Hi Lal,
Replying to all this time....
On Tue, Aug 24, 2010 at 10:18 PM, Lal wrote:
> Is below code race free?
>
> void process_context_function(void)
> {
> /* interrupts enabled here */
You need to call spin_lock_irqsave
> spin_lock(&lock);
> modify_critical_section_list();
And spin_lock_irqrestore here.
> spin_unlock(&lock);
On a uniprocessor machine spin_lock is a no-op (unless spinlock
debugging is enabled).
> }
>
> irqreturn_t a2091_intr (int irq, void *_instance)
> {
Why us trylock? Why not just use spin_lock? If the only interrupt
which modifies the critical section list is this one, then you can use
plain spin_lock/unlock, otherwise you may need to use
spin_lock_irqsave/restore (depends on how the call to request_irq is
made).
> if(spin_trylock(&lock)) {
> modify_critical_section_list();
> spin_unlock(&lock);
> }
> else {
> /* skip */
> }
>
> return IRQ_HANDLED;
> }
> More specifically, can spin_lock & spin_trylock have race?
Anything can have a race. The trick is to figure out if you're
protected against the race. Just because you use spin_lock's doesn't
mean that something will be racy or that you won't have any races.
Unfortunately, it's a case of how they're used.
--
Dave Hylands
Shuswap, BC, Canada
http://www.DaveHylands.com/
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to [email protected]
Please read the FAQ at http://kernelnewbies.org/FAQ
