diff mbox series

[1/4] ARM: dts: omap3-n900: fix LCD reset line polarity

Message ID 20221004213503.848262-1-dmitry.torokhov@gmail.com
State New
Headers show
Series [1/4] ARM: dts: omap3-n900: fix LCD reset line polarity | expand

Commit Message

Dmitry Torokhov Oct. 4, 2022, 9:35 p.m. UTC
The LCD driver (panel-sony-acx565akm), when probing, starts with line
driven low, and then toggles it to high and keeps it there. Also, the
line is driven low when powering off the device, and ls released when
powering it back on. This means that the reset line should be described
as "active low" in DTS. This will be important when the driver is
converted to gpiod API which respects the polarity declared in DTS.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
---
 arch/arm/boot/dts/omap3-n900.dts | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Tony Lindgren Oct. 11, 2022, 5:45 a.m. UTC | #1
Hi,

* Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
> The LCD driver (panel-sony-acx565akm), when probing, starts with line
> driven low, and then toggles it to high and keeps it there. Also, the
> line is driven low when powering off the device, and ls released when
> powering it back on. This means that the reset line should be described
> as "active low" in DTS. This will be important when the driver is
> converted to gpiod API which respects the polarity declared in DTS.

We should ensure these patches get merged together with the driver
change to avoid breaking LCD for booting. Probably no need to have
the driver quirk handling for inverted polartity in this case.

It's probably easiest to have an immutable branch for the driver
changes I can base the dts changes on. Or I can ack the dts changes
if they get merged with the driver.

Regards,

Tony
Sebastian Reichel Oct. 11, 2022, 12:37 p.m. UTC | #2
Hi,

On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
> > The LCD driver (panel-sony-acx565akm), when probing, starts with line
> > driven low, and then toggles it to high and keeps it there. Also, the
> > line is driven low when powering off the device, and ls released when
> > powering it back on. This means that the reset line should be described
> > as "active low" in DTS. This will be important when the driver is
> > converted to gpiod API which respects the polarity declared in DTS.
> 
> We should ensure these patches get merged together with the driver
> change to avoid breaking LCD for booting. Probably no need to have
> the driver quirk handling for inverted polartity in this case.
> 
> It's probably easiest to have an immutable branch for the driver
> changes I can base the dts changes on. Or I can ack the dts changes
> if they get merged with the driver.

Both drivers are already using gpiod API:

drivers/gpu/drm/panel/panel-sony-acx565akm.c
drivers/gpu/drm/panel/panel-dsi-cm.c

So this just breaks things.

-- Sebastian
Dmitry Torokhov Oct. 11, 2022, 2:06 p.m. UTC | #3
Hi Sebastian,

On Tue, Oct 11, 2022 at 02:37:26PM +0200, Sebastian Reichel wrote:
> Hi,
> 
> On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
> > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
> > > The LCD driver (panel-sony-acx565akm), when probing, starts with line
> > > driven low, and then toggles it to high and keeps it there. Also, the
> > > line is driven low when powering off the device, and ls released when
> > > powering it back on. This means that the reset line should be described
> > > as "active low" in DTS. This will be important when the driver is
> > > converted to gpiod API which respects the polarity declared in DTS.
> > 
> > We should ensure these patches get merged together with the driver
> > change to avoid breaking LCD for booting. Probably no need to have
> > the driver quirk handling for inverted polartity in this case.
> > 
> > It's probably easiest to have an immutable branch for the driver
> > changes I can base the dts changes on. Or I can ack the dts changes
> > if they get merged with the driver.
> 
> Both drivers are already using gpiod API:
> 
> drivers/gpu/drm/panel/panel-sony-acx565akm.c
> drivers/gpu/drm/panel/panel-dsi-cm.c

I was looking at

drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c

which are not using gpiod. Should they be retired?

> 
> So this just breaks things.

I missed the drivers in drivers/gpu/... and I see that they essentially
abuse gpiod API as gpiod_set_value() operates on logical level
(active/inactive) and not absolute (high/low). They should either use
the gpiod_*_raw() variants, or they should be adjusted to do the proper
thing together with the accompanying DTS change.

What are your preferences?

