[Letux-kernel] DSI panels with omapdrm on 4.20-rc

[H. Nikolaus Schaller]

Hi,

> Am 12.11.2018 um 21:09 schrieb Sebastian Reichel <sre at kernel.org>:
> 
> Hi,
> 
> On Mon, Nov 12, 2018 at 10:30:21AM -0800, Tony Lindgren wrote:
>> * H. Nikolaus Schaller <hns at goldelico.com> [181112 17:35]:
>>> 4.20-rc1 seems to have broken our DSI panel driver for the OMAP5 Pyra.
> 
> Laurent's patches break omapdrm DSI support in 4.20 (please read as
> factual information, not as insult). Note, that there is no working
> mainline user of this, so technically its not a user visible
> regression.

Yes, but makes the same problems for out-of tree users :)

The Pyra is planned to become a mainline user of omapdrm.dsi, but as
long as there are only a handful of prototypes it is still too early to
submit (parts of) the DTS. I think it is more convincing for inclusion
if mass production of the hardware is already running.

> 
>>> There were some omapdrm API changes which I have tried to fix by inspecting
>>> panel-dsi-cm.c and also comparing with panel-tpo-td028ttec1.c (which is still working
>>> on our omap3 systems).
>>> 
>>> I also have removed the tiler patches by Matthijs to make sure that
>>> they do not have an influence.
>>> 
>>> The strange thing is that the panel driver module is not even loaded
>>> during probe time, although we have defined the display alias and not
>>> changed that compared to 4.19.
>>> 
>>> The other strange thing is that I can manually modprobe the driver
>>> and it is loaded and shown by lsmod, but the probe function is never
>>> called.
>>> 
>>> Any hints what we should check and compare and verify?
> 
> This is probably fixed by the following patch, which depends on
> Laurent's fixes from a few days ago:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-n900.git/commit/?h=droid4-display-omapdrm-4.20-next&id=9aedb0adc243180c08d000d9d9f811da229857c5

Yeah! It now loads the panel driver again:

[   10.344961] dsi: w677l_probe_of()
[   10.628239] dsi: w677l_connect()

The latter ends in a NULL pointer deref, but this is obviously my fault
when converting the driver to the new API. And I can debug that now.

> 
>>> Did the panel driver compatible matching mechanism change for dsi panels
>>> without a bus (dpi panels are usually loaded as an spi bus client)?
>>> 
>>> Or is there a new CONFIG that must be set?
>> 
>> Test with Laurent's fixes, and also looks like Sebastian also has some
>> recent yet to be posted probe order changes on top of Laurent's fixes
>> in his linux-n900 branch if Laurent's fixes alone don't help.
>> 
>> Laurent's patches are "[PATCH v3 0/4] omapdrm: Fix runtime PM issues
>> at module load and unload time" and Sebastian's git tree is at
>> https://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-n900.git/.
>> 
>> Adding Laurent and Sebastian to Cc too.
> 
> I hope to send a new patchset fixing panel-dsi-cm support in omapdrm
> later this week (assuming, that nothing unexpected happens).
> 
> -- Sebastian

BR and thanks,
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/20181112/0440fac3/attachment.asc>