[Letux-kernel] weird problem with pwm_bl on omap3

[H. Nikolaus Schaller]

Hi,

> Am 29.05.2018 um 22:45 schrieb Andreas Kemnade <andreas at kemnade.info>:
> 
> Hi,
> 
> On Tue, 29 May 2018 22:03:52 +0200
> "H. Nikolaus Schaller" <hns at goldelico.com> wrote:
> 
>> Ah,
>> more info below
>> 
>>> Am 29.05.2018 um 21:55 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
>>> 
>>> Hi,
>>> 
>>>>> GTA04A5 that a GTA04A4.
>>>>> 
>>>> Reproducible with letux-4.17-rc5
>>> 
>>> Good (even if bad :)
>>> 
>>>> not reproducible with  mainline 4.17-rc7 + letux-4.17-rc5 dtb
>> 
>> Here is an excerpt from by bisect script which makes it appear if I
>> pick this minimal set of letux branches/commits on top of mainline 4.16..4.17:
>> 
>> 			picklatest work/letux-base/dt-gta04-common || return
> 
> dt only, implicitly used by me, too

ok.

> 
>> 			picklatest work/letux-base/arm_defconfig || return
> 
> defconfig, I am using it too implicitly

ok.

> 
>> 			picklatest work/letux-base/Letux || return
> that branch looks a bit strange, latest commit is
> 
> commit 315ed0ad432764ba1598dcd0581e10e34c830d3a
> Author: H. Nikolaus Schaller <hns at goldelico.com>
> Date:   Sun Sep 17 09:08:19 2017 +0200
> 
>    Merge tag 'v4.14-rc1' into letux-base
> 
>    Linux 4.14-rc1
> 
> 
> on top of something related to 4.17

Yes, that comes from the automatic deduction of branch points by git blame.

> 
>> 			picklatest work/letux-base/compile-fixes || return
> 
> nothing interesting for me

Yes, should not make a difference since we both see the problem.

>> 			picklatest work/letux-base/dt-gta04 || return
> dt, same as above

ok.

> 
>> 			pick 18ee03dc6f6b || return	# full serdev patchset may end in merge conflict
> bluetooth, should change anything without uartpm stuff

Should not make a difference, especially if we are using gta04a5.dtb

> 
> 
>> 			picklatest work/letux-base/dt-gta04a5 || return
> dt again, same as above
> 
> 
> 
>> 			picklatest work/letux-base/marek/generic-adc-battery || return	# because we include some .dtsi
> 
> oh, here we have actual code... which could be evil

Yes, it is a strong suspect, as I have seen similar worker_thread issues with it.

But I also get the pwm_bl problems when blacklisting the generic-adc-battery
driver.

And I got the same failure with generic-adc-battery once from the bq27xxx driver.

> 
>> 			picklatest work/letux-base/hacks || return
> hmm, why the display stuff is still in there? Actual code here too...
> which could be evil

Indeed... This branch collects hacks suggested by others after reporting
problems.

It appears that we never got a feedback that things have been solved upstream
and we can remove the patch.

Or the patch simply did not arrive upstream (which would have removed it by
rebasing). Some people are very helpful by providing hacks but forget to
upstream.

But I have the pwm_bl problem also with blacklisting omapdss so that the
display patches are dead code.

If we go through them:

9be4cf656074 media: omap3isp: fix unbalanced dma_iommu_mapping

	forward-port from linux-next - but omap3isp is blacklisted

4633366b1b3e ARM: dts: omap3-beagle-xm: Add fixed 26MHz clock as fck for twl

	forward-port from linux-next

7318d0352d47 mfd: twl-core: Fix clock initialization

	forward-port from linux-next

4e3996d3a048 drm/omap: work-around for omap3 display enable

	omapdss blacklisted

a20e01a7877d Revert "[PATCH] dispc hackfix"
15abba214c63 [PATCH] dispc hackfix

	the above two are empty in combination

5518e5b3554b drivers: omap: dmtimer: this hack solves L3 custom errors

	this is pwm_bl related

d6d9fdf0fbdd mmc: hack pwrseq to fix gta04 bt problems

	not used on gta04a5 (uses different WLAN/BT combo)

> 
>> 			picklatest work/letux-base/ubi-fs || return
> 
> we fix here a problem created by changing nand config in dt (which is
> incompatible to <=3.12)

Should not be part of the problem - it is just to keep the boot log clean
of ubifs errors.

So in summary, if our observations are correct, we should look deeper into

>> work/letux-base/marek/generic-adc-battery
>> work/letux-base/hacks

BR,
Nikolaus

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.goldelico.com/pipermail/letux-kernel/attachments/20180530/73239fdd/attachment.asc>