Andreas Gruenbacher <agruenba@xxxxxxxxxx> 于2024年8月20日周二 22:22写道： > > Hi, > > On Tue, Aug 20, 2024 at 5:32 AM Julian Sun <sunjunchao2870@xxxxxxxxx> wrote: > > When gfs2_fill_super() fails, destroy_workqueue() > > is called within gfs2_gl_hash_clear(), and the > > subsequent code path calls destroy_workqueue() > > on the same work queue again. > > > > This issue can be fixed by setting the work > > queue pointer to NULL after the first > > destroy_workqueue() call and checking for > > a NULL pointer before attempting to destroy > > the work queue again. > > > > Reported-by: syzbot+d34c2a269ed512c531b0@xxxxxxxxxxxxxxxxxxxxxxxxx > > Closes: https://syzkaller.appspot.com/bug?extid=d34c2a269ed512c531b0 > > Fixes: 30e388d57367 ("gfs2: Switch to a per-filesystem glock workqueue") > > Cc: stable@xxxxxxxxxxxxxxx > > Signed-off-by: Julian Sun <sunjunchao2870@xxxxxxxxx> > > --- > > fs/gfs2/glock.c | 1 + > > fs/gfs2/ops_fstype.c | 3 ++- > > 2 files changed, 3 insertions(+), 1 deletion(-) > > > > diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c > > index 12a769077ea0..4775c2cb8ae1 100644 > > --- a/fs/gfs2/glock.c > > +++ b/fs/gfs2/glock.c > > @@ -2249,6 +2249,7 @@ void gfs2_gl_hash_clear(struct gfs2_sbd *sdp) > > gfs2_free_dead_glocks(sdp); > > glock_hash_walk(dump_glock_func, sdp); > > destroy_workqueue(sdp->sd_glock_wq); > > + sdp->sd_glock_wq = NULL; > > Here, sdp->sd_glock_wq is set to NULL, > > > } > > > > static const char *state2str(unsigned state) > > diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c > > index ff1f3e3dc65c..c1a7ff713c84 100644 > > --- a/fs/gfs2/ops_fstype.c > > +++ b/fs/gfs2/ops_fstype.c > > @@ -1305,7 +1305,8 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc) > > gfs2_delete_debugfs_file(sdp); > > gfs2_sys_fs_del(sdp); > > fail_delete_wq: > > - destroy_workqueue(sdp->sd_delete_wq); > > + if (sdp->sd_delete_wq) > > + destroy_workqueue(sdp->sd_delete_wq); > > > > but here, we check if sdp->sd_delete_wq is NULL? That doesn't make sense. I'm not sure if I have missed anything important. My understanding is that in gfs2_fill_super(), if execution jumps to the fail_lm label, gfs2_gl_hash_clear() is called first, which internally calls destroy_workqueue(sdp->sd_glock_wq). Subsequently, the code reaches the fail_delete_wq label, where destroy_workqueue(sdp->sd_glock_wq) is called again, leading to a bug. If there is anything important I'm missing, please let me know. > > > fail_glock_wq: > > destroy_workqueue(sdp->sd_glock_wq); > > fail_free: > > -- > > 2.39.2 > > > > Thanks, > Andreas > Thanks, -- Julian Sun <sunjunchao2870@xxxxxxxxx>
