| Message ID | 1455168112-12803-1-git-send-email-devesh.sharma@broadcom.com (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
On Thu, Feb 11, 2016 at 10:51 AM, Devesh Sharma <devesh.sharma@broadcom.com> wrote: > CQ is never armed before calling poll on it. Most of the kernel > ULPs have already moved to use this new API or have a code where > poll is called before arming the CQ. You might want to wait for NFS whose patches are in pipe to migrate to newer APIs before merging this? -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Parav, I have tested this patch with nfsrdma + linux-4.4 (i.e. without Chuck's patches to use new CQ API) iozone on nfsrdma mount point passed. -Regards Devesh On Fri, Feb 12, 2016 at 2:43 AM, Parav Pandit <pandit.parav@gmail.com> wrote: > On Thu, Feb 11, 2016 at 10:51 AM, Devesh Sharma > <devesh.sharma@broadcom.com> wrote: >> CQ is never armed before calling poll on it. Most of the kernel >> ULPs have already moved to use this new API or have a code where >> poll is called before arming the CQ. > > You might want to wait for NFS whose patches are in pipe to migrate to > newer APIs before merging this? -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 02/11/2016 10:46 PM, Devesh Sharma wrote: > Hi Parav, > > I have tested this patch with nfsrdma + linux-4.4 (i.e. without > Chuck's patches to use new CQ API) > iozone on nfsrdma mount point passed. > > -Regards > Devesh > > On Fri, Feb 12, 2016 at 2:43 AM, Parav Pandit <pandit.parav@gmail.com> wrote: >> On Thu, Feb 11, 2016 at 10:51 AM, Devesh Sharma >> <devesh.sharma@broadcom.com> wrote: >>> CQ is never armed before calling poll on it. Most of the kernel >>> ULPs have already moved to use this new API or have a code where >>> poll is called before arming the CQ. >> >> You might want to wait for NFS whose patches are in pipe to migrate to >> newer APIs before merging this? Thanks, I've picked this up for 4.5-rc.
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma.h b/drivers/infiniband/hw/ocrdma/ocrdma.h index 040bb8b..12503f1 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma.h @@ -323,9 +323,6 @@ struct ocrdma_cq { */ u32 max_hw_cqe; bool phase_change; - bool deferred_arm, deferred_sol; - bool first_arm; - spinlock_t cq_lock ____cacheline_aligned; /* provide synchronization * to cq polling */ diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c index 37620b4..12420e4 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c @@ -1094,7 +1094,6 @@ struct ib_cq *ocrdma_create_cq(struct ib_device *ibdev, spin_lock_init(&cq->comp_handler_lock); INIT_LIST_HEAD(&cq->sq_head); INIT_LIST_HEAD(&cq->rq_head); - cq->first_arm = true; if (ib_ctx) { uctx = get_ocrdma_ucontext(ib_ctx); @@ -2910,12 +2909,9 @@ expand_cqe: } stop_cqe: cq->getp = cur_getp; - if (cq->deferred_arm || polled_hw_cqes) { - ocrdma_ring_cq_db(dev, cq->id, cq->deferred_arm, - cq->deferred_sol, polled_hw_cqes); - cq->deferred_arm = false; - cq->deferred_sol = false; - } + + if (polled_hw_cqes) + ocrdma_ring_cq_db(dev, cq->id, false, false, polled_hw_cqes); return i; } @@ -2999,13 +2995,7 @@ int ocrdma_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags cq_flags) if (cq_flags & IB_CQ_SOLICITED) sol_needed = true; - if (cq->first_arm) { - ocrdma_ring_cq_db(dev, cq_id, arm_needed, sol_needed, 0); - cq->first_arm = false; - } - - cq->deferred_arm = true; - cq->deferred_sol = sol_needed; + ocrdma_ring_cq_db(dev, cq_id, arm_needed, sol_needed, 0); spin_unlock_irqrestore(&cq->cq_lock, flags); return 0;
Today ocrdma driver defer arming the CQ till poll is called. This was used to prevent calling poll-cq on an armed CQ. Recently a set of new CQ API has been introduced into the linux kernel. The implementation of this API guarantees that a given CQ is never armed before calling poll on it. Most of the kernel ULPs have already moved to use this new API or have a code where poll is called before arming the CQ. Thus, the above workaround in ocrdma is not needed anymore. This patch removes the additional logic to deffer arm till poll is called. This patch adds a simple scheme where ib_req_notify_cq() will actually arm the cq. Signed-off-by: Devesh Sharma <devesh.sharma@broadcom.com> --- drivers/infiniband/hw/ocrdma/ocrdma.h | 3 --- drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 18 ++++-------------- 2 files changed, 4 insertions(+), 17 deletions(-)