| Message ID | 20211023164557.7921-1-palok@marvell.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v2,for-rc] RDMA/qedr: qedr crash while running rdma-tool | expand |
On 10/23/21 19:45, Alok Prasad wrote: > This patch fixes crash caused by querying qp. > Also corrects the state of gsi qp. > > Below call trace is generated while using iproute2 utility > "rdma res show -dd qp" on rdma interface. > ========================================================================== > [ 302.569794] BUG: kernel NULL pointer dereference, address: 0000000000000034 > .. > [ 302.570378] Hardware name: Dell Inc. PowerEdge R720/0M1GCR, BIOS 1.2.6 05/10/2012 > [ 302.570500] RIP: 0010:qed_rdma_query_qp+0x33/0x1a0 [qed] > [ 302.570861] RSP: 0018:ffffba560a08f580 EFLAGS: 00010206 > [ 302.570979] RAX: 0000000200000000 RBX: ffffba560a08f5b8 RCX: 0000000000000000 > [ 302.571100] RDX: ffffba560a08f5b8 RSI: 0000000000000000 RDI: ffff9807ee458090 > [ 302.571221] RBP: ffffba560a08f5a0 R08: 0000000000000000 R09: ffff9807890e7048 > [ 302.571342] R10: ffffba560a08f658 R11: 0000000000000000 R12: 0000000000000000 > [ 302.571462] R13: ffff9807ee458090 R14: ffff9807f0afb000 R15: ffffba560a08f7ec > [ 302.571583] FS: 00007fbbf8bfe740(0000) GS:ffff980aafa00000(0000) knlGS:0000000000000000 > [ 302.571729] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 302.571847] CR2: 0000000000000034 CR3: 00000001720ba001 CR4: 00000000000606f0 > [ 302.571968] Call Trace: > [ 302.572083] qedr_query_qp+0x82/0x360 [qedr] > [ 302.572211] ib_query_qp+0x34/0x40 [ib_core] > [ 302.572361] ? ib_query_qp+0x34/0x40 [ib_core] > [ 302.572503] fill_res_qp_entry_query.isra.26+0x47/0x1d0 [ib_core] > [ 302.572670] ? __nla_put+0x20/0x30 > [ 302.572788] ? nla_put+0x33/0x40 > [ 302.572901] fill_res_qp_entry+0xe3/0x120 [ib_core] > [ 302.573058] res_get_common_dumpit+0x3f8/0x5d0 [ib_core] > [ 302.573213] ? fill_res_cm_id_entry+0x1f0/0x1f0 [ib_core] > [ 302.573377] nldev_res_get_qp_dumpit+0x1a/0x20 [ib_core] > [ 302.573529] netlink_dump+0x156/0x2f0 > [ 302.573648] __netlink_dump_start+0x1ab/0x260 > [ 302.573765] rdma_nl_rcv+0x1de/0x330 [ib_core] > [ 302.573918] ? nldev_res_get_cm_id_dumpit+0x20/0x20 [ib_core] > [ 302.574074] netlink_unicast+0x1b8/0x270 > [ 302.574191] netlink_sendmsg+0x33e/0x470 > [ 302.574307] sock_sendmsg+0x63/0x70 > [ 302.574421] __sys_sendto+0x13f/0x180 > [ 302.574536] ? setup_sgl.isra.12+0x70/0xc0 > [ 302.574655] __x64_sys_sendto+0x28/0x30 > [ 302.574769] do_syscall_64+0x3a/0xb0 > [ 302.574884] entry_SYSCALL_64_after_hwframe+0x44/0xae > ========================================================================== > Signed-off-by: Ariel Elior <aelior@marvell.com> > Signed-off-by: Michal Kalderon <mkalderon@marvell.com> > Signed-off-by: Shai Malin <smalin@marvell.com> > Signed-off-by: Alok Prasad <palok@marvell.com> > --- > v2 (from [1]): > - Change description. > - Corrected enum type. > [1] https://patchwork.kernel.org/project/linux-rdma/patch/20210821074339.16614-1-palok@marvell.com/ > --- > drivers/infiniband/hw/qedr/verbs.c | 13 +++++++++---- > 1 file changed, 9 insertions(+), 4 deletions(-) > > diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c > index dcb3653db72d..85baa4f730df 100644 > --- a/drivers/infiniband/hw/qedr/verbs.c > +++ b/drivers/infiniband/hw/qedr/verbs.c > @@ -2744,15 +2744,20 @@ int qedr_query_qp(struct ib_qp *ibqp, > int rc = 0; > > memset(¶ms, 0, sizeof(params)); > + memset(qp_attr, 0, sizeof(*qp_attr)); > + memset(qp_init_attr, 0, sizeof(*qp_init_attr)); > > - rc = dev->ops->rdma_query_qp(dev->rdma_ctx, qp->qed_qp, ¶ms); > + if (qp->qed_qp) I suggest to use "if (qp->qp_type != IB_QPT_GSI)" to match the handling of GSI QPs in the QEDR