Re: [PATCH 2/4] acpi: allow for an override to set _REV

["Rafael J. Wysocki" <rjw@rjwysocki.net>]

On Thursday, May 14, 2015 03:31:26 PM Dominik Brodowski wrote:
> By using a module parameter named acpi.supported_rev=<value>, the BIOS
> may be told a different supported ACPI revision compared to the default
> (which currently is 5, but will be modified to 2 when the revert of
> b1ef29725865 is reverted).
> 
> Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
> ---
>  Documentation/kernel-parameters.txt | 14 ++++++++++++++
>  drivers/acpi/osl.c                  | 16 ++++++++++++++++
>  2 files changed, 30 insertions(+)
> 
> diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
> index 61ab162..75f1f8e 100644
> --- a/Documentation/kernel-parameters.txt
> +++ b/Documentation/kernel-parameters.txt
> @@ -335,6 +335,20 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
>  			to assume that this machine's pmtimer latches its value
>  			and always returns good values.
>  
> +	acpi.rev= [HW,ACPI]
> +			Tell ACPI BIOS the supported ACPI REV
> +			Format: <int> in range 0..5
> +			Up to and including Linux v4.1, the BIOS was told which
> +			ACPI revision the ACPICA subsystem in Linux actually
> +			supports (which was 5 at the time); from v4.2 on, this
> +			value will be set statically to 2 to match the behavior
> +			of other ACPI implementations. As some BIOS may operate
> +			differently depending on which value _REV is set to, this
> +			parameter offers the capability to specify what to export
> +			to the BIOS. Note that such changes in the behavior of
> +			the BIOS may only be visible after cold booting the
> +			system with this parameter _twice_.
> +
>  	acpi_sci=	[HW,ACPI] ACPI System Control Interrupt trigger mode
>  			Format: { level | edge | high | low }
>  
> diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
> index db14a66..caa52f7 100644
> --- a/drivers/acpi/osl.c
> +++ b/drivers/acpi/osl.c
> @@ -538,6 +538,11 @@ acpi_os_get_physical_address(void *virt, acpi_physical_address * phys)
>  
>  static char acpi_os_name[ACPI_MAX_OVERRIDE_LEN];
>  
> +/* acpi_supported_rev is 0 in case of no override; overrides are limited to
> + *  values between 1 to 5. To simplify casting, use an unsigned long */
> +static unsigned long acpi_supported_rev;
> +module_param_named(rev, acpi_supported_rev, ulong, 0);
> +
>  acpi_status
>  acpi_os_predefined_override(const struct acpi_predefined_names *init_val,
>  			    char **new_val)
> @@ -552,6 +557,17 @@ acpi_os_predefined_override(const struct acpi_predefined_names *init_val,
>  		*new_val = acpi_os_name;
>  	}
>  
> +	if (!memcmp(init_val->name, "_REV", 4) && (acpi_supported_rev > 0)) {
> +		if (acpi_supported_rev <= 5) {

So the only value that would really make sense here is 5.

1 should never ever be used with Linux, 2 is the default, 3 is equivalent to 5
for all practical purposes and 4 has never been used in practice, so it is
meaningless.

I'd be better to rename the command line switch to acpi.rev_override and
simply do "acpi_supported_rev = 5" for it as well as in acpi_set_supported_rev()
in [3/4].

> +			printk(KERN_INFO PREFIX
> +				"Overriding _REV definition to %lu\n",
> +				acpi_supported_rev);
> +			*new_val = (char *) acpi_supported_rev;
> +		} else
> +			printk(KERN_INFO PREFIX
> +				"_REV override must be between 1 to 5");
> +	}
> +
>  	return AE_OK;
>  }

Rafael