On Thu, Sep 06, 2018 at 06:55:44PM +0200, Hans de Goede wrote:
> Commit eebb3e8d8aaf ("ACPI / LPSS: override power state for LPSS DMA
> device") was added to work around a shutdown / reboot problem reported
> on some Bay Trail devices. Specifically this was reported on a HP
> Pavilion x360 11" and on an Asus T100TA.
>
> Before Commit 12864ff8545f ("ACPI / LPSS: Avoid PM quirks on suspend and
> resume from hibernation") this code would force on the DMA controllers
> and never turn them off again on systems where the conditions for
> lpss_iosf_enter_d3_state() did not become true which is actually the case
> on a lot of systems.
>
> It turns out this was blocking entering of S0ix state on these systems,
> now lpss_iosf_exit_d3_state() only forces the DMA controllers to D0
> after the conditions in enter_d3_state() have become true at least once.
>
> The Pavilion x360 11" uses S3 suspend, not suspend to idle, so
> lpss_iosf_enter_d3_state() never gets called there and the fix has
> effectively been disabled for this device.
>
> On the Asus T100TA the conditions do become true, but the forcing to D0
> state of the DMA controllers may still not happen when the system is booted
> and then directly shutdown without a suspend/resume in between.
>
> This made me wonder if this workaround is still necessary at all, or if
> other fixes which have been done then means that this is no longer
> necessary.
And this makes me wonder which patches may have fixed the issue ;-)
> I've tested running a kernel with the workaround removed on an Asus T100TA
> and on a HP stream x360 - 11, which is the same hardware (same mainboard)
> as the Pavilion x360 11" with a lower-power version of the SoC.
>
> I've further tested this on an Asus T200TA and a Toshiba Click Mini L9W-B
> which are both devices of the same generation as the other 2.
>
> Shutdown and reboot work fine on all these devices without the workaround.
>
> So it seems that the workaround is no longer necessary and therefor this
> commit removes it, leading to a substantial cleanup of the acpi_lpss code.
>
> Signed-off-by: Hans de Goede <hdegoede@xxxxxxxxxx>
Nice cleanup!
Acked-by: Mika Westerberg <mika.westerberg@xxxxxxxxxxxxxxx>
