Missing calls to mntget() (or equivalently, too many calls to mntput()) are hard to detect because mntput() delays freeing mounts using task_work_add(), then again using call_rcu(). As a result, mnt_count can often be decremented to -1 without getting a KASAN use-after-free report. Such cases are still bugs though, and they point to real use-after-frees being possible. For an example of this, see the bug fixed by commit 1b0b9cc8d379 ("vfs: fsmount: add missing mntget()"), discussed at https://lkml.kernel.org/linux-fsdevel/20190605135401.GB30925@xxxxxxxxxxxxxxxxxxxxxxxxx/T/#u. This bug *should* have been trivial to find. But actually, it wasn't found until syzkaller happened to use fchdir() to manipulate the reference count just right for the bug to be noticeable. Address this by making mntput_no_expire() issue a WARN if mnt_count has become negative.
Unified · Split
| @@ -156,10 +156,10 @@ static inline void mnt_add_count(struct mount *mnt, int n) · 155 unmodified lines | |||
| 156 | /* | 156 | /* |
| 157 | * vfsmount lock must be held for write | 157 | * vfsmount lock must be held for write |
| 158 | */ | 158 | */ |
| 159 | unsigned int mnt_get_count(struct mount *mnt) | 159 | int mnt_get_count(struct mount *mnt) |
| 160 | { | 160 | { |
| 161 | #ifdef CONFIG_SMP | 161 | #ifdef CONFIG_SMP |
| 162 | unsigned int count = 0; | 162 | int count = 0; |
| 163 | int cpu; | 163 | int cpu; |
| 164 | 164 | ||
| 165 | for_each_possible_cpu(cpu) { | 165 | for_each_possible_cpu(cpu) { |
| @@ -1139,6 +1139,7 @@ static DECLARE_DELAYED_WORK(delayed_mntput_work, delayed_mntput); · 973 unmodified lines | |||
| 1139 | static void mntput_no_expire(struct mount *mnt) | 1139 | static void mntput_no_expire(struct mount *mnt) |
| 1140 | { | 1140 | { |
| 1141 | LIST_HEAD(list); | 1141 | LIST_HEAD(list); |
| 1142 | int count; | ||
| 1142 | 1143 | ||
| 1143 | rcu_read_lock(); | 1144 | rcu_read_lock(); |
| 1144 | if (likely(READ_ONCE(mnt->mnt_ns))) { | 1145 | if (likely(READ_ONCE(mnt->mnt_ns))) { |
| @@ -1162,7 +1163,9 @@ static void mntput_no_expire(struct mount *mnt) · 17 unmodified lines | |||
| 1162 | */ | 1163 | */ |
| 1163 | smp_mb(); | 1164 | smp_mb(); |
| 1164 | mnt_add_count(mnt, -1); | 1165 | mnt_add_count(mnt, -1); |
| 1165 | if (mnt_get_count(mnt)) { | 1166 | count = mnt_get_count(mnt); |
| 1167 | if (count != 0) { | ||
| 1168 | WARN_ON(count < 0); | ||
| 1166 | rcu_read_unlock(); | 1169 | rcu_read_unlock(); |
| 1167 | unlock_mount_hash(); | 1170 | unlock_mount_hash(); |
| 1168 | return; | 1171 | return; |
| @@ -44,7 +44,7 @@ int propagate_mount_busy(struct mount *, int); · 43 unmodified lines | |||
| 44 | void propagate_mount_unlock(struct mount *); | 44 | void propagate_mount_unlock(struct mount *); |
| 45 | void mnt_release_group_id(struct mount *); | 45 | void mnt_release_group_id(struct mount *); |
| 46 | int get_dominating_id(struct mount *mnt, const struct path *root); | 46 | int get_dominating_id(struct mount *mnt, const struct path *root); |
| 47 | unsigned int mnt_get_count(struct mount *mnt); | 47 | int mnt_get_count(struct mount *mnt); |
| 48 | void mnt_set_mountpoint(struct mount *, struct mountpoint *, | 48 | void mnt_set_mountpoint(struct mount *, struct mountpoint *, |
| 49 | struct mount *); | 49 | struct mount *); |
| 50 | void mnt_change_mountpoint(struct mount *parent, struct mountpoint *mp, | 50 | void mnt_change_mountpoint(struct mount *parent, struct mountpoint *mp, |