Re: [PATCH v6 1/2] Ethernet driver for the WIZnet W5300 chip

[Mike Sinkovsky <msink@permonline.ru>]

04.04.2012 16:01, Mark Brown написал:
> On Wed, Apr 04, 2012 at 12:18:12PM +0600, Mike Sinkovsky wrote:
>> 03.04.2012 19:36, Mark Brown написал:
>
>>> ...you use devm_request_threaded_irq() here and rely on it for cleanup.
>>> Are you sure there's no possibility of the interrupt firing after you
>>> start to tear down the device?
>
>>> By using a specifically threaded IRQ you're also adding a performance
>>> overhead for no good reason if you can call netdev_carrier_*() from IRQ
>>> context and the GPIO is capable of generating a hard IRQ.  If you use
>>> request_any_context_irq() instead then the driver will get a hard IRQ if
>>> that's supported.
>
>> There isn't devm* variant of request_any_context_irq(), and using
>> plain version looks inconsistent with other resources handling.
>> Anyway, this is not performance critical procedure, and latency
>> around 100 millisecond is acceptable. Some our boards even don't
>> have this gpio at all, and nothing bad happens, just userspace
>> doesn't know is carrier on or off.
>
> None of this addresses the primary concern which is that because you're
> not (as far as I can tell) ensuring that the interrupt won't fire the
> driver might crash if the interrupt fires in between the resources it
> needs to handle the interrupt being deallocated and the interrupt being
> unregistered.  Managed interrupts are relatively tricky to use because
> of this issue, with most things like memory it doesn't matter exactly
> when it's deallocated but interrupts can potentially trigger actions
> themselves.
>
> If you can use it and devm_request_any_context_irq() doesn't exist it'd
> be better to add it.

Ok, now I understand - better don't use devm* functions for interrupts, 
use plain request* in probe() and free_irq() in remove().

Will post v8.


-- 
Mike