Linux 2.6.39-rc5

Hey,

* Michael Chang -- Friday 13 May 2011:
> But there's more questions in my mind, made me feel not
> able to proceed any further..

No problem. The reason for inconsistencies in my reports is
simply that I've realized some properties only later. So here's
a new error description, based on ddb503b42960, which is current
HEAD.


On this "Acer Travelmate 5735Z-452G32Mnss" the following happens
since after 2.6.37-rc8, with acpi_osi=Linux:

- the backlight goes dark as soon as KMS takes over early in the
boot process (the screen contents aren't corrupted, though, and
under appropriate lighting conditions I can even see the
(very dark) uncorrupted contents.)

- when I press the "backlight darker" key, the backlight is turned
on. (No other key does that AFAICS, including "display toggle"
and "backlight brighter.)

Everything works correctly after that, including backlight adjustment,
BUT:

- when I close the lid and open it again, the backlight stays black
again, just like before. Backlight adjustment turns it on, and
now even the "brighter" key does it sometimes, but not always.)



> 1. is_backlight_combination_mode() really returns 0x4 ..?

From [bugzilla.kernel.org], comment #8:
"It turned out that on this machine INTEL_INFO(dev)->gen equals 4,
and is_backlight_combination_mode() returns 0x40000000."

That was true when I reported it. But lpbc was always 0 in my tests
back then, and now it's not.



> I understand your words, but if this is true then Joey Lee provided
> patch should log out some information about changing the LPBC
>
> set backlight lbpc = ...
>
> but I didn't see any of the log output ..

You hadn't told me that these patches were meant to be incremental,
so I backed the first logging patch out before applying the other.
Here is the result of ddb503b42960 with both logging patches applied:

