Re: [PATCH 2/2] hso: fix deadlock when receiving bursts of data

[Olivier Sobrie <olivier@sobrie.be>]

On Mon, Jul 07, 2014 at 12:55:44PM +0000, David Laight wrote:
> From: Olivier Sobrie 
> > Hi David,
> > 
> > On Mon, Jul 07, 2014 at 09:13:53AM +0000, David Laight wrote:
> > > From: Olivier Sobrie
> > > > When the module sends bursts of data, sometimes a deadlock happens in
> > > > the hso driver when the tty buffer doesn't get the chance to be flushed
> > > > quickly enough.
> > > >
> > > > To avoid this, first, we remove the endless while loop in
> > > > put_rx_bufdata() which is the root cause of the deadlock.
> > > > Secondly, when there is no room anymore in the tty buffer, we set up a
> > > > timer of 100 msecs to give a chance to the upper layer to flush the tty
> > > > buffer and make room for new data.
> > >
> > > What is the timer for?
> > > You need to get the sending code woken up by the urb completion.
> > 
> > In put_rxbuf_data() (which can be called under irq disabled),
> > tty_flip_buffer_push() is called and schedules a push of the tty buffer.
> > When the buffer is full, I give some time to the above layer in order
> > to flush it.
> > The timer is used to recall put_rxbuf_data_and_resubmit_bulk_urb()
> > later in order to read the remaining data stored in
> > "urb->transfer_buffer" and then to resubmit the urb to receive more data
> > from the gsm module.
> > I don't understand what you mean by "getting the sending code woken up".
> > Calling tty_port_tty_wakeup()?? We are in the receive path...
> 
> Sorry, it isn't at all clear from the general description whether you
> are referring to the transmit or receive path.
> 
> 'flush' can mean all sorts of things ...
> 
> In either case a 100ms delay to data doesn't seem right at all.

An option to avoid the delay, is to replace the timer by a workqueue
and to schedule work in put_rxbuf_data_and_resubmit_bulk_urb()
when put_rxbuf_data() returns something greater than 0.
If you have better ideas, let me know.

Thanks,

-- 
Olivier