Thanks.
Tony Lindgren Oct. 11, 2022, 2:30 p.m. UTC | #4
* Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 13:57]:
> Hi Sebastian,
> 
> On Tue, Oct 11, 2022 at 02:37:26PM +0200, Sebastian Reichel wrote:
> > Hi,
> > 
> > On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
> > > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
> > > > The LCD driver (panel-sony-acx565akm), when probing, starts with line
> > > > driven low, and then toggles it to high and keeps it there. Also, the
> > > > line is driven low when powering off the device, and ls released when
> > > > powering it back on. This means that the reset line should be described
> > > > as "active low" in DTS. This will be important when the driver is
> > > > converted to gpiod API which respects the polarity declared in DTS.
> > > 
> > > We should ensure these patches get merged together with the driver
> > > change to avoid breaking LCD for booting. Probably no need to have
> > > the driver quirk handling for inverted polartity in this case.
> > > 
> > > It's probably easiest to have an immutable branch for the driver
> > > changes I can base the dts changes on. Or I can ack the dts changes
> > > if they get merged with the driver.
> > 
> > Both drivers are already using gpiod API:
> > 
> > drivers/gpu/drm/panel/panel-sony-acx565akm.c
> > drivers/gpu/drm/panel/panel-dsi-cm.c
> 
> I was looking at
> 
> drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
> drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c

Ah OK that explains :)

> which are not using gpiod. Should they be retired?

Yes we should just get rid of them with omapdrm working just fine.

> > So this just breaks things.
> 
> I missed the drivers in drivers/gpu/... and I see that they essentially
> abuse gpiod API as gpiod_set_value() operates on logical level
> (active/inactive) and not absolute (high/low). They should either use
> the gpiod_*_raw() variants, or they should be adjusted to do the proper
> thing together with the accompanying DTS change.
> 
> What are your preferences?

Seems like high/low at the connected device end is what we should use,
right? Otherwise things will misbehave if the panel is connected to
some other SoC possibly.

Regards,

Tony
Dmitry Torokhov Oct. 11, 2022, 2:38 p.m. UTC | #5
On Tue, Oct 11, 2022 at 05:30:12PM +0300, Tony Lindgren wrote:
> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 13:57]:
> > Hi Sebastian,
> > 
> > On Tue, Oct 11, 2022 at 02:37:26PM +0200, Sebastian Reichel wrote:
> > > Hi,
> > > 
> > > On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
> > > > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
> > > > > The LCD driver (panel-sony-acx565akm), when probing, starts with line
> > > > > driven low, and then toggles it to high and keeps it there. Also, the
> > > > > line is driven low when powering off the device, and ls released when
> > > > > powering it back on. This means that the reset line should be described
> > > > > as "active low" in DTS. This will be important when the driver is
> > > > > converted to gpiod API which respects the polarity declared in DTS.
> > > > 
> > > > We should ensure these patches get merged together with the driver
> > > > change to avoid breaking LCD for booting. Probably no need to have
> > > > the driver quirk handling for inverted polartity in this case.
> > > > 
> > > > It's probably easiest to have an immutable branch for the driver
> > > > changes I can base the dts changes on. Or I can ack the dts changes
> > > > if they get merged with the driver.
> > > 
> > > Both drivers are already using gpiod API:
> > > 
> > > drivers/gpu/drm/panel/panel-sony-acx565akm.c
> > > drivers/gpu/drm/panel/panel-dsi-cm.c
> > 
> > I was looking at
> > 
> > drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
> > drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c
> 
> Ah OK that explains :)
> 
> > which are not using gpiod. Should they be retired?
> 
> Yes we should just get rid of them with omapdrm working just fine.

Will you be submitting such patches? I'd like to get rid of
of_get_named_gpio() and friends if I can...

> 
> > > So this just breaks things.
> > 
> > I missed the drivers in drivers/gpu/... and I see that they essentially
> > abuse gpiod API as gpiod_set_value() operates on logical level
> > (active/inactive) and not absolute (high/low). They should either use
> > the gpiod_*_raw() variants, or they should be adjusted to do the proper
> > thing together with the accompanying DTS change.
> > 
> > What are your preferences?
> 
> Seems like high/low at the connected device end is what we should use,
> right? Otherwise things will misbehave if the panel is connected to
> some other SoC possibly.

It is exactly because of this case the driver should use active/inactive
and follow polarity described in DTS. If the driver does:

	gpiod_set_value_cansleep(d->reset, 1);

then if DTS is saying that the reset line is active low, under the wraps
the line will be driven to "0", but if DTS is saying that the line is
active high, then the very same call will drive the line to "1".