boot phase:
[ 2.620237] [drm] Initialized drm 1.1.0 20060810
[ 2.664806] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[ 2.664812] i915 0000:00:02.0: setting latency timer to 64
[ 2.726569] [drm:intel_opregion_setup], graphic opregion physical addr: 0x7ba8c018
[ 2.726612] [drm:intel_opregion_setup], Public ACPI methods supported
[ 2.726615] [drm:intel_opregion_setup], SWSCI supported
[ 2.726617] [drm:intel_opregion_setup], ASLE supported
[ 2.726650] i915 0000:00:02.0: irq 44 for MSI/MSI-X
[ 2.726655] [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
[ 2.726656] [drm] Driver supports precise vblank timestamp query.
[ 2.753294] [drm:init_status_page], render ring hws offset: 0x00000000
[ 2.753431] [drm:init_status_page], bsd ring hws offset: 0x00021000
[ 2.753547] [drm:intel_parse_bios], Using VBT from OpRegion: $VBT CANTIGA d
[ 2.797315] [drm:intel_panel_get_backlight], get backlight val = 2890
[ 2.797319] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 2.797322] [drm:intel_panel_get_backlight], get backlight PWM = 0
[ 2.797328] vgaarb: device changed decodes: PCI:0000:00:02.0,olddecodes=io+mem,decodes=io+memwns=io+mem
[ 2.797658] [drm:intel_panel_set_backlight], set backlight PWM = 0
[ 2.797661] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 2.797664] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 2.797667] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 2.797669] [drm:intel_panel_set_backlight], set backlight max = 736950
[ 2.797671] [drm:intel_panel_set_backlight], set backlight lbpc = 1
[ 2.797678] [drm:intel_opregion_asle_intr], non asle set request??
[ 2.797681] [drm:intel_panel_set_backlight], set backlight tmp(1) = 189401930
[ 2.797683] [drm:intel_panel_set_backlight], set backlight tmp(2) = 189399040
[ 2.797685] [drm:intel_panel_set_backlight], set backlight level = 0
[ 2.905962] scsi 4:0:0:0: Direct-Access Generic- Multi-Card 1.00 PQ: 0 ANSI: 0 CCS
[ 3.040041] [drm:gm45_get_vblank_counter], trying to get vblank count for disabled pipe A
[ 3.040047] [drm:gm45_get_vblank_counter], trying to get vblank count for disabled pipe A
[ 3.090368] checking generic (80000000 3ff0000) vs hw (80000000 10000000)
[ 3.090373] fb: conflicting fb hw usage inteldrmfb vs VESA VGA - removing generic driver
[ 3.090392] Console: switching to colour dummy device 80x25
[ 3.091058] fbcon: inteldrmfb (fb0) is primary device
[ 3.125522] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 3.125525] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 3.125527] [drm:intel_panel_set_backlight], set backlight PWM = 736950
[ 3.125529] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 3.125531] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 3.125533] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 3.125535] [drm:intel_panel_set_backlight], set backlight max = 736950
[ 3.125537] [drm:intel_panel_set_backlight], set backlight lbpc = 255
[ 3.125541] [drm:intel_panel_set_backlight], set backlight tmp(1) = 189399040
[ 3.125543] [drm:intel_panel_set_backlight], set backlight tmp(2) = 189399040
[ 3.125544] [drm:intel_panel_set_backlight], set backlight level = 2890
[ 3.125557] [drm:intel_opregion_asle_intr],
[ 3.125559] Console: switching non asle set request??
[ 3.125562] to colour frame buffer device 170x48
[ 3.128264] fb0: inteldrmfb frame buffer device
[ 3.128265] drm: registered panic notifier
[ 3.429079] acpi device:07: registered as cooling_device2
[ 3.429425] input: Video Bus as /devices/LNXSYSTM:00/device:00/PNP0A08:00/LNXVIDEO:01/input/input6
[ 3.429521] ACPI: Video Device [OVGA] (multi-head: yes rom: no post: no)
[ 3.429980] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0
[ 3.806479] sd 4:0:0:0: [sdb] 15720448 512-byte logical blocks: (8.04 GB/7.49 Gi
[ 3.807220] sd 4:0:0:0: [sdb] Write Protect is off
[ 3.807222] sd 4:0:0:0: [sdb] Mode Sense: 03 00 00 00
[ 3.807225] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[ 3.809345] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[ 3.813728] sdb: sdb1
[ 3.815475] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[ 3.815513] sd 4:0:0:0: [sdb] Attached SCSI removable disk
[ 4.786309] md: linear personality registered for level -1
[ 84.936637] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 84.936641] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 84.936644] [drm:intel_panel_set_backlight], set backlight PWM = 72250
[ 84.936646] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 84.936649] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 84.936652] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 84.936654] [drm:intel_panel_set_backlight], set backlight max = 736950
[ 84.936656] [drm:intel_panel_set_backlight], set backlight lbpc = 25
[ 84.936660] [drm:intel_panel_set_backlight], set backlight tmp(1) = 189401930
[ 84.936662] [drm:intel_panel_set_backlight], set backlight tmp(2) = 189399040
[ 84.936664] [drm:intel_panel_set_backlight], set backlight level = 2890
[ 84.936669] [drm:intel_opregion_asle_intr], non asle set request??
[ 84.936672] [drm:intel_opregion_asle_intr], non asle set request??

brighter:
[ 325.356259] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 325.356268] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 325.356273] [drm:intel_panel_set_backlight], set backlight PWM = 0
[ 325.356278] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 325.356283] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 325.356288] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 325.356292] [drm:intel_panel_set_backlight], set backlight max = 736950
[ 325.356297] [drm:intel_panel_set_backlight], set backlight lbpc = 1
[ 325.356302] [drm:intel_panel_set_backlight], set backlight tmp(1) = 189401930
[ 325.356307] [drm:intel_panel_set_backlight], set backlight tmp(2) = 189399040
[ 325.356311] [drm:intel_panel_set_backlight], set backlight level = 0
[ 325.356318] [drm:intel_opregion_asle_intr], non asle set request??
[ 325.356324] [drm:intel_opregion_asle_intr], non asle set request??

darker:
[ 359.642417] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 359.642424] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 359.642429] [drm:intel_panel_set_backlight], set backlight PWM = 72250
[ 359.642434] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 359.642439] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 359.642443] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 359.642448] [drm:intel_panel_set_backlight], set backlight max = 736950
[ 359.642452] [drm:intel_panel_set_backlight], set backlight lbpc = 25
[ 359.642458] [drm:intel_panel_set_backlight], set backlight tmp(1) = 189399040
[ 359.642463] [drm:intel_panel_set_backlight], set backlight tmp(2) = 189399040
[ 359.642467] [drm:intel_panel_set_backlight], set backlight level = 2890
[ 359.642474] [drm:intel_opregion_asle_intr], non asle set request??
[ 359.642480] [drm:intel_opregion_asle_intr], non asle set request??



