Quoting r. Shirley Ma [EMAIL PROTECTED]:
Subject: Re: Re: [PATCH] rdma_lat-09 and results
./rdma_lat
libibverbs: Warning: no userspace device-specific driver found for uverbs0
driver search path: /usr/local/lib/infiniband
No IB devices found
I have one node working, one node
On Wed, Jun 01, 2005 at 07:24:18PM -0400, Hal Rosenstock wrote:
Hi Roland,
I can run ibv_devices but ibv_asyncwatch seems to fail:
ibv_devices
device node GUID
--
mthca0 0008f10403960558
ibv_asyncwatch
On Thu, Jun 02, 2005 at 08:29:34AM +0300, Michael S. Tsirkin wrote:
Quoting r. Grant Grundler [EMAIL PROTECTED]:
Subject: [PATCH] rdma_lat-09 and results
Michael,
Good news:
My next cleanup of rdma_lat.c is working and patch is appended.
Summary of changes below.
Bad
Quoting r. Gleb Natapov [EMAIL PROTECTED]:
Subject: Re: Re: [PATCH] rdma_lat-09 and results
On Thu, Jun 02, 2005 at 08:29:34AM +0300, Michael S. Tsirkin wrote:
Quoting r. Grant Grundler [EMAIL PROTECTED]:
Subject: [PATCH] rdma_lat-09 and results
Michael,
Good news:
My
libibverbs cant find the libmthca plugin.
ibv_pingpong, ibv_asyncwatch ibv_devices work
OK.
Shirley Ma
IBM Linux Technology Center
15300 SW Koll Parkway
Beaverton, OR 97006-6063
Phone(Fax): (503) 578-7638
___
openib-general mailing list
I hadn't heard that insmod was being removed from Linux.
In fact the DAPL Plugfest successfully used kernel
daemons and kdapltest to demonstrate DAT interoperability
across multiple vendors: kernel to kernel, kernel to user
and user to user.
These are existing applications already deployed. I
On Thu, Jun 02, 2005 at 05:22:26AM -0700, Caitlin Bestler wrote:
I hadn't heard that insmod was being removed from Linux.
No one claimed that.
In fact the DAPL Plugfest successfully used kernel
daemons and kdapltest to demonstrate DAT interoperability
across multiple vendors: kernel to
in order to enable kdapltest -T P i needed to remark attr that are not set by
openib gen2
1) ia_attr.max_evd_qlen
2) ia_attr.max_rdma_read_per_ep_in
3) ia_attr.max_rdma_read_per_ep_out
also there was bug in kdapltest memory registration (file dapl_bpool.c)
and there was bug where we free memory
On Tue, 31 May 2005, Tom Duffy wrote:
On Tue, 2005-05-31 at 14:17 -0400, James Lentini wrote:
Here's the specification's exact description:
timeout: Duration of time, in microseconds, that a consumer waits for
connection establishment. The value of DAT_TIMEOUT_INFINITE
On Tue, 31 May 2005, Tom Duffy wrote:
On Tue, 2005-05-31 at 14:34 -0700, Sean Hefty wrote:
Sean,
Is there any way of requesting an infinite number of retries?
There is not, but nothing prevents a user from simply re-issuing a request
after it times out.
Infinite retries inside the
Quoting r. Roland Dreier [EMAIL PROTECTED]:
Subject: [PATCH] mthca: fix registration for giant MRs
Here's a patch that allows mthca to break up registration of giant
userspace MRs into multiple firmware commands. The net effect of this
is that there should no longer be any limit (beyond
Sorry about that. Should be fixed now.
