acpi_video_device_lcd_get_level_current() called acpi_video_bqc_value_to_level()
with "*level" as a second argument, resulting in level being returned based on
initial input, not current brightness, breaking backlight controls.
Signed-off-by: Artem Savkov <[email protected]>
---
drivers/acpi/video.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/acpi/video.c b/drivers/acpi/video.c
index 3cdd047..68a777b 100644
--- a/drivers/acpi/video.c
+++ b/drivers/acpi/video.c
@@ -505,7 +505,8 @@ acpi_video_device_lcd_get_level_current(struct acpi_video_device *device,
return 0;
}
- *level = acpi_video_bqc_value_to_level(device, *level);
+ *level = acpi_video_bqc_value_to_level(device,
+ device->brightness->curr);
for (i = 2; i < device->brightness->count; i++)
if (device->brightness->levels[i] == *level) {
--
1.8.2
On Saturday, March 30, 2013 05:01:41 PM Artem Savkov wrote:
> acpi_video_device_lcd_get_level_current() called acpi_video_bqc_value_to_level()
> with "*level" as a second argument, resulting in level being returned based on
> initial input, not current brightness, breaking backlight controls.
This looks like a valid fix, but (1) is the problem being fixed a regression
and if so, which was the last working kernel and (2) do you have any pointers
to bug reports/BZ entries related to that?
Also, do I assume correctly that you've tested it?
Rafael
> Signed-off-by: Artem Savkov <[email protected]>
> ---
> drivers/acpi/video.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/acpi/video.c b/drivers/acpi/video.c
> index 3cdd047..68a777b 100644
> --- a/drivers/acpi/video.c
> +++ b/drivers/acpi/video.c
> @@ -505,7 +505,8 @@ acpi_video_device_lcd_get_level_current(struct acpi_video_device *device,
> return 0;
> }
>
> - *level = acpi_video_bqc_value_to_level(device, *level);
> + *level = acpi_video_bqc_value_to_level(device,
> + device->brightness->curr);
>
> for (i = 2; i < device->brightness->count; i++)
> if (device->brightness->levels[i] == *level) {
>
--
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
Artem Savkov <[email protected]> schrieb:
>acpi_video_device_lcd_get_level_current() called
>acpi_video_bqc_value_to_level()
>with "*level" as a second argument, resulting in level being returned
>based on
>initial input, not current brightness, breaking backlight controls.
I don't think this change is correct. As level was passed as parameter into the evaluation of _BQC, *level contains the AML returned brightness level afterwards, so it's correct to use it as an input to acpi_video_bqc_value_to_level(). Actually, the whole point of acpi_video_device_lcd_get_level_current() is to update device->brightness->curr, so it doesn't make sense to me to use it in that function.
What's the exact problem this patch tries to solve?
Regards,
Danny
Hi,
On Sun, Mar 31, 2013 at 08:15:14AM +0200, Danny Baumann wrote:
> Artem Savkov <[email protected]> schrieb:
> >acpi_video_device_lcd_get_level_current() called
> >acpi_video_bqc_value_to_level()
> >with "*level" as a second argument, resulting in level being returned
> >based on
> >initial input, not current brightness, breaking backlight controls.
> I don't think this change is correct. As level was passed as parameter into the evaluation of _BQC, *level contains the AML returned brightness level afterwards, so it's correct to use it as an input to acpi_video_bqc_value_to_level(). Actually, the whole point of acpi_video_device_lcd_get_level_current() is to update device->brightness->curr, so it doesn't make sense to me to use it in that function.
>
> What's the exact problem this patch tries to solve?
I'm running a -next kernel on my laptop and couple of days ago keyboard
backlight controls stopped working: only 2 lower brightness levels. I've
debugged it a bit and found out that acpi_video_switch_brightness() calls
acpi_video_device_lcd_get_level_current() with level uninitialized and
always gets lowest posible value.
--
Regards,
Artem
Hi,
Artem Savkov <[email protected]> schrieb:
>On Sun, Mar 31, 2013 at 08:15:14AM +0200, Danny Baumann wrote:
>> Artem Savkov <[email protected]> schrieb:
>> >acpi_video_device_lcd_get_level_current() called
>> >acpi_video_bqc_value_to_level()
>> >with "*level" as a second argument, resulting in level being
>returned
>> >based on
>> >initial input, not current brightness, breaking backlight controls.
>> I don't think this change is correct. As level was passed as
>parameter into the evaluation of _BQC, *level contains the AML returned
>brightness level afterwards, so it's correct to use it as an input to
>acpi_video_bqc_value_to_level(). Actually, the whole point of
>acpi_video_device_lcd_get_level_current() is to update
>device->brightness->curr, so it doesn't make sense to me to use it in
>that function.
