linux-kernel - Re: [PATCH v2 1/6] gpio: i8255: Introduce the i8255 module

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 11 Jul 2022 22:31:09 -0400
From:   William Breathitt Gray <william.gray@...aro.org>
To:     Andy Shevchenko <andy.shevchenko@...il.com>
Cc:     Linus Walleij <linus.walleij@...aro.org>,
        Bartosz Golaszewski <brgl@...ev.pl>,
        "open list:GPIO SUBSYSTEM" <linux-gpio@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Fred Eckert <Frede@...laser.com>,
        John Hentges <jhentges@...esio.com>,
        Jay Dolan <jay.dolan@...esio.com>
Subject: Re: [PATCH v2 1/6] gpio: i8255: Introduce the i8255 module

On Fri, Jul 08, 2022 at 04:40:01PM +0200, Andy Shevchenko wrote:
> On Fri, Jul 8, 2022 at 1:16 AM William Breathitt Gray
> <william.gray@...aro.org> wrote:
> >
> > Exposes consumer functions providing support for Intel 8255 Programmable
> > Peripheral Interface devices. A CONFIG_GPIO_I8255 Kconfig option is
> > introduced; modules wanting access to these functions should select this
> > Kconfig option.
> 
> Spent much time with these chips in my teenage times :-)
> 
> Very good written library, see my comments below.
> 
> ...
> 
> > +#include <linux/compiler_types.h>
> 
> Should be simple types.h as you are using u8, etc.

Ack.

> > +#include <linux/err.h>
> > +#include <linux/export.h>
> 
> > +#include <linux/gpio/i8255.h>
> 
> gpio/driver.h ?
> 
> And since it belongs to GPIO, I would group them and move after all
> other include/* to emphasize the relationship.
> 
> Also, why is it in the global header folder? Do you expect some
> drivers outside of drivers/gpio/? Maybe we can move after when the
> user comes?

I think gpio/driver.h does make more sense for now since all the users
of library are located under drivers/gpio/. I'll move the header code
into gpio/driver.h then and adjust the includes accordingly.

> > +#include <linux/io.h>
> > +#include <linux/module.h>
> 
> ...
> 
> > +#define I8255_CONTROL_PORTCLOWER_DIRECTION BIT(0)
> > +#define I8255_CONTROL_PORTCUPPER_DIRECTION BIT(3)
> 
> Missed underscore.

Ack.

> ...
> 
> > +static u8 i8255_direction_mask(const unsigned long offset)
> > +{
> > +       const unsigned long io_port = offset / 8;
> > +       const unsigned long ppi_port = io_port % 3;
> > +
> > +       switch (ppi_port) {
> > +       case I8255_PORTA:
> > +               return I8255_CONTROL_PORTA_DIRECTION;
> > +       case I8255_PORTB:
> > +               return I8255_CONTROL_PORTB_DIRECTION;
> > +       case I8255_PORTC:
> > +               /* Port C can be configured by nibble */
> 
> > +               if (offset % 8 > 3)
> 
> I would move it to the local definition block close to offset/8. On
> some architectures this may give one assembly instruction for two
> variables.

Ack.

> > +                       return I8255_CONTROL_PORTCUPPER_DIRECTION;
> > +               return I8255_CONTROL_PORTCLOWER_DIRECTION;
> > +       default:
> > +               /* Should never reach this path */
> > +               return 0;
> > +       }
> > +}
> 
> > +void i8255_direction_input(struct i8255 __iomem *const ppi,
> > +                          u8 *const control_state, const unsigned long offset)
> > +{
> > +       const unsigned long io_port = offset / 8;
> > +       const unsigned long group = io_port / 3;
> > +
> > +       control_state[group] |= I8255_CONTROL_MODE_SET;
> > +       control_state[group] |= i8255_direction_mask(offset);
> > +
> > +       iowrite8(control_state[group], &ppi[group].control);
> 
> No I/O serialization? Can this be accessed during interrupt? (It may
> be that the code is correct, but please go through it and check with a
> question "can this register be accessed during IRQ and if yes, will
> the user get the correct / meaningful data?")

Writing to the 8255 control register for the device shouldn't be a
problem, but we do have a race condition with the control_state[group]
value. This value is accessed and modified in other functions (e.g. the
i8255_direction_output() right below) so if an interrupt occurs the
value could end up clobbered before it's written.

I'm not sure what the best approach would be here. In the subsequent
patches I have the GPIO drivers take a lock before calling these i8255_*
functions in order to synchronize access to those state arrays. Do you
think it would be better to move the sychronization lock acquisition
internally to the i8255_* functions here?

> > +}
> > +EXPORT_SYMBOL_GPL(i8255_direction_input);
> 
> Make it with a namespace. Ditto for the rest.

Ack.

William Breathitt Gray

> -- 
> With Best Regards,
> Andy Shevchenko

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)