Re: [syzbot] Re: [syzbot] [bcachefs?] possible deadlock in bch2_gc_mark_key

[syzbot <syzbot+050e797ad21ccc3f5d1a@syzkaller.appspotmail.com>]

For archival purposes, forwarding an incoming command email to
linux-kernel@vger.kernel.org.

***

Subject: Re: [syzbot] [bcachefs?] possible deadlock in bch2_gc_mark_key
Author: lizhi.xu@windriver.com

#syz test https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 50736169ecc8

diff --git a/fs/bcachefs/sb-clean.c b/fs/bcachefs/sb-clean.c
index 47f10ab57f40..7a75615ba745 100644
--- a/fs/bcachefs/sb-clean.c
+++ b/fs/bcachefs/sb-clean.c
@@ -373,6 +373,7 @@ void bch2_fs_mark_clean(struct bch_fs *c)
 
 	entry = sb_clean->start;
 	bch2_journal_super_entries_add_common(c, &entry, 0);
+	mutex_unlock(&c->sb_lock);
 	entry = bch2_btree_roots_to_journal_entries(c, entry, 0);
 	BUG_ON((void *) entry > vstruct_end(&sb_clean->field));
 
@@ -383,6 +384,7 @@ void bch2_fs_mark_clean(struct bch_fs *c)
 	 * this should be in the write path, and we should be validating every
 	 * superblock section:
 	 */
+	mutex_lock(&c->sb_lock);
 	ret = bch2_sb_clean_validate_late(c, sb_clean, WRITE);
 	if (ret) {
 		bch_err(c, "error writing marking filesystem clean: validate error");