Quoting r. Ralph Campbell <[EMAIL PROTECTED]>:
> Subject: [PATCH] fix minor typo in SDP
>
> This patch fixes a minor misspelling in SDP.
Thanks! Applied.
--
Michael S. Tsirkin
Staff Engineer, Mellanox Technologies
___
openib-general mailing list
openi
commited in r5228
added -EAGAIN scheme to the tx posting and completion code. When the QP is
full and we can't post to it, TX is suspended till there's room.
Signed-off-by: Or Gerlitz <[EMAIL PROTECTED]>
Index: ulp/iser/iser_conn.c
Hi jeff.
just a little obvious thing: It doesn't matter if you are working over gen1 /
gen2 /windows / linux: the behavior will be the same.
> I'm on Windows running gen1 code and after successfully
> sending a number
> Of packets my call to ib_poll_cq is failing with a status of
> IB_WCS_RNR_
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: Re: [PATCH 0/4] SA path record caching
>
> Michael S. Tsirkin wrote:
> >>The cache is updated every 15 minutes, but this value is user configurable
> >>at
> >>module load time. The cache is also updated if a local event occurs, such
> >>as
Sean Hefty wrote:
The CMA will eventually be extended to query the SA directly if the
local look-up fails. The intent is to keep these failures hidden from
the end-user.
I see, it makes sense.
The SA cache does not use a failed lookup as a trigger to update its
cache, but that could be adde
We have run into the following problem: if a task receives a signal
while in the process of e.g. destroying a resource (which could be
because the relevant file was closed) mthca could bail out from trying
to take a command interface semaphore without performing the
appropriate command to tell hard
Convert semaphores to mutexes in mthca. Leave firmware command
interface poll_sem and event_sem as semaphores.
Signed-off-by: Roland Dreier <[EMAIL PROTECTED]>
---
drivers/infiniband/hw/mthca/mthca_cmd.c |6 ++--
drivers/infiniband/hw/mthca/mthca_dev.h |8 --
drivers/infi
Convert srp_host->target_mutex from a semaphore to a mutex.
Signed-off-by: Ingo Molnar <[EMAIL PROTECTED]>
Signed-off-by: Roland Dreier <[EMAIL PROTECTED]>
---
drivers/infiniband/ulp/srp/ib_srp.c | 14 +++---
drivers/infiniband/ulp/srp/ib_srp.h |5 ++---
2 files changed, 9 inserti
There are some cards around that have UAR (user access region) size
different from 8 MB. Relax our sanity check to make sure that the PCI
BAR is big enough to access the UAR size reported by the device
firmware instead.
Signed-off-by: Michael S. Tsirkin <[EMAIL PROTECTED]>
Signed-off-by: Roland D
err, with this fixed:
> + down_interruptible(&dev->mcg_table.sem);
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-gen
Thanks, applied along with the mthca_mcg change:
--- infiniband/hw/mthca/mthca_cmd.c (revision 5216)
+++ infiniband/hw/mthca/mthca_cmd.c (working copy)
@@ -199,8 +199,7 @@ static int mthca_cmd_post(struct mthca_d
{
int err = 0;
- if (down_interruptible(&dev->cmd.hcr_sem))
Arlin Davis wrote:
> Kanevsky, Arkady wrote:
>
>> Arlin,
>> I am not convinced we need a new recv for immediate data.
>> But what is needed is change in normative text in many places.
>> Recv, RDMA Write, DTO completion events, error behavior.
>> Sure you can define immed data in extension but it
Kanevsky, Arkady wrote:
Arlin,
I am not convinced we need a new recv for immediate data.
But what is needed is change in normative text in many places.
Recv, RDMA Write, DTO completion events, error behavior.
Sure you can define immed data in extension but it still effects
behavior of the normat
Here's an updated version that actually does the device checks correctly, and
hopefully clarifies how node_type is formatted. I include two helper functions
to extract the transport and device information from the node_type.
