Re: new locking in change_termios breaks USB serial drivers

[Greg KH <greg@kroah.com>]

On Fri, Oct 01, 2004 at 12:36:09PM +0100, Alan Cox wrote:
> On Gwe, 2004-10-01 at 11:40, Al Borchers wrote:
> > Unfortunately, many USB serial drivers' set_termios functions
> > send an urb to change the termios settings and sleep waiting for
> > it to complete.
> > 
> > I just looked quickly, but it seems belkin_sa.c, digi_acceleport.c,
> > ftdi_sio.c, io_ti.c, kl5usb105.c, mct_u232.c, pl2303.c, and whiteheat.c
> > all sleep in their set_termios functions.
> > 
> > If this locking in change_termios() stays, we are going to have to
> > fix set_termios in all of these drivers.  I am updating io_ti.c right
> > now.
> 
> How much of a problem is this, would it make more sense to make the
> termios locking also include a semaphore to serialize driver side events
> and not the spin lock ?

It would make the usb-serial drivers much simpler if this was turned
into a semaphore.

> We need some kind of locking there otherwise multiple parallel termios
> setters resulting in truely strange occurences because driver authors
> don't think about 64 parallel executions of ->change_termios()
> 
> I can switch the lock around if you want.

I'm all for it, if the tty core isn't messing with any line settings
from interrupts.

thanks,

greg k-h