[PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

[PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

[Steve Wise]

I forgot to add the stable Cc tag when I submitted this, but it is important to
get this patch into 4.9.x:

commit b414fa01c31318383ae29d9d23cb9ca4184bbd86
Author: Steve Wise <swise@opengridcomputing.com>
Date:   Thu Dec 15 08:09:35 2016 -0800

    iw_cxgb4: set correct FetchBurstMax for QPs

    The current QP FetchBurstMax value is 256B, which
    is incorrect since a WR can exceed that value.  The
    result being a partial WR fetched by hardware, and
    a fatal "bad WR" error posted by the SGE.

    So bump the FetchBurstMax to 512B.

    Signed-off-by: Steve Wise <swise@opengridcomputing.com>
    Signed-off-by: Doug Ledford <dledford@redhat.com>

I'm not sure if this is the correct way to request this, but I think I'm
following _option_2 from Documentation/stable_kernel_rules.txt.


Thanks,

Steve.

Re: [PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

[Greg KH]

On Fri, Feb 03, 2017 at 11:04:13AM -0600, Steve Wise wrote:
> I forgot to add the stable Cc tag when I submitted this, but it is important to
> get this patch into 4.9.x:
> 
> commit b414fa01c31318383ae29d9d23cb9ca4184bbd86
> Author: Steve Wise <swise@opengridcomputing.com>
> Date:   Thu Dec 15 08:09:35 2016 -0800
> 
>     iw_cxgb4: set correct FetchBurstMax for QPs
> 
>     The current QP FetchBurstMax value is 256B, which
>     is incorrect since a WR can exceed that value.  The
>     result being a partial WR fetched by hardware, and
>     a fatal "bad WR" error posted by the SGE.
> 
>     So bump the FetchBurstMax to 512B.
> 
>     Signed-off-by: Steve Wise <swise@opengridcomputing.com>
>     Signed-off-by: Doug Ledford <dledford@redhat.com>
> 
> I'm not sure if this is the correct way to request this, but I think I'm
> following _option_2 from Documentation/stable_kernel_rules.txt.

Now queued up.  Are you sure this isn't also needed in the 4.4-stable
tree?

thanks,

greg k-h

RE: [PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

[Steve Wise]

> > commit b414fa01c31318383ae29d9d23cb9ca4184bbd86
> > Author: Steve Wise <swise@opengridcomputing.com>
> > Date:   Thu Dec 15 08:09:35 2016 -0800
> >
> >     iw_cxgb4: set correct FetchBurstMax for QPs
> >
> >     The current QP FetchBurstMax value is 256B, which
> >     is incorrect since a WR can exceed that value.  The
> >     result being a partial WR fetched by hardware, and
> >     a fatal "bad WR" error posted by the SGE.
> >
> >     So bump the FetchBurstMax to 512B.
> >
> >     Signed-off-by: Steve Wise <swise@opengridcomputing.com>
> >     Signed-off-by: Doug Ledford <dledford@redhat.com>
> >
> > I'm not sure if this is the correct way to request this, but I think I'm
> > following _option_2 from Documentation/stable_kernel_rules.txt.
> 
> Now queued up.  Are you sure this isn't also needed in the 4.4-stable
> tree?
>

Hey Greg,  At first, I didn't think it is needed because, while the bug is
indeed present in that code, it is only exposed (and causes problems) with
the new T6 hardware.   But now I think I'd like the fix pulled in 4.4-stable
to avoid any issues if T6 support is pulled into that base.

Thanks!

Steve
 

Re: [PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

['Greg KH']

