Hi,

On 8/9/21 4:39 PM, Limonciello, Mario wrote:

This thread is a mix of top posting and bottom posting and confusing to follow.  Some inline comments below.

On 8/2/2021 21:42, Liang, Prike wrote:

quoted

[Public]

In the AMD existing S0ix system seems needn't monitor SFH idle state, meanwhile SFH is powered by S5 rail and the rail keeps ON when SOC is in S3/S0i3 state. Regards to SFH stop working after S3 resume maybe caused by power rail and device context tear down during S3 suspend. In addition, we also need implement SFH suspend callback in amd-sfh-hid like as other vendor.

I think PM callbacks for SFH will resolve the issue at hand and are the most likely the correct solution in this instance.

Ack.

quoted

quoted

-----Original Message-----
From: Huang, Ray <Ray.Huang@amd.com>
Sent: Tuesday, August 3, 2021 8:20 AM
To: Hans de Goede <redacted>
Cc: Natikar, Basavaraj <Basavaraj.Natikar@amd.com>; Shah, Nehal-
bakulchandra [off-list ref]; Stephen MacNeil
[off-list ref]; Limonciello, Mario [off-list ref];
Rafael J . Wysocki [off-list ref]; Linux PM <linux-
pm@vger.kernel.org>; linux-acpi [off-list ref]; Natikar,
Basavaraj [off-list ref]; S-k, Shyam-sundar <Shyam-
sundar.S-k@amd.com>; Liang, Prike [off-list ref]
Subject: Re: AMD laptops defaulting to S3 instead of S0ix (was amd_sfh)

+ Prike

On Mon, Aug 02, 2021 at 01:43:01PM +0200, Hans de Goede wrote:

quoted

Hi,

On 8/2/21 11:34 AM, Basavaraj Natikar wrote:

quoted

On 8/1/2021 5:15 PM, Shah, Nehal-bakulchandra wrote:

quoted

[AMD Official Use Only]

Adding few more folks

-----Original Message-----
From: Hans de Goede <redacted>
Sent: Sunday, August 1, 2021 3:17 PM
To: Stephen MacNeil <redacted>; Limonciello, Mario
[off-list ref]; Rafael J . Wysocki [off-list ref]
Cc: Linux PM <redacted>; linux-acpi
[off-list ref]
Subject: AMD laptops defaulting to S3 instead of S0ix (was amd_sfh)

Hi Rafael, Mario,

Stephen is having an issue with a recent AMD laptop (a Lenovo Ideapad

model) where Linux defaults to using S3/deep suspend instead of S0ix/s2idle.

quoted

quoted

Hi Hans, Et al.

Looks like the Lenovo platform Stephen MacNeil is using does not support

the S2Idle as the FADT flags as not set (looking at the output of the script).

So Lenovo intends the platform to be S3 this means.

Ok.

quoted

quoted

quoted

I believe it does, if it would not support s2idle at all, then this
would not be offered as an option in the "cat /sys/power/mem_sleep"

output.

I don't believe this to be correct.  s2idle is always added.

You can look at the comments (and in the code) to see this:
https://github.com/torvalds/linux/blob/5bfc75d92efd494db37f5c4c173d3639d4772966/kernel/power/suspend.c#L174

Ah, I was under the impression that s2idle would only be shown and be supported 
by the kernel when advertised as supported by the ACPI tables. So this is all just
my bad and in this case it indeed seems the intention of the vendor (Lenovo)
that the laptop will use S3 for suspend.

Sorry for the noise / confusion.

Regards,

Hans