On Wed, 1 Jun 2005, Itamar Rabenstein wrote:
itamar Hi James,
itamar
itamar you replied with Committed in revision 2514.
itamar But I am checking and you have committed my patch with a change.
itamar Please whenever you decide to change a patch please
On Wed, 1 Jun 2005, Tom Duffy wrote:
On Wed, 2005-06-01 at 12:04 +0200, Christoph Hellwig wrote:
That beeing said, one of the first thing you should get rid of if you
want to be able to take code from kdapl to the generic rdma code is way
it deals with handles. The kdapl code gives up
On Thu, 2005-06-02 at 11:16 -0400, James Lentini wrote:
- in dat.h, create a public structure for each object type:
struct dat_ep {
struct dat_provider *provider;
};
- in the transport provider (dapl.h) have a private structure that
contains the public one:
struct
On Thu, Jun 02, 2005 at 08:29:34AM +0300, Michael S. Tsirkin wrote:
...
I changed the timestamping strategy. I used to:
...
This meant that tstamp instruction was out of the data path,
while we did polling.
On the negative side, although the average (and likely median) delta
between tstamps
Thanks, I found a problem with that patch and checked in a fix. Can
you let me know if it works better now?
- R.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit
Michael I'd like to suggest we keep passing struct mthca_buddy
Michael *to mthca_alloc_mtt, instead of passing around and
Michael keeping in memory the binary fmr flag, since all that
Michael this flag does is select the right allocator, and callers
Michael of
On Thu, 2 Jun 2005, Tom Duffy wrote:
On Thu, 2005-06-02 at 11:16 -0400, James Lentini wrote:
- in dat.h, create a public structure for each object type:
struct dat_ep {
struct dat_provider *provider;
};
- in the transport provider (dapl.h) have a private structure that
Please see some comments below which Im
offering on my own but I hope speaking fairly and responsibly for the steering
committee (more correctly the Board) of OpenIB.
Jim Ryan, Chairman, OpenIB
From:
[EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Venkata Jagana
On 6/2/05, Tom Duffy [EMAIL PROTECTED] wrote:
On Thu, 2005-06-02 at 11:16 -0400, James Lentini wrote:
- in dat.h, create a public structure for each object type:
struct dat_ep {
struct dat_provider *provider;
};
- in the transport provider (dapl.h) have a private structure
Quoting r. Roland Dreier [EMAIL PROTECTED]:
Subject: Re: [PATCH] mthca: fix registration for giant MRs
Michael Why dont we keep mthca_mtt by instance in struct
Michael mthca_mr, like this: struct mthca_mr { struct ib_mr ibmr;
Michael struct mthca_mtt mtt; };
Michael Saves
Why is it that you believe that the DAT registry does not support
plug and play? The interface was most specifically designed
to allow that.
When a device is plugged in, the driver is loaded by existing
OS mechanisms. It can then load the provider code (if needed).
Whenever the provider module is
On Wed, Jun 01, 2005 at 05:52:28PM -0700, Grant Grundler wrote:
Here's with the new rdma_lat.c:
Here are some pfmon31 perf results for the new rdma_lat.
The first set is just L3 Cache misses.
The second set shows all cache misses.
The sample rate on the second set was a bit high - note the
On Thu, 2005-06-02 at 10:17 -0400, James Lentini wrote:
On Tue, 31 May 2005, Tom Duffy wrote:
On Tue, 2005-05-31 at 14:17 -0400, James Lentini wrote:
Here's the specification's exact description:
timeout: Duration of time, in microseconds, that a consumer waits for
I haven't been able to determine the cause of the MAD crashes, so I was
going to separate the CQs for QP 0 and 1 to give me more context/state
information when a completion occurs.
Is this a change that we want permanently? If so, I will cleanup the change
for submission, otherwise, I'll
On Wed, 2005-06-01 at 20:45 -0400, Hal Rosenstock wrote:
On Wed, 2005-06-01 at 16:51, Tom Duffy wrote:
I am putting together a network with a dumb IB switch, a couple of Linux
OpenIB boxes, a Solaris 10 box, a Solaris Nevada box, etc. I fired up
opensm on one of the Linux nodes, tried to
Sean I haven't been able to determine the cause of the MAD
Sean crashes, so I was going to separate the CQs for QP 0 and 1
Sean to give me more context/state information when a completion
Sean occurs.