I tried to include related changes to all files with this patch, so a fe
Thanks, applied.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Sean Hefty wrote:
This would be true if node_type = IB_NODE_IB | IB_NODE_CA. Okay, if the
check for IB_NODE_IB is there, I think that all that's needed is:
if (node_type & IB_NODE_SWITCH)
Never mind - this doesn't work in a generic case, since IB_NODE_CA=1 and
IB_NODE_ROUTER = 3.
_
Sean> I understand. This sort of test works since IB_NODE_IB = 0,
Sean> but wouldn't work for IB_NODE_IWARP, where node_type =
Sean> IB_NODE_IWARP | IB_NODE_CA.
Oh, I see. I was thinking that we would have something like
IB_NODE_RNIC = IB_NODE_IWARP | whatever
in the enum (
Thanks, applied.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Michael S. Tsirkin wrote:
The cache is updated every 15 minutes, but this value is user configurable at
module load time. The cache is also updated if a local event occurs, such as
port up/down or SM LID change.
Unfortunately port down on any link in the path has the same effect
but wont inva
Caitlin> For those devices the best support is provided when the
Caitlin> consumer allocates the maximum, any logical resizing to a
Caitlin> smaller size is ignored, and there is *no* extra checking
Caitlin> (let alone a spinlock) for each cq poll.
There is always a spinlock for a
Grant Grundler wrote:
If node_type were a bitmask mask, I'd expect
if (node_type & (IB_NODE_IB | IB_NODE_SWITCH))
This would be true if node_type = IB_NODE_IB | IB_NODE_CA. Okay, if the check
for IB_NODE_IB is there, I think that all that's needed is:
if (node_type & IB_NODE_SWITCH)
Roland Dreier wrote:
That's still silly. I wasn't very clear in my original email. But my
point was that if a function does
if ((device->node_type & IB_NODE_IB) != IB_NODE_IB)
return;
then we should just write tests like
if (device->node_type == IB_NODE_SWITCH
> static inline int ib_poll_cq(struct ib_cq *cq, int num_entries,
> struct ib_wc *wc)
> {
> unsigned long flags;
> int ret;
> spin_lock_irqsave(cq->lock, flags);
> ret = cq->poll_cq(cq, num_entries, wc);
> spin_unlock_irqrestore(cq->l
Michael> Might be a good idea to test err == -EAGAIN first: err is
Michael> likely a register, and once you get a valid CQE returning
Michael> it to ULP ASAP has direct latency impact.
Makes sense. The x86_64 asm looks slightly better at least, so I put
this change in my tree.
Mi
[EMAIL PROTECTED] wrote:
> Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
>> I'm not sure I got it right but it looks OK to me.
>>
>> I don't really see any other way to do this, because the RESIZE_CQ
>> command can block, so we can't lock out poll CQ operations while the
>> firmware command is exe
On Mon, Jan 30, 2006 at 01:37:39PM -0800, Sean Hefty wrote:
> As long as IB_NODE_IB remains 0, we can eliminate this change.
> Or we could go with something like:
>
> if (device->node_type == IB_NODE_IB | IB_NODE_SWITCH)
If node_type were a bitmask mask, I'd expect
if (node_type & (IB_NOD
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> I'm not sure I got it right but it looks OK to me.
>
> I don't really see any other way to do this, because the RESIZE_CQ
> command can block, so we can't lock out poll CQ operations while the
> firmware command is executing.
I thought about this so
Sean> As long as IB_NODE_IB remains 0, we can eliminate this
Sean> change. Or we could go with something like:
Sean> if (device->node_type == IB_NODE_IB | IB_NODE_SWITCH)
That's still silly. I wasn't very clear in my original email. But my
point was that if a function does
Roland Dreier wrote:
The changes like this in ipoib, srp, sa_query, etc:
> - if (device->node_type == IB_NODE_SWITCH) {
> + if ((device->node_type & IB_NODE_SWITCH) == IB_NODE_SWITCH) {
seem rather silly. We've already checked that the device is an IB
device, so the change is pure obfuscatio
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: Re: Re: [PATCH 0/4] SA path record caching
>
> Michael S. Tsirkin wrote:
> > When is the cache invalidated?
