linux-kernel - Re: [PATCH 7/7] gpio/tegra: convert to use linear irqdomain

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: <CACRpkdbknYsVKTMRzeDRo5DPfnSUSUwCNd2GKVRO_YOwrNdBFA@mail.gmail.com>
Date:	Fri, 19 Oct 2012 12:17:09 +0200
From:	Linus Walleij <linus.walleij@...aro.org>
To:	Stephen Warren <swarren@...dotorg.org>,
	Rob Herring <rob.herring@...xeda.com>,
	Grant Likely <grant.likely@...retlab.ca>
Cc:	linux-kernel@...r.kernel.org, Stephen Warren <swarren@...dia.com>
Subject: Re: [PATCH 7/7] gpio/tegra: convert to use linear irqdomain

On Wed, Oct 17, 2012 at 8:32 PM, Stephen Warren <swarren@...dotorg.org> wrote:
> On 10/16/2012 04:33 PM, Stephen Warren wrote:
>> On 10/16/2012 01:23 PM, Linus Walleij wrote:
>>> The MXS driver tries to do the work of irq_domain_add_linear()
>>> by reserving a bunch of descriptors somewhere and keeping track
>>> of the base offset, then calling irq_domain_add_legacy(). Let's
>>> stop doing that and simply use the linear IRQ domain.
>>
>> This /looks/ fine, but appears to break users of GPIOs from this module,
>> and causes a backtrace when cat /sys/kernel/debug/gpio:
> ...
>
> The following additional diff makes it work:
>
>> diff --git a/drivers/gpio/gpio-tegra.c b/drivers/gpio/gpio-tegra.c
>> index 0234162..c7c175a 100644
>> --- a/drivers/gpio/gpio-tegra.c
>> +++ b/drivers/gpio/gpio-tegra.c
>> @@ -460,7 +460,7 @@ static int __devinit tegra_gpio_probe(struct platform_device *pdev)
>>         gpiochip_add(&tegra_gpio_chip);
>>
>>         for (gpio = 0; gpio < tegra_gpio_chip.ngpio; gpio++) {
>> -               int irq = irq_find_mapping(irq_domain, gpio);
>> +               int irq = irq_create_mapping(irq_domain, gpio);
>>                 /* No validity check; all Tegra GPIOs are valid IRQs */
>>
>>                 bank = &tegra_gpio_banks[GPIO_BANK(gpio)];
>
> I wonder if perhaps the entirety of that loop and perhaps the one after
> it should be in the IRQ domain's map op - is that how all this is
> intended to work?

I've been asking the same kind of question...

Basically as far as I understand:

- If you use a linear IRQ domain you need to call irq_create_mapping()
 atleast once, and at that point the IRQ descriptor will be dynamically
 allocated. On subsequent calls, irq_find_mapping() can be used.

- If using legacy IRQ domains, descriptors are supposed to be
 allocated elsewhere, and you can just use irq_find_mapping(),
 but irq_create_mapping() does not hurt, because that call will
 check if a descriptor is already available.

- With the simple domain (as augmented by myself) the mappins
  will be dynamically created if the passed IRQ base is <= 0
  else it falls through to creating a linear domain, so in this case
  to be certain irq_create_mapping() should also be called at
  least once.

I'm not overly happy with these semantics :-(

Basically irq_find_mapping() may be a fragile and dangerous
optimization path.

I've folded the above into my patch.

Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/