Re: [patch 3/9] acpi: adjust register handling

["Jan Beulich" <jbeulich@novell.com>]

>>> Zhao Yakui <yakui.zhao@intel.com> 22.04.08 13:15 >>>
>On Fri, 2008-04-18 at 13:27 -0700, akpm@linux-foundation.org wrote:
>> From: "Jan Beulich" <jbeulich@novell.com>
>> 
>> acpi_hw_low_level_{read,write}() have no need to accept a NULL reg argument
>> anymore (all callers use addresses of or derived from ACPI globals), and it
>> really should always have been considered an error to call these functions in
>> such a way.
>IMO this is inappropriate. On some laptops there are some optional
>register blocks . For example: PM1b_event_block, PM1B_control_block. 
>And the optional register block will be accessed in many functions. (For
>example: sleep flowchart, power_off flowchart).
>If the AE_BAD_PARAMETER is returned by acpi_hw_low_level_read when the
>reg doesn't exist, there will be many regressions.

No, that's not the case: as said in the description, the register structures
are static (hence the addresses of them are never NULL) - an optional,
not implemented register is identified by the address field inside the
register structure being zero (which is still being accounted for, as it
was before the change).

Jan