block: disable iopoll for split bio
PATCH v5 1/1 block: disable iopoll for split bio
iopoll is initially for small size, latency sensitive IO. It doesn't work well for big IO, especially when it needs to be split to multiple bios. In this case, the returned cookie of __submit_bio_noacct_mq() is indeed the cookie of the last split bio. The completion of *this* last split bio done by iopoll doesn't mean the whole original bio has completed. Callers of iopoll still need to wait for completion of other split bios.
Besides bio splitting may cause more trouble for iopoll which isn't supposed to be used in case of big IO.
iopoll for split bio may cause potential race if CPU migration happens during bio submission. Since the returned cookie is that of the last split bio, polling on the corresponding hardware queue doesn't help complete other split bios, if these split bios are enqueued into different hardware queues. Since interrupts are disabled for polling queues, the completion of these other split bios depends on timeout mechanism, thus causing a potential hang.
iopoll for split bio may also cause hang for sync polling. Currently both the blkdev and iomap-based fs (ext4/xfs, etc) support sync polling in direct IO routine. These routines will submit bio without REQ_NOWAIT flag set, and then start sync polling in current process context. The process may hang in blk_mq_get_tag() if the submitted bio has to be split into multiple bios and can rapidly exhaust the queue depth. The process are waiting for the completion of the previously allocated requests, which should be reaped by the following polling, and thus causing a deadlock.
To avoid these subtle trouble described above, just disable iopoll for split bio.
--- block/bio.c | 2 ++ block/blk-merge.c | 10 ++++++++++ block/blk-mq.c | 2 +- include/linux/blk_types.h | 1 + 4 files changed, 14 insertions(+), 1 deletion(-)
block/bio.c +2 -0
@@ -684,6 +684,8 @@ void __bio_clone_fast(struct bio *bio, struct bio *bio_src)bio_set_flag (bio,BIO_CLONED );if (bio_flagged (bio_src,BIO_THROTTLED ))bio_set_flag (bio,BIO_THROTTLED ); +if (bio_flagged (bio_src,BIO_NONE_COOKIE )) +bio_set_flag (bio,BIO_NONE_COOKIE ); bio->bi_opf = bio_src->bi_opf; bio->bi_ioprio = bio_src->bi_ioprio; bio->bi_write_hint = bio_src->bi_write_hint;
block/blk-merge.c +10 -0
@@ -279,6 +279,16 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,return NULL ; split: *segs = nsegs; + +/* +* Bio splitting may cause subtle trouble such as hang +* when doing sync iopoll in direct IO routine. Given +* performance gain of iopoll for big IO can be trival, +* disable iopoll when split needed. +*/ + bio->bi_opf &= ~REQ_HIPRI ; +bio_set_flag (bio,BIO_NONE_COOKIE ); +return bio_split (bio, sectors,GFP_NOIO , bs); }
block/blk-mq.c +1 -1
@@ -2265,7 +2265,7 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)blk_mq_sched_insert_request (rq, false, true, true); } -return cookie; +return bio_flagged (bio,BIO_NONE_COOKIE ) ?BLK_QC_T_NONE : cookie; queue_exit:blk_queue_exit (q);return BLK_QC_T_NONE ;
include/linux/blk_types.h +1 -0
@@ -284,6 +284,7 @@ enum {* of this bio. */ BIO_CGROUP_ACCT ,/* has been accounted to a cgroup */ BIO_TRACKED ,/* set if bio goes through the rq_qos path */ +BIO_NONE_COOKIE ,/* disable iopoll for split bio */ BIO_FLAG_LAST };
Signature · 1 line
2.27.0
This looks generally food:
Some cosmetic nitpicks below:
block/blk-merge.c static struct bio *blk_bio_segment_split(struct request_queue *q, · line 288
Quoted 6 lines
block/blk-merge.c +6
-0+/* +* Bio splitting may cause subtle trouble such as hang +* when doing sync iopoll in direct IO routine. Given +* performance gain of iopoll for big IO can be trival, +* disable iopoll when split needed. +*/
You can use up all 80 characters for comments, and doing so generally helps readability.
block/blk-mq.c blk_qc_t blk_mq_submit_bio(struct bio *bio) · line 2268
Quoted 2 lines
block/blk-mq.c +1
-1-return cookie; +return bio_flagged (bio,BIO_NONE_COOKIE ) ?BLK_QC_T_NONE : cookie;
I'd write this a little more easily flowing as:
if (bio_flagged(bio, BIO_NONE_COOKIE)) return BLK_QC_T_NONE; return cookie;
include/linux/blk_types.h enum { · line 287
Quoted 12 lines
queue_exit:block/blk-mq.c +0
-0blk_queue_exit (q);return BLK_QC_T_NONE ;include/linux/blk_types.h +1
-0@@ -284,6 +284,7 @@ enum {* of this bio. */ BIO_CGROUP_ACCT ,/* has been accounted to a cgroup */ BIO_TRACKED ,/* set if bio goes through the rq_qos path */ + BIO_NONE_COOKIE, /* disable iopoll for split bio */
I'd rename this to BIO_SPLIT and update the comment. The rationale is BIO_SPLIT is what happened to the bio. The fact that blk_mq_submit_bio doesn't return the cookie is just one good use of such information.
Quoted 6 lines · Christoph Hellwig
This looks generally food:
Reviewed-by Christoph Hellwig <[email protected]>Some cosmetic nitpicks below:
Quoted 10 lines
block/blk-merge.c static struct bio *blk_bio_segment_split(struct request_queue *q, · line 288
Quoted 6 lines
block/blk-merge.c +6
-0+/* +* Bio splitting may cause subtle trouble such as hang +* when doing sync iopoll in direct IO routine. Given +* performance gain of iopoll for big IO can be trival, +* disable iopoll when split needed. +*/ You can use up all 80 characters for comments, and doing so generally helps readability.
Thanks. I will be more careful about this. I'm gonna set related constraint in my vim, checkpatch etc. ;)
Quoted 9 lines
block/blk-mq.c blk_qc_t blk_mq_submit_bio(struct bio *bio) · line 2268
Quoted 2 lines
block/blk-mq.c +1
-1-return cookie; +return bio_flagged (bio,BIO_NONE_COOKIE ) ?BLK_QC_T_NONE : cookie;I'd write this a little more easily flowing as:
if (bio_flagged(bio, BIO_NONE_COOKIE)) return BLK_QC_T_NONE; return cookie;
Regards.
Quoted 9 lines
Quoted 9 lines
queue_exit: blk_queue_exit(q); return BLK_QC_T_NONE;include/linux/blk_types.h +0
-0@@ -284,6 +284,7 @@ enum {* of this bio. */
Quoted 9 lines
include/linux/blk_types.h enum { · line 287
Quoted 3 lines
include/linux/blk_types.h +1
-0BIO_CGROUP_ACCT ,/* has been accounted to a cgroup */ BIO_TRACKED ,/* set if bio goes through the rq_qos path */ + BIO_NONE_COOKIE, /* disable iopoll for split bio */I'd rename this to
BIO_SPLITand update the comment. The rationale isBIO_SPLITis what happened to the bio. The fact thatblk_mq_submit_biodoesn't return the cookie is just one good use of such information.
Actually why I didn't choose 'BIO_SPLIT' at the very beginning is that, currently only normal IO in mq routine gets marked when they gets split, neither abnormal IO nor normal IO in dm/md routine, though this is reasonable since currently only normal IO in mq routine is capable of polling and gets stuck in this issue.
But BIO_SPLIT may be more clear? I could explain the above constraint in the comment instead.
Thanks for reviewing.
Signature · 2 lines
Thanks, Jeffle