linux-kernel - Re: [PATCH v2 0/2] Add GPIO level-sensitive interrupt support

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 for Android: free password hash cracker in your pocket

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

Message-ID: <66520848-f630-e210-aecb-96c8828605b7@st.com>
Date:   Thu, 20 Feb 2020 14:09:31 +0100
From:   Alexandre Torgue <alexandre.torgue@...com>
To:     Marek Vasut <marex@...x.de>, Thomas Gleixner <tglx@...utronix.de>,
        Jason Cooper <jason@...edaemon.net>,
        Marc Zyngier <marc.zyngier@....com>,
        Linus Walleij <linus.walleij@...aro.org>
CC:     <linux-arm-kernel@...ts.infradead.org>,
        <linux-kernel@...r.kernel.org>, <linux-gpio@...r.kernel.org>
Subject: Re: [PATCH v2 0/2] Add GPIO level-sensitive interrupt support

Hi Marek

On 2/19/20 6:24 PM, Marek Vasut wrote:
> On 2/19/20 10:20 AM, Alexandre Torgue wrote:
> Hi,
> [...]
>>>>>> This series adds the possibility to handle gpio interrupts on level.
>>>>>>
>>>>>> GPIO hardware block is directly linked to EXTI block but EXTI handles
>>>>>> external interrupts only on edge. To be able to handle GPIO
>>>>>> interrupt on
>>>>>> level a "hack" is done in gpio irq chip: parent interrupt (exti irq
>>>>>> chip)
>>>>>> is retriggered following interrupt type and gpio line value.
>>>>>>
>>>>>> In exti irq chip, retrigger ops function is added.
>>>>>
>>>>> btw. this might be unrelated, but is it possible to have e.g. gpioC2
>>>>> set
>>>>> as trigger-level-low and gpioD2 set as trigger-edge-falling ? It seems
>>>>> 8eb2dfee9fb1 ("pinctrl: stm32: add lock mechanism for irqmux
>>>>> selection")
>>>>> prevents that.
>>>>>
>>>>
>>>> No it's not possible. Each gpio line doesn't have a dedicated Exti line
>>>> Each Exti line is muxing between gpio banks.
>>>
>>> OK, that confirms my assumption.
>>>
>>>> Mapping is done as following:
>>>>
>>>> EXTI0 = A0 or B0 or C0 .... or Z0 : selected by Mux
>>>> EXTI1 = A1 or B1 or C1 ....or Z1 : selected by Mux
>>>> EXTI2 = A2 or B2 or C2 ....or Z2 : selected by Mux
>>>> ...
>>>
>>> Is it at least possible to have IRQs of the same type on the same exti
>>> line? E.g. gpioA2 of trigger-edge-falling and gpioB2
>>> trigger-edge-falling ?
>>>
>>
>> Sorry I don't catch your point. If you already succeed to get gpioA2,
>> then you will failed to get gpioB2 but looking at function call stack we
>> could get an other issue.
> 
> Considering the EXTI line limitations, I'd like to know what kind of IRQ
> input configuration is allowed/valid and what kind of configuration is
> not valid.

As a mux is used to select which GPIO[A..Z]_X has to be mapped on exti_X 
line, only one GPIO can be used on the EXTI line.

For example, on EXTI2 you could map either gpioa2 or gpiob2 or 
....gpioz2 but not gpioa2 and gpiob2 in the same time.


> 
>> Lets take example where you succeed to get gpioa2 as interrupt (using
>> interrupt bindings) and now you try to do the same for gpiob2, you will
>> have (roughly):
>>
>> stm32_gpio_irq_request_resources (for gpiob2) --> succeed
>>
>> stm32_gpio_set_type
>>   |
>>   |--> stm32_exti_set_type type -> change exti line 2 trigger registers
>>                                    with gpiob2 binding.
>>
>> stm32_gpio_domain_activate --> failed as exti line2 is already used
>>                     by gpioa2.
>>
>> So as stm32_gpio_set_type is called before checking than exti line is
>> available, type could be changed and behavior of gpioa2 interrupt broken.
>>
>> Solution would be to move the exti line mux check from
>> stm32_gpio_domain_activate to  stm32_gpio_irq_request_resources callback.
> 
> So the hardware does support using both gpioA2 and gpioB2 as an
> interrupt source, for different drivers, if they are of the same
> interrupt type. Except the current implementation does not permit that.
> 

No hardware doesn't allow it for reason explain above.

> If the interrupt types are different, that is not supported by the hardware.
> 
> Correct ?
>