Re: [linux-pm] [RFC PATCH 0/4] save and restore ACPI NVS memory during hibernation

["Yinghai Lu" <yhlu.kernel@gmail.com>]

On Fri, Jul 11, 2008 at 1:59 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> On Friday, 11 of July 2008, Pavel Machek wrote:
>> On Fri 2008-07-11 21:18:34, Rafael J. Wysocki wrote:
>> > On Friday, 11 of July 2008, Pavel Machek wrote:
>> > > On Wed 2008-07-09 21:37:18, Rafael J. Wysocki wrote:
>> > > > On Tuesday, 8 of July 2008, Pavel Machek wrote:
>> > > > > Hi!
>> > > > >
>> > > > > > According to the ACPI spec, ACPI NVS memory region is required to
>> > > > > > be saved/restored by OS during hibernation.
>> > > > > >
>> > > > > > Section 15.3.2 ACPI Spec 3.0b,
>> > > > > > "OSPM will call the _PTS control method some time before entering
>> > > > > > a sleeping state, to allow the platform???s AML code to update
>> > > > > > this memory image before entering the sleeping state.
>> > > > > > After the system awakes from an S4 state, OSPM will restore this
>> > > > > > memory area and call the _WAK control method to enable the BIOS
>> > > > > > to reclaim its memory image."
>> > > > > >
>> > > > > > This patch set add the mechanism to save/restore ACPI NVS memory
>> > > > > > during hibernation.
>> > > > > >
>> > > > > > Patch 01: call platform_begin before swsusp_shrink_memory.
>> > > > > >     So that we can allocate enough pages for ACPI NVS memory
>> > > > > >     before shrink the memory.
>> > > > >
>> > > > > Why is it neccessary to allocate memory for a copy? We should be able
>> > > > > to save ACPI NVS area same way we are saving kernel pages, no?
>> > > >
>> > > > Because we want to restore it from the hibernated kernel (when it gets control
>> > > > back again).
>> > >
>> > > Why is that important? So we can run some ACPI methods from hibernated
>> > > kernel before restoring it?
>> >
>> > We're supposed to restore it right prior to executing _WAK, which is after
>> > we've executed _BFS.  This is a bit theoretical, because no one seems to
>> > actaully implement _BFS, but well.
>>
>> Could we just call _BFS from the "loading" kernel, instead? That would
>> save us that copying code...
>
> I'd prefer not to, because that would require the boot kernel to support ACPI,
> which need not (and in fact should not IMO) be the case at present.
>
> Anyway, we'd have to either copy the NVS data into memory which is mapped and
> can be saved, or to map the NVS area in order to save/restore it, which I
> wouldn't like to do.

io_remap doesn't work at that time?

YH