[PATCH 2.6.36] RDMA/cxgb3: Don't exceed the max HW CQ depth.

[PATCH 2.6.36] RDMA/cxgb3: Don't exceed the max HW CQ depth.

[Steve Wise]

The max depth supported by T3 is 64K entries.

This bug will cause stalls and possibly crashes in large MPI clusters.

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
---

 drivers/infiniband/hw/cxgb3/cxio_hal.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.h b/drivers/infiniband/hw/cxgb3/cxio_hal.h
index 8f0caf7..78fbe9f 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.h
@@ -53,7 +53,7 @@
 #define T3_MAX_PBL_SIZE 256
 #define T3_MAX_RQ_SIZE 1024
 #define T3_MAX_QP_DEPTH (T3_MAX_RQ_SIZE-1)
-#define T3_MAX_CQ_DEPTH 262144
+#define T3_MAX_CQ_DEPTH 65536
 #define T3_MAX_NUM_STAG (1<<15)
 #define T3_MAX_MR_SIZE 0x100000000ULL
 #define T3_PAGESIZE_MASK 0xffff000  /* 4KB-128MB */

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2.6.36] RDMA/cxgb3: Don't exceed the max HW CQ depth.

[Steve Wise]

  On 8/28/2010 8:35 AM, Steve Wise wrote:
> The max depth supported by T3 is 64K entries.
>
> This bug will cause stalls and possibly crashes in large MPI clusters.
>
> Signed-off-by: Steve Wise<swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
> ---

I would also like to get this into 2.6.35 stable.  What's the process 
for that?

Steve.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2.6.36] RDMA/cxgb3: Don't exceed the max HW CQ depth.

[Bart Van Assche]

On Sat, Aug 28, 2010 at 3:40 PM, Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+v7I7tHvgBF7@public.gmane.orgm> wrote:
>
>  On 8/28/2010 8:35 AM, Steve Wise wrote:
>>
>> The max depth supported by T3 is 64K entries.
>>
>> This bug will cause stalls and possibly crashes in large MPI clusters.
>>
>> Signed-off-by: Steve Wise<swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
>> ---
>
> I would also like to get this into 2.6.35 stable.  What's the process for that?

The easiest way is to add "CC: <stable-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>" when submitting a
patch. An alternative is after the patch has been accepted in Linus'
tree to submit it to stable-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org with a reference to the commit
ID in Linus' tree. More information can be found in
Documentation/stable_kernel_rules.txt.

Bart.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2.6.36] RDMA/cxgb3: Don't exceed the max HW CQ depth.

[Roland Dreier]

 > The easiest way is to add "CC: <stable-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>" when submitting a
 > patch. An alternative is after the patch has been accepted in Linus'
 > tree to submit it to stable-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org with a reference to the commit
 > ID in Linus' tree. More information can be found in
 > Documentation/stable_kernel_rules.txt.

Yep, this is exactly right.  In this case I'll add the cc to the commit
message so you don't need to do anything more.

 - R.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2.6.36] RDMA/cxgb3: Don't exceed the max HW CQ depth.

[Roland Dreier]

thanks, applied with cc to stable added.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html