Sean Is this a change that we want permanently? If so, I will
Sean
Roland Dreier wrote:
Sean I haven't been able to determine the cause of the MAD
Sean crashes, so I was going to separate the CQs for QP 0 and 1
Sean to give me more context/state information when a completion
Sean occurs.
Sean Is this a change that we want permanently? If
OK, I respun the patch.
I didn't change the calculation of mtt-order to use fls, because the
formula fls(max(size, MTHCA_MTT_SEG_SIZE / 8) - 1) should really be
fls(max(size, MTHCA_MTT_SEG_SIZE / 8)) - fls(MTHCA_MTT_SEG_SIZE / 8)
or something like that, and I ended up confusing myself and so I
Hi,
The kdapltest program wedges consistently wedges a node. I just tried
it with Revision 2523, and it runs fine the first couple of times. The
third time took a while, and the fourth time killed the server.
The server is running:
kdapltest -T S -D mthca0a
The client runs this several times
On Thu, 2005-06-02 at 13:31, Tom Duffy wrote:
On Wed, 2005-06-01 at 20:45 -0400, Hal Rosenstock wrote:
On Wed, 2005-06-01 at 16:51, Tom Duffy wrote:
I am putting together a network with a dumb IB switch, a couple of Linux
OpenIB boxes, a Solaris 10 box, a Solaris Nevada box, etc. I
I committed everything in revision 2526, except changing the name of
dat_dictionary_key_is_equal to dat_dict_key_is_equal. I assumed that
you did this to fit as much of the function declartion on one line as
possible. I think I achieved the result you wanted by giving the
parameters shorter
Thanks , ping is working now (;-)
Itamar
-Original Message-
From: Roland Dreier [mailto:[EMAIL PROTECTED]
Sent: Thursday, June 02, 2005 7:05 PM
To: Itamar Rabenstein
Cc: openib-general@openib.org
Subject: Re: station hang with the last patch to
infiniband/core/sa_query.c
Hi ,
I am not an expert to pci hot plug but as far as know we currently dont have
away in kdapl
to inform the consumer of kdapl that a provider is going to be unloaded.
pci hot plug should enable any low level driver (openib) to inform high
level driver(kdapl)
that the devide is going down.
In
Hi Josh,
Can you run it with debug option ( -d )
and tell me what is the last debug print on the server side ?
when you say kill can you hit ^c or is it hanging the station ?
what is system configuration ? (arch , Os ,kernel)
Itamar
-Original Message-
From: Josh England
The ucm code in ib_ucm_event_handler() reads the cm_id-state information in
response to an event occurrence. The cm_id state can change dynamically in
response to an event, and in general should not be accessed. (I.e. the
current cm_id state may not match up with the event being reported.)
Patch to add basic content to the CM library
README file. Add build instructions, install
directory, and device node configuration.
Signed-off-by: Bill Jordan [EMAIL PROTECTED]
Index: README
===
--- README (revision 2523)
+++
On Thu, 2005-06-02 at 16:18 -0400, James Lentini wrote:
I'd recommend these changes (see attached):
- use the new microsecond conversion function you sent
great
- keep a default DAT timeout constant (I propose DAT_TIMEOUT_MAX)
fair enough.
- set the timeout a little differently. Instead
On Thu, 2 Jun 2005, Tom Duffy wrote:
On Thu, 2005-06-02 at 16:18 -0400, James Lentini wrote:
I'd recommend these changes (see attached):
- use the new microsecond conversion function you sent
great
- keep a default DAT timeout constant (I propose DAT_TIMEOUT_MAX)
fair enough.
- set
Libor,
Ignore this patch. I'll submit a different one to change
the ib_ucm device node from /dev/infiniband/cm
to /dev/infiniband_cm. This makes the udev setup
unnecessary.