> [marc.info]
>
> If you rebase this patch to your currently git commit, will it fixes
> all problems [...]

This patch didn't fix my problems back then, and since an always zeroed lbpc
is apparently not the case, I assume this test wouldn't be very helpful. Of
course, if you insist, then I'll try it again.

Thanks for your help, and sorry for the confusion. :-)
m.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at [vger.kernel.org]
Please read the FAQ at [www.tux.org]

* Takashi Iwai -- Sunday 15 May 2011:
> IIRC, you reported that the backlight gets normal when you revert my
> commit in 2.6.38.x. So, this was regarded as a regression at first.

Yes. And it *is* a regression, which is the whole point of my
initial complaint, as reported by Maciej in
[bugzilla.kernel.org]



> But, one question remains: whether the backlight level control worked
> with the reverted kernel?

Good point. Turns out it didn't work with 2.6.38-rc8 either. But it
did work at some time before. (I use this notebook mainly as a
terminal ATM, so I didn't care much for backlight level. This came
up later during investigation.)

So the only thing that 2.6.38 broke was that the backlight
was initially off. Adjustment had already been broken before
(and works now again; sigh ... confused? I am! :-).



> If you can still change the level without
> LBPC, the former analysis was incorrect.

I don't even know what an LBPC is, other than a variable named like that.
So I'd need a hint for how to test that.



> Also, with the latest 2.6.38.x, you found that the backlight gets back
> when you adjust the level down.

When I reported this, it was about 2.6.39-* and HEAD, not stable
versions. But I tried now, and openSuSE's 2.6.38.6 behaves the same.



> Another question now is what happens if you again turn it up to the
> max level. Is the backlight still on?

Yes. If the backlight was on at one point, then increasing the level
to maximum never turned if off.



> If the backlight is kept on even with the max level, it implies that
> the problem is only the initial value; once when set correctly, it'll
> work fine after that.

Yes, that's the case. (Except that after closing the lid it's off again.)

m.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at [vger.kernel.org]
Please read the FAQ at [www.tux.org]

2011/5/16 Melchior FRANZ <melchior.franz@gmail.com>:

> I don't even know what an LBPC is, other than a variable named like that.
> So I'd need a hint for how to test that.

It resides at your vga adapter's pci configuration space, offset 0xf4.

You can know your LPBC value by:
$ lspci -xxx -s 00:02.0 | awk '/^f0:/ {print $6}'

And alter it's value via setpci (assuming set it to max)
$ setpci -s 00:02.0 F4.B=ff

I assume you've tried this .. as you report setpci works for you.

From your new attached log, In booting phase your LPBC is set to 255
and actual brightness is maximum (255*2890=736950). I see no other
change after it until manual brightness change. This is a totally
reversed to your problem.

Could you help to get the LPBC value in the initial dark condition? If
it's not corresponds to 255, maybe something sneaks and changes it but
not logged.

[ 3.125522] [drm:is_backlight_combination_mode],
BLM_COMBINATION_MODE = 1073741824
[ 3.125525] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 3.125527] [drm:intel_panel_set_backlight], set backlight PWM = 736950
[ 3.125529] [drm:is_backlight_combination_mode],
BLM_COMBINATION_MODE = 1073741824
[ 3.125531] [drm:is_backlight_combination_mode],
BLM_COMBINATION_MODE = 1073741824
[ 3.125533] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 3.125535] [drm:intel_panel_set_backlight], set backlight max = 736950
[ 3.125537] [drm:intel_panel_set_backlight], set backlight lbpc = 255
[ 3.125544] [drm:intel_panel_set_backlight], set backlight level = 2890

> Yes, that's the case. (Except that after closing the lid it's off again.)

Maybe this is another clue for root cause. Would it be convenient for
you to gather the log?

