On Thu, Dec 10, 2020 at 04:00:16AM +0000, liweihang wrote:
> On 2020/12/10 5:09, Jason Gunthorpe wrote:
> > On Fri, Dec 04, 2020 at 06:40:30PM +0800, Weihang Li wrote:
> >> According to the RoCE v1 specification, the sl (service level) 0-7 are
> >> mapped directly to priorities 0-7 respectively, sl 8-15 are reserved. The
> >> driver should verify whether the value of sl is larger than 7, if so, an
> >> exception should be returned.
> >>
> >> Fixes: 172505cfa3a8 ("RDMA/hns: Add check for the validity of sl configuration")
> >> Fixes: d6a3627e311c ("RDMA/hns: Optimize wqe buffer set flow for post send")
> >> Signed-off-by: Weihang Li <liweihang@xxxxxxxxxx>
> >> drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 10 +++++-----
> >> 1 file changed, 5 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
> >> index 7a0c1ab..15e1313 100644
> >> +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
> >> @@ -433,6 +433,10 @@ static int fill_ud_av(struct hns_roce_v2_ud_send_wqe *ud_sq_wqe,
> >> V2_UD_SEND_WQE_BYTE_36_TCLASS_S, ah->av.tclass);
> >> roce_set_field(ud_sq_wqe->byte_40, V2_UD_SEND_WQE_BYTE_40_FLOW_LABEL_M,
> >> V2_UD_SEND_WQE_BYTE_40_FLOW_LABEL_S, ah->av.flowlabel);
> >> +
> >> + if (WARN_ON(ah->av.sl > MAX_SERVICE_LEVEL))
> >> + return -EINVAL;
> >> +
> >> roce_set_field(ud_sq_wqe->byte_40, V2_UD_SEND_WQE_BYTE_40_SL_M,
> >> V2_UD_SEND_WQE_BYTE_40_SL_S, ah->av.sl);
> >>
> >> @@ -4609,12 +4613,8 @@ static int hns_roce_v2_set_path(struct ib_qp *ibqp,
> >> memset(qpc_mask->dgid, 0, sizeof(grh->dgid.raw));
> >>
> >> hr_qp->sl = rdma_ah_get_sl(&attr->ah_attr);
> >> - if (unlikely(hr_qp->sl > MAX_SERVICE_LEVEL)) {
> >> - ibdev_err(ibdev,
> >> - "failed to fill QPC, sl (%d) shouldn't be larger than %d.\n",
> >> - hr_qp->sl, MAX_SERVICE_LEVEL);
> >> + if (WARN_ON(hr_qp->sl > MAX_SERVICE_LEVEL))
> >> return -EINVAL;
> >> - }
> >>
> >> roce_set_field(context->byte_28_at_fl, V2_QPC_BYTE_28_SL_M,
> >> V2_QPC_BYTE_28_SL_S, hr_qp->sl);
> >
> > Can any of these warn_on's be triggered by user space? That would not
> > be OK
> >
> > Jason
> >
>
> Hi Jason,
>
> Thanks for your comments, I understand that error that can be triggered by
> userspace shouldn't use WARN_ON(). So I shouldn't use WARN_ON() in
> hns_roce_v2_set_path().
>
> As for the error in process of post_send, you suggested me to warn_on if
> a kernel user try to pass in an illegal opcode. So I guess I should use
> WARN_ON() too in sl's check when filling a UD WQE. Am I right?
Userspace should not be able to trigger warn_on
Bad kernel ULPs are OK to trigger warn_on
Jason