And, this probably wasn't the right README file for the udev
rules anyway. They belong with the module, not with the
A couple of questions below:
On Thu, 2 Jun 2005, Itamar wrote:
in order to enable kdapltest -T P i needed to remark attr that are
not set by openib gen2
1) ia_attr.max_evd_qlen
2) ia_attr.max_rdma_read_per_ep_in
3) ia_attr.max_rdma_read_per_ep_out
also there was bug in kdapltest memory
On 05/31/05 16:30, Sean Hefty wrote:
Has anyone seen ib_send_cm_req() return -22?
I believe that this is a timeout error, possibly indicating that the
server
side of the connection wasn't running. You may also want to verify the
slid
and dlid are correct for your configuration.
On Thu, Jun 02, 2005 at 11:16:53AM -0400, James Lentini wrote:
This could be an improvement. We just need to be careful that we don't
expose anything transport specific. Off the top of my head, I can
think of one way to do this:
- in dat.h, create a public structure for each object type:
On Thu, 2005-06-02 at 18:09, Jeff Carr wrote:
Is there a simple way to discover the lid values of other systems?
Locally, you can run /usr/local/ib/bin/ibstatus or ibstat.
Remotely, you would need to know the remote GID (subnet prefix + GUID)
and ask the SA what the LID was for that (by getting
Michael Wrote
Patches are located in the SVN tree under
gen2/trunk/src/linux-kernel/patches/backport-to-2.6.9/
infiniband-backport-svn2425-to-2.6.9-kernel-fixups-01.diff
infiniband-backport-svn2425-to-2.6.9-openib-drivers-02.diff
Jeff Carr wrote:
Is there a simple way to discover the lid values of other systems?
Simple? Not really. You could query the SA to obtain a list of path
records to all systems, and then extract the LIDs from those. I use:
cat /sys/class/infiniband/mthca0/ports/1/lid
to get the LIDs of
I'm having intermittent problems with opensm.. It seems after a while
IPoIB stops working and if I restart opensm, it starts spitting out
errors. Do I have a misbehaving switch somewhere?
ibnetdiscover seems to work fine.
(this is from running 'opensm -v -o -r')
On 06/01/05 14:43, William Jordan wrote:
Has anyone seen ib_send_cm_req() return -22?
I'm not sure what you are testing with, Jeff, but I ran into the same
problem the first time I tried to use ucm_simple. The source and
destination lid and guid are embedded in the source, and need to be
On 06/02/05 15:27, Hal Rosenstock wrote:
On Thu, 2005-06-02 at 18:09, Jeff Carr wrote:
Is there a simple way to discover the lid values of other systems?
Locally, you can run /usr/local/ib/bin/ibstatus or ibstat.
Remotely, you would need to know the remote GID (subnet prefix + GUID)
and
On Fri, Jun 03, 2005 at 02:45:19AM +0300, Michael S. Tsirkin wrote:
And PLEASE, if you reply, please delete quoted text you are not responding
to from your reply. I'm getting tired of wading through 5 pages
of quotes to get to a 3 line comment.
Michael,
Could you please at least attempt to
On Thu, Jun 02, 2005 at 07:25:51PM -0500, Troy Benjegerdes wrote:
Some of my problems seem to be from intermittent cables..
Is there anything for OpenIB that can read error counters?
cat /sys/class/infiniband/mthca0/ports/1/counters/*errors
What I'd really like to see is something that I
In my work going through trying to get rid of the opaque dat handles, I
came across what looks like a bug in kdapltest. I don't think
DT_Performance_Test_Create() and DT_Performance_Test_Client() should
take a DAT_IA_HANDLE * as an argument as this would be a pointer to a
pointer.
Of course, the
*/
Could we initialize the ia_attr.max_evd_qlen value correctly in the
provider?
Current openib gen2 code is not reporting the max cq size and i dont think
that we should put
a fix number .
if we want to get the number we need Roland to fill this number
53 matches
Mail list logo
