On Sun, Dec 23, 2007 at 03:17:35PM +0300, Anton Vorontsov wrote:
> On Sun, Dec 23, 2007 at 02:59:35PM +0300, Anton Vorontsov wrote:
> [..]
> > > > +static inline void fsl_upm_start_pattern(struct fsl_upm *upm, u32
> > > > pat_offset)
> > > > +{
> > > > + spin_lock_irqsave(&upm_lock, upm_lock_flags);
> > >
> > > I may be wrong, but don't we need the "flags" argument to
> > > spin_lock_irqsave to be on the stack? And the save and restore to be in
> > > the same function?
> >
> > In general case, yes. Here, not exactly. We have to grab a lock at the
> > start(), do runs(), and release a lock at the end():
>
> Ugh, that's stupid of course. flags are indeed should be on the stack.
> So, what I can use here is a mutex, and thus forbid using these
> routines from the isrs. Another option would be disabling interrupts
> and getting plain lock, but that is ugly. So will use a mutex.
Ignore me please, I should had more sleep today. For God's sake,
why I've just said a "mutex"?.. I should just get a plain lock and
forget about isrs.
--
Anton Vorontsov
email: [EMAIL PROTECTED]
backup email: [EMAIL PROTECTED]
irc://irc.freenode.net/bd2
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
