linux-kernel - Re: [PATCH] gpiolib: Show correct direction from the beginning

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] [day] [month] [year] [list]

Message-ID: <CACRpkdYV+A9rUZvYN_Vh3wD6Vb73qj8=SkHJCJ0ZiTWP5Ojw1A@mail.gmail.com>
Date:   Fri, 21 Sep 2018 09:07:38 -0700
From:   Linus Walleij <linus.walleij@...aro.org>
To:     timur@...nel.org
Cc:     Ricardo Ribalda Delgado <ricardo.ribalda@...il.com>,
        Stephen Boyd <swboyd@...omium.org>,
        "open list:GPIO SUBSYSTEM" <linux-gpio@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] gpiolib: Show correct direction from the beginning

On Thu, Sep 20, 2018 at 7:05 PM Timur Tabi <timur@...nel.org> wrote:
> On 9/20/18 5:36 PM, Linus Walleij wrote:
> > What I mean is that $SUBJECT patch might not hurt Qualcomms
> > GPIOs (not crash the platform) if and only if it is augmented to not
> > try to get the initial direction from lines masked off in .valid_mask
> > if .need_valid_mask is true.
> >
> > Whether it makes sense semantically is a different debate, but it
> > seems possible to reintroduce calling .get_direction() without
> > hurting anyone.
>
> That means that all the logic for checking valid_mask needs to be added
> to the chip driver's .get_direction() function.  We can add that logic
> to msm_gpio_get_direction (at one point, I had a patch that did that,
> but it was rejected).

Nah, what is in patch v2 is better, just checking it when we need to.

> My concern is: what if a driver depends on a .request call being made
> (in order to configure muxes, for example) before touching the hardware?

Hm. That is a good question.

I wonder if we have that problem in practice. If this happens, maybe the
driver needs to keep track of stuff a bit. I think if we just loop request
over everything we could disturb other mux set-up.

> I wonder if this is something that really should be handled in the
> driver's .probe function.  The driver should collect that information
> and pass it to add_data.

I see the idea, but it seems complicated compared to just calling
the callbacks. Let's try the v2 patch approach first.

Yours,
Linus Walleij