Re: [bug report] gpio: remove gpio_lock

[Dan Carpenter <dan.carpenter@xxxxxxxxxx>]

On Wed, Feb 14, 2024 at 09:55:17AM +0100, Bartosz Golaszewski wrote:
> On Wed, Feb 14, 2024 at 8:31 AM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
> >
> > Hello Bartosz Golaszewski,
> >
> > The patch 35b545332b80: "gpio: remove gpio_lock" from Jan 12, 2024
> > (linux-next), leads to the following Smatch static checker warning:
> >
> >         drivers/gpio/gpiolib.c:120 desc_set_label()
> >         warn: sleeping in atomic context
> >
> > drivers/gpio/gpiolib.c
> >     115 static int desc_set_label(struct gpio_desc *desc, const char *label)
> >     116 {
> >     117         const char *new = NULL, *old;
> >     118
> >     119         if (label) {
> > --> 120                 new = kstrdup_const(label, GFP_KERNEL);
> >                                                    ^^^^^^^^^^
> > This used to be GFP_ATOMIC
> >
> 
> This not being GFP_ATOMIC is the least of the problems, we could
> simply change it back...
> 
> >     121                 if (!new)
> >     122                         return -ENOMEM;
> >     123         }
> >     124
> >     125         old = rcu_replace_pointer(desc->label, new, 1);
> >     126         synchronize_srcu(&desc->srcu);
> 
> ... but this synchronize_srcu() call will trigger the same issue.

Hm...  Smatch only triggers a warning for the first sleep.  But then
Smatch doesn't consider synchronize_srcu() as a sleeping function
because of the first return before the might_sleep() so if it did
trigger a static checker warning it would be inside the synchronize_srcu()
function.  And unfortunately I ignore synchronize_srcu() warnings
because there are some kind of old bugs which trigger that warning...

Sadness all around on that count, I guess.

regards,
dan carpenter