From: Matthew Garrett on
On Thu, Mar 04, 2010 at 01:25:30AM +0100, Rafael J. Wysocki wrote:
> There is a problem with the ACPI video resume routine that it's
> executed before the GPU that may be accessed by it. To fix this
> issue, move the ACPI video resume to a power management notifier,
> so that it's executed after resuming all devices, including the GPU.
>
> Signed-off-by: Rafael J. Wysocki <rjw(a)sisk.pl>
Acked-by: Matthew Garrett <mjg(a)redhat.com>

--
Matthew Garrett | mjg59(a)srcf.ucam.org
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Rafał Miłecki on
2010/3/4 Rafael J. Wysocki <rjw(a)sisk.pl>:
> There is a problem with the ACPI video resume routine that it's
> executed before the GPU that may be accessed by it.  To fix this
> issue, move the ACPI video resume to a power management notifier,
> so that it's executed after resuming all devices, including the GPU.
>
> Signed-off-by: Rafael J. Wysocki <rjw(a)sisk.pl>
> ---

Tested-by: Rafał Miłecki <zajec5(a)gmail.com>

It partially fixes bugzilla bug #15096 and it first needed step to fix
it completely.

Can we send this to stable for .32 and .33?

--
Rafał
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Rafael J. Wysocki on
On Thursday 04 March 2010, Rafał Miłecki wrote:
> 2010/3/4 Rafael J. Wysocki <rjw(a)sisk.pl>:
> > There is a problem with the ACPI video resume routine that it's
> > executed before the GPU that may be accessed by it. To fix this
> > issue, move the ACPI video resume to a power management notifier,
> > so that it's executed after resuming all devices, including the GPU.
> >
> > Signed-off-by: Rafael J. Wysocki <rjw(a)sisk.pl>
> > ---
>
> Tested-by: Rafał Miłecki <zajec5(a)gmail.com>

Thanks!

> It partially fixes bugzilla bug #15096 and it first needed step to fix
> it completely.

To fix that completely we'd need to avoid calling acpi_video_resume() in the
non-KMS case, but it's not clear how to tell the ACPI driver which case we're
in. It's not certain how to reliably distinguish the KMS and non-KMS cases at
this level in the first place.

> Can we send this to stable for .32 and .33?

That depends on Len.

Rafael
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Rafał Miłecki on
W dniu 4 marca 2010 20:12 użytkownik Rafael J. Wysocki <rjw(a)sisk.pl> napisał:
>> Can we send this to stable for .32 and .33?
>
> That depends on Len.

Len?

--
Rafał
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Rafael J. Wysocki on
Hi Len,

On Thursday 04 March 2010, Rafael J. Wysocki wrote:
> There is a problem with the ACPI video resume routine that it's
> executed before the GPU that may be accessed by it. To fix this
> issue, move the ACPI video resume to a power management notifier,
> so that it's executed after resuming all devices, including the GPU.
>
> Signed-off-by: Rafael J. Wysocki <rjw(a)sisk.pl>

Please replace the $subject patch with the appended one.

The problem is that the notifier is also executed while we're checking for
the presence of hibernation image during boot and in that case we'll set the
brightness to a wrong value if device->backlight->props.brightness is not
initialized by acpi_video_device_find_cap().

Rafael

---
From: Rafael J. Wysocki <rjw(a)sisk.pl>
Subject: ACPI / PM: Move ACPI video resume to a PM notifier

There is a problem with the ACPI video resume routine that it's
executed before the GPU that may be accessed by it. To fix this
issue, move the ACPI video resume to a power management notifier,
so that's executed after resuming all devices, including the GPU.

Fixes https://bugzilla.kernel.org/show_bug.cgi?id=15096, which is
a listed regression from 2.6.31.

