summaryrefslogtreecommitdiffstats
path: root/drivers/infiniband/sw/rdmavt/trace_tx.h
diff options
context:
space:
mode:
authorMike Marciniszyn2017-05-12 18:20:31 +0200
committerDoug Ledford2017-06-27 22:56:33 +0200
commit14fe13fcd3afb96b06809f280b586be1c998332c (patch)
treecf713910a88e99f3264dd3a28c041659b7b5c983 /drivers/infiniband/sw/rdmavt/trace_tx.h
parentIB/hfi1: Setup common IB fields in hfi1_packet struct (diff)
downloadkernel-qcow2-linux-14fe13fcd3afb96b06809f280b586be1c998332c.tar.gz
kernel-qcow2-linux-14fe13fcd3afb96b06809f280b586be1c998332c.tar.xz
kernel-qcow2-linux-14fe13fcd3afb96b06809f280b586be1c998332c.zip
IB/rdmavt: Compress adjacent SGEs in rvt_lkey_ok()
SGEs that are contiguous needlessly consume driver dependent TX resources. The lkey validation logic is enhanced to compress the SGE that ends up in the send wqe when consecutive addresses are detected. The lkey validation API used to return 1 (success) or 0 (fail). The return value is now an -errno, 0 (compressed), or 1 (uncompressed). A additional argument is added to pass the last SQE for the compression. Loopback callers always pass a NULL to last_sge since the optimization is of little benefit in that situation. Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Brian Welty <brian.welty@intel.com> Signed-off-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Diffstat (limited to 'drivers/infiniband/sw/rdmavt/trace_tx.h')
-rw-r--r--drivers/infiniband/sw/rdmavt/trace_tx.h11
1 files changed, 7 insertions, 4 deletions
diff --git a/drivers/infiniband/sw/rdmavt/trace_tx.h b/drivers/infiniband/sw/rdmavt/trace_tx.h
index a613a2223751..0ef25fc49f25 100644
--- a/drivers/infiniband/sw/rdmavt/trace_tx.h
+++ b/drivers/infiniband/sw/rdmavt/trace_tx.h
@@ -84,12 +84,12 @@ __print_symbolic(opcode, \
wr_opcode_name(RESERVED10))
#define POS_PRN \
-"[%s] wqe %p wr_id %llx send_flags %x qpn %x qpt %u psn %x lpsn %x ssn %x length %u opcode 0x%.2x,%s size %u avail %u head %u last %u pid %u num_sge %u"
+"[%s] wqe %p wr_id %llx send_flags %x qpn %x qpt %u psn %x lpsn %x ssn %x length %u opcode 0x%.2x,%s size %u avail %u head %u last %u pid %u num_sge %u wr_num_sge %u"
TRACE_EVENT(
rvt_post_one_wr,
- TP_PROTO(struct rvt_qp *qp, struct rvt_swqe *wqe),
- TP_ARGS(qp, wqe),
+ TP_PROTO(struct rvt_qp *qp, struct rvt_swqe *wqe, int wr_num_sge),
+ TP_ARGS(qp, wqe, wr_num_sge),
TP_STRUCT__entry(
RDI_DEV_ENTRY(ib_to_rvt(qp->ibqp.device))
__field(u64, wr_id)
@@ -108,6 +108,7 @@ TRACE_EVENT(
__field(int, send_flags)
__field(pid_t, pid)
__field(int, num_sge)
+ __field(int, wr_num_sge)
),
TP_fast_assign(
RDI_DEV_ASSIGN(ib_to_rvt(qp->ibqp.device))
@@ -127,6 +128,7 @@ TRACE_EVENT(
__entry->ssn = wqe->ssn;
__entry->send_flags = wqe->wr.send_flags;
__entry->num_sge = wqe->wr.num_sge;
+ __entry->wr_num_sge = wr_num_sge;
),
TP_printk(
POS_PRN,
@@ -146,7 +148,8 @@ TRACE_EVENT(
__entry->head,
__entry->last,
__entry->pid,
- __entry->num_sge
+ __entry->num_sge,
+ __entry->wr_num_sge
)
);