cifs: fix use-after-free caused by invalid pointer `hostname`
authorZeng Heng <zengheng4@huawei.com>
Thu, 27 Oct 2022 12:45:28 +0000 (20:45 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 14 Dec 2022 10:37:18 +0000 (11:37 +0100)
[ Upstream commit 153695d36ead0ccc4d0256953c751cabf673e621 ]

`hostname` needs to be set as null-pointer after free in
`cifs_put_tcp_session` function, or when `cifsd` thread attempts
to resolve hostname and reconnect the host, the thread would deref
the invalid pointer.

Here is one of practical backtrace examples as reference:

Task 477
---------------------------
 do_mount
  path_mount
   do_new_mount
    vfs_get_tree
     smb3_get_tree
      smb3_get_tree_common
       cifs_smb3_do_mount
        cifs_mount
         mount_put_conns
          cifs_put_tcp_session
          --> kfree(server->hostname)

cifsd
---------------------------
 kthread
  cifs_demultiplex_thread
   cifs_reconnect
    reconn_set_ipaddr_from_hostname
    --> if (!server->hostname)
    --> if (server->hostname[0] == '\0')  // !! UAF fault here

CIFS: VFS: cifs_mount failed w/return code = -112
mount error(112): Host is down
BUG: KASAN: use-after-free in reconn_set_ipaddr_from_hostname+0x2ba/0x310
Read of size 1 at addr ffff888108f35380 by task cifsd/480
CPU: 2 PID: 480 Comm: cifsd Not tainted 6.1.0-rc2-00106-gf705792f89dd-dirty #25
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-1ubuntu1.1 04/01/2014
Call Trace:
 <TASK>
 dump_stack_lvl+0x68/0x85
 print_report+0x16c/0x4a3
 kasan_report+0x95/0x190
 reconn_set_ipaddr_from_hostname+0x2ba/0x310
 __cifs_reconnect.part.0+0x241/0x800
 cifs_reconnect+0x65f/0xb60
 cifs_demultiplex_thread+0x1570/0x2570
 kthread+0x2c5/0x380
 ret_from_fork+0x22/0x30
 </TASK>
Allocated by task 477:
 kasan_save_stack+0x1e/0x40
 kasan_set_track+0x21/0x30
 __kasan_kmalloc+0x7e/0x90
 __kmalloc_node_track_caller+0x52/0x1b0
 kstrdup+0x3b/0x70
 cifs_get_tcp_session+0xbc/0x19b0
 mount_get_conns+0xa9/0x10c0
 cifs_mount+0xdf/0x1970
 cifs_smb3_do_mount+0x295/0x1660
 smb3_get_tree+0x352/0x5e0
 vfs_get_tree+0x8e/0x2e0
 path_mount+0xf8c/0x1990
 do_mount+0xee/0x110
 __x64_sys_mount+0x14b/0x1f0
 do_syscall_64+0x3b/0x90
 entry_SYSCALL_64_after_hwframe+0x63/0xcd
Freed by task 477:
 kasan_save_stack+0x1e/0x40
 kasan_set_track+0x21/0x30
 kasan_save_free_info+0x2a/0x50
 __kasan_slab_free+0x10a/0x190
 __kmem_cache_free+0xca/0x3f0
 cifs_put_tcp_session+0x30c/0x450
 cifs_mount+0xf95/0x1970
 cifs_smb3_do_mount+0x295/0x1660
 smb3_get_tree+0x352/0x5e0
 vfs_get_tree+0x8e/0x2e0
 path_mount+0xf8c/0x1990
 do_mount+0xee/0x110
 __x64_sys_mount+0x14b/0x1f0
 do_syscall_64+0x3b/0x90
 entry_SYSCALL_64_after_hwframe+0x63/0xcd
The buggy address belongs to the object at ffff888108f35380
 which belongs to the cache kmalloc-16 of size 16
The buggy address is located 0 bytes inside of
 16-byte region [ffff888108f35380ffff888108f35390)
The buggy address belongs to the physical page:
page:00000000333f8e58 refcount:1 mapcount:0 mapping:0000000000000000 index:0xffff888108f350e0 pfn:0x108f35
flags: 0x200000000000200(slab|node=0|zone=2)
raw: 0200000000000200 0000000000000000 dead000000000122 ffff8881000423c0
raw: ffff888108f350e0 000000008080007a 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
 ffff888108f35280: fa fb fc fc fa fb fc fc fa fb fc fc fa fb fc fc
 ffff888108f35300: fa fb fc fc fa fb fc fc fa fb fc fc fa fb fc fc
>ffff888108f35380: fa fb fc fc fa fb fc fc fa fb fc fc fa fb fc fc
                   ^
 ffff888108f35400: fa fb fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 ffff888108f35480: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc

Fixes: 7be3248f3139 ("cifs: To match file servers, make sure the server hostname matches")
Signed-off-by: Zeng Heng <zengheng4@huawei.com>
Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
fs/cifs/connect.c

index c6e2a0f..a4284c4 100644 (file)
@@ -1392,6 +1392,7 @@ cifs_put_tcp_session(struct TCP_Server_Info *server, int from_reconnect)
        server->session_key.response = NULL;
        server->session_key.len = 0;
        kfree(server->hostname);
+       server->hostname = NULL;
 
        task = xchg(&server->tsk, NULL);
        if (task)