-stable review patch. If anyone has any objections, please let us know. ------------------ From: Arthur Kepner We discovered a problem when running IPoIB applications on multiple CPUs on an Altix system. Many messages such as: ib_mthca 0002:01:00.0: SQ 000014 full (19941644 head, 19941707 tail, 64 max, 0 nreq) appear in syslog, and the driver wedges up. Apparently this is because writes to the doorbells from different CPUs reach the device out of order. The following patch adds mmiowb() calls after doorbell rings to ensure the doorbell writes are ordered. Signed-off-by: Arthur Kepner Signed-off-by: Roland Dreier Signed-off-by: Greg Kroah-Hartman Signed-off-by: Chris Wright --- drivers/infiniband/hw/mthca/mthca_cq.c | 7 +++++++ drivers/infiniband/hw/mthca/mthca_qp.c | 19 +++++++++++++++++++ drivers/infiniband/hw/mthca/mthca_srq.c | 8 ++++++++ 3 files changed, 34 insertions(+) --- linux-2.6.18.1.orig/drivers/infiniband/hw/mthca/mthca_cq.c +++ linux-2.6.18.1/drivers/infiniband/hw/mthca/mthca_cq.c @@ -39,6 +39,8 @@ #include #include +#include + #include #include "mthca_dev.h" @@ -210,6 +212,11 @@ static inline void update_cons_index(str mthca_write64(doorbell, dev->kar + MTHCA_CQ_DOORBELL, MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); + /* + * Make sure doorbells don't leak out of CQ spinlock + * and reach the HCA out of order: + */ + mmiowb(); } } --- linux-2.6.18.1.orig/drivers/infiniband/hw/mthca/mthca_qp.c +++ linux-2.6.18.1/drivers/infiniband/hw/mthca/mthca_qp.c @@ -39,6 +39,8 @@ #include #include +#include + #include #include #include @@ -1730,6 +1732,11 @@ out: mthca_write64(doorbell, dev->kar + MTHCA_SEND_DOORBELL, MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); + /* + * Make sure doorbells don't leak out of SQ spinlock + * and reach the HCA out of order: + */ + mmiowb(); } qp->sq.next_ind = ind; @@ -1849,6 +1856,12 @@ out: qp->rq.next_ind = ind; qp->rq.head += nreq; + /* + * Make sure doorbells don't leak out of RQ spinlock and reach + * the HCA out of order: + */ + mmiowb(); + spin_unlock_irqrestore(&qp->rq.lock, flags); return err; } @@ -2110,6 +2123,12 @@ out: MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); } + /* + * Make sure doorbells don't leak out of SQ spinlock and reach + * the HCA out of order: + */ + mmiowb(); + spin_unlock_irqrestore(&qp->sq.lock, flags); return err; } --- linux-2.6.18.1.orig/drivers/infiniband/hw/mthca/mthca_srq.c +++ linux-2.6.18.1/drivers/infiniband/hw/mthca/mthca_srq.c @@ -35,6 +35,8 @@ #include #include +#include + #include "mthca_dev.h" #include "mthca_cmd.h" #include "mthca_memfree.h" @@ -593,6 +595,12 @@ int mthca_tavor_post_srq_recv(struct ib_ MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); } + /* + * Make sure doorbells don't leak out of SRQ spinlock and + * reach the HCA out of order: + */ + mmiowb(); + spin_unlock_irqrestore(&srq->lock, flags); return err; } -- - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/