This allows accommodating different designs without having to change the
driver code.
Tony Lindgren Oct. 11, 2022, 2:47 p.m. UTC | #6
* Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 14:30]:
> On Tue, Oct 11, 2022 at 05:30:12PM +0300, Tony Lindgren wrote:
> > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 13:57]:
> > > Hi Sebastian,
> > > 
> > > On Tue, Oct 11, 2022 at 02:37:26PM +0200, Sebastian Reichel wrote:
> > > > Hi,
> > > > 
> > > > On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
> > > > > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
> > > > > > The LCD driver (panel-sony-acx565akm), when probing, starts with line
> > > > > > driven low, and then toggles it to high and keeps it there. Also, the
> > > > > > line is driven low when powering off the device, and ls released when
> > > > > > powering it back on. This means that the reset line should be described
> > > > > > as "active low" in DTS. This will be important when the driver is
> > > > > > converted to gpiod API which respects the polarity declared in DTS.
> > > > > 
> > > > > We should ensure these patches get merged together with the driver
> > > > > change to avoid breaking LCD for booting. Probably no need to have
> > > > > the driver quirk handling for inverted polartity in this case.
> > > > > 
> > > > > It's probably easiest to have an immutable branch for the driver
> > > > > changes I can base the dts changes on. Or I can ack the dts changes
> > > > > if they get merged with the driver.
> > > > 
> > > > Both drivers are already using gpiod API:
> > > > 
> > > > drivers/gpu/drm/panel/panel-sony-acx565akm.c
> > > > drivers/gpu/drm/panel/panel-dsi-cm.c
> > > 
> > > I was looking at
> > > 
> > > drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
> > > drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c
> > 
> > Ah OK that explains :)
> > 
> > > which are not using gpiod. Should they be retired?
> > 
> > Yes we should just get rid of them with omapdrm working just fine.
> 
> Will you be submitting such patches? I'd like to get rid of
> of_get_named_gpio() and friends if I can...

Adding Tomi to Cc, my guess is he already has such patches and knows
better which ones can go :)

> > > > So this just breaks things.
> > > 
> > > I missed the drivers in drivers/gpu/... and I see that they essentially
> > > abuse gpiod API as gpiod_set_value() operates on logical level
> > > (active/inactive) and not absolute (high/low). They should either use
> > > the gpiod_*_raw() variants, or they should be adjusted to do the proper
> > > thing together with the accompanying DTS change.
> > > 
> > > What are your preferences?
> > 
> > Seems like high/low at the connected device end is what we should use,
> > right? Otherwise things will misbehave if the panel is connected to
> > some other SoC possibly.
> 
> It is exactly because of this case the driver should use active/inactive
> and follow polarity described in DTS. If the driver does:
> 
> 	gpiod_set_value_cansleep(d->reset, 1);
> 
> then if DTS is saying that the reset line is active low, under the wraps
> the line will be driven to "0", but if DTS is saying that the line is
> active high, then the very same call will drive the line to "1".
> 
> This allows accommodating different designs without having to change the
> driver code.

OK

Thanks,

Tony
Tomi Valkeinen Oct. 12, 2022, 10:58 a.m. UTC | #7
Hi,

On 11/10/2022 17:47, Tony Lindgren wrote:
> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 14:30]:
>> On Tue, Oct 11, 2022 at 05:30:12PM +0300, Tony Lindgren wrote:
>>> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 13:57]:
>>>> Hi Sebastian,
>>>>
>>>> On Tue, Oct 11, 2022 at 02:37:26PM +0200, Sebastian Reichel wrote:
>>>>> Hi,
>>>>>
>>>>> On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
>>>>>> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
>>>>>>> The LCD driver (panel-sony-acx565akm), when probing, starts with line
>>>>>>> driven low, and then toggles it to high and keeps it there. Also, the
>>>>>>> line is driven low when powering off the device, and ls released when
>>>>>>> powering it back on. This means that the reset line should be described
>>>>>>> as "active low" in DTS. This will be important when the driver is
>>>>>>> converted to gpiod API which respects the polarity declared in DTS.
>>>>>>
>>>>>> We should ensure these patches get merged together with the driver
>>>>>> change to avoid breaking LCD for booting. Probably no need to have
>>>>>> the driver quirk handling for inverted polartity in this case.
>>>>>>
>>>>>> It's probably easiest to have an immutable branch for the driver
>>>>>> changes I can base the dts changes on. Or I can ack the dts changes
>>>>>> if they get merged with the driver.
>>>>>
>>>>> Both drivers are already using gpiod API:
>>>>>
>>>>> drivers/gpu/drm/panel/panel-sony-acx565akm.c
>>>>> drivers/gpu/drm/panel/panel-dsi-cm.c
>>>>
>>>> I was looking at
>>>>
>>>> drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
>>>> drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c
>>>
>>> Ah OK that explains :)
>>>
>>>> which are not using gpiod. Should they be retired?
>>>
>>> Yes we should just get rid of them with omapdrm working just fine.
>>
>> Will you be submitting such patches? I'd like to get rid of
>> of_get_named_gpio() and friends if I can...
> 
> Adding Tomi to Cc, my guess is he already has such patches and knows
> better which ones can go :)

