ALSA: seq: Protect in-kernel ioctl calls with mutex
authorTakashi Iwai <tiwai@suse.de>
Tue, 9 Apr 2019 15:35:22 +0000 (17:35 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 19 Jun 2019 06:20:56 +0000 (08:20 +0200)
[ Upstream commit feb689025fbb6f0aa6297d3ddf97de945ea4ad32 ]

ALSA OSS sequencer calls the ioctl function indirectly via
snd_seq_kernel_client_ctl().  While we already applied the protection
against races between the normal ioctls and writes via the client's
ioctl_mutex, this code path was left untouched.  And this seems to be
the cause of still remaining some rare UAF as spontaneously triggered
by syzkaller.

For the sake of robustness, wrap the ioctl_mutex also for the call via
snd_seq_kernel_client_ctl(), too.

Reported-by: syzbot+e4c8abb920efa77bace9@syzkaller.appspotmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
sound/core/seq/seq_clientmgr.c

index 3bcd7a2f039451863bb23f800224fb9ce6a74486..692631bd4a358b7bfe785becd623afe7e6f356a0 100644 (file)
@@ -2348,14 +2348,19 @@ int snd_seq_kernel_client_ctl(int clientid, unsigned int cmd, void *arg)
 {
        const struct ioctl_handler *handler;
        struct snd_seq_client *client;
+       int err;
 
        client = clientptr(clientid);
        if (client == NULL)
                return -ENXIO;
 
        for (handler = ioctl_handlers; handler->cmd > 0; ++handler) {
-               if (handler->cmd == cmd)
-                       return handler->func(client, arg);
+               if (handler->cmd == cmd) {
+                       mutex_lock(&client->ioctl_mutex);
+                       err = handler->func(client, arg);
+                       mutex_unlock(&client->ioctl_mutex);
+                       return err;
+               }
        }
 
        pr_debug("ALSA: seq unknown ioctl() 0x%x (type='%c', number=0x%02x)\n",