block: fix trace completion for chained bio
PATCH v3 1/1 block: fix trace completion for chained bio
From: Edward Hsieh <[email protected]>
For chained bio, trace_block_bio_complete in bio_endio is currently called only by the parent bio once upon all chained bio completed. However, the sector and size for the parent bio are modified in bio_split. Therefore, the size and sector of the complete events might not match the queue events in blktrace.
The original fix of bio completion trace <fbbaf700e7b1> ("block: trace completion of all bios.") wants multiple complete events to correspond to one queue event but missed this.
The issue can be reproduced by md/raid5 read with bio cross chunks.
To fix, move trace completion into the loop for every chained bio to call.
To make sense of the tail call optimization, the bio_chained_endio handing should be at the end of bio_endio(), move blk_throtl_bio_endio() and bio_uninit() to the else closure of bio_chain_endio condition. blk_throtl_bio_endio updates latency for the throttle group, and considering the throttle group tracks the latency of the current device, we should record the latency after all chained bio finish. And since the resources of the chained bio are released by bio_put() in __bio_chain_endio(), calling bio_uninit() for chained bio is not necessary.
--- block/bio.c | 25 ++++++++++++------------- 1 file changed, 12 insertions(+), 13 deletions(-)
block/bio.c +12 -13
@@ -1375,8 +1375,7 @@ static inline bool bio_remaining_done(struct bio *bio)* * bio_endio() can be called several times on a bio that has been chained * using bio_chain(). The ->bi_end_io() function will only be called the -* last time. At this point the BLK_TA_COMPLETE tracing event will be -* generated if BIO_TRACE_COMPLETION is set. +* last time. **/ void bio_endio (struct bio *bio) {
@@ -1389,6 +1388,11 @@ void bio_endio(struct bio *bio)if (bio->bi_bdev)rq_qos_done_bio (bio->bi_bdev->bd_disk->queue, bio); +if (bio->bi_bdev &&bio_flagged (bio,BIO_TRACE_COMPLETION )) { +trace_block_bio_complete (bio->bi_bdev->bd_disk->queue, bio); +bio_clear_flag (bio,BIO_TRACE_COMPLETION ); + } +/* * Need to have a real endio function for chained bios, otherwise * various corner cases will break (like stacking block devices that
@@ -1400,18 +1404,13 @@ void bio_endio(struct bio *bio)if (bio->bi_end_io == bio_chain_endio) { bio =__bio_chain_endio (bio); goto again; + }else { +blk_throtl_bio_endio (bio); +/* release cgroup info */ +bio_uninit (bio); +if (bio->bi_end_io) + bio->bi_end_io (bio); } - -if (bio->bi_bdev &&bio_flagged (bio,BIO_TRACE_COMPLETION )) { -trace_block_bio_complete (bio->bi_bdev->bd_disk->queue, bio); -bio_clear_flag (bio,BIO_TRACE_COMPLETION ); - } - -blk_throtl_bio_endio (bio); -/* release cgroup info */ -bio_uninit (bio); -if (bio->bi_end_io) - bio->bi_end_io (bio); }EXPORT_SYMBOL (bio_endio);
Signature · 1 line
2.31.1
block/bio.c void bio_endio(struct bio *bio) · line 1412
Quoted 10 lines · edwardh
block/bio.c +6
-0@@ -1400,18 +1404,13 @@ void bio_endio(struct bio *bio)if (bio->bi_end_io == bio_chain_endio) { bio =__bio_chain_endio (bio); goto again; + }else { +blk_throtl_bio_endio (bio); +/* release cgroup info */ +bio_uninit (bio); +if (bio->bi_end_io) + bio->bi_end_io (bio);
No need for an else after a goto.
Quoted 14 lines · Christoph Hellwig
block/bio.c void bio_endio(struct bio *bio) · line 1412
Quoted 10 lines · edwardh
block/bio.c +6
-0@@ -1400,18 +1404,13 @@ void bio_endio(struct bio *bio)if (bio->bi_end_io == bio_chain_endio) { bio =__bio_chain_endio (bio); goto again; + }else { +blk_throtl_bio_endio (bio); +/* release cgroup info */ +bio_uninit (bio); +if (bio->bi_end_io) + bio->bi_end_io (bio);No need for an else after a goto.
We are suggested by Neil Brown in the last version that from the comment, the bio_chain_endio handling is used *only* to avoid deep recursion so it should be at the end of the function. Therefore, the position of blk_throtl_bio_endio() and bio_uninit() are a bit odd.
We believe that blk_throtl_bio_endio() and bio_uninit() is in the correct position now. And adding an else closure is our attempt to make it more clear.
If it's not necessary then V2 patch should work to fix the missing completion traces. Should we resend PATCH V2?
Thank you,
Edward
Quoted 5 lines · Edward Hsieh
We are suggested by Neil Brown in the last version that from the comment, the
bio_chain_endiohandling is used *only* to avoid deep recursion so it should be at the end of the function. Therefore, the position ofblk_throtl_bio_endio()andbio_uninit()are a bit odd.
This is BS. Placement of an else after a returning statement does not change code generation.
Quoted 2 lines
If it's not necessary then V2 patch should work to fix the missing completion traces. Should we resend PATCH V2?
Please send an updated version.