To be honest, I haven't really even had a glance towards fbdev for a 
long time.

There is one thing that omapdrm doesn't support, which is VRFB rotation. 
I cannot say if the users of those above-mentioned panels require VRFB.

>>>>> So this just breaks things.
>>>>
>>>> I missed the drivers in drivers/gpu/... and I see that they essentially
>>>> abuse gpiod API as gpiod_set_value() operates on logical level
>>>> (active/inactive) and not absolute (high/low). They should either use
>>>> the gpiod_*_raw() variants, or they should be adjusted to do the proper
>>>> thing together with the accompanying DTS change.
>>>>
>>>> What are your preferences?
>>>
>>> Seems like high/low at the connected device end is what we should use,
>>> right? Otherwise things will misbehave if the panel is connected to
>>> some other SoC possibly.
>>
>> It is exactly because of this case the driver should use active/inactive
>> and follow polarity described in DTS. If the driver does:
>>
>> 	gpiod_set_value_cansleep(d->reset, 1);
>>
>> then if DTS is saying that the reset line is active low, under the wraps
>> the line will be driven to "0", but if DTS is saying that the line is
>> active high, then the very same call will drive the line to "1".
>>
>> This allows accommodating different designs without having to change the
>> driver code.

Isn't breaking an old dts file quite a bad thing? Why not just add a 
comment to the .dts and to the driver about the situation. I don't quite 
see that the fixing the dts (And, if done properly, adding a boot time 
fixup for old dtbs) and changing the drivers is worth the hassle.

Unless we see new users for these drivers, which would require the new 
users to write broken dts files.

  Tomi
Dmitry Torokhov Oct. 12, 2022, 7:30 p.m. UTC | #8
On Wed, Oct 12, 2022 at 01:58:15PM +0300, Tomi Valkeinen wrote:
> Hi,
> 
> On 11/10/2022 17:47, Tony Lindgren wrote:
> > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 14:30]:
> > > On Tue, Oct 11, 2022 at 05:30:12PM +0300, Tony Lindgren wrote:
> > > > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 13:57]:
> > > > > Hi Sebastian,
> > > > > 
> > > > > On Tue, Oct 11, 2022 at 02:37:26PM +0200, Sebastian Reichel wrote:
> > > > > > Hi,
> > > > > > 
> > > > > > On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
> > > > > > > * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
> > > > > > > > The LCD driver (panel-sony-acx565akm), when probing, starts with line
> > > > > > > > driven low, and then toggles it to high and keeps it there. Also, the
> > > > > > > > line is driven low when powering off the device, and ls released when
> > > > > > > > powering it back on. This means that the reset line should be described
> > > > > > > > as "active low" in DTS. This will be important when the driver is
> > > > > > > > converted to gpiod API which respects the polarity declared in DTS.
> > > > > > > 
> > > > > > > We should ensure these patches get merged together with the driver
> > > > > > > change to avoid breaking LCD for booting. Probably no need to have
> > > > > > > the driver quirk handling for inverted polartity in this case.
> > > > > > > 
> > > > > > > It's probably easiest to have an immutable branch for the driver
> > > > > > > changes I can base the dts changes on. Or I can ack the dts changes
> > > > > > > if they get merged with the driver.
> > > > > > 
> > > > > > Both drivers are already using gpiod API:
> > > > > > 
> > > > > > drivers/gpu/drm/panel/panel-sony-acx565akm.c
> > > > > > drivers/gpu/drm/panel/panel-dsi-cm.c
> > > > > 
> > > > > I was looking at
> > > > > 
> > > > > drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
> > > > > drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c
> > > > 
> > > > Ah OK that explains :)
> > > > 
> > > > > which are not using gpiod. Should they be retired?
> > > > 
> > > > Yes we should just get rid of them with omapdrm working just fine.
> > > 
> > > Will you be submitting such patches? I'd like to get rid of
> > > of_get_named_gpio() and friends if I can...
> > 
> > Adding Tomi to Cc, my guess is he already has such patches and knows
> > better which ones can go :)
> 
> To be honest, I haven't really even had a glance towards fbdev for a long
> time.
> 
> There is one thing that omapdrm doesn't support, which is VRFB rotation. I
> cannot say if the users of those above-mentioned panels require VRFB.
> 
> > > > > > So this just breaks things.
> > > > > 
> > > > > I missed the drivers in drivers/gpu/... and I see that they essentially
> > > > > abuse gpiod API as gpiod_set_value() operates on logical level
> > > > > (active/inactive) and not absolute (high/low). They should either use
> > > > > the gpiod_*_raw() variants, or they should be adjusted to do the proper
> > > > > thing together with the accompanying DTS change.
> > > > > 
> > > > > What are your preferences?
> > > > 
> > > > Seems like high/low at the connected device end is what we should use,
> > > > right? Otherwise things will misbehave if the panel is connected to
> > > > some other SoC possibly.
> > > 
> > > It is exactly because of this case the driver should use active/inactive
> > > and follow polarity described in DTS. If the driver does:
> > > 
> > > 	gpiod_set_value_cansleep(d->reset, 1);
> > > 
> > > then if DTS is saying that the reset line is active low, under the wraps
> > > the line will be driven to "0", but if DTS is saying that the line is
> > > active high, then the very same call will drive the line to "1".
> > > 
> > > This allows accommodating different designs without having to change the
> > > driver code.
> 
> Isn't breaking an old dts file quite a bad thing? Why not just add a comment
> to the .dts and to the driver about the situation. I don't quite see that
> the fixing the dts (And, if done properly, adding a boot time fixup for old
> dtbs) and changing the drivers is worth the hassle.
> 
> Unless we see new users for these drivers, which would require the new users
> to write broken dts files.

