On Wed, 2012-10-31 at 10:15 -0400, Steven Rostedt wrote:
> On Wed, 2012-10-31 at 23:02 +0900, anish kumar wrote:
> > From: anish kumar <anish198519851...@gmail.com>
> >
> > As no one is using the return value of irq_work_queue function
> > it is better to just make it void.
> >
> > This patch is just a way to understand if there is some future
> > plan to use it but in any case please let me know the reason.
> > ---
> > kernel/irq_work.c | 21 ++++++++++-----------
> > 1 files changed, 10 insertions(+), 11 deletions(-)
> >
> > diff --git a/kernel/irq_work.c b/kernel/irq_work.c
> > index 1588e3b..4a9a44c 100644
> > --- a/kernel/irq_work.c
> > +++ b/kernel/irq_work.c
> > @@ -32,21 +32,21 @@ static DEFINE_PER_CPU(struct llist_head, irq_work_list);
> > /*
> > * Claim the entry so that no one else will poke at it.
> > */
> > -static bool irq_work_claim(struct irq_work *work)
> > +static void irq_work_claim(struct irq_work *work)
> > {
> > unsigned long flags, nflags;
> >
> > for (;;) {
> > flags = work->flags;
> > if (flags & IRQ_WORK_PENDING)
> > - return false;
> > + return;
> > nflags = flags | IRQ_WORK_FLAGS;
> > if (cmpxchg(&work->flags, flags, nflags) == flags)
> > break;
> > cpu_relax();
> > }
> >
> > - return true;
> > + return;
> > }
> >
> > void __weak arch_irq_work_raise(void)
> > @@ -79,15 +79,14 @@ static void __irq_work_queue(struct irq_work *work)
> > *
> > * Can be re-enqueued while the callback is still in progress.
> > */
> > -bool irq_work_queue(struct irq_work *work)
> > +void irq_work_queue(struct irq_work *work)
> > {
> > - if (!irq_work_claim(work)) {
> > - /*
> > - * Already enqueued, can't do!
> > - */
> > - return false;
> > - }
> > -
> > + /*
> > + * This function either will claim the entry to queue
> > + * the work or if the work is already queued and is in
> > + * pending state then it will simply return.
> > + */
> > + irq_work_claim(work)
>
> Um, no.
>
> If the state was already pending, we will corrupt the llist node of the
> work if we call irq_work_queue(). You must check the return value of
> irq_work_claim() and return if it fails. You can not call
> __irq_work_queue() if irq_work_claim() does not succeed.
>
> The return value of irq_work_queue() can be ignored, but not
> irq_work_claim().
Oh I didn't see that logic properly and rightly pointed out by you, we
should _just_ return instead of queuing the work if the state was
already pending.
>
> -- Steve
>
> > __irq_work_queue(work);
> > return true;
> > }
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