Signed-off-by: Rafael J. Wysocki <rjw(a)sisk.pl>
---
drivers/acpi/video.c | 38 ++++++++++++++++++++++++++++++--------
1 file changed, 30 insertions(+), 8 deletions(-)

Index: linux-2.6/drivers/acpi/video.c
===================================================================
--- linux-2.6.orig/drivers/acpi/video.c
+++ linux-2.6/drivers/acpi/video.c
@@ -43,6 +43,7 @@
#include <linux/dmi.h>
#include <acpi/acpi_bus.h>
#include <acpi/acpi_drivers.h>
+#include <linux/suspend.h>

#define PREFIX "ACPI: "

@@ -88,7 +89,6 @@ module_param(allow_duplicates, bool, 064
static int register_count = 0;
static int acpi_video_bus_add(struct acpi_device *device);
static int acpi_video_bus_remove(struct acpi_device *device, int type);
-static int acpi_video_resume(struct acpi_device *device);
static void acpi_video_bus_notify(struct acpi_device *device, u32 event);

static const struct acpi_device_id video_device_ids[] = {
@@ -104,7 +104,6 @@ static struct acpi_driver acpi_video_bus
.ops = {
.add = acpi_video_bus_add,
.remove = acpi_video_bus_remove,
- .resume = acpi_video_resume,
.notify = acpi_video_bus_notify,
},
};
@@ -159,6 +158,7 @@ struct acpi_video_bus {
struct proc_dir_entry *dir;
struct input_dev *input;
char phys[32]; /* for input device */
+ struct notifier_block pm_nb;
};

struct acpi_video_device_flags {
@@ -1020,6 +1020,13 @@ static void acpi_video_device_find_cap(s
if (IS_ERR(device->backlight))
return;

+ /*
+ * Save current brightness level in case we have to restore it
+ * before acpi_video_device_lcd_set_level() is called next time.
+ */
+ device->backlight->props.brightness =
+ acpi_video_get_brightness(device->backlight);
+
result = sysfs_create_link(&device->backlight->dev.kobj,
&device->dev->dev.kobj, "device");
if (result)
@@ -2235,24 +2242,31 @@ static void acpi_video_device_notify(acp
return;
}

-static int instance;
-static int acpi_video_resume(struct acpi_device *device)
+static int acpi_video_resume(struct notifier_block *nb,
+ unsigned long val, void *ign)
{
struct acpi_video_bus *video;
struct acpi_video_device *video_device;
int i;

- if (!device || !acpi_driver_data(device))
- return -EINVAL;
+ switch (val) {
+ case PM_HIBERNATION_PREPARE:
+ case PM_SUSPEND_PREPARE:
+ case PM_RESTORE_PREPARE:
+ return NOTIFY_DONE;
+ }

- video = acpi_driver_data(device);
+ video = container_of(nb, struct acpi_video_bus, pm_nb);
+
+ dev_info(&video->device->dev, "Restoring backlight state\n");

for (i = 0; i < video->attached_count; i++) {
video_device = video->attached_array[i].bind_info;
if (video_device && video_device->backlight)
acpi_video_set_brightness(video_device->backlight);
}
- return AE_OK;
+
+ return NOTIFY_OK;
}

static acpi_status
@@ -2276,6 +2290,8 @@ acpi_video_bus_match(acpi_handle handle,
return AE_OK;
}

+static int instance;
+
static int acpi_video_bus_add(struct acpi_device *device)
{
struct acpi_video_bus *video;
@@ -2369,6 +2385,10 @@ static int acpi_video_bus_add(struct acp
video->flags.rom ? "yes" : "no",
video->flags.post ? "yes" : "no");

+ video->pm_nb.notifier_call = acpi_video_resume;
+ video->pm_nb.priority = 0;
+ register_pm_notifier(&video->pm_nb);
+
return 0;

err_free_input_dev:
@@ -2395,6 +2415,8 @@ static int acpi_video_bus_remove(struct

video = acpi_driver_data(device);

+ unregister_pm_notifier(&video->pm_nb);
+
acpi_video_bus_stop_devices(video);
acpi_video_bus_put_devices(video);
acpi_video_bus_remove_fs(device);
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/