Drivers taking different actions depending on sleep state

[f.fainelli@gmail.com (Florian Fainelli)]

On 06/09/2017 03:53 PM, Rafael J. Wysocki wrote:
> Hi,
> 
> On Fri, Jun 9, 2017 at 5:20 PM, Mason <slash.tmp@free.fr> wrote:
>> Hello,
>>
>> I read the "Sleep States" documentation:
>> https://www.kernel.org/doc/Documentation/power/states.txt
>>
>> It mentions /sys/power/mem_sleep but I don't have that in 4.9
>> # ll /sys/power/
>> -rw-r--r--    1 root     root          4096 Jan  1 00:31 pm_async
>> -rw-r--r--    1 root     root          4096 Jan  1 00:31 pm_freeze_timeout
>> -rw-r--r--    1 root     root          4096 Jan  1 00:31 state
>> -rw-r--r--    1 root     root          4096 Jan  1 00:31 wakeup_count
>>
>> # cat /sys/power/state
>> freeze mem
>>
>> Currently my platform's "mem" is a true suspend-to-RAM trigger,
>> where drivers are supposed to save their state (register values
>> will be lost), then Linux hands control over to firmware which
>> enables RAM self-refresh and powers the chip down. When the system
>> resumes, drivers restore their state from their copy in memory.
>>
>> One driver is responsible for loading/unloading microcode running
>> on the DSPs. This operation is required only when powering down
>> the chip, but it should be avoided for "low-latency" sleeps.
>>
>> The problem is that, if I understand correctly, drivers have no way
>> of knowing which sleep state is being entered/exited?
>>
>> How can I have the microcode driver take different decisions
>> based on the sleep state?
> 
> The cleanest way would be to run that code from one of the platform
> suspend hooks that receive information on what sleep state is to be
> entered.

I am not sure this would be cleaner, because that would create a tighter
dependency between different drivers, each of them having their
suspend/resume routings and the driver that implements the
platform_suspend_ops, that could also create some nice layering
violations and some difficult to solve dependencies.

> 
> Alternatively, those hooks can set/clear flags that can be accessed by
> drivers, but that of course may your drivers depend on the platform
> (still, in the microcode case the driver seems to be
> platform-dependent anyway).

It may be platform dependent, but the actual system-wide suspend/resume
implementations can vary a lot. For example you may have started with
some particular CPU architecture on your platforms, with one driver
implementing an instance of platform_suspend_ops, and then as you moved
to another CPU architecture, some of that could be managed by a generic
driver (e.g: ARM SCPI, ACPI etc. etc.).

The same HW blocks are likely to be present on these different SoCs, and
have the same requirements where they need to see a slightly different
path taken on suspend/resume. If we have to patch both the "legacy"
platform_suspend_ops, and the "new" platform_suspend_ops that does not
really scale.

Would it be that much of a stretch if we reflected e.g:
PM_SUSPEND_STANDBY, PM_SUSPEND_MEM into the pm_message_t that is
communicated to platform_driver::suspend and platform_driver::resume?

Thanks!
-- 
Florian