1. $ dmesg -c
2. Close the lid
3. Open the lid
4. $ dmesg > dmesg.lid

Thanks a lot,
Michael Chang
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at [vger.kernel.org]
Please read the FAQ at [www.tux.org]

* Michael Chang -- Tuesday 17 May 2011:
[LBPC]
> You can know your LPBC value by:
> $ lspci -xxx -s 00:02.0 | awk '/^f0:/ {print $6}'
>
> And alter it's value via setpci (assuming set it to max)
> $ setpci -s 00:02.0 F4.B=ff
>
> I assume you've tried this .. as you report setpci works for you.

Ahh, yes, indeed. This lspci line (almost) always shows a value that
corresponds with the actual backlight brightness, and both lines work
with and without acpi_osi=Linux. (Whereas adjustment via keys only works
*with* acpi_osi=Linux.) (Tested with 2.6.39-rc7/bd1a643e10)



> Could you help to get the LPBC value in the initial dark condition? If
> it's not corresponds to 255, maybe something sneaks and changes it but
> not logged.

Initially (with acpi_osi=Linux) LPBC=0xff. After "brightness down"
it's 0x19, and after then pressing "brightness up" it's 0x01. The lowest
value I can achieve after pressing "brightness down" a few times is 0xe5,
the highest is 0x01. Neither 0xff nor 0x00 can be reached via keys, but
can (most of the time -- see below!) be set with the setpci line and give
the expected results, backlight off and fully on. ("display toggle" has
no effect on LBPC, btw.)

Setting 0xFF doesn't always turn the backlight off, though! If LBPC is
at the (key) maximum 0x01, then setting 0xFF via setpci changes only the
LBPC value as shown by lspci, but the backlight isn't actually turned off!
If I then press "backlight darker" once (LBPC=0x19), then setting 0xFF
via setpci works again.



> 1. $ dmesg -c
> 2. Close the lid
> 3. Open the lid
> 4. $ dmesg > dmesg.lid

The results are:

[ 188.571040] [drm:intel_panel_set_backlight], set backlight PWM = 736950
[ 188.571048] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 188.571054] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
[ 188.571059] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[ 188.571064] [drm:intel_panel_set_backlight], set backlight max = 736950
[ 188.571069] [drm:intel_panel_set_backlight], set backlight lbpc = 255
[ 188.571075] [drm:intel_panel_set_backlight], set backlight tmp(1) = 189399040
[ 188.571080] [drm:intel_panel_set_backlight], set backlight tmp(2) = 189399040
[ 188.571084] [drm:intel_panel_set_backlight],
[ 188.571088] [drm:intel_opregion_asle_intr], non asle set request??
[ 188.571093] set backlight level = 2890

The backlight is off now, as always after closing/reopening the lid,
and LPBC=0xff.

m.



PS: I'd be willing to join an irc.freenode.net channel in case you or
anyone else wants quicker ("live" responses to such queries.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at [vger.kernel.org]
Please read the FAQ at [www.tux.org]

Melchior,

I want to double check with you. If you change brightness from lowest
to highest, the LPBC value changes this way

Highest Lowest
0x10 , 0x19 ...... 0xe5

And vice versa, from Lowest to Highest

Lowest Highest
0xe5 , ...... 0x19 0x10

If above is true, then you have a very particular notebook that
reverse the sense of the LPBC values .... I don't know who ever came
across such device. wow.

Regarding that your brightness control key, due to that it is
controlled by acer_wmi and likely that BIOS aware this, it would do
brightness in a reversed way. I think Joey would check DSDT and see
any clue or evidence for this. OTOH the i915 driver didn't aware of
this and operates as usual sense that means falling into the trap.

PS. I see your log always reverse the sense of this LPBC value, now I
know why.

