commit 41fa722239b4 ("blk-mq: do not include passthrough requests in I/O
accounting")' disables I/O accounting for passthrough requests. Since tools
like 'iostat' do not show anything useful for passthrough I/O, it's
wasteful to do start/end time-stamping. So do away with that.
Avoiding the time-stamping improves the I/O performance by ~7%
Signed-off-by: Kundan Kumar <kundan.kumar@xxxxxxxxxxx>
Signed-off-by: Kanchan Joshi <joshi.k@xxxxxxxxxxx>
---
Changes compared to v1:
- Modified commit description.
- Made code cleaner with bit of separation.
include/linux/blk-mq.h | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 1ab3081c82ed..a676e116085f 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -830,6 +830,12 @@ void blk_mq_end_request_batch(struct io_comp_batch *ib);
*/
static inline bool blk_mq_need_time_stamp(struct request *rq)
{
+ /*
+ * passthrough io doesn't use iostat accounting, cgroup stats
+ * and io scheduler functionalities.
+ */
+ if (blk_rq_is_passthrough(rq))
+ return false;
return (rq->rq_flags & (RQF_IO_STAT | RQF_STATS | RQF_USE_SCHED));
}
--
2.25.1