>>
>> What's the exact problem this patch tries to solve?
>
>I'm running a -next kernel on my laptop and couple of days ago keyboard
>backlight controls stopped working: only 2 lower brightness levels.
>I've
>debugged it a bit and found out that acpi_video_switch_brightness()
>calls
>acpi_video_device_lcd_get_level_current() with level uninitialized and
>always gets lowest posible value.
The point is: after the acpi_evaluate_object call, *level must contain the current brightness level, otherwise your BIOS is broken (this happens: e.g. my laptop always returns 100 from _BQC in Windows 8 mode). You can verify this easily by initializing level_current to some invalid value and checking it again after the call to _get_level_current. I'd be pretty surprised if the value didn't change. Also, if you look at [1], you'll see that the code operated on *level before as well.
This problem may have been obscured by the fact that the brightness device wasn't initialized properly before my patches went in. Does acpi_video_switch_brightness actually do anything for you when reverting the 3 newest commits of video.c?
Regards,
Danny
[1] https://git.kernel.org/cgit/linux/kernel/git/rafael/linux-pm.git/commit/drivers/acpi/video.c?h=linux-next&id=994fa63c5b126df6e9f31ef4e09000e2e243234b
On Sun, Mar 31, 2013 at 09:04:21AM +0200, Danny Baumann wrote:
> Hi,
>
> Artem Savkov <[email protected]> schrieb:
>
> >On Sun, Mar 31, 2013 at 08:15:14AM +0200, Danny Baumann wrote:
> >> Artem Savkov <[email protected]> schrieb:
> >> >acpi_video_device_lcd_get_level_current() called
> >> >acpi_video_bqc_value_to_level()
> >> >with "*level" as a second argument, resulting in level being
> >returned
> >> >based on
> >> >initial input, not current brightness, breaking backlight controls.
> >> I don't think this change is correct. As level was passed as
> >parameter into the evaluation of _BQC, *level contains the AML returned
> >brightness level afterwards, so it's correct to use it as an input to
> >acpi_video_bqc_value_to_level(). Actually, the whole point of
> >acpi_video_device_lcd_get_level_current() is to update
> >device->brightness->curr, so it doesn't make sense to me to use it in
> >that function.
> >>
> >> What's the exact problem this patch tries to solve?
> >
> >I'm running a -next kernel on my laptop and couple of days ago keyboard
> >backlight controls stopped working: only 2 lower brightness levels.
> >I've
> >debugged it a bit and found out that acpi_video_switch_brightness()
> >calls
> >acpi_video_device_lcd_get_level_current() with level uninitialized and
> >always gets lowest posible value.
>
> The point is: after the acpi_evaluate_object call, *level must contain the current brightness level, otherwise your BIOS is broken (this happens: e.g. my laptop always returns 100 from _BQC in Windows 8 mode). You can verify this easily by initializing level_current to some invalid value and checking it again after the call to _get_level_current. I'd be pretty surprised if the value didn't change. Also, if you look at [1], you'll see that the code operated on *level before as well.
> This problem may have been obscured by the fact that the brightness device wasn't initialized properly before my patches went in. Does acpi_video_switch_brightness actually do anything for you when reverting the 3 newest commits of video.c?
Yes, you are right. Both my patch and the version before your patches
result in initial acpi_video_device_lcd_get_level_current call ending
with invalid level being returned and bqc disabled, so they both just
use *level = brightness->curr after that.
--
Regards,
Artem
On 03/31/2013 04:16 PM, Artem Savkov wrote:
> On Sun, Mar 31, 2013 at 09:04:21AM +0200, Danny Baumann wrote:
>> Hi,
>>
>> Artem Savkov <[email protected]> schrieb:
>>
>>> On Sun, Mar 31, 2013 at 08:15:14AM +0200, Danny Baumann wrote:
>>>> Artem Savkov <[email protected]> schrieb:
>>>>> acpi_video_device_lcd_get_level_current() called
>>>>> acpi_video_bqc_value_to_level()
>>>>> with "*level" as a second argument, resulting in level being
>>> returned
>>>>> based on
>>>>> initial input, not current brightness, breaking backlight controls.
>>>> I don't think this change is correct. As level was passed as
>>> parameter into the evaluation of _BQC, *level contains the AML returned
>>> brightness level afterwards, so it's correct to use it as an input to
>>> acpi_video_bqc_value_to_level(). Actually, the whole point of
>>> acpi_video_device_lcd_get_level_current() is to update
>>> device->brightness->curr, so it doesn't make sense to me to use it in
>>> that function.