Regards,
Michael Chang

2011/5/17 Melchior FRANZ <melchior.franz@gmail.com>:
> * Michael Chang -- Tuesday 17 May 2011:
> [LBPC]
>> You can know your LPBC value by:
>>   $ lspci -xxx -s 00:02.0 | awk '/^f0:/ {print $6}'
>>
>> And alter it's value via setpci (assuming set it to max)
>>   $ setpci -s 00:02.0 F4.B=ff
>>
>> I assume you've tried this .. as you report setpci works for you.
>
> Ahh, yes, indeed. This lspci line (almost) always shows a value that
> corresponds with the actual backlight brightness, and both lines work
> with and without acpi_osi=Linux. (Whereas adjustment via keys only works
> *with* acpi_osi=Linux.)  (Tested with 2.6.39-rc7/bd1a643e10)
>
>
>
>> Could you help to get the LPBC value in the initial dark condition? If
>> it's not corresponds to 255, maybe something sneaks and changes it but
>> not logged.
>
> Initially (with acpi_osi=Linux) LPBC=0xff. After "brightness down"
> it's 0x19, and after then pressing "brightness up" it's 0x01. The lowest
> value I can achieve after pressing "brightness down" a few times is 0xe5,
> the highest is 0x01. Neither 0xff nor 0x00 can be reached via keys, but
> can (most of the time -- see below!) be set with the setpci line and give
> the expected results, backlight off and fully on. ("display toggle" has
> no effect on LBPC, btw.)
>
> Setting 0xFF doesn't always turn the backlight off, though! If LBPC is
> at the (key) maximum 0x01, then setting 0xFF via setpci changes only the
> LBPC value as shown by lspci, but the backlight isn't actually turned off!
> If I then press "backlight darker" once (LBPC=0x19), then setting 0xFF
> via setpci works again.
>
>
>
>>  1. $ dmesg -c
>>  2. Close the lid
>>  3. Open the lid
>>  4. $ dmesg > dmesg.lid
>
> The results are:
>
> [  188.571040] [drm:intel_panel_set_backlight], set backlight PWM = 736950
> [  188.571048] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
> [  188.571054] [drm:is_backlight_combination_mode], BLM_COMBINATION_MODE = 1073741824
> [  188.571059] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
> [  188.571064] [drm:intel_panel_set_backlight], set backlight max = 736950
> [  188.571069] [drm:intel_panel_set_backlight], set backlight lbpc = 255
> [  188.571075] [drm:intel_panel_set_backlight], set backlight tmp(1) = 189399040
> [  188.571080] [drm:intel_panel_set_backlight], set backlight tmp(2) = 189399040
> [  188.571084] [drm:intel_panel_set_backlight],
> [  188.571088] [drm:intel_opregion_asle_intr], non asle set request??
> [  188.571093] set backlight level = 2890
>
> The backlight is off now, as always after closing/reopening the lid,
> and LPBC=0xff.
>
> m.
>
>
>
> PS: I'd be willing to join an irc.freenode.net channel in case you or
>    anyone else wants quicker ("live" responses to such queries.
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at [vger.kernel.org]
Please read the FAQ at [www.tux.org]

Hey,

* Michael Chang -- Tuesday 17 May 2011:
> If you change brightness from lowest to highest, the LPBC value changes
> this way
>
> Highest Lowest
> 0x10 , 0x19 ...... 0xe5

Yes. (Though it's 0x01, not 0x10.)

---0xFF ... initial value and after closing/reopening lid
| 0xE5 \ ... dark
| 0xCC |
| 0xB2 |
(1)| 0x99 |
| 0x7F |_____key adjustable range
| 0x66 |
| 0x4C |
| 0x33 |
-->0x19 |
0x01 / ... bright
0x00 ... also bright, only settable with setpci


(1) is the jump that resurrects the backlight from darkness by
pressing the "brightness down" key

And again, if fully turned up via keys (and therefore at 0x01),
then setting to 0xff with setpci doesn't turn it off. At all other
key steps it does. :-)



> If above is true, then you have a very particular notebook that
> reverse the sense of the LPBC values .... I don't know who ever came
> across such device. wow.

Maybe they screwed it up and therefore sold it with Linux ("linpus"
and a custom driver preinstalled ...



> Regarding that your brightness control key, due to that it is
> controlled by acer_wmi and likely that BIOS aware this, it would do
> brightness in a reversed way.

acer_wmi doesn't do the brightness adjustment. But maybe under "linpus"
some variant of it did. (BTW: there are two keys on this notebook that
aren't handled by the kernel, by X11 or even acer_wmi. There are just no
events for them. Looks like another ACER brain damage err.. "capability". :-)

m.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at [vger.kernel.org]
Please read the FAQ at [www.tux.org]