fs/gfs2: Fix slab-use-after-free in gfs2_qd_dealloc

[PATCH] fs/gfs2: Fix slab-use-after-free in gfs2_qd_dealloc

Posted by Juntong Deng 2 years, 2 months ago

In gfs2_put_super(), whether withdrawn or not, the quota should
be cleaned up by gfs2_quota_cleanup() and wait for all quota to
be deallocated.

Otherwise struct gfs2_sbd has been freed before gfs2_qd_dealloc
(rcu callback) is completed, resulting in use-after-free.

Also, gfs2_destroy_threads() and gfs2_quota_cleanup() have
already been called in gfs2_make_fs_ro(), so there is no need to
call them again subsequently, whether withdrawn or not.

Reported-by: syzbot+29c47e9e51895928698c@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=29c47e9e51895928698c
Signed-off-by: Juntong Deng <juntong.deng@outlook.com>
---
 fs/gfs2/super.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index a4b3655c5361..d21c04a22d73 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -602,13 +602,15 @@ static void gfs2_put_super(struct super_block *sb)
 	}
 	spin_unlock(&sdp->sd_jindex_spin);
 
-	if (!sb_rdonly(sb)) {
+	if (!sb_rdonly(sb))
 		gfs2_make_fs_ro(sdp);
-	}
-	if (gfs2_withdrawn(sdp)) {
-		gfs2_destroy_threads(sdp);
+	else {
+		if (gfs2_withdrawn(sdp))
+			gfs2_destroy_threads(sdp);
+
 		gfs2_quota_cleanup(sdp);
 	}
+
 	WARN_ON(gfs2_withdrawing(sdp));
 
 	/*  At this point, we're through modifying the disk  */
-- 
2.39.2

Re: [PATCH] fs/gfs2: Fix slab-use-after-free in gfs2_qd_dealloc

Posted by Andreas Gruenbacher 2 years, 2 months ago

Hello Juntong,

On Sun, Oct 29, 2023 at 10:11 PM Juntong Deng <juntong.deng@outlook.com> wrote:
> In gfs2_put_super(), whether withdrawn or not, the quota should
> be cleaned up by gfs2_quota_cleanup() and wait for all quota to
> be deallocated.
>
> Otherwise struct gfs2_sbd has been freed before gfs2_qd_dealloc
> (rcu callback) is completed, resulting in use-after-free.
>
> Also, gfs2_destroy_threads() and gfs2_quota_cleanup() have
> already been called in gfs2_make_fs_ro(), so there is no need to
> call them again subsequently, whether withdrawn or not.
>
> Reported-by: syzbot+29c47e9e51895928698c@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=29c47e9e51895928698c
> Signed-off-by: Juntong Deng <juntong.deng@outlook.com>
> ---
>  fs/gfs2/super.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
> index a4b3655c5361..d21c04a22d73 100644
> --- a/fs/gfs2/super.c
> +++ b/fs/gfs2/super.c
> @@ -602,13 +602,15 @@ static void gfs2_put_super(struct super_block *sb)
>         }
>         spin_unlock(&sdp->sd_jindex_spin);
>
> -       if (!sb_rdonly(sb)) {
> +       if (!sb_rdonly(sb))
>                 gfs2_make_fs_ro(sdp);
> -       }
> -       if (gfs2_withdrawn(sdp)) {
> -               gfs2_destroy_threads(sdp);
> +       else {
> +               if (gfs2_withdrawn(sdp))
> +                       gfs2_destroy_threads(sdp);
> +
>                 gfs2_quota_cleanup(sdp);
>         }
> +
>         WARN_ON(gfs2_withdrawing(sdp));
>
>         /*  At this point, we're through modifying the disk  */
> --
> 2.39.2

thanks for this fix, I've merged it.

Andreas