>
> The cache is updated every 15 minutes, but this value is user configurable at
> module load time. The cache is also updated if a
Seems OK.
The changes like this in ipoib, srp, sa_query, etc:
> -if (device->node_type == IB_NODE_SWITCH) {
> +if ((device->node_type & IB_NODE_SWITCH) == IB_NODE_SWITCH) {
seem rather silly. We've already checked that the device is an IB
device, so the change is pure obfuscation.
Al
The following patch extends the node_type associated with an ib_device
to support iWarp without breaking the ABI. Comments on this approach?
Signed-off-by: Sean Hefty <[EMAIL PROTECTED]>
---
Index: ulp/ipoib/ipoib_main.c
===
--- ul
Hi There -- I'm making progress on my IB transport system, but ran into
A bit of a snag and I can't seem to find an answer in the documentation,
Sample apps, or google.
I'm on Windows running gen1 code and after successfully sending a number
Of packets my call to ib_poll_cq is failing with a statu
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: [PATCH 0/4] SA path record caching
>
> The following patch series adds caching of path records with the local system.
> I divided the changes up into 4 patches to make the review easier. The
> patches
> are arranged as follows:
>
> 1. Add a
Michael S. Tsirkin wrote:
When is the cache invalidated?
The cache is updated every 15 minutes, but this value is user configurable at
module load time. The cache is also updated if a local event occurs, such as
port up/down or SM LID change.
In the presence of failures in the network,
an
Or Gerlitz wrote:
I see. The CMA code (cma_resolve_ib_route) just calls ib_get_path_rec
and if the latter fails returns error to its consumer. Two questions
come into my mind here: a) is it the final version or you plan to extend
the CMA to directly query the SA in that case? b) the code of
ib
[EMAIL PROTECTED] wrote:
>
> I think I see a problem with this approach: if a ULP performs
> CQ poll while mthca_RESIZE_CQ is in progress, it might get a
> false indication that the CQ is empty since CQEs are being
> written to the new buffer already.
>
> As a result e.g. a ULP that does:
>
> a
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> Subject: Re: [PATCH 2/6] [RFC] mthca kernel changes for resizeCQ
>
> Michael> I think I see a problem with this approach: if a ULP
> Michael> performs CQ poll while mthca_RESIZE_CQ is in progress, it
> Michael> might get a false indicatio
Michael> I think I see a problem with this approach: if a ULP
Michael> performs CQ poll while mthca_RESIZE_CQ is in progress, it
Michael> might get a false indication that the CQ is empty since
Michael> CQEs are being written to the new buffer already.
I tried to handle this in the
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> +static int mthca_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata
> *udata)
> +{
> + struct mthca_dev *dev = to_mdev(ibcq->device);
> + struct mthca_cq *cq = to_mcq(ibcq);
> + struct mthca_resize_cq ucmd;
> + u32 lkey;
> +
And here's a simple test kernel module I wrote to test resizing kernel
CQs. It is similar to the userspace test program, and should produce
endless output like
Resizing to 800... resized to 1023
2832800 writes done
when loaded.
If either the "Resizing" or "xxx writes done" lines s
Here's a simple test program I wrote that just creates a loopback QP
on the first port of the first HCA it finds, and tries to keep 100
RDMA writes queued all the time. In another thread, it resizes the CQ
at random intervals. The first thread makes sure that all the
completions expected do arriv
libmthca implementation of resizing CQs. This is the real guts of
it -- there is some slightly tricky code to handle the transition from
the old CQ buffer to the new one.
---
Index: libmthca/src/verbs.c
===
--- libmthca/src/verbs.c
libibverbs changes to handle resizing CQs. Essentially just adding
API and support for passing the call through to provider plug-ins.
