Re: git: a25cfcab9a9d - main - linuxkpi: Support s2idle in `pm_suspend_target_state`

On Mon, 6 Oct 2025, Aymeric Wibo wrote:

> The branch main has been updated by obiwac:
>
> URL: https://cgit.FreeBSD.org/src/commit/?id=a25cfcab9a9d2c7d733b85bf04b504369623de1c
>
> commit a25cfcab9a9d2c7d733b85bf04b504369623de1c
> Author:     Aymeric Wibo <obiwac@FreeBSD.org>
> AuthorDate: 2025-10-06 21:38:58 +0000
> Commit:     Aymeric Wibo <obiwac@FreeBSD.org>
> CommitDate: 2025-10-06 21:49:54 +0000
>
>    linuxkpi: Support s2idle in `pm_suspend_target_state`
>
>    When going to s2idle, set Linux' pm_suspend_target_state to
>    PM_SUSPEND_TO_IDLE instead of setting PM_SUSPEND_MEM (for S3) in all
>    cases.
>
>    This is necessary because more recent versions of the amdgpu DRM drivers
>    have different paths for when entering s2idle & S0ix is supported than
>    S3.
>
>    Reviewed by:    bz
>    Approved by:    bz

I cannot apporve them; I am neither a mentor, srcmgr, re, or other entity in this case.

But thank you for splitting them up into two logical chunks (with the only bit that adding
the enum here should have been part of the other half so each of them would have compiled
independently).


>    Sponsored by:   The FreeBSD Foundation
>    Differential Revision:  https://reviews.freebsd.org/D51591
> ---
> sys/compat/linuxkpi/common/src/linux_acpi.c | 30 +++++++++++++++++++----------
> 1 file changed, 20 insertions(+), 10 deletions(-)
>
> diff --git a/sys/compat/linuxkpi/common/src/linux_acpi.c b/sys/compat/linuxkpi/common/src/linux_acpi.c
> index 60e14cfc98c5..c7d62c745c7e 100644
> --- a/sys/compat/linuxkpi/common/src/linux_acpi.c
> +++ b/sys/compat/linuxkpi/common/src/linux_acpi.c
> @@ -119,17 +119,27 @@ acpi_evaluate_dsm(ACPI_HANDLE ObjHandle, const guid_t *guid,
> }
>
> static void
> -linux_handle_power_suspend_event(void *arg __unused,
> -    enum power_stype stype __unused)
> +linux_handle_power_suspend_event(void *arg __unused, enum power_stype stype)
> {
> -	/*
> -	 * Only support S3 for now.
> -	 * acpi_sleep_event isn't always called so we use power_suspend_early
> -	 * instead which means we don't know what state we're switching to.
> -	 * TODO: Make acpi_sleep_event consistent
> -	 */
> -	linux_acpi_target_sleep_state = ACPI_STATE_S3;
> -	pm_suspend_target_state = PM_SUSPEND_MEM;
> +	switch (stype) {
> +	case POWER_STYPE_SUSPEND_TO_IDLE:
> +		/*
> +		 * XXX: obiwac Not 100% sure this is correct, but
> +		 * acpi_target_sleep_state does seem to be set to
> +		 * ACPI_STATE_S3 during s2idle on Linux.
> +		 */
> +		linux_acpi_target_sleep_state = ACPI_STATE_S3;
> +		pm_suspend_target_state = PM_SUSPEND_TO_IDLE;
> +		break;
> +	case POWER_STYPE_SUSPEND_TO_MEM:
> +		linux_acpi_target_sleep_state = ACPI_STATE_S3;
> +		pm_suspend_target_state = PM_SUSPEND_MEM;
> +		break;
> +	default:
> +		printf("%s: sleep type %d not yet supported\n",
> +		    __func__, stype);
> +		break;
> +	}
> }
>
> static void
>

-- 
Bjoern A. Zeeb                                                     r15:7