Quoting Rafael J. Wysocki (2020-02-21 00:46:18)
> On Thursday, February 20, 2020 11:41:22 PM CET Rafael J. Wysocki wrote:
> > On Monday, February 17, 2020 10:29:35 PM CET Chris Wilson wrote:
> > > Quoting Linus Torvalds (2020-02-17 21:20:27)
> > > > On Mon, Feb 17, 2020 at 8:22 AM Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> wrote:
> > > > >
> > > > > Quoting Linus Torvalds (2020-02-16 21:32:32)
> > > > > > Rafael J. Wysocki (4):
> > > > > > ACPI: EC: Fix flushing of pending work
> > > > > > ACPI: PM: s2idle: Avoid possible race related to the EC GPE
> > > > > > ACPICA: Introduce acpi_any_gpe_status_set()
> > > > > > ACPI: PM: s2idle: Prevent spurious SCIs from waking up the system
> > > > >
> > > > > Our S0 testing broke on all platforms, so we've reverted
> > > > > e3728b50cd9b ("ACPI: PM: s2idle: Avoid possible race related to the EC GPE")
> > > > > fdde0ff8590b ("ACPI: PM: s2idle: Prevent spurious SCIs from waking up the system")
> > > > >
> > > > > There wasn't much in the logs, for example,
> > > > > https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5445/fi-kbl-7500u/igt@gem_exec_suspend@xxxxxxxxxxxxx
> > > >
> > > > So the machine suspends, but never comes back?
> > > >
> > > > Do you need to revert both for it to work for you? Or is the revert of
> > > > fdde0ff8590b just to avoid the conflict?
> > >
> > > fdde0ff85 was just to avoid conflicts.
> > >
> > > > I'm assuming you bisected this, and the bisect indicated e3728b50cd9b,
> > > > and then to revert it you reverted the other commit too..
> > >
> > > Lucky guess based on diff rc1..rc2. Bisect was going to be painful, but
> > > could be done if this is not enough clue for Rafael.
> >
> > Sorry for the delayed response, was away.
> >
> > I'm guessing that you are using rtcwake for wakeup, in which case reverting
> > fdde0ff85 alone should unbreak it.
> >
> > Can you please double check that?
>
> And below is a patch that should fix it if I'm not mistaken (verified on my
> system where I was able to reproduce the issue), so it would suffice to test
> this one on top of the -rc2.
Correct on both accounts. Reverting fdde0ff85 alone was enough, and
replacing the reverts with the suggested patch works.
> ---
> From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> Subject: [PATCH] ACPI: PM: s2idle: Check fixed wakeup events in acpi_s2idle_wake()
>
> Commit fdde0ff8590b ("ACPI: PM: s2idle: Prevent spurious SCIs from
> waking up the system") overlooked the fact that fixed events can wake
> up the system too and broke RTC wakeup from suspend-to-idle as a
> result.
>
> Fix this issue by checking the fixed events in acpi_s2idle_wake() in
> addition to checking wakeup GPEs and break out of the suspend-to-idle
> loop if the status bits of any enabled fixed events are set then.
>
> Fixes: fdde0ff8590b ("ACPI: PM: s2idle: Prevent spurious SCIs from waking up the system")
> Reported-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> Cc: 5.4+ <stable@xxxxxxxxxxxxxxx> # 5.4+
> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
Tested-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
-Chris
