bcache fixes for Linux v5.18-rc3
Subsystems (2)
cover bcache fixes for Linux v5.18-rc3
Hi Jens,
There are two regressions introduced in the generic block layer changes from Linux v5.18-rc1. Both of them may panic the kernel, please take them for -rc4.
Thanks in advance.
Coly Li
---
Coly Li (2):
bcache: put bch_bio_map() back to correct location in
journal_write_unlocked()
bcache: fix wrong bdev parameter when calling bio_alloc_clone() in
do_bio_hook()
drivers/md/bcache/journal.c | 2 +- drivers/md/bcache/request.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
Signature · 1 line
2.34.1
PATCH 1/2 bcache: put bch_bio_map() back to correct location in journal_write_unlocked()
Commit a7c50c940477 ("block: pass a block_device and opf to bio_reset") moves bch_bio_map() inside journal_write_unlocked() next to the location where the modified bio_reset() was called.
This change is wrong because calling bch_bio_map() immediately after bio_reset(), a BUG_ON(!bio->bi_iter.bi_size) inside bch_bio_map() will be triggered and panic the kernel.
This patch puts bch_bio_map() back to its original correct location in journal_write_unlocked() and avoid the BUG_ON().
--- drivers/md/bcache/journal.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/md/bcache/journal.c +1 -1
@@ -771,12 +771,12 @@ static void journal_write_unlocked(struct closure *cl)bio_reset (bio, ca->bdev,REQ_OP_WRITE |REQ_SYNC |REQ_META |REQ_PREFLUSH |REQ_FUA ); -bch_bio_map (bio, w->data); bio->bi_iter.bi_sector =PTR_OFFSET (k, i); bio->bi_iter.bi_size = sectors <<9 ; bio->bi_end_io = journal_write_endio; bio->bi_private = w; +bch_bio_map (bio, w->data);trace_bcache_journal_write (bio, w->data->keys);bio_list_add (&list, bio);
Signature · 1 line
2.34.1
PATCH 2/2 bcache: fix wrong bdev parameter when calling bio_alloc_clone() in do_bio_hook()
In reply to Coly LiCommit abfc426d1b2f ("block: pass a block_device to bio_clone_fast") calls the modified bio_alloc_clone() in bcache code as:
bio_init_clone(bio->bi_bdev, bio, orig_bio, GFP_NOIO);
But the first parameter is wrong, where bio->bi_bdev should be orig_bio->bi_bdev. The wrong bi_bdev panics the kernel when submitting cache bio.
This patch fixes the wrong bdev parameter usage and avoid the panic.
--- drivers/md/bcache/request.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/md/bcache/request.c +1 -1
@@ -685,7 +685,7 @@ static void do_bio_hook(struct search *s, {struct bio *bio = &s->bio.bio; -bio_init_clone (bio->bi_bdev, bio, orig_bio,GFP_NOIO ); +bio_init_clone (orig_bio->bi_bdev, bio, orig_bio,GFP_NOIO );/* * bi_end_io can be set separately somewhere else, e.g. the * variants in,
Signature · 1 line
2.34.1
Quoted 9 lines · Coly Li
There are two regressions introduced in the generic block layer changes from Linux v5.18-rc1. Both of them may panic the kernel, please take them for -rc4.
Thanks in advance.
Coly Li
[...]
Applied, thanks!
[1/2] bcache: put bch_bio_map() back to correct location in journal_write_unlocked()
commit: ff2695e52c9936febf65aa36a1769881da71bec5
[2/2] bcache: fix wrong bdev parameter when calling bio_alloc_clone() in do_bio_hook()
commit: 9dca4168a37c9cfe182f077f0d2289292e9e3656Best regards,
Signature · 1 line
Jens Axboe
Looks good: