On Tue, Mar 12, 2019 at 03:50:49PM -0700, James Bottomley wrote: > On Tue, 2019-03-12 at 15:42 -0700, Tadeusz Struk wrote: > > @@ -202,13 +196,15 @@ __poll_t tpm_common_poll(struct file *file, > > poll_table *wait) > > struct file_priv *priv = file->private_data; > > __poll_t mask = 0; > > > > + mutex_lock(&priv->buffer_mutex); > > poll_wait(file, &priv->async_wait, wait); > > > > - if (!priv->response_read || priv->response_length) > > + if (priv->response_length) > > mask = EPOLLIN | EPOLLRDNORM; > > else > > mask = EPOLLOUT | EPOLLWRNORM; > > > > + mutex_unlock(&priv->buffer_mutex); > > Isn't this setting us up for a deadlock? If the work queued by write > hasn't run by the time you call poll, you take the buffer mutex and > sleep. Now if the work runs, it blocks acquiring the mutex in > tpm_async_work and never calls wakeup on async_wait. Yes, the lock should be taken after poll_wait(). /Jarkko