On Sun, Feb 05, 2017 at 07:41:45AM -0600, Steve Wise wrote:
> > > commit b414fa01c31318383ae29d9d23cb9ca4184bbd86
> > > Author: Steve Wise <swise@opengridcomputing.com>
> > > Date:   Thu Dec 15 08:09:35 2016 -0800
> > >
> > >     iw_cxgb4: set correct FetchBurstMax for QPs
> > >
> > >     The current QP FetchBurstMax value is 256B, which
> > >     is incorrect since a WR can exceed that value.  The
> > >     result being a partial WR fetched by hardware, and
> > >     a fatal "bad WR" error posted by the SGE.
> > >
> > >     So bump the FetchBurstMax to 512B.
> > >
> > >     Signed-off-by: Steve Wise <swise@opengridcomputing.com>
> > >     Signed-off-by: Doug Ledford <dledford@redhat.com>
> > >
> > > I'm not sure if this is the correct way to request this, but I think I'm
> > > following _option_2 from Documentation/stable_kernel_rules.txt.
> > 
> > Now queued up.  Are you sure this isn't also needed in the 4.4-stable
> > tree?
> >
> 
> Hey Greg,  At first, I didn't think it is needed because, while the bug is
> indeed present in that code, it is only exposed (and causes problems) with
> the new T6 hardware.   But now I think I'd like the fix pulled in 4.4-stable
> to avoid any issues if T6 support is pulled into that base.

How would I pull "T6" support into that kernel?  I'm guessing that's a
much larger issue and isn't something that would normally happen in a
stable kernel release?

thanks,

greg k-h

Re: [PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

[Steve Wise]

>>> Now queued up.  Are you sure this isn't also needed in the 4.4-stable
>>> tree?
>>>
>> Hey Greg,  At first, I didn't think it is needed because, while the bug is
>> indeed present in that code, it is only exposed (and causes problems) with
>> the new T6 hardware.   But now I think I'd like the fix pulled in 4.4-stable
>> to avoid any issues if T6 support is pulled into that base.
> How would I pull "T6" support into that kernel?  I'm guessing that's a
> much larger issue and isn't something that would normally happen in a
> stable kernel release?
>

I guess that won't happen.  To enable the T6 adapter several commits to 
drivers/net/ethernet/chelsio/cxgb4 and drivers/infiniband/hw/cxgb4 would 
have to flow into stable, and I guess that will not happen.    Ok, so no 
4.4-stable.

Thanks,

Steve.

Re: [PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

[Doug Ledford]

[-- Attachment #1.1: Type: text/plain, Size: 592 bytes --]

On 12/15/2016 11:09 AM, Steve Wise wrote:
> The current QP FetchBurstMax value is 256B, which
> is incorrect since a WR can exceed that value.  The
> result being a partial WR fetched by hardware, and
> a fatal "bad WR" error posted by the SGE.
> 
> So bump the FetchBurstMax to 512B.
> 
> Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

Thanks, queued for -rc.

-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
    GPG Key ID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]

[PATCH] iw_cxgb4: set correct FetchBurstMax for QPs

[Steve Wise]

The current QP FetchBurstMax value is 256B, which
is incorrect since a WR can exceed that value.  The
result being a partial WR fetched by hardware, and
a fatal "bad WR" error posted by the SGE.

So bump the FetchBurstMax to 512B.

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/qp.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c
index b7ac97b..cda5542 100644
--- a/drivers/infiniband/hw/cxgb4/qp.c
+++ b/drivers/infiniband/hw/cxgb4/qp.c
@@ -321,7 +321,8 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq,
 		FW_RI_RES_WR_DCAEN_V(0) |
 		FW_RI_RES_WR_DCACPU_V(0) |
 		FW_RI_RES_WR_FBMIN_V(2) |
-		FW_RI_RES_WR_FBMAX_V(2) |
+		(t4_sq_onchip(&wq->sq) ? FW_RI_RES_WR_FBMAX_V(2) :
+					 FW_RI_RES_WR_FBMAX_V(3)) |
 		FW_RI_RES_WR_CIDXFTHRESHO_V(0) |
 		FW_RI_RES_WR_CIDXFTHRESH_V(0) |
 		FW_RI_RES_WR_EQSIZE_V(eqsize));
@@ -345,7 +346,7 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq,
 		FW_RI_RES_WR_DCAEN_V(0) |
 		FW_RI_RES_WR_DCACPU_V(0) |
 		FW_RI_RES_WR_FBMIN_V(2) |
-		FW_RI_RES_WR_FBMAX_V(2) |
+		FW_RI_RES_WR_FBMAX_V(3) |
 		FW_RI_RES_WR_CIDXFTHRESHO_V(0) |
 		FW_RI_RES_WR_CIDXFTHRESH_V(0) |
 		FW_RI_RES_WR_EQSIZE_V(eqsize));
-- 
2.7.0

--
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