linux-kernel - Re: [PATCH v2 5/9] gpiolib: Get rid of ARCH_NR

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:   Fri, 2 Sep 2022 15:22:20 +0000
From:   Christophe Leroy <christophe.leroy@...roup.eu>
To:     Andy Shevchenko <andy.shevchenko@...il.com>
CC:     Linus Walleij <linus.walleij@...aro.org>,
        Bartosz Golaszewski <brgl@...ev.pl>,
        Geert Uytterhoeven <geert+renesas@...der.be>,
        Keerthy <j-keerthy@...com>, Russell King <linux@...linux.org.uk>,
        Arnd Bergmann <arnd@...db.de>,
        Jonathan Corbet <corbet@....net>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        "H. Peter Anvin" <hpa@...or.com>,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>,
        Davide Ciminaghi <ciminaghi@...dd.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        "open list:GPIO SUBSYSTEM" <linux-gpio@...r.kernel.org>,
        linux-arm Mailing List <linux-arm-kernel@...ts.infradead.org>,
        Linux-Arch <linux-arch@...r.kernel.org>,
        Linux Documentation List <linux-doc@...r.kernel.org>,
        "maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)" <x86@...nel.org>
Subject: Re: [PATCH v2 5/9] gpiolib: Get rid of ARCH_NR_GPIOS



Le 02/09/2022 à 16:58, Andy Shevchenko a écrit :
> On Fri, Sep 2, 2022 at 4:57 PM Christophe Leroy
> <christophe.leroy@...roup.eu> wrote:
>>
>> Since commit 14e85c0e69d5 ("gpio: remove gpio_descs global array")
>> there is no limitation on the number of GPIOs that can be allocated
>> in the system since the allocation is fully dynamic.
>>
>> ARCH_NR_GPIOS is today only used in order to provide downwards
>> gpiobase allocation from that value, while static allocation is
>> performed upwards from 0. However that has the disadvantage of
>> limiting the number of GPIOs that can be registered in the system.
>>
>> To overcome this limitation without requiring each and every
>> platform to provide its 'best-guess' maximum number, rework the
>> allocation to allocate upwards, allowing approx 2 millions of
>> GPIOs.
>>
>> In order to still allow static allocation for legacy drivers, define
>> GPIO_DYNAMIC_BASE with the value 512 as the start for dynamic
>> allocation. The 512 value is chosen because it is the end of
>> the current default range so all current static allocations are
>> expected to be below that value. Of course that's just a rough
>> estimate based on the default value, but assuming static
>> allocations come first, even if there are more static allocations
>> it should fit under the 512 value.
>>
>> In the future, it is expected that all static allocations go away
>> and then dynamic allocation will be patched to start at 0.
> 
> Eventually we have to get rid of gpio_is_valid() completely...
> But this is another story.
> Reviewed-by: Andy Shevchenko <andy.shevchenko@...il.com>

Yes that could be done as a follow-up.

There are about 300 call sites.

Should simply replace gpio_is_valid(gpio) by gpio >= 0. And then verify 
that the check is really required. But needs to check signness of gpio 
at every place.

First look seems already promissing:


int gpio_request_one(unsigned gpio, unsigned long flags, const char *label)
{
	struct gpio_desc *desc;
	int err;

	desc = gpio_to_desc(gpio);

	/* Compatibility: assume unavailable "valid" GPIOs will appear later */
	if (!desc && gpio_is_valid(gpio))
		return -EPROBE_DEFER;