[2/2] gpiolib-acpi: Don't set GPIOs for wakeup in S3 mode

Message ID 20230121134812.16637-3-mario.limonciello@amd.com
State New
Headers
Series Fix some more fallout from GPIOs from _CRS |

Commit Message

Mario Limonciello Jan. 21, 2023, 1:48 p.m. UTC
  commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
adjusted the policy to enable wakeup by default if the ACPI tables
indicated that a device was wake capable.

It was reported however that this broke suspend on at least two System76
systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
When the machines are set to s2idle, wakeup behaves properly.

Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
set it when the system supports low power idle.

Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set wake_irq")
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
Suggested-by: Raul Rangel <rrangel@chromium.org>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
---
 drivers/gpio/gpiolib-acpi.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
  

Comments

Andy Shevchenko Jan. 23, 2023, 12:23 p.m. UTC | #1
On Sat, Jan 21, 2023 at 07:48:11AM -0600, Mario Limonciello wrote:
> commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> adjusted the policy to enable wakeup by default if the ACPI tables
> indicated that a device was wake capable.
> 
> It was reported however that this broke suspend on at least two System76
> systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
> When the machines are set to s2idle, wakeup behaves properly.
> 
> Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
> set it when the system supports low power idle.

Fine by me,
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

> Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set wake_irq")
> Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
> Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
> Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
> Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
> Suggested-by: Raul Rangel <rrangel@chromium.org>
> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> ---
>  drivers/gpio/gpiolib-acpi.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
> index 9ef0f5641b521..17c53f484280f 100644
> --- a/drivers/gpio/gpiolib-acpi.c
> +++ b/drivers/gpio/gpiolib-acpi.c
> @@ -1104,7 +1104,8 @@ int acpi_dev_gpio_irq_wake_get_by(struct acpi_device *adev, const char *name, in
>  				dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
>  			}
>  
> -			if (wake_capable)
> +			/* avoid suspend issues with GPIOs when systems are using S3 */
> +			if (wake_capable && acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)
>  				*wake_capable = info.wake_capable;
>  
>  			return irq;
> -- 
> 2.34.1
>
  
Bartosz Golaszewski Jan. 23, 2023, 3:02 p.m. UTC | #2
On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
<mario.limonciello@amd.com> wrote:
>
> commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> adjusted the policy to enable wakeup by default if the ACPI tables
> indicated that a device was wake capable.
>
> It was reported however that this broke suspend on at least two System76
> systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
> When the machines are set to s2idle, wakeup behaves properly.
>
> Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
> set it when the system supports low power idle.
>
> Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set wake_irq")
> Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
> Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
> Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
> Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
> Suggested-by: Raul Rangel <rrangel@chromium.org>
> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> ---
>  drivers/gpio/gpiolib-acpi.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
> index 9ef0f5641b521..17c53f484280f 100644
> --- a/drivers/gpio/gpiolib-acpi.c
> +++ b/drivers/gpio/gpiolib-acpi.c
> @@ -1104,7 +1104,8 @@ int acpi_dev_gpio_irq_wake_get_by(struct acpi_device *adev, const char *name, in
>                                 dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
>                         }
>
> -                       if (wake_capable)
> +                       /* avoid suspend issues with GPIOs when systems are using S3 */
> +                       if (wake_capable && acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)
>                                 *wake_capable = info.wake_capable;
>
>                         return irq;
> --
> 2.34.1
>

Applied, thanks!

Bart
  
Raul Rangel Jan. 23, 2023, 3:55 p.m. UTC | #3
On Mon, Jan 23, 2023 at 8:03 AM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
>
> On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
> <mario.limonciello@amd.com> wrote:
> >
> > commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > adjusted the policy to enable wakeup by default if the ACPI tables
> > indicated that a device was wake capable.
> >
> > It was reported however that this broke suspend on at least two System76
> > systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
> > When the machines are set to s2idle, wakeup behaves properly.
> >
> > Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
> > set it when the system supports low power idle.
> >
> > Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set wake_irq")
> > Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
> > Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
> > Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
> > Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
> > Suggested-by: Raul Rangel <rrangel@chromium.org>
> > Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> > ---
> >  drivers/gpio/gpiolib-acpi.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
> > index 9ef0f5641b521..17c53f484280f 100644
> > --- a/drivers/gpio/gpiolib-acpi.c
> > +++ b/drivers/gpio/gpiolib-acpi.c
> > @@ -1104,7 +1104,8 @@ int acpi_dev_gpio_irq_wake_get_by(struct acpi_device *adev, const char *name, in
> >                                 dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
> >                         }
> >
> > -                       if (wake_capable)
> > +                       /* avoid suspend issues with GPIOs when systems are using S3 */
> > +                       if (wake_capable && acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)
> >                                 *wake_capable = info.wake_capable;
> >
> >                         return irq;
> > --
> > 2.34.1
> >
>
> Applied, thanks!
>
> Bart


