When mounting ocfs2 and then remounting it as read-only, a
slab-use-after-free occurs after the user uses a syscall to
quota_getnextquota. Specifically, sb_dqinfo(sb, type)->dqi_priv is the
dangling pointer.

During the remounting process, the pointer dqi_priv is freed but is
never set as null leaving it to to be accessed. Additionally, the
read-only option for remounting sets the DQUOT_SUSPENDED flag instead of
setting the DQUOT_USAGE_ENABLED flags. Moreover, later in the process of
getting the next quota, the function ocfs2_get_next_id is called and
only checks the quota usage flags and not the quota suspended flags.

To fix this, I set dqi_priv to null when it is freed after remounting
with read-only and put a check for DQUOT_SUSPENDED in ocfs2_get_next_id.

Signed-off-by: Dennis Lam <dennis.lamerice@gmail.com>
Reported-by: syzbot+d173bf8a5a7faeede34c@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/6731d26f.050a0220.1fb99c.014b.GAE@google.com/T/
---
 fs/ocfs2/quota_global.c | 3 ++-
 fs/ocfs2/quota_local.c  | 1 +
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/ocfs2/quota_global.c b/fs/ocfs2/quota_global.c
index 2b0daced98eb..c6d38340d6d4 100644
--- a/fs/ocfs2/quota_global.c
+++ b/fs/ocfs2/quota_global.c
@@ -893,7 +893,8 @@ static int ocfs2_get_next_id(struct super_block *sb, struct kqid *qid)
 	int status = 0;
 
 	trace_ocfs2_get_next_id(from_kqid(&init_user_ns, *qid), type);
-	if (!sb_has_quota_loaded(sb, type)) {
+	if (!sb_has_quota_loaded(sb, type) ||
+			sb_dqopt(sb)->flags & DQUOT_SUSPENDED) {
 		status = -ESRCH;
 		goto out;
 	}
diff --git a/fs/ocfs2/quota_local.c b/fs/ocfs2/quota_local.c
index 73d3367c533b..2956d888c131 100644
--- a/fs/ocfs2/quota_local.c
+++ b/fs/ocfs2/quota_local.c
@@ -867,6 +867,7 @@ static int ocfs2_local_free_info(struct super_block *sb, int type)
 	brelse(oinfo->dqi_libh);
 	brelse(oinfo->dqi_lqi_bh);
 	kfree(oinfo);
+	info->dqi_priv = NULL;
 	return status;
 }
 
-- 
2.47.0

On 2024/12/15 11:58, Dennis Lam wrote:
> When mounting ocfs2 and then remounting it as read-only, a
> slab-use-after-free occurs after the user uses a syscall to
> quota_getnextquota. Specifically, sb_dqinfo(sb, type)->dqi_priv is the
> dangling pointer.
> 
> During the remounting process, the pointer dqi_priv is freed but is
> never set as null leaving it to to be accessed. Additionally, the
> read-only option for remounting sets the DQUOT_SUSPENDED flag instead of
> setting the DQUOT_USAGE_ENABLED flags. Moreover, later in the process of
> getting the next quota, the function ocfs2_get_next_id is called and
> only checks the quota usage flags and not the quota suspended flags.
> 
> To fix this, I set dqi_priv to null when it is freed after remounting
> with read-only and put a check for DQUOT_SUSPENDED in ocfs2_get_next_id.
> 
> Signed-off-by: Dennis Lam <dennis.lamerice@gmail.com>
> Reported-by: syzbot+d173bf8a5a7faeede34c@syzkaller.appspotmail.com
> Closes: https://lore.kernel.org/all/6731d26f.050a0220.1fb99c.014b.GAE@google.com/T/
> ---
>  fs/ocfs2/quota_global.c | 3 ++-
>  fs/ocfs2/quota_local.c  | 1 +
>  2 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/ocfs2/quota_global.c b/fs/ocfs2/quota_global.c
> index 2b0daced98eb..c6d38340d6d4 100644
> --- a/fs/ocfs2/quota_global.c
> +++ b/fs/ocfs2/quota_global.c
> @@ -893,7 +893,8 @@ static int ocfs2_get_next_id(struct super_block *sb, struct kqid *qid)
>  	int status = 0;
>  
>  	trace_ocfs2_get_next_id(from_kqid(&init_user_ns, *qid), type);
> -	if (!sb_has_quota_loaded(sb, type)) {
> +	if (!sb_has_quota_loaded(sb, type) ||
> +			sb_dqopt(sb)->flags & DQUOT_SUSPENDED) {

Looks reasonable.
We can just use !sb_has_quota_active() instead.

Thanks,
Joseph

>  		status = -ESRCH;
>  		goto out;
>  	}
> diff --git a/fs/ocfs2/quota_local.c b/fs/ocfs2/quota_local.c
> index 73d3367c533b..2956d888c131 100644
> --- a/fs/ocfs2/quota_local.c
> +++ b/fs/ocfs2/quota_local.c
> @@ -867,6 +867,7 @@ static int ocfs2_local_free_info(struct super_block *sb, int type)
>  	brelse(oinfo->dqi_libh);
>  	brelse(oinfo->dqi_lqi_bh);
>  	kfree(oinfo);
> +	info->dqi_priv = NULL;
>  	return status;
>  }
>

>
> Looks reasonable.
> We can just use !sb_has_quota_active() instead.
>

Thanks for the feedback. I'll be sure to send a new patch with the
updated function.

Regards,
Dennis