On Thu, Mar 4, 2010 at 2:16 AM, Frederic Weisbecker <fweisbec@xxxxxxxxx> wrote:
> On Tue, Mar 02, 2010 at 09:32:16AM +0530, Manish Katiyar wrote:
>> Hello all,
>>
>> Is the following code valid ???
>>
>> foo() {
>> ...........
>> spin_lock(lock);
>> ptr = kzalloc(size, GFP_KERNEL);
>> spin_unlock(lock);
>> .....
>> }
>
>
>
> You can't because GFP_KERNEL involves the fact
> kzalloc might sleep while looking for memory
> somewhere (swapping out pages to get some free
> rooms).
>
> And sleeping while holding a spinlock may
> lead to a deadlock:
>
> Task A holds spinlock in cpu0. Task A sleeps.
> Task B is scheduled in cpu0, tries to take the
> spinlock and then deadlock for ever.
>
> You can use GFP_ATOMIC instead, but it's usually
> not recommended as the GFP_ATOMIC pool is a pretty
> limited resource.
Hi,
Thanks for the clarification. That is what I had suspected too and
wanted to clarify.
>
> The best is to allocate before you take the spinlock.
Yes, that is what I ended coding up.
Thanks again -
Manish
>
>
--
Thanks -
Manish
==================================
[$\*.^ -- I miss being one of them
==================================
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
