[PATCH for-next 00/16] IB/hfi1, qib, rdmavt: patches for next 08/21/2017

[PATCH for-next 00/16] IB/hfi1, qib, rdmavt: patches for next 08/21/2017

[Dennis Dalessandro]

Hi Doug,

Another week another set of patches. I meant to send on Friday but was waiting
for 0-day builds to complete. I still haven't gotten any notification, the
service may be down but I wanted to get these on the list anyway for review.
If I get any bad result I will reply but there is nothing particularly scary
in here.

Content wise this is more clean ups including a patch for rdmavt to use
the put_swqe/clear_mr_ref helpers that support was added for in qib/hfi1
previously. 

We have some minor bug fixes as well. A large portion of the patch set is
a clean up/refactoring exercise by Harish that I took because it makes the code
cleaner and easier to read.

Note, one of the fixes from Mike is marked stable. I don't know if it's something
that you want to take into this late of an RC so it may be best to just go to
next and flow through the stable process.

Patches can can also be found in my GitHub repo at:
https://github.com/ddalessa/kernel/tree/for-4.14

---

Grzegorz Morys (1):
      IB/hfi1: Ratelimit prints from sdma_interrupt

Harish Chegondi (7):
      IB/hfi1: Clean up hfi1_user_exp_rcv_setup function
      IB/hfi1: Clean up user_sdma_send_pkts() function
      IB/hfi1: Clean up pin_vector_pages() function
      IB/hfi1: Fix the bail out code in pin_vector_pages() function
      IB/hfi1: Remove duplicate definitions of num_user_pages() function
      IB/hfi1: Move structure definitions from user_exp_rcv.c to user_exp_rcv.h
      IB/hfi1: Move structure and MACRO definitions in user_sdma.c to user_sdma.h

Kaike Wan (1):
      IB/hfi1: Add received request info to qp_stats

Kamenee Arumugam (3):
      IB/qib: Remove unnecessary memory allocation for boardname
      IB/qib: Stricter bounds checking for copy and array access
      IB/hfi1: Fix whitespace alignment issue for MAD

Michael J. Ruhl (1):
      IB/hfi1: Improve local kmem_cache_alloc performance

Mike Marciniszyn (3):
      IB/rdmavt: Use rvt_put_swqe() in rvt_clear_mr_ref()
      IB/{qib,hfi1}: Avoid flow control testing for RDMA write operation
      IB/hfi1: Add opcode states to qp_stats


 drivers/infiniband/hw/hfi1/chip.c         |    4 
 drivers/infiniband/hw/hfi1/file_ops.c     |    2 
 drivers/infiniband/hw/hfi1/hfi.h          |    6 
 drivers/infiniband/hw/hfi1/mad.c          |    2 
 drivers/infiniband/hw/hfi1/qp.c           |   13 +
 drivers/infiniband/hw/hfi1/rc.c           |    3 
 drivers/infiniband/hw/hfi1/user_exp_rcv.c |  255 ++++++++++-------
 drivers/infiniband/hw/hfi1/user_exp_rcv.h |   34 ++
 drivers/infiniband/hw/hfi1/user_sdma.c    |  434 +++++++++--------------------
 drivers/infiniband/hw/hfi1/user_sdma.h    |  166 +++++++++++
 drivers/infiniband/hw/hfi1/verbs_txreq.c  |   11 -
 drivers/infiniband/hw/hfi1/vnic_sdma.c    |   13 -
 drivers/infiniband/hw/qib/qib_iba6120.c   |   20 -
 drivers/infiniband/hw/qib/qib_iba7220.c   |   22 +
 drivers/infiniband/hw/qib/qib_iba7322.c   |   40 +--
 drivers/infiniband/hw/qib/qib_init.c      |    1 
 drivers/infiniband/hw/qib/qib_mad.c       |    4 
 drivers/infiniband/hw/qib/qib_rc.c        |    4 
 drivers/infiniband/hw/qib/qib_ruc.c       |    6 
 drivers/infiniband/sw/rdmavt/qp.c         |    6 
 20 files changed, 556 insertions(+), 490 deletions(-)

