snd_seq_poll() calls snd_seq_write_pool_allocated() that reads out a
field in client->pool object, while it can be updated concurrently via
ioctls, as reported by syzbot. The data race itself is harmless, as
it's merely a poll() call, and the state is volatile. OTOH, the read
out of poll object info from the caller side is fragile, and we can
leave it better in snd_seq_pool_poll_wait() alone.
A similar pattern is seen in snd_seq_kernel_client_write_poll(), too,
which is called from the OSS sequencer.
This patch drops the pool checks from the caller side and add the
pool->lock in snd_seq_pool_poll_wait() for better data consistency.
Reported-by: syzbot+2d373c9936c00d7e120c@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/67c88903.050a0220.15b4b9.0028.GAE@google.com
Link: https://patch.msgid.link/20250307084246.29271-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
if (snd_seq_file_flags(file) & SNDRV_SEQ_LFLG_OUTPUT) {
/* check if data is available in the pool */
- if (!snd_seq_write_pool_allocated(client) ||
- snd_seq_pool_poll_wait(client->pool, file, wait))
+ if (snd_seq_pool_poll_wait(client->pool, file, wait))
mask |= EPOLLOUT | EPOLLWRNORM;
}
if (client == NULL)
return -ENXIO;
- if (! snd_seq_write_pool_allocated(client))
- return 1;
if (snd_seq_pool_poll_wait(client->pool, file, wait))
return 1;
return 0;
poll_table *wait)
{
poll_wait(file, &pool->output_sleep, wait);
+ guard(spinlock_irq)(&pool->lock);
return snd_seq_output_ok(pool);
}