---
--- libibverbs/include/infiniband/driver.h (revision 5201)
+++ libibverbs/include/infiniband/driver.h (working copy)
@@ -1,6 +1,6 @@
/*
* Copyrig
Add low-level driver support for resizing CQs (both kernel and
userspace) to mthca.
---
--- infiniband/hw/mthca/mthca_user.h(revision 5213)
+++ infiniband/hw/mthca/mthca_user.h(working copy)
@@ -1,6 +1,6 @@
/*
* Copyright (c) 2005 Topspin Communications. All rights reserved.
- * Copyr
Add support to uverbs to handle resizing userspace CQs (completion
queues), including adding an ABI for marshalling requests and
responses. The kernel midlayer already has ib_resize_cq().
---
--- infiniband/include/rdma/ib_user_verbs.h (revision 5213)
+++ infiniband/include/rdma/ib_user_verb
Here is a series of patches that adds support for resize CQ operations
to both the core uverbs/libibverbs code as well as implementing it in
the device-specific mthca/libmthca code.
This is a provider ABI breaking change to libibverbs, because it
changes the layout of struct ibv_context_ops. Sour
Fix typo in overflow test in libmthca.
Noted by Yossi Leybovich.
Signed-off-by: Michael S. Tsirkin <[EMAIL PROTECTED]>
Index: openib/src/userspace/libmthca/src/qp.c
===
--- openib/src/userspace/libmthca/src/qp.c (revision 5134)
Dotan> i can't see the code that handles this event in the kernel
Dotan> part of the mthca.
Dotan> we wrote a test that tries to create this event in user
Dotan> level, but the modify QP from RTS to SQD with this flag
Dotan> (IBV_QP_EN_SQD_ASYNC_NOTIFY) enabled fails.
You're r
On Mon, 2006-01-30 at 09:53, Michael S. Tsirkin wrote:
> Quoting r. Eitan Zahavi <[EMAIL PROTECTED]>:
> > Subject: [PATCH] osm: support osm_svn_revision.h in case ofSVN export
> >
> > Hi Hal
> >
> > We are using SVN export when building standalone OpenSM packages.
> > During the SVN export we ove
>
> Are you talking about the IB_SQ_DRAINED event? If so the code is
> there but I never tested it.
>
Yes, this is the event i'm talking about.
i can't see the code that handles this event in the kernel part of the mthca.
we wrote a test that tries to create this event in user level, but the
Dotan> Hi. does the mthca driver support the request of an async
Dotan> event when changing the QP state from RTS -> SQD? (the
Dotan> event is actually a notification when the HCA changes the
Dotan> QP state from SQD:draining -> SQD:drained)
Are you talking about the IB_SQ_DRAINE
Quoting r. Eitan Zahavi <[EMAIL PROTECTED]>:
> Subject: [PATCH] osm: support osm_svn_revision.h in case ofSVN export
>
> Hi Hal
>
> We are using SVN export when building standalone OpenSM packages.
> During the SVN export we overwrite the osm_svn_revision.h with the
> SVN version used for the ex
test
-Original Message-
From: [EMAIL PROTECTED] on behalf of Michael S. Tsirkin
Sent: Mon 1/30/2006 4:53 PM
To: Hal Rosenstock
Cc: OPENIB
Subject:[openib-general] Re: [PATCH] osm: support osm_svn_revision.h in
case ofSVN export
Quoting r. Eitan Zahavi <[EMAIL PROTECT
Hi Amith,
On Sat, 2006-01-28 at 16:18, amith rajith mamidala wrote:
> Hi Hal,
>
> There is only one application running on a node. I am running opensm on
> a different node. I am also listing the other processes I observed on
> doing a "ps":
>
> root 356411 0 Jan26 ?00:00:00 [i
IBGold 1.8.0 does include access layer supplied with it.
Our FAE will contact you shortly to explain how to use the access layer
of gen1.