We still need to figure out a proper fix for this. If you read my post
here: https://gitlab.freedesktop.org/drm/amd/-/issues/2357#note_1732372
I think we misinterpreted what the SharedAndWake bit is used for. To
me it sounds like it's only valid for HW Reduced ACPI platforms, and
S0ix. My changes made it so we call `dev_pm_set_wake_irq` when the
Wake bit is set. Does anyone have any additional context on the Wake
bit? I think we either need to make `dev_pm_set_wake_irq` (or a
variant) only enable the wake on S0i3, or we can teach the ACPI
subsystem to manage arming the IRQ's wake bit. Kind of like we already
manage the GPE events for the device.
  
Mario Limonciello Jan. 23, 2023, 4:06 p.m. UTC | #4
[Public]



> -----Original Message-----
> From: Raul Rangel <rrangel@chromium.org>
> Sent: Monday, January 23, 2023 09:55
> To: Bartosz Golaszewski <brgl@bgdev.pl>
> Cc: Limonciello, Mario <Mario.Limonciello@amd.com>; Mika Westerberg
> <mika.westerberg@linux.intel.com>; Andy Shevchenko
> <andriy.shevchenko@linux.intel.com>; Linus Walleij
> <linus.walleij@linaro.org>; Dmitry Torokhov <dmitry.torokhov@gmail.com>;
> Benjamin Tissoires <benjamin.tissoires@redhat.com>; Wolfram Sang
> <wsa@kernel.org>; Rafael J. Wysocki <rafael.j.wysocki@intel.com>; Nathan
> Smythe <ncsmythe@scruboak.org>; linux-gpio@vger.kernel.org; linux-
> acpi@vger.kernel.org; linux-kernel@vger.kernel.org; Mark Hasemeyer
> <markhas@chromium.org>
> Subject: Re: [PATCH 2/2] gpiolib-acpi: Don't set GPIOs for wakeup in S3 mode
> 
> On Mon, Jan 23, 2023 at 8:03 AM Bartosz Golaszewski <brgl@bgdev.pl>
> wrote:
> >
> > On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
> > <mario.limonciello@amd.com> wrote:
> > >
> > > commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > adjusted the policy to enable wakeup by default if the ACPI tables
> > > indicated that a device was wake capable.
> > >
> > > It was reported however that this broke suspend on at least two
> System76
> > > systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
> > > When the machines are set to s2idle, wakeup behaves properly.
> > >
> > > Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
> > > set it when the system supports low power idle.
> > >
> > > Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set
> wake_irq")
> > > Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
> > > Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
> > > Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > Suggested-by: Raul Rangel <rrangel@chromium.org>
> > > Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> > > ---
> > >  drivers/gpio/gpiolib-acpi.c | 3 ++-
> > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
> > > index 9ef0f5641b521..17c53f484280f 100644
> > > --- a/drivers/gpio/gpiolib-acpi.c
> > > +++ b/drivers/gpio/gpiolib-acpi.c
> > > @@ -1104,7 +1104,8 @@ int acpi_dev_gpio_irq_wake_get_by(struct
> acpi_device *adev, const char *name, in
> > >                                 dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
> > >                         }
> > >
> > > -                       if (wake_capable)
> > > +                       /* avoid suspend issues with GPIOs when systems are using
> S3 */
> > > +                       if (wake_capable && acpi_gbl_FADT.flags &
> ACPI_FADT_LOW_POWER_S0)
> > >                                 *wake_capable = info.wake_capable;
> > >
> > >                         return irq;
> > > --
> > > 2.34.1
> > >
> >
> > Applied, thanks!
> >
> > Bart
> 
> 
> We still need to figure out a proper fix for this. If you read my post
> here: https://gitlab.freedesktop.org/drm/amd/-/issues/2357#note_1732372
> I think we misinterpreted what the SharedAndWake bit is used for. To
> me it sounds like it's only valid for HW Reduced ACPI platforms, and
> S0ix. My changes made it so we call `dev_pm_set_wake_irq` when the
> Wake bit is set. Does anyone have any additional context on the Wake
> bit? I think we either need to make `dev_pm_set_wake_irq` (or a
> variant) only enable the wake on S0i3, or we can teach the ACPI
> subsystem to manage arming the IRQ's wake bit. Kind of like we already
> manage the GPE events for the device.

There is an FADT flag for HW reduced (ACPI_FADT_HW_REDUCED).  So
maybe something on top of my change to look at that too?

IE:
if (wake_capable && (acpi_gbl_FADT.flags & (ACPI_FADT_LOW_POWER_S0 | ACPI_FADT_HW_REDUCED)
  
Raul Rangel Jan. 23, 2023, 4:34 p.m. UTC | #5
On Mon, Jan 23, 2023 at 9:07 AM Limonciello, Mario
<Mario.Limonciello@amd.com> wrote:
>
> [Public]
>
>
>
> > -----Original Message-----
> > From: Raul Rangel <rrangel@chromium.org>
> > Sent: Monday, January 23, 2023 09:55
> > To: Bartosz Golaszewski <brgl@bgdev.pl>
> > Cc: Limonciello, Mario <Mario.Limonciello@amd.com>; Mika Westerberg
> > <mika.westerberg@linux.intel.com>; Andy Shevchenko
> > <andriy.shevchenko@linux.intel.com>; Linus Walleij
> > <linus.walleij@linaro.org>; Dmitry Torokhov <dmitry.torokhov@gmail.com>;
> > Benjamin Tissoires <benjamin.tissoires@redhat.com>; Wolfram Sang
> > <wsa@kernel.org>; Rafael J. Wysocki <rafael.j.wysocki@intel.com>; Nathan
> > Smythe <ncsmythe@scruboak.org>; linux-gpio@vger.kernel.org; linux-
> > acpi@vger.kernel.org; linux-kernel@vger.kernel.org; Mark Hasemeyer
> > <markhas@chromium.org>
> > Subject: Re: [PATCH 2/2] gpiolib-acpi: Don't set GPIOs for wakeup in S3 mode
> >
> > On Mon, Jan 23, 2023 at 8:03 AM Bartosz Golaszewski <brgl@bgdev.pl>
> > wrote:
> > >
> > > On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
> > > <mario.limonciello@amd.com> wrote:
> > > >
> > > > commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > > adjusted the policy to enable wakeup by default if the ACPI tables
> > > > indicated that a device was wake capable.
> > > >
> > > > It was reported however that this broke suspend on at least two
> > System76
> > > > systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
> > > > When the machines are set to s2idle, wakeup behaves properly.
> > > >
> > > > Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
> > > > set it when the system supports low power idle.
> > > >
> > > > Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > > Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set
> > wake_irq")
> > > > Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
> > > > Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
> > > > Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > > Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > > Suggested-by: Raul Rangel <rrangel@chromium.org>
> > > > Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> > > > ---
> > > >  drivers/gpio/gpiolib-acpi.c | 3 ++-
> > > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
> > > > index 9ef0f5641b521..17c53f484280f 100644
> > > > --- a/drivers/gpio/gpiolib-acpi.c
> > > > +++ b/drivers/gpio/gpiolib-acpi.c
> > > > @@ -1104,7 +1104,8 @@ int acpi_dev_gpio_irq_wake_get_by(struct
> > acpi_device *adev, const char *name, in
> > > >                                 dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
> > > >                         }
> > > >
> > > > -                       if (wake_capable)
> > > > +                       /* avoid suspend issues with GPIOs when systems are using
> > S3 */
> > > > +                       if (wake_capable && acpi_gbl_FADT.flags &
> > ACPI_FADT_LOW_POWER_S0)
> > > >                                 *wake_capable = info.wake_capable;
> > > >
> > > >                         return irq;
> > > > --
> > > > 2.34.1
> > > >
> > >
> > > Applied, thanks!
> > >
> > > Bart
> >
> >
> > We still need to figure out a proper fix for this. If you read my post
> > here: https://gitlab.freedesktop.org/drm/amd/-/issues/2357#note_1732372
> > I think we misinterpreted what the SharedAndWake bit is used for. To
> > me it sounds like it's only valid for HW Reduced ACPI platforms, and
> > S0ix. My changes made it so we call `dev_pm_set_wake_irq` when the
> > Wake bit is set. Does anyone have any additional context on the Wake
> > bit? I think we either need to make `dev_pm_set_wake_irq` (or a
> > variant) only enable the wake on S0i3, or we can teach the ACPI
> > subsystem to manage arming the IRQ's wake bit. Kind of like we already
> > manage the GPE events for the device.
>
> There is an FADT flag for HW reduced (ACPI_FADT_HW_REDUCED).  So
> maybe something on top of my change to look at that too?
>
> IE:
> if (wake_capable && (acpi_gbl_FADT.flags & (ACPI_FADT_LOW_POWER_S0 | ACPI_FADT_HW_REDUCED)

The problem with the ACPI_FADT_LOW_POWER_S0 FADT flag is that it
defines if S0ix is supported. That's not mutually exclusive with S3.
So we really need a runtime check to see which suspend mode we are
entering.
  
Andy Shevchenko Jan. 23, 2023, 5:30 p.m. UTC | #6
On Mon, Jan 23, 2023 at 08:55:02AM -0700, Raul Rangel wrote:
> On Mon, Jan 23, 2023 at 8:03 AM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> >
> > On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
> > <mario.limonciello@amd.com> wrote:
> > >
> > > commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > adjusted the policy to enable wakeup by default if the ACPI tables
> > > indicated that a device was wake capable.
> > >
> > > It was reported however that this broke suspend on at least two System76
> > > systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
> > > When the machines are set to s2idle, wakeup behaves properly.
> > >
> > > Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
> > > set it when the system supports low power idle.
> > >
> > > Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set wake_irq")
> > > Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
> > > Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
> > > Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > Suggested-by: Raul Rangel <rrangel@chromium.org>
> > > Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> > > ---
> > >  drivers/gpio/gpiolib-acpi.c | 3 ++-
> > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
> > > index 9ef0f5641b521..17c53f484280f 100644
> > > --- a/drivers/gpio/gpiolib-acpi.c
> > > +++ b/drivers/gpio/gpiolib-acpi.c
> > > @@ -1104,7 +1104,8 @@ int acpi_dev_gpio_irq_wake_get_by(struct acpi_device *adev, const char *name, in
> > >                                 dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
> > >                         }
> > >
> > > -                       if (wake_capable)
> > > +                       /* avoid suspend issues with GPIOs when systems are using S3 */
> > > +                       if (wake_capable && acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)
> > >                                 *wake_capable = info.wake_capable;
> > >
> > >                         return irq;
> > > --
> > > 2.34.1
> > >
> >
> > Applied, thanks!
> >
> > Bart
> 
> 
> We still need to figure out a proper fix for this. If you read my post
> here: https://gitlab.freedesktop.org/drm/amd/-/issues/2357#note_1732372
> I think we misinterpreted what the SharedAndWake bit is used for. To
> me it sounds like it's only valid for HW Reduced ACPI platforms, and
> S0ix. My changes made it so we call `dev_pm_set_wake_irq` when the
> Wake bit is set. Does anyone have any additional context on the Wake
> bit? I think we either need to make `dev_pm_set_wake_irq` (or a
> variant) only enable the wake on S0i3, or we can teach the ACPI
> subsystem to manage arming the IRQ's wake bit. Kind of like we already
> manage the GPE events for the device.

From the spec:

Shared is an optional argument and can be one of Shared, Exclusive,
SharedAndWake or ExclusiveAndWake. If not specified, Exclusive is assumed.
The “Wake” designation indicates that the interrupt is capable of waking
the system from a low-power idle state or a system sleep state. The bit
field name _SHR is automatically created to refer to this portion of
the resource descriptor.


Note: "...a low-power idle state or a system sleep state.". I believe it
applies to both.
  
Andy Shevchenko Jan. 23, 2023, 5:33 p.m. UTC | #7
On Mon, Jan 23, 2023 at 04:06:59PM +0000, Limonciello, Mario wrote:
> > From: Raul Rangel <rrangel@chromium.org>
> > Sent: Monday, January 23, 2023 09:55
> > On Mon, Jan 23, 2023 at 8:03 AM Bartosz Golaszewski <brgl@bgdev.pl>
> > wrote:
> > > On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
> > > <mario.limonciello@amd.com> wrote:

...

> > > > +                       /* avoid suspend issues with GPIOs when systems are using
> > S3 */
> > > > +                       if (wake_capable && acpi_gbl_FADT.flags &
> > ACPI_FADT_LOW_POWER_S0)
> > > >                                 *wake_capable = info.wake_capable;
> > > >
> > > >                         return irq;

...

> > We still need to figure out a proper fix for this. If you read my post
> > here: https://gitlab.freedesktop.org/drm/amd/-/issues/2357#note_1732372
> > I think we misinterpreted what the SharedAndWake bit is used for. To
> > me it sounds like it's only valid for HW Reduced ACPI platforms, and
> > S0ix. My changes made it so we call `dev_pm_set_wake_irq` when the
> > Wake bit is set. Does anyone have any additional context on the Wake
> > bit? I think we either need to make `dev_pm_set_wake_irq` (or a
> > variant) only enable the wake on S0i3, or we can teach the ACPI
> > subsystem to manage arming the IRQ's wake bit. Kind of like we already
> > manage the GPE events for the device.
> 
> There is an FADT flag for HW reduced (ACPI_FADT_HW_REDUCED).  So
> maybe something on top of my change to look at that too?
> 
> IE:
> if (wake_capable && (acpi_gbl_FADT.flags & (ACPI_FADT_LOW_POWER_S0 | ACPI_FADT_HW_REDUCED)

I'm not sure why we are talking about HW reduced case?
In HP reduced case IIRC the GPE are absent as a class.
  
Raul Rangel Jan. 23, 2023, 5:54 p.m. UTC | #8
On Mon, Jan 23, 2023 at 10:30 AM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Mon, Jan 23, 2023 at 08:55:02AM -0700, Raul Rangel wrote:
> > On Mon, Jan 23, 2023 at 8:03 AM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> > >
> > > On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
> > > <mario.limonciello@amd.com> wrote:
> > > >
> > > > commit 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > > adjusted the policy to enable wakeup by default if the ACPI tables
> > > > indicated that a device was wake capable.
> > > >
> > > > It was reported however that this broke suspend on at least two System76
> > > > systems in S3 mode and two Lenovo Gen2a systems, but only with S3.
> > > > When the machines are set to s2idle, wakeup behaves properly.
> > > >
> > > > Configuring the GPIOs for wakeup with S3 doesn't work properly, so only
> > > > set it when the system supports low power idle.
> > > >
> > > > Fixes: 1796f808e4bb ("HID: i2c-hid: acpi: Stop setting wakeup_capable")
> > > > Fixes: b38f2d5d9615c ("i2c: acpi: Use ACPI wake capability bit to set wake_irq")
> > > > Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2357
> > > > Link: https://bugzilla.redhat.com/show_bug.cgi?id=2162013
> > > > Reported-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > > Tested-by: Nathan Smythe <ncsmythe@scruboak.org>
> > > > Suggested-by: Raul Rangel <rrangel@chromium.org>
> > > > Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> > > > ---
> > > >  drivers/gpio/gpiolib-acpi.c | 3 ++-
> > > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
> > > > index 9ef0f5641b521..17c53f484280f 100644
> > > > --- a/drivers/gpio/gpiolib-acpi.c
> > > > +++ b/drivers/gpio/gpiolib-acpi.c
> > > > @@ -1104,7 +1104,8 @@ int acpi_dev_gpio_irq_wake_get_by(struct acpi_device *adev, const char *name, in
> > > >                                 dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
> > > >                         }
> > > >
> > > > -                       if (wake_capable)
> > > > +                       /* avoid suspend issues with GPIOs when systems are using S3 */
> > > > +                       if (wake_capable && acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)
> > > >                                 *wake_capable = info.wake_capable;
> > > >
> > > >                         return irq;
> > > > --
> > > > 2.34.1
> > > >
> > >
> > > Applied, thanks!
> > >
> > > Bart
> >
> >
> > We still need to figure out a proper fix for this. If you read my post
> > here: https://gitlab.freedesktop.org/drm/amd/-/issues/2357#note_1732372
> > I think we misinterpreted what the SharedAndWake bit is used for. To
> > me it sounds like it's only valid for HW Reduced ACPI platforms, and
> > S0ix. My changes made it so we call `dev_pm_set_wake_irq` when the
> > Wake bit is set. Does anyone have any additional context on the Wake
> > bit? I think we either need to make `dev_pm_set_wake_irq` (or a
> > variant) only enable the wake on S0i3, or we can teach the ACPI
> > subsystem to manage arming the IRQ's wake bit. Kind of like we already
> > manage the GPE events for the device.
>
> From the spec:
>
> Shared is an optional argument and can be one of Shared, Exclusive,
> SharedAndWake or ExclusiveAndWake. If not specified, Exclusive is assumed.
> The “Wake” designation indicates that the interrupt is capable of waking
> the system from a low-power idle state or a system sleep state. The bit
> field name _SHR is automatically created to refer to this portion of
> the resource descriptor.
>
>
> Note: "...a low-power idle state or a system sleep state.". I believe it
> applies to both.

Without the _PRW, how do we determine the valid system sleep states
the device can wake the system from?

>
> --
> With Best Regards,
> Andy Shevchenko
>
>
  
Andy Shevchenko Jan. 23, 2023, 6:21 p.m. UTC | #9
On Mon, Jan 23, 2023 at 10:54:29AM -0700, Raul Rangel wrote:
> On Mon, Jan 23, 2023 at 10:30 AM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Mon, Jan 23, 2023 at 08:55:02AM -0700, Raul Rangel wrote:
> > > On Mon, Jan 23, 2023 at 8:03 AM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> > > > On Sat, Jan 21, 2023 at 2:48 PM Mario Limonciello
> > > > <mario.limonciello@amd.com> wrote:

...

> > > We still need to figure out a proper fix for this. If you read my post
> > > here: https://gitlab.freedesktop.org/drm/amd/-/issues/2357#note_1732372
> > > I think we misinterpreted what the SharedAndWake bit is used for. To
> > > me it sounds like it's only valid for HW Reduced ACPI platforms, and
> > > S0ix. My changes made it so we call `dev_pm_set_wake_irq` when the
> > > Wake bit is set. Does anyone have any additional context on the Wake
> > > bit? I think we either need to make `dev_pm_set_wake_irq` (or a
> > > variant) only enable the wake on S0i3, or we can teach the ACPI
> > > subsystem to manage arming the IRQ's wake bit. Kind of like we already
> > > manage the GPE events for the device.
> >
> > From the spec:
> >
> > Shared is an optional argument and can be one of Shared, Exclusive,
> > SharedAndWake or ExclusiveAndWake. If not specified, Exclusive is assumed.
> > The “Wake” designation indicates that the interrupt is capable of waking
> > the system from a low-power idle state or a system sleep state. The bit
> > field name _SHR is automatically created to refer to this portion of
> > the resource descriptor.
> >
> >
> > Note: "...a low-power idle state or a system sleep state.". I believe it
> > applies to both.
> 
> Without the _PRW, how do we determine the valid system sleep states
> the device can wake the system from?

Good question. I believe you need to ask somebody from ASWG for the
clarifications.
  

Patch

diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
index 9ef0f5641b521..17c53f484280f 100644
--- a/drivers/gpio/gpiolib-acpi.c
+++ b/drivers/gpio/gpiolib-acpi.c
@@ -1104,7 +1104,8 @@  int acpi_dev_gpio_irq_wake_get_by(struct acpi_device *adev, const char *name, in
 				dev_dbg(&adev->dev, "IRQ %d already in use\n", irq);
 			}
 
-			if (wake_capable)
+			/* avoid suspend issues with GPIOs when systems are using S3 */
+			if (wake_capable && acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)
 				*wake_capable = info.wake_capable;
 
 			return irq;