ALSA: rawmidi: Fix possible deadlock with virmidi registration
When a seq-virmidi driver is initialized, it registers a rawmidi instance with its callback to create an associated seq kernel client. Currently it's done throughly in rawmidi's register_mutex context. Recently it was found that this may lead to a deadlock another rawmidi device that is being attached with the sequencer is accessed, as both open with the same register_mutex. This was actually triggered by syzkaller, as Dmitry Vyukov reported: ====================================================== [ INFO: possible circular locking dependency detected ] 4.8.0-rc1+ #11 Not tainted ------------------------------------------------------- syz-executor/7154 is trying to acquire lock: (register_mutex#5){+.+.+.}, at: [<ffffffff84fd6d4b>] snd_rawmidi_kernel_open+0x4b/0x260 sound/core/rawmidi.c:341 but task is already holding lock: (&grp->list_mutex){++++.+}, at: [<ffffffff850138bb>] check_and_subscribe_port+0x5b/0x5c0 sound/core/seq/seq_ports.c:495 which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&grp->list_mutex){++++.+}: [<ffffffff8147a3a8>] lock_acquire+0x208/0x430 kernel/locking/lockdep.c:3746 [<ffffffff863f6199>] down_read+0x49/0xc0 kernel/locking/rwsem.c:22 [< inline >] deliver_to_subscribers sound/core/seq/seq_clientmgr.c:681 [<ffffffff85005c5e>] snd_seq_deliver_event+0x35e/0x890 sound/core/seq/seq_clientmgr.c:822 [<ffffffff85006e96>] > snd_seq_kernel_client_dispatch+0x126/0x170 sound/core/seq/seq_clientmgr.c:2418 [<ffffffff85012c52>] snd_seq_system_broadcast+0xb2/0xf0 sound/core/seq/seq_system.c:101 [<ffffffff84fff70a>] snd_seq_create_kernel_client+0x24a/0x330 sound/core/seq/seq_clientmgr.c:2297 [< inline >] snd_virmidi_dev_attach_seq sound/core/seq/seq_virmidi.c:383 [<ffffffff8502d29f>] snd_virmidi_dev_register+0x29f/0x750 sound/core/seq/seq_virmidi.c:450 [<ffffffff84fd208c>] snd_rawmidi_dev_register+0x30c/0xd40 sound/core/rawmidi.c:1645 [<ffffffff84f816d3>] __snd_device_register.part.0+0x63/0xc0 sound/core/device.c:164 [< inline >] __snd_device_register sound/core/device.c:162 [<ffffffff84f8235d>] snd_device_register_all+0xad/0x110 sound/core/device.c:212 [<ffffffff84f7546f>] snd_card_register+0xef/0x6c0 sound/core/init.c:749 [<ffffffff85040b7f>] snd_virmidi_probe+0x3ef/0x590 sound/drivers/virmidi.c:123 [<ffffffff833ebf7b>] platform_drv_probe+0x8b/0x170 drivers/base/platform.c:564 ...... -> #0 (register_mutex#5){+.+.+.}: [< inline >] check_prev_add kernel/locking/lockdep.c:1829 [< inline >] check_prevs_add kernel/locking/lockdep.c:1939 [< inline >] validate_chain kernel/locking/lockdep.c:2266 [<ffffffff814791f4>] __lock_acquire+0x4d44/0x4d80 kernel/locking/lockdep.c:3335 [<ffffffff8147a3a8>] lock_acquire+0x208/0x430 kernel/locking/lockdep.c:3746 [< inline >] __mutex_lock_common kernel/locking/mutex.c:521 [<ffffffff863f0ef1>] mutex_lock_nested+0xb1/0xa20 kernel/locking/mutex.c:621 [<ffffffff84fd6d4b>] snd_rawmidi_kernel_open+0x4b/0x260 sound/core/rawmidi.c:341 [<ffffffff8502e7c7>] midisynth_subscribe+0xf7/0x350 sound/core/seq/seq_midi.c:188 [< inline >] subscribe_port sound/core/seq/seq_ports.c:427 [<ffffffff85013cc7>] check_and_subscribe_port+0x467/0x5c0 sound/core/seq/seq_ports.c:510 [<ffffffff85015da9>] snd_seq_port_connect+0x2c9/0x500 sound/core/seq/seq_ports.c:579 [<ffffffff850079b8>] snd_seq_ioctl_subscribe_port+0x1d8/0x2b0 sound/core/seq/seq_clientmgr.c:1480 [<ffffffff84ffe9e4>] snd_seq_do_ioctl+0x184/0x1e0 sound/core/seq/seq_clientmgr.c:2225 [<ffffffff84ffeae8>] snd_seq_kernel_client_ctl+0xa8/0x110 sound/core/seq/seq_clientmgr.c:2440 [<ffffffff85027664>] snd_seq_oss_midi_open+0x3b4/0x610 sound/core/seq/oss/seq_oss_midi.c:375 [<ffffffff85023d67>] snd_seq_oss_synth_setup_midi+0x107/0x4c0 sound/core/seq/oss/seq_oss_synth.c:281 [<ffffffff8501b0a8>] snd_seq_oss_open+0x748/0x8d0 sound/core/seq/oss/seq_oss_init.c:274 [<ffffffff85019d8a>] odev_open+0x6a/0x90 sound/core/seq/oss/seq_oss.c:138 [<ffffffff84f7040f>] soundcore_open+0x30f/0x640 sound/sound_core.c:639 ...... other info that might help us debug this: Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(&grp->list_mutex); lock(register_mutex#5); lock(&grp->list_mutex); lock(register_mutex#5); *** DEADLOCK *** ====================================================== The fix is to simply move the registration parts in snd_rawmidi_dev_register() to the outside of the register_mutex lock. The lock is needed only to manage the linked list, and it's not necessarily to cover the whole initialization process. Reported-by: Dmitry Vyukov <dvyukov@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Takashi Iwai <tiwai@suse.de>
This commit is contained in:
Родитель
9f8a7658bc
Коммит
816f318b23
|
@ -1633,11 +1633,13 @@ static int snd_rawmidi_dev_register(struct snd_device *device)
|
|||
return -EBUSY;
|
||||
}
|
||||
list_add_tail(&rmidi->list, &snd_rawmidi_devices);
|
||||
mutex_unlock(®ister_mutex);
|
||||
err = snd_register_device(SNDRV_DEVICE_TYPE_RAWMIDI,
|
||||
rmidi->card, rmidi->device,
|
||||
&snd_rawmidi_f_ops, rmidi, &rmidi->dev);
|
||||
if (err < 0) {
|
||||
rmidi_err(rmidi, "unable to register\n");
|
||||
mutex_lock(®ister_mutex);
|
||||
list_del(&rmidi->list);
|
||||
mutex_unlock(®ister_mutex);
|
||||
return err;
|
||||
|
@ -1645,6 +1647,7 @@ static int snd_rawmidi_dev_register(struct snd_device *device)
|
|||
if (rmidi->ops && rmidi->ops->dev_register &&
|
||||
(err = rmidi->ops->dev_register(rmidi)) < 0) {
|
||||
snd_unregister_device(&rmidi->dev);
|
||||
mutex_lock(®ister_mutex);
|
||||
list_del(&rmidi->list);
|
||||
mutex_unlock(®ister_mutex);
|
||||
return err;
|
||||
|
@ -1677,7 +1680,6 @@ static int snd_rawmidi_dev_register(struct snd_device *device)
|
|||
}
|
||||
}
|
||||
#endif /* CONFIG_SND_OSSEMUL */
|
||||
mutex_unlock(®ister_mutex);
|
||||
sprintf(name, "midi%d", rmidi->device);
|
||||
entry = snd_info_create_card_entry(rmidi->card, name, rmidi->card->proc_root);
|
||||
if (entry) {
|
||||
|
|
Загрузка…
Ссылка в новой задаче