driver. Thanks, Kamal > + rc = dev->ops->rdma_query_qp(dev->rdma_ctx, > + qp->qed_qp, ¶ms); > if (rc) > goto err; > > - memset(qp_attr, 0, sizeof(*qp_attr)); > - memset(qp_init_attr, 0, sizeof(*qp_init_attr)); > + if (qp->qp_type == IB_QPT_GSI) > + qp_attr->qp_state = qedr_get_ibqp_state(QED_ROCE_QP_STATE_RTS); > + else > + qp_attr->qp_state = qedr_get_ibqp_state(params.state); > > - qp_attr->qp_state = qedr_get_ibqp_state(params.state); > qp_attr->cur_qp_state = qedr_get_ibqp_state(params.state); > qp_attr->path_mtu = ib_mtu_int_to_enum(params.mtu); > qp_attr->path_mig_state = IB_MIG_MIGRATED; >
On Sun, Oct 24, 2021 at 01:46:03PM +0300, Kamal Heib wrote: > > diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c > > index dcb3653db72d..85baa4f730df 100644 > > +++ b/drivers/infiniband/hw/qedr/verbs.c > > @@ -2744,15 +2744,20 @@ int qedr_query_qp(struct ib_qp *ibqp, > > int rc = 0; > > memset(¶ms, 0, sizeof(params)); > > + memset(qp_attr, 0, sizeof(*qp_attr)); > > + memset(qp_init_attr, 0, sizeof(*qp_init_attr)); > > - rc = dev->ops->rdma_query_qp(dev->rdma_ctx, qp->qed_qp, ¶ms); > > + if (qp->qed_qp) > > I suggest to use "if (qp->qp_type != IB_QPT_GSI)" to match the handling of > GSI QPs in the QEDR driver. Alok? Time is closing to get this in before the merge window. Jason
> -----Original Message----- > From: Jason Gunthorpe <jgg@nvidia.com> > Sent: 27 October 2021 22:31 > To: Kamal Heib <kheib@redhat.com> > Cc: Alok Prasad <palok@marvell.com>; Michal Kalderon <mkalderon@marvell.com>; Ariel Elior > <aelior@marvell.com>; linux-rdma@vger.kernel.org; Shai Malin <smalin@marvell.com>; Ariel > Elior <aelior@marvell.com>; alok.prasad7@gmail.com; Michal Kalderon > <mkalderon@marvell.com>; dledford@redhat.com > Subject: [EXT] Re: [v2,for-rc] RDMA/qedr: qedr crash while running rdma-tool > > External Email > > ---------------------------------------------------------------------- > On Sun, Oct 24, 2021 at 01:46:03PM +0300, Kamal Heib wrote: > > > > diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c > > > index dcb3653db72d..85baa4f730df 100644 > > > +++ b/drivers/infiniband/hw/qedr/verbs.c > > > @@ -2744,15 +2744,20 @@ int qedr_query_qp(struct ib_qp *ibqp, > > > int rc = 0; > > > memset(¶ms, 0, sizeof(params)); > > > + memset(qp_attr, 0, sizeof(*qp_attr)); > > > + memset(qp_init_attr, 0, sizeof(*qp_init_attr)); > > > - rc = dev->ops->rdma_query_qp(dev->rdma_ctx, qp->qed_qp, ¶ms); > > > + if (qp->qed_qp) > > > > I suggest to use "if (qp->qp_type != IB_QPT_GSI)" to match the handling of > > GSI QPs in the QEDR driver. > > Alok? Time is closing to get this in before the merge window. > > Jason Done now with v3, Sorry for delaying on this. - Alok
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c index dcb3653db72d..85baa4f730df 100644 --- a/drivers/infiniband/hw/qedr/verbs.c +++ b/drivers/infiniband/hw/qedr/verbs.c @@ -2744,15 +2744,20 @@ int qedr_query_qp(struct ib_qp *ibqp, int rc = 0; memset(¶ms, 0, sizeof(params)); + memset(qp_attr, 0, sizeof(*qp_attr)); + memset(qp_init_attr, 0, sizeof(*qp_init_attr)); - rc = dev->ops->rdma_query_qp(dev->rdma_ctx, qp->qed_qp, ¶ms); + if (qp->qed_qp) + rc = dev->ops->rdma_query_qp(dev->rdma_ctx, + qp->qed_qp, ¶ms); if (rc) goto err; - memset(qp_attr, 0, sizeof(*qp_attr)); - memset(qp_init_attr, 0, sizeof(*qp_init_attr)); + if (qp->qp_type == IB_QPT_GSI) + qp_attr->qp_state = qedr_get_ibqp_state(QED_ROCE_QP_STATE_RTS); + else + qp_attr->qp_state = qedr_get_ibqp_state(params.state); - qp_attr->qp_state = qedr_get_ibqp_state(params.state); qp_attr->cur_qp_state = qedr_get_ibqp_state(params.state); qp_attr->path_mtu = ib_mtu_int_to_enum(params.mtu); qp_attr->path_mig_state = IB_MIG_MIGRATED;