afs: Fix NULL deref in afs_dynroot_depopulate()
authorDavid Howells <dhowells@redhat.com>
Fri, 21 Aug 2020 09:15:12 +0000 (10:15 +0100)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 21 Aug 2020 17:56:40 +0000 (10:56 -0700)
If an error occurs during the construction of an afs superblock, it's
possible that an error occurs after a superblock is created, but before
we've created the root dentry.  If the superblock has a dynamic root
(ie.  what's normally mounted on /afs), the afs_kill_super() will call
afs_dynroot_depopulate() to unpin any created dentries - but this will
oops if the root hasn't been created yet.

Fix this by skipping that bit of code if there is no root dentry.

This leads to an oops looking like:

general protection fault, ...
KASAN: null-ptr-deref in range [0x0000000000000068-0x000000000000006f]
...
RIP: 0010:afs_dynroot_depopulate+0x25f/0x529 fs/afs/dynroot.c:385
...
Call Trace:
 afs_kill_super+0x13b/0x180 fs/afs/super.c:535
 deactivate_locked_super+0x94/0x160 fs/super.c:335
 afs_get_tree+0x1124/0x1460 fs/afs/super.c:598
 vfs_get_tree+0x89/0x2f0 fs/super.c:1547
 do_new_mount fs/namespace.c:2875 [inline]
 path_mount+0x1387/0x2070 fs/namespace.c:3192
 do_mount fs/namespace.c:3205 [inline]
 __do_sys_mount fs/namespace.c:3413 [inline]
 __se_sys_mount fs/namespace.c:3390 [inline]
 __x64_sys_mount+0x27f/0x300 fs/namespace.c:3390
 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

which is oopsing on this line:

inode_lock(root->d_inode);

presumably because sb->s_root was NULL.

Fixes: 0da0b7fd73e4 ("afs: Display manually added cells in dynamic root mount")
Reported-by: syzbot+c1eff8205244ae7e11a6@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/afs/dynroot.c

index b79879aacc02e80ddaa7cfdcd1706d547eca9a6d..7b784af604fd9fb125c71ab41720828448647e9c 100644 (file)
@@ -382,15 +382,17 @@ void afs_dynroot_depopulate(struct super_block *sb)
                net->dynroot_sb = NULL;
        mutex_unlock(&net->proc_cells_lock);
 
-       inode_lock(root->d_inode);
-
-       /* Remove all the pins for dirs created for manually added cells */
-       list_for_each_entry_safe(subdir, tmp, &root->d_subdirs, d_child) {
-               if (subdir->d_fsdata) {
-                       subdir->d_fsdata = NULL;
-                       dput(subdir);
+       if (root) {
+               inode_lock(root->d_inode);
+
+               /* Remove all the pins for dirs created for manually added cells */
+               list_for_each_entry_safe(subdir, tmp, &root->d_subdirs, d_child) {
+                       if (subdir->d_fsdata) {
+                               subdir->d_fsdata = NULL;
+                               dput(subdir);
+                       }
                }
-       }
 
-       inode_unlock(root->d_inode);
+               inode_unlock(root->d_inode);
+       }
 }