bcachefs: slab-use-after-free Read in bch2_sb_errors_from_cpu
PATCH 1/1 bcachefs: slab-use-after-free Read in bch2_sb_errors_from_cpu
Acquire fsck_error_counts_lock before accessing the critical section protected by this lock.
syzbot has tested the proposed patch and the reproducer did not trigger any issue.
---
Syzbot detected we are accessing free'd memory inbch2_sb_errors_from_cpu().
It is caused by race condition when another task is freeing the array protected by fsck_error_counts_lock.
This patch acquires fsck_error_counts_lock before accessing the entries and get the current number of elements in the array.
syzbot has tested the proposed patch and the reproducer did not trigger any issue:
Tested on:
commit: 55027e68 Merge tag 'input-for-v6.10-rc5' of git://git... git tree: upstream console output: https://syzkaller.appspot.com/x/log.txt?x=1790a501980000 kernel config: https://syzkaller.appspot.com/x/.config?x=d6b9ee98d841760c dashboard link: https://syzkaller.appspot.com/bug?extid=a2bc0e838efd7663f4d9 compiler: gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40 patch: https://syzkaller.appspot.com/x/patch.diff?x=1272faae980000
Note: testing is done by a robot and is best-effort only.
--- fs/bcachefs/sb-errors.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-)
fs/bcachefs/sb-errors.c +10 -4
@@ -110,19 +110,25 @@ void bch2_sb_error_count(struct bch_fs *c, enum bch_sb_error_id err)void bch2_sb_errors_from_cpu (struct bch_fs *c) {bch_sb_errors_cpu *src = &c->fsck_error_counts; -struct bch_sb_field_errors *dst = -bch2_sb_field_resize (&c->disk_sb, errors, -bch2_sb_field_errors_u64s (src->nr)); +struct bch_sb_field_errors *dst;unsigned i; +mutex_lock (&c->fsck_error_counts_lock); + + dst =bch2_sb_field_resize (&c->disk_sb, errors, +bch2_sb_field_errors_u64s (src->nr)); +if (!dst) -return ; + goto err;for (i =0 ; i < src->nr; i++) {SET_BCH_SB_ERROR_ENTRY_ID (&dst->entries[i], src->data[i].id);SET_BCH_SB_ERROR_ENTRY_NR (&dst->entries[i], src->data[i].nr); dst->entries[i].last_error_time =cpu_to_le64 (src->data[i].last_error_time); } + +err: +mutex_unlock (&c->fsck_error_counts_lock); }static int bch2_sb_errors_to_cpu (struct bch_fs *c)
---
base-commit: 66cc544fd75c70b5ee74df87ab99acc45b835e69
change-id: 20240625-bug2-94405201c4c8
Best regards,
Signature · 1 line
Pei Li <[email protected]>
Quoted 31 lines · Pei Li
Acquire
fsck_error_counts_lockbefore accessing the critical section protected by this lock.syzbot has tested the proposed patch and the reproducer did not trigger any issue.
Reported-by [email protected]Signed-off-by Pei Li <[email protected]>---
Syzbot detected we are accessing free'd memory inbch2_sb_errors_from_cpu().It is caused by race condition when another task is freeing the array protected by
fsck_error_counts_lock.This patch acquires
fsck_error_counts_lockbefore accessing the entries and get the current number of elements in the array.syzbot has tested the proposed patch and the reproducer did not trigger any issue:
Tested on:
commit: 55027e68 Merge tag 'input-for-v6.10-rc5' of git://git... git tree: upstream console output: https://syzkaller.appspot.com/x/log.txt?x=1790a501980000 kernel config: https://syzkaller.appspot.com/x/.config?x=d6b9ee98d841760c dashboard link: https://syzkaller.appspot.com/bug?extid=a2bc0e838efd7663f4d9 compiler: gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40 patch: https://syzkaller.appspot.com/x/patch.diff?x=1272faae980000
Quoted 1 line
Note: testing is done by a robot and is best-effort only.
Looks good - applied