On Mon, Nov 27, 2017 at 11:37:55AM -0700, Jason Gunthorpe wrote:
> On Sun, Nov 26, 2017 at 08:23:54PM +0200, Leon Romanovsky wrote:
> > From: Moni Shoua <monis@xxxxxxxxxxxx>
> >
> > This patch limits the initial value for PSN to 24 bits as
> > spec requires.
>
> Since this is for-rc, what is the user visible bug this corrects?
>
> I think it is because the QP setup can fail if reserved bits in the
> CMA message happen to be set?
It won't fail, but will print nasty warning down the road after commit
from Majd "IB/core: Make sure that PSN does not overflow" will be
accepted.
The flow is PSN ->
seq_num ->
rq_psn ->
rdma_init_qp_attr() ->
ib_modify_qp ->
ib_modify_qp_with_udata ->
1334 if (attr_mask & IB_QP_RQ_PSN && attr->rq_psn & ~0xffffff) {
1335 pr_warn("%s: %s rq_psn overflow, masking to 24 bits\n",
1336 __func__, qp->device->name);
1337 attr->rq_psn &= 0xffffff;
1338 }
Thanks
>
> Jason
Attachment:
signature.asc
Description: PGP signature