--
-Denny

[PATCH for-next 02/16] IB/{qib, hfi1}: Avoid flow control testing for RDMA write operation

[Dennis Dalessandro]

From: Mike Marciniszyn <mike.marciniszyn@intel.com>

Section 9.7.7.2.5 of the 1.3 IBTA spec clearly says that receive
credits should never apply to RDMA write.

qib and hfi1 were doing that.  The following situation will result
in a QP hang:
- A prior SEND or RDMA_WRITE with immmediate consumed the last
  credit for a QP using RC receive buffer credits
- The prior op is acked so there are no more acks
- The peer ULP fails to post receive for some reason
- An RDMA write sees that the credits are exhausted and waits
- The peer ULP posts receive buffers
- The ULP posts a send or RDMA write that will be hung

The fix is to avoid the credit test for the RDMA write operation.

Cc: <stable@vger.kernel.org>
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
---
 drivers/infiniband/hw/hfi1/rc.c    |    3 ++-
 drivers/infiniband/hw/qib/qib_rc.c |    4 ++--
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c
index 99defcc..e1cf0c0 100644
--- a/drivers/infiniband/hw/hfi1/rc.c
+++ b/drivers/infiniband/hw/hfi1/rc.c
@@ -442,7 +442,7 @@ int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 		case IB_WR_RDMA_WRITE:
 			if (newreq && !(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 				qp->s_lsn++;
-			/* FALLTHROUGH */
+			goto no_flow_control;
 		case IB_WR_RDMA_WRITE_WITH_IMM:
 			/* If no credit, return. */
 			if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT) &&
@@ -450,6 +450,7 @@ int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 				qp->s_flags |= RVT_S_WAIT_SSN_CREDIT;
 				goto bail;
 			}
+no_flow_control:
 			put_ib_reth_vaddr(
 				wqe->rdma_wr.remote_addr,
 				&ohdr->u.rc.reth);
diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
index 4ddbcac..e9a9173 100644
--- a/drivers/infiniband/hw/qib/qib_rc.c
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -348,7 +348,7 @@ int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags)
 		case IB_WR_RDMA_WRITE:
 			if (newreq && !(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 				qp->s_lsn++;
-			/* FALLTHROUGH */
+			goto no_flow_control;
 		case IB_WR_RDMA_WRITE_WITH_IMM:
 			/* If no credit, return. */
 			if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT) &&
@@ -356,7 +356,7 @@ int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags)
 				qp->s_flags |= RVT_S_WAIT_SSN_CREDIT;
 				goto bail;
 			}
-
+no_flow_control:
 			ohdr->u.rc.reth.vaddr =
 				cpu_to_be64(wqe->rdma_wr.remote_addr);
 			ohdr->u.rc.reth.rkey =

Re: [PATCH for-next 00/16] IB/hfi1, qib, rdmavt: patches for next 08/21/2017

[Doug Ledford]

On Mon, 2017-08-21 at 18:26 -0700, Dennis Dalessandro wrote:
> Hi Doug,
> 
> Another week another set of patches. I meant to send on Friday but
> was waiting
> for 0-day builds to complete. I still haven't gotten any
> notification, the
> service may be down but I wanted to get these on the list anyway for
> review.
> If I get any bad result I will reply but there is nothing
> particularly scary
> in here.
> 
> Content wise this is more clean ups including a patch for rdmavt to
> use
> the put_swqe/clear_mr_ref helpers that support was added for in
> qib/hfi1
> previously. 
> 
> We have some minor bug fixes as well. A large portion of the patch
> set is
> a clean up/refactoring exercise by Harish that I took because it
> makes the code
> cleaner and easier to read.
> 
> Note, one of the fixes from Mike is marked stable. I don't know if
> it's something
> that you want to take into this late of an RC so it may be best to
> just go to
> next and flow through the stable process.
> 
> Patches can can also be found in my GitHub repo at:
> https://github.com/ddalessa/kernel/tree/for-4.14
> 

Hi Denny, 

I went ahead and took these, but please make sure you follow up with
the review requests from Leon.  Preferably by the time I'm back from
PTO the incremental patches will be waiting for me ;-)

-- 
Doug Ledford <dledford@redhat.com>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD