[v1] ACPI: scan: Avoid enumerating devices with clearly invalid _STA values

Commit Message

Rafael J. Wysocki April 26, 2024, 4:56 p.m. UTC

From: Rafael J. Wysocki <[e-mail]>

The return value of _STA with the "present" bit unset and the "enabled"
bit set is clearly invalid as per the ACPI specification, Section 6.3.7
"_STA (Device Status)", so make the ACPI device enumeration code
disregard devices with such _STA return values.

Also, because this implies that status.enabled will only be set if
status.present is set too, acpi_device_is_enabled() can be modified
to simply return the value of the former.

Link: https://uefi.org/specs/ACPI/6.5/06_Device_Configuration.html#sta-device-status
Link: https://lore.kernel.org/linux-acpi/[e-mail]/
Suggested-by: Salil Mehta <[e-mail]>
Signed-off-by: Rafael J. Wysocki <[e-mail]>
---
 drivers/acpi/bus.c  |   11 +++++++++++
 drivers/acpi/scan.c |    2 +-
 2 files changed, 12 insertions(+), 1 deletion(-)

Comments

Zhang Rui April 28, 2024, 4:17 a.m. UTC | #1

On Fri, 2024-04-26 at 18:56 +0200, Rafael J. Wysocki wrote:
> From: Rafael J. Wysocki <[e-mail]>
> 
> The return value of _STA with the "present" bit unset and the
> "enabled"
> bit set is clearly invalid as per the ACPI specification, Section
> 6.3.7
> "_STA (Device Status)", so make the ACPI device enumeration code
> disregard devices with such _STA return values.
> 
> Also, because this implies that status.enabled will only be set if
> status.present is set too, acpi_device_is_enabled() can be modified
> to simply return the value of the former.
> 
> Link:
> https://uefi.org/specs/ACPI/6.5/06_Device_Configuration.html#sta-device-status
> Link:
> https://lore.kernel.org/linux-acpi/[e-mail]/
> Suggested-by: Salil Mehta <[e-mail]>
> Signed-off-by: Rafael J. Wysocki <[e-mail]>
> ---
>  drivers/acpi/bus.c  |   11 +++++++++++
>  drivers/acpi/scan.c |    2 +-
>  2 files changed, 12 insertions(+), 1 deletion(-)
> 
> Index: linux-pm/drivers/acpi/bus.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/bus.c
> +++ linux-pm/drivers/acpi/bus.c
> @@ -112,6 +112,17 @@ int acpi_bus_get_status(struct acpi_devi
>         if (ACPI_FAILURE(status))
>                 return -ENODEV;
>  
> +       if (!device->status.present && device->status.enabled) {
> +               pr_info(FW_BUG "Device [%s] status [%08x]: not
> present and enabled\n",
> +                       device->pnp.bus_id, (u32)sta);
> +               device->status.enabled = 0;
> +               /*
> +                * The status is clearly invalid, so clear the
> enabled bit as
> +                * well to avoid attempting to use the device.
> +                */

seems that this comment is for the line above?

thanks,
rui
> +               device->status.functional = 0;
> +       }
> +
>         acpi_set_device_status(device, sta);
>  
>         if (device->status.functional && !device->status.present) {
> Index: linux-pm/drivers/acpi/scan.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/scan.c
> +++ linux-pm/drivers/acpi/scan.c
> @@ -1962,7 +1962,7 @@ bool acpi_device_is_present(const struct
>  
>  bool acpi_device_is_enabled(const struct acpi_device *adev)
>  {
> -       return adev->status.present && adev->status.enabled;
> +       return adev->status.enabled;
>  }
>  
>  static bool acpi_scan_handler_matching(struct acpi_scan_handler
> *handler,
> 
> 
> 
>

Jonathan Cameron April 29, 2024, 8:29 a.m. UTC | #2

On Fri, 26 Apr 2024 18:56:21 +0200
"Rafael J. Wysocki" <[e-mail]> wrote:

> From: Rafael J. Wysocki <[e-mail]>
> 
> The return value of _STA with the "present" bit unset and the "enabled"
> bit set is clearly invalid as per the ACPI specification, Section 6.3.7
> "_STA (Device Status)", so make the ACPI device enumeration code
> disregard devices with such _STA return values.
> 
> Also, because this implies that status.enabled will only be set if
> status.present is set too, acpi_device_is_enabled() can be modified
> to simply return the value of the former.
> 
> Link: https://uefi.org/specs/ACPI/6.5/06_Device_Configuration.html#sta-device-status
> Link: https://lore.kernel.org/linux-acpi/[e-mail]/
> Suggested-by: Salil Mehta <[e-mail]>
> Signed-off-by: Rafael J. Wysocki <[e-mail]>
Seems a sensible tidying up.  Hopefully nothing was relying on
this looser behavior.  One trivial thing inline.

Reviewed-by: Jonathan Cameron <[e-mail]>

> ---
>  drivers/acpi/bus.c  |   11 +++++++++++
>  drivers/acpi/scan.c |    2 +-
>  2 files changed, 12 insertions(+), 1 deletion(-)
> 
> Index: linux-pm/drivers/acpi/bus.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/bus.c
> +++ linux-pm/drivers/acpi/bus.c
> @@ -112,6 +112,17 @@ int acpi_bus_get_status(struct acpi_devi
>  	if (ACPI_FAILURE(status))
>  		return -ENODEV;
>  
> +	if (!device->status.present && device->status.enabled) {
> +		pr_info(FW_BUG "Device [%s] status [%08x]: not present and enabled\n",
> +			device->pnp.bus_id, (u32)sta);
> +		device->status.enabled = 0;
> +		/*
> +		 * The status is clearly invalid, so clear the enabled bit as
> +		 * well to avoid attempting to use the device.
> +		 */

Comment seems to be in a slightly odd place.  Perhaps one line earlier makes
more sense?  Or was the intent to mention functional here?

> +		device->status.functional = 0;
> +	}
> +
>  	acpi_set_device_status(device, sta);
>  
>  	if (device->status.functional && !device->status.present) {
> Index: linux-pm/drivers/acpi/scan.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/scan.c
> +++ linux-pm/drivers/acpi/scan.c
> @@ -1962,7 +1962,7 @@ bool acpi_device_is_present(const struct
>  
>  bool acpi_device_is_enabled(const struct acpi_device *adev)
>  {
> -	return adev->status.present && adev->status.enabled;
> +	return adev->status.enabled;
>  }
>  
>  static bool acpi_scan_handler_matching(struct acpi_scan_handler *handler,
> 
> 
>

Rafael J. Wysocki April 29, 2024, 4:06 p.m. UTC | #3

On Sun, Apr 28, 2024 at 6:17 AM Zhang, Rui <[e-mail]> wrote:
>
> On Fri, 2024-04-26 at 18:56 +0200, Rafael J. Wysocki wrote:
> > From: Rafael J. Wysocki <[e-mail]>
> >
> > The return value of _STA with the "present" bit unset and the
> > "enabled"
> > bit set is clearly invalid as per the ACPI specification, Section
> > 6.3.7
> > "_STA (Device Status)", so make the ACPI device enumeration code
> > disregard devices with such _STA return values.
> >
> > Also, because this implies that status.enabled will only be set if
> > status.present is set too, acpi_device_is_enabled() can be modified
> > to simply return the value of the former.
> >
> > Link:
> > https://uefi.org/specs/ACPI/6.5/06_Device_Configuration.html#sta-device-status
> > Link:
> > https://lore.kernel.org/linux-acpi/[e-mail]/
> > Suggested-by: Salil Mehta <[e-mail]>
> > Signed-off-by: Rafael J. Wysocki <[e-mail]>
> > ---
> >  drivers/acpi/bus.c  |   11 +++++++++++
> >  drivers/acpi/scan.c |    2 +-
> >  2 files changed, 12 insertions(+), 1 deletion(-)
> >
> > Index: linux-pm/drivers/acpi/bus.c
> > ===================================================================
> > --- linux-pm.orig/drivers/acpi/bus.c
> > +++ linux-pm/drivers/acpi/bus.c
> > @@ -112,6 +112,17 @@ int acpi_bus_get_status(struct acpi_devi
> >         if (ACPI_FAILURE(status))
> >                 return -ENODEV;
> >
> > +       if (!device->status.present && device->status.enabled) {
> > +               pr_info(FW_BUG "Device [%s] status [%08x]: not
> > present and enabled\n",
> > +                       device->pnp.bus_id, (u32)sta);
> > +               device->status.enabled = 0;
> > +               /*
> > +                * The status is clearly invalid, so clear the
> > enabled bit as
> > +                * well to avoid attempting to use the device.
> > +                */
>
> seems that this comment is for the line above?

No, I meant "functional" and wrote "enabled".  Not sure why really.

> > +               device->status.functional = 0;
> > +       }
> > +
> >         acpi_set_device_status(device, sta);
> >
> >         if (device->status.functional && !device->status.present) {
> > Index: linux-pm/drivers/acpi/scan.c
> > ===================================================================
> > --- linux-pm.orig/drivers/acpi/scan.c
> > +++ linux-pm/drivers/acpi/scan.c
> > @@ -1962,7 +1962,7 @@ bool acpi_device_is_present(const struct
> >
> >  bool acpi_device_is_enabled(const struct acpi_device *adev)
> >  {
> > -       return adev->status.present && adev->status.enabled;
> > +       return adev->status.enabled;
> >  }
> >
> >  static bool acpi_scan_handler_matching(struct acpi_scan_handler
> > *handler,
> >
> >
> >
> >
>

Rafael J. Wysocki April 29, 2024, 4:07 p.m. UTC | #4

On Mon, Apr 29, 2024 at 10:29 AM Jonathan Cameron
<[e-mail]> wrote:
>
> On Fri, 26 Apr 2024 18:56:21 +0200
> "Rafael J. Wysocki" <[e-mail]> wrote:
>
> > From: Rafael J. Wysocki <[e-mail]>
> >
> > The return value of _STA with the "present" bit unset and the "enabled"
> > bit set is clearly invalid as per the ACPI specification, Section 6.3.7
> > "_STA (Device Status)", so make the ACPI device enumeration code
> > disregard devices with such _STA return values.
> >
> > Also, because this implies that status.enabled will only be set if
> > status.present is set too, acpi_device_is_enabled() can be modified
> > to simply return the value of the former.
> >
> > Link: https://uefi.org/specs/ACPI/6.5/06_Device_Configuration.html#sta-device-status
> > Link: https://lore.kernel.org/linux-acpi/[e-mail]/
> > Suggested-by: Salil Mehta <[e-mail]>
> > Signed-off-by: Rafael J. Wysocki <[e-mail]>
> Seems a sensible tidying up.  Hopefully nothing was relying on
> this looser behavior.  One trivial thing inline.
>
> Reviewed-by: Jonathan Cameron <[e-mail]>

Thanks!

> > ---
> >  drivers/acpi/bus.c  |   11 +++++++++++
> >  drivers/acpi/scan.c |    2 +-
> >  2 files changed, 12 insertions(+), 1 deletion(-)
> >
> > Index: linux-pm/drivers/acpi/bus.c
> > ===================================================================
> > --- linux-pm.orig/drivers/acpi/bus.c
> > +++ linux-pm/drivers/acpi/bus.c
> > @@ -112,6 +112,17 @@ int acpi_bus_get_status(struct acpi_devi
> >       if (ACPI_FAILURE(status))
> >               return -ENODEV;
> >
> > +     if (!device->status.present && device->status.enabled) {
> > +             pr_info(FW_BUG "Device [%s] status [%08x]: not present and enabled\n",
> > +                     device->pnp.bus_id, (u32)sta);
> > +             device->status.enabled = 0;
> > +             /*
> > +              * The status is clearly invalid, so clear the enabled bit as
> > +              * well to avoid attempting to use the device.
> > +              */
>
> Comment seems to be in a slightly odd place.  Perhaps one line earlier makes
> more sense?  Or was the intent to mention functional here?

Rui has noticed this already.

I thought "functional" and wrote "enabled".  Oh well, I'll send a v2.

> > +             device->status.functional = 0;
> > +     }
> > +
> >       acpi_set_device_status(device, sta);
> >
> >       if (device->status.functional && !device->status.present) {
> > Index: linux-pm/drivers/acpi/scan.c
> > ===================================================================
> > --- linux-pm.orig/drivers/acpi/scan.c
> > +++ linux-pm/drivers/acpi/scan.c
> > @@ -1962,7 +1962,7 @@ bool acpi_device_is_present(const struct
> >
> >  bool acpi_device_is_enabled(const struct acpi_device *adev)
> >  {
> > -     return adev->status.present && adev->status.enabled;
> > +     return adev->status.enabled;
> >  }
> >
> >  static bool acpi_scan_handler_matching(struct acpi_scan_handler *handler,
> >
> >
> >
>
>

Patch

Index: linux-pm/drivers/acpi/bus.c
===================================================================
--- linux-pm.orig/drivers/acpi/bus.c
+++ linux-pm/drivers/acpi/bus.c
@@ -112,6 +112,17 @@  int acpi_bus_get_status(struct acpi_devi
 	if (ACPI_FAILURE(status))
 		return -ENODEV;
 
+	if (!device->status.present && device->status.enabled) {
+		pr_info(FW_BUG "Device [%s] status [%08x]: not present and enabled\n",
+			device->pnp.bus_id, (u32)sta);
+		device->status.enabled = 0;
+		/*
+		 * The status is clearly invalid, so clear the enabled bit as
+		 * well to avoid attempting to use the device.
+		 */
+		device->status.functional = 0;
+	}
+
 	acpi_set_device_status(device, sta);
 
 	if (device->status.functional && !device->status.present) {
Index: linux-pm/drivers/acpi/scan.c
===================================================================
--- linux-pm.orig/drivers/acpi/scan.c
+++ linux-pm/drivers/acpi/scan.c
@@ -1962,7 +1962,7 @@  bool acpi_device_is_present(const struct
 
 bool acpi_device_is_enabled(const struct acpi_device *adev)
 {
-	return adev->status.present && adev->status.enabled;
+	return adev->status.enabled;
 }
 
 static bool acpi_scan_handler_matching(struct acpi_scan_handler *handler,