>>>>
>>>> What's the exact problem this patch tries to solve?
>>>
>>> I'm running a -next kernel on my laptop and couple of days ago keyboard
>>> backlight controls stopped working: only 2 lower brightness levels.
>>> I've
>>> debugged it a bit and found out that acpi_video_switch_brightness()
>>> calls
>>> acpi_video_device_lcd_get_level_current() with level uninitialized and
>>> always gets lowest posible value.
>>
>> The point is: after the acpi_evaluate_object call, *level must contain the current brightness level, otherwise your BIOS is broken (this happens: e.g. my laptop always returns 100 from _BQC in Windows 8 mode). You can verify this easily by initializing level_current to some invalid value and checking it again after the call to _get_level_current. I'd be pretty surprised if the value didn't change. Also, if you look at [1], you'll see that the code operated on *level before as well.
>> This problem may have been obscured by the fact that the brightness device wasn't initialized properly before my patches went in. Does acpi_video_switch_brightness actually do anything for you when reverting the 3 newest commits of video.c?
>
> Yes, you are right. Both my patch and the version before your patches
> result in initial acpi_video_device_lcd_get_level_current call ending
> with invalid level being returned and bqc disabled, so they both just
> use *level = brightness->curr after that.
>
OK, this suggests the _BQC control method of your system is broken.
Please kindly file a bug in bugzilla.kernel.org, against
ACPI/Power-Video and attach the output of acpidump and dmidecode there,
I want to take a look, thanks.
-Aaron
On Sun, Mar 31, 2013 at 06:07:57PM +0800, Aaron Lu wrote:
> On 03/31/2013 04:16 PM, Artem Savkov wrote:
> >On Sun, Mar 31, 2013 at 09:04:21AM +0200, Danny Baumann wrote:
> >>Hi,
> >>
> >>Artem Savkov <[email protected]> schrieb:
> >>
> >>>On Sun, Mar 31, 2013 at 08:15:14AM +0200, Danny Baumann wrote:
> >>>>Artem Savkov <[email protected]> schrieb:
> >>>>>acpi_video_device_lcd_get_level_current() called
> >>>>>acpi_video_bqc_value_to_level()
> >>>>>with "*level" as a second argument, resulting in level being
> >>>returned
> >>>>>based on
> >>>>>initial input, not current brightness, breaking backlight controls.
> >>>>I don't think this change is correct. As level was passed as
> >>>parameter into the evaluation of _BQC, *level contains the AML returned
> >>>brightness level afterwards, so it's correct to use it as an input to
> >>>acpi_video_bqc_value_to_level(). Actually, the whole point of
> >>>acpi_video_device_lcd_get_level_current() is to update
> >>>device->brightness->curr, so it doesn't make sense to me to use it in
> >>>that function.
> >>>>
> >>>>What's the exact problem this patch tries to solve?
> >>>
> >>>I'm running a -next kernel on my laptop and couple of days ago keyboard
> >>>backlight controls stopped working: only 2 lower brightness levels.
> >>>I've
> >>>debugged it a bit and found out that acpi_video_switch_brightness()
> >>>calls
> >>>acpi_video_device_lcd_get_level_current() with level uninitialized and
> >>>always gets lowest posible value.
> >>
> >>The point is: after the acpi_evaluate_object call, *level must contain the current brightness level, otherwise your BIOS is broken (this happens: e.g. my laptop always returns 100 from _BQC in Windows 8 mode). You can verify this easily by initializing level_current to some invalid value and checking it again after the call to _get_level_current. I'd be pretty surprised if the value didn't change. Also, if you look at [1], you'll see that the code operated on *level before as well.
> >>This problem may have been obscured by the fact that the brightness device wasn't initialized properly before my patches went in. Does acpi_video_switch_brightness actually do anything for you when reverting the 3 newest commits of video.c?
> >
> >Yes, you are right. Both my patch and the version before your patches
> >result in initial acpi_video_device_lcd_get_level_current call ending
> >with invalid level being returned and bqc disabled, so they both just
> >use *level = brightness->curr after that.
> >
>
> OK, this suggests the _BQC control method of your system is broken.
> Please kindly file a bug in bugzilla.kernel.org, against
> ACPI/Power-Video and attach the output of acpidump and dmidecode there,
> I want to take a look, thanks.
>
> -Aaron
>
Submitted: https://bugzilla.kernel.org/show_bug.cgi?id=56011
--
Regards,
Artem