rare. However its when you
first boot the fabric, reboot the SM or other similar start up actions
that things get real busy.
Hmm... I need to think more about how to handle the start up scenario.
In general I have found that only a
Steve Smaldone wrote:
Hi,
Sorry for replying to myself, but I loaded rdma_ucm and the rdma_cm
device appears. However, it now fails with the following:
$ ./dapltest -T S -D IB1
...
DAT Registry: dat_ia_openv (IB1,1:2,0) called
DAT Registry: IA IB1, trying to load library
Arlin Davis wrote:
Steve Smaldone wrote:
Hi,
Sorry for replying to myself, but I loaded rdma_ucm and the rdma_cm
device appears. However, it now fails with the following:
$ ./dapltest -T S -D IB1
...
DAT Registry: dat_ia_openv (IB1,1:2,0) called
DAT Registry: IA IB1, trying to load
Ah, yes, that works.
Thanks,
Ira
On Tue, 17 Oct 2006 00:21:06 +0200
Sasha Khapyorsky [EMAIL PROTECTED] wrote:
On 14:46 Mon 16 Oct , Ira Weiny wrote:
Sorry, I don't know much about git...
I tried to git the module code for OFED 1.1 rc7 this using the following
command. What am I
We tested RC7, but fork() does not work:
1. system() causes IB to fail.
2. fork(), child calling exit(0) immediately also causes IB to fail.
Anyone has tested fork() related issue ?
--CQ Tang, HP-MPI
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of
The plan is OK with Silverstorm.
I have a question though.What aretheplans to support SRP-HA featureon RHEL4 kernels ?
Thanks
Karun
From: [EMAIL PROTECTED] on behalf of Tziporet KorenSent: Mon 10/16/2006 1:03 PMTo: Open FabricsCc: openibSubject: [openfabrics-ewg] OFED 1.1 release schedule
In queue_req(), use time_after_eq() instead of time_after()
for following reasons :
- Improves insert time if multiple entries with same time are
present.
- set_timeout need not be called if entry with same time
is added to the list (and that happens to be the entry
with the smallest time),
Fixes following problems in process_req() relating to
cancellation :
- Function is wrongly doing another addr_remote() when
cancelled, which is not required.
- Make failure reporting immediate by using time_after_eq().
- On cancellation, -ETIMEDOUT was returned to the callback
routine instead
Rewrite cma_req_handler error handling case to encapsulate
common code.
Signed-off-by: Krishna Kumar [EMAIL PROTECTED]
diff -ruNp org/drivers/infiniband/core/cma.c new/drivers/infiniband/core/cma.c
--- org/drivers/infiniband/core/cma.c 2006-10-09 16:57:26.0 +0530
+++
rdma_bind_addr leaks a cma_dev reference count in failure
case.
Signed-off-by: Krishna Kumar [EMAIL PROTECTED]
diff -ruNp org/drivers/infiniband/core/cma.c new/drivers/infiniband/core/cma.c
--- org/drivers/infiniband/core/cma.c 2006-10-09 17:13:41.0 +0530
+++
cma_new_id() does not require to do destroy_id(), instead
it can kfree(), since nothing is allocated on that id.
Posting this as an RFC in case anyone feels that create_id
should be cleaned up by destroy_id (even if redundant).
Signed-off-by: Krishna Kumar [EMAIL PROTECTED]
diff -ruNp
Re-send ARP, since earlier ARP request could have got
dropped/lost. This should be done in addr_resolve_remote()
as doing it in rdma_resolve_ip() means sending ARP only
once.
Signed-off-by: Krishna Kumar [EMAIL PROTECTED]
diff -ruNp org/drivers/infiniband/core/addr.c
If addr_handler() got invoked with an error status,
do not set id_priv-state to success followed by
resettting it to the old value (redundant code).
Also encapsulate some common code.
Signed-off-by: Krishna Kumar [EMAIL PROTECTED]
diff -ruNp org/drivers/infiniband/core/cma.c
Quoting r. Doug Ledford [EMAIL PROTECTED]:
Dough, would it be possible to update this + libmthca?
Possibly. What's the justification? What's in 1.0.4 that is the
primary reason for wanting to update from 1.0.3?
Hmm, Roland, I went to look into Changelog and I note that you don't
label svn
Quoting r. Krishna Kumar [EMAIL PROTECTED]:
Subject: [PATCH] rdma_bind_addr() leaks a cma_dev reference count
rdma_bind_addr leaks a cma_dev reference count in failure
case.
Signed-off-by: Krishna Kumar [EMAIL PROTECTED]
diff -ruNp org/drivers/infiniband/core/cma.c
Michael Hmm, Roland, I went to look into Changelog and I note
Michael that you don't label svn versions or release versions
Michael there. So it's hard to see what was fixed in what
Michael version, or to map to svn versions.
Michael I'll dig that info up but I thin we want it
look up Documentation/CodingStyle Chapter 7: Centralized exiting of
functions
which says unconditional statements are easier to understand and follow,
and note that with this style nesting is reduced.
Hmmm, OK, I will re-phrase this patch to reduce nesting.
thanks,
- KK
Michael S. Tsirkin
Quoting r. Roland Dreier [EMAIL PROTECTED]:
I do put the releases into the changelog. There has not been a
libibverbs 1.0.4 release yet, but you can find the entry
2006-05-02 Roland Dreier [EMAIL PROTECTED]
* Release version 1.0.3.
in the libibverbs ChangeLog. Everything after
76 matches
Mail list logo