On Tue, Dec 06, 2005 at 01:24:03PM +0100, Arjan van de Ven wrote:
> On Tue, 2005-12-06 at 09:57 -0200, Luiz Fernando Capitulino wrote:
> >  Introduces URB write locking macros.
> 
> ugh.. WHY ?

Maybe we need a better description for the patch: the locking is already
there (the 'lock' field in struct sub_serial_port), and there is no
change of behaviour, just replacing a (spin_lock_t, int) by an atomic_t.

The purpose of the changes is removing the spinlock, because it is used
only to protect write_urb_busy, and if someday we need locking on other
parts, we already have a semaphore (introduced by Capitulino some time
ago, to fix a open()/close() race condition).

-- 
Eduardo

Attachment: pgpm7rW968gck.pgp
Description: PGP signature

Reply via email to