Re: [PATCH v5 8/8] gpiolib: notify user-space about in-kernel line state changes

[Bartosz Golaszewski <brgl@xxxxxxxx>]

On Thu, Oct 24, 2024 at 2:00 PM Klara Modin <klarasmodin@xxxxxxxxx> wrote:
>
> On 2024-10-24 13:45, Bartosz Golaszewski wrote:
> > On Thu, 24 Oct 2024 13:34:11 +0200, Klara Modin <klarasmodin@xxxxxxxxx> said:
> >>
> >> I think I hit the same, or a similar bug, on my Edgerouter 6P (Cavium
> >> Octeon III):
> >>
> >> CPU 3 Unable to handle kernel paging request at virtual address
> >> 0000000000000000, epc == ffffffff817a40c8, ra == ffffffff817a40a4
> >> Oops[#1]:
> >> CPU: 3 UID: 0 PID: 1 Comm: swapper/0 Not tainted
> >> 6.12.0-rc4-next-20241023-00032-g01c8e35f8d89 #84
> >> ...
> >> Call Trace:
> >> gpiod_direction_output (drivers/gpio/gpiolib.c:4164
> >> drivers/gpio/gpiolib.c:2700 drivers/gpio/gpiolib.c:2694)
> >> i2c_register_adapter (drivers/i2c/i2c-core-base.c:396
> >> drivers/i2c/i2c-core-base.c:422 drivers/i2c/i2c-core-base.c:434
> >> drivers/i2c/i2c-core-base.c:1574)
> >> octeon_i2c_probe (drivers/i2c/busses/i2c-octeon-platdrv.c:247)
> >>
> >> (the complete log is attached)
> >>
> >> Unfortunately the oops remains after applying the diff and the call
> >> trace looks to be the same.
> >>
> >> Please let me know if there's anything else you need.
> >>
> >
> > Hmm so it's desc->gdev that is NULL... Could you try the following:
> >
> > diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> > index ae758ba6dc3d..c95c79ea2b49 100644
> > --- a/drivers/gpio/gpiolib.c
> > +++ b/drivers/gpio/gpiolib.c
> > @@ -1026,6 +1026,9 @@ int gpiochip_add_data_with_key(struct gpio_chip
> > *gc, void *data,
> >               }
> >       }
> >
> > +     for (desc_index = 0; desc_index < gc->ngpio; desc_index++)
> > +             gdev->descs[desc_index].gdev = gdev;
> > +
> >       ATOMIC_INIT_NOTIFIER_HEAD(&gdev->line_state_notifier);
> >       BLOCKING_INIT_NOTIFIER_HEAD(&gdev->device_notifier);
> >
> > @@ -1055,8 +1058,6 @@ int gpiochip_add_data_with_key(struct gpio_chip
> > *gc, void *data,
> >       for (desc_index = 0; desc_index < gc->ngpio; desc_index++) {
> >               struct gpio_desc *desc = &gdev->descs[desc_index];
> >
> > -             desc->gdev = gdev;
> > -
> >               if (gc->get_direction && gpiochip_line_is_valid(gc, desc_index)) {
> >                       assign_bit(FLAG_IS_OUT,
> >                                  &desc->flags, !gc->get_direction(gc, desc_index));
> >
> > I'm wondering if this is not an earlier commit.
> >
> > Bart
>
> That doesn't seem to change anything significantly:
>

I managed to reproduce it by using optional GPIOs. I sent out a fix
with both of you in Cc, please give it a try.

Bart