linux-kernel - Re: [PATCH] gpiolib: acpi: Add gpiolib_acpi_run_edge_events_on

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]

Date:   Tue, 27 Aug 2019 16:41:10 +0200
From:   Hans de Goede <hdegoede@...hat.com>
To:     Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
Cc:     Mika Westerberg <mika.westerberg@...ux.intel.com>,
        Linus Walleij <linus.walleij@...aro.org>,
        Bartosz Golaszewski <bgolaszewski@...libre.com>,
        Benjamin Tissoires <benjamin.tissoires@...hat.com>,
        linux-gpio@...r.kernel.org, linux-acpi@...r.kernel.org,
        linux-kernel@...r.kernel.org, stable@...r.kernel.org,
        Daniel Drake <drake@...lessm.com>,
        Ian W MORRISON <ianwmorrison@...il.com>
Subject: Re: [PATCH] gpiolib: acpi: Add gpiolib_acpi_run_edge_events_on_boot
 option and blacklist

Hi Andy,

On 26-08-19 11:11, Andy Shevchenko wrote:
> On Fri, Aug 23, 2019 at 11:52:55PM +0200, Hans de Goede wrote:
>> Another day; another DSDT bug we need to workaround...
>>
>> Since commit ca876c7483b6 ("gpiolib-acpi: make sure we trigger edge events
>> at least once on boot") we call _AEI edge handlers at boot.
>>
>> In some rare cases this causes problems. One example of this is the Minix
>> Neo Z83-4 mini PC, this device has a clear DSDT bug where it has some copy
>> and pasted code for dealing with Micro USB-B connector host/device role
>> switching, while the mini PC does not even have a micro-USB connector.
>> This code, which should not be there, messes with the DDC data pin from
>> the HDMI connector (switching it to GPIO mode) breaking HDMI support.
>>
>> To avoid problems like this, this commit adds a new
>> gpiolib_acpi_run_edge_events_on_boot kernel commandline option which
>> can be "on", "off", or "auto" (default).
>>
>> In auto mode the default is on and a DMI based blacklist is used,
>> the initial version of this blacklist contains the Minix Neo Z83-4
>> fixing the HDMI being broken on this device.
> 
>> +static int gpiolib_acpi_run_edge_events_on_boot = -1;
>> +
>> +static int __init gpiolib_acpi_run_edge_events_on_boot_setup(char *arg)
>> +{
> 
>> +	if (!strcmp(arg, "on"))
>> +		gpiolib_acpi_run_edge_events_on_boot = 1;
>> +	else if (!strcmp(arg, "off"))
>> +		gpiolib_acpi_run_edge_events_on_boot = 0;
> 
> kstrtobool() ?
> 
>> +	else if (!strcmp(arg, "auto"))
>> +		gpiolib_acpi_run_edge_events_on_boot = -1;
>> +
>> +	return 1;
>> +}
> 
>> +
>> +__setup("gpiolib_acpi_run_edge_events_on_boot=",
>> +        gpiolib_acpi_run_edge_events_on_boot_setup);
> 
> Can't we use module_param() ?
> The resulting option would be 'gpiolib_acpi.foo=...'

I was expecting that would not work, since gpiolib is a bool
not a tristate, but it seems that if there is no module-name
to use as prefix for module-parameters the kernel simply uses
the .c file name, so this works and yes, this is better then
using __setup, will fix for v2.

>> +{
> 
>> +	if (gpiolib_acpi_run_edge_events_on_boot == -1) {
>> +		if (dmi_check_system(run_edge_events_on_boot_blacklist))
>> +			gpiolib_acpi_run_edge_events_on_boot = 0;
>> +		else
>> +			gpiolib_acpi_run_edge_events_on_boot = 1;
>> +	}
> 
> Can we run this at an initcall once and use variable instead of calling a
> method below?

I was a bit worried about init ordering, but I've checked and dmi_setup()
is done as a core_initcall, so we can do this once as a postcore_initcall
which should be early enough, will fix for v2.

> 
>> +	return gpiolib_acpi_run_edge_events_on_boot;
>> +}
>> +
>>   static void acpi_gpiochip_request_irq(struct acpi_gpio_chip *acpi_gpio,
>>   				      struct acpi_gpio_event *event)
>>   {
>> @@ -170,10 +211,13 @@ static void acpi_gpiochip_request_irq(struct acpi_gpio_chip *acpi_gpio,
>>   	event->irq_requested = true;
>>   
>>   	/* Make sure we trigger the initial state of edge-triggered IRQs */
>> -	value = gpiod_get_raw_value_cansleep(event->desc);
>> -	if (((event->irqflags & IRQF_TRIGGER_RISING) && value == 1) ||
>> -	    ((event->irqflags & IRQF_TRIGGER_FALLING) && value == 0))
>> -		event->handler(event->irq, event);
>> +	if (acpi_gpiochip_run_edge_events_on_boot() &&
>> +	    (event->irqflags & (IRQF_TRIGGER_RISING | IRQF_TRIGGER_FALLING))) {
>> +		value = gpiod_get_raw_value_cansleep(event->desc);
>> +		if (((event->irqflags & IRQF_TRIGGER_RISING) && value == 1) ||
>> +		    ((event->irqflags & IRQF_TRIGGER_FALLING) && value == 0))
>> +			event->handler(event->irq, event);
>> +	}
> 

Regards,

Hans