Skip to contentSkip to conversation

blk: optimization for classic polling

PATCH New Nitesh Shetty 1 participant 1 message Feb 9, 2018 1 message not in the archive Block layer

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): CGME20180209104227epcas5p26e19e332c00996a777a2c3b73a347405@epcas5p2.samsung.com

#1 lore raw

PATCH v2 1/1 blk: optimization for classic polling

From: Nitesh Shetty <[email protected]>

This removes the dependency on interrupts to wake up task. Set task state as TASK_RUNNING, if need_resched() returns true,
while polling for IO completion.
Earlier, polling task used to sleep, relying on interrupt to wake it up. This made some IO take very long when interrupt-coalescing is enabled in NVMe.

Changes v1->v2:
-setting task state once in blk_poll, instead of multiple callers.

---
 block/blk-mq.c | 1 +
 1 file changed, 1 insertion(+)
block/blk-mq.c +1 -0
index df93102..40285fe 100644
@@ -3164,6 +3164,7 @@ static bool __blk_mq_poll(struct blk_mq_hw_ctx *hctx, struct request *rq)
 		cpu_relax();
 	}
 
+	set_current_state(TASK_RUNNING);
 	return false;
 }
 
Signature · 1 line
2.7.4