Tziporet
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Bub Thomas
Sent: Monday, January 30, 2006 2:30 PM
To: openib
Tziporet
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Michael S.
Tsirkin
Sent: Monday, January 23, 2006 7:17 PM
To: openib-general@openib.org
Subject: [openib-general] ANNOUNCE: mstflint update
Hi!
I have updated mstflint tool with code from mellanox MFT 1.0.1
On Mon, 2006-01-30 at 08:23, Eitan Zahavi wrote:
> Hi Hal
>
> We are using SVN export when building standalone OpenSM packages.
> During the SVN export we overwrite the osm_svn_revision.h with the
> SVN version used for the export.
>
> However the makefile override that. This patch avoids this b
Sean Hefty wrote:
The implementation is complete. The interface to the cache operates
synchronously. If an item is found in the cache, it is returned. If no
item is found, an error is returned. The caller can query the SA
directly in this case.(If we wanted to be fancy, the results of t
Hi Hal
We are using SVN export when building standalone OpenSM packages.
During the SVN export we overwrite the osm_svn_revision.h with the
SVN version used for the export.
However the makefile override that. This patch avoids this by checking
if the svnversion is "exported".
Eitan
Signed-off
Quoting r. Yael Kalka <[EMAIL PROTECTED]>:
> ===
> --- include/opensm/osm_svn_revision.h (revision 5203)
> +++ include/opensm/osm_svn_revision.h (working copy)
> @@ -1 +1 @@
> -#define OSM_SVN_REVISION ""
> +#define OSM_SVN_REVISION "5
Hi Yael,
On Mon, 2006-01-30 at 06:18, Yael Kalka wrote:
> Hi Hal,
>
> The following patch adds the use of the OSM_DEFAULT_CACHE_DIR instead
> of full name usage. Also - add the "/" at the end of the
> OSM_DEFAULT_CACHE_DIR definition, and refrain from adding it in the
> code, to avoid problems in
Title: gen1 drivers as rpm or tgz?
Hi,
we are planning to replace a HighSpeed GSN (Gigabyte System Network) data interface with IB.
We’d like to to use the verbs layer with the Access Layer on top.
Unfortunately Mellanox, our Hardware provider, does not supply an Access layer with their IB
Hi Hal,
The following patch adds the use of the OSM_DEFAULT_CACHE_DIR instead
of full name usage. Also - add the "/" at the end of the
OSM_DEFAULT_CACHE_DIR definition, and refrain from adding it in the
code, to avoid problems in Windows.
Thanks,
Yael
Signed-off-by: Yael Kalka <[EMAIL PROTECT
Hi,
I don't know when exactly the i_sem to i_mutex conversion was done. In
2.6.16 this patch is needed to build the ipath driver.
Signed-off-by: Bernhard Fischer <[EMAIL PROTECTED]>
Please apply.
Index: infiniband/hw/ipath/ipath_driver.c
Quoting r. Bernhard Fischer <[EMAIL PROTECTED]>:
> Add note about linking the rdma includes and linux/*h to the kernel
> sources.
>
> Signed-off-by: Bernhard Fischer <[EMAIL PROTECTED]>
>
> Please apply if appropriate,
> Bernhard
This has been on wiki for a while now
https://openib.org/tiki/tik
On Fri, Dec 23, 2005 at 07:31:21AM -0500, Hal Rosenstock wrote:
>On Fri, 2005-12-23 at 07:25, Devesh Sharma wrote:
>> Hi I have downloaded openib stack from svn with Revision 4595 and
>> trying to compile it with 2.6.14.4. I am getting following errors
>>
>>
>> [EMAIL PROTECTED] linux-2.6.14.4]#
Title: does the mthca driver support RTS->SQD event request?
Hi.
does the mthca driver support the request of an async event when changing the QP state from RTS -> SQD?
(the event is actually a notification when the HCA changes the QP state from SQD:draining -> SQD:drained)
if the answe
69 matches
Mail list logo