On Wed, Feb 20, 2013 at 05:05:10PM +0900, Jingoo Han wrote:
> On Wednesday, February 20, 2013 4:31 PM, Dmitry Torokhov wrote:
> >
> > Hi Jongoo,
> >
> > On Wed, Feb 20, 2013 at 03:12:38PM +0900, Jingoo Han wrote:
> > > Use devm_request_irq() and devm_kzalloc() to make cleanup paths
> > > more simple.
> > >
> >
> > ...
> >
> > > @@ -1269,9 +1266,7 @@ static int __exit menelaus_remove(struct i2c_client
> > > *client)
> > > {
> > > struct menelaus_chip *menelaus = i2c_get_clientdata(client);
> > >
> > > - free_irq(client->irq, menelaus);
> > > flush_work(&menelaus->work);
> > > - kfree(menelaus);
> > > the_menelaus = NULL;
> > > return 0;
> >
> > This conversion is certainly wrong - you really want to disable IRQ and
> > then wait for the scheduled work to finish before freeing memory. Here
> > you flush work but nothing stops IRQ from firing and scheduling that
> > work again.
>
> Yes, you're right.
> I will use devm_free_irq() before flush_work().
Why change it at all if you have to call it manually in both error
unwinding and menelaus_remove() cases?
BTW, that __exit markup on menelaus_remove() is surprising... I am
pretty sure it can be unbound via sysfs and so there will be a nasty
oops.
--
Dmitry
--
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/