Or maybe there are devices with fixed DTSes and fixed up kernels but the
fixes have not been contributed upstream. I don't know...

My personal opinion is that we pay too much attention to DTS
compatibility in cases when it is not totally clear if there are devices
that use DTSes that are not bundled with the kernel and also have a
chance to have their kernel updated (and be lucky enough for the
upstream kernel to work on such device without extensive work).

Anyway, my goal is to stop exposing of_get_named_gpio() and its
derivatives, so please let me know your preference. Should I:

- mirror in omapfb drivers what gpu drivers do and use inverted
  polarity
- use gpiod_set_value_raw() and essentially ignore polarity described
  in DTS
- continue pushing polarity fixes throughout
- something else?

Thanks!
Tomi Valkeinen Oct. 13, 2022, 6:22 a.m. UTC | #9
On 12/10/2022 22:30, Dmitry Torokhov wrote:
> On Wed, Oct 12, 2022 at 01:58:15PM +0300, Tomi Valkeinen wrote:
>> Hi,
>>
>> On 11/10/2022 17:47, Tony Lindgren wrote:
>>> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 14:30]:
>>>> On Tue, Oct 11, 2022 at 05:30:12PM +0300, Tony Lindgren wrote:
>>>>> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221011 13:57]:
>>>>>> Hi Sebastian,
>>>>>>
>>>>>> On Tue, Oct 11, 2022 at 02:37:26PM +0200, Sebastian Reichel wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On Tue, Oct 11, 2022 at 08:45:54AM +0300, Tony Lindgren wrote:
>>>>>>>> * Dmitry Torokhov <dmitry.torokhov@gmail.com> [221004 21:26]:
>>>>>>>>> The LCD driver (panel-sony-acx565akm), when probing, starts with line
>>>>>>>>> driven low, and then toggles it to high and keeps it there. Also, the
>>>>>>>>> line is driven low when powering off the device, and ls released when
>>>>>>>>> powering it back on. This means that the reset line should be described
>>>>>>>>> as "active low" in DTS. This will be important when the driver is
>>>>>>>>> converted to gpiod API which respects the polarity declared in DTS.
>>>>>>>>
>>>>>>>> We should ensure these patches get merged together with the driver
>>>>>>>> change to avoid breaking LCD for booting. Probably no need to have
>>>>>>>> the driver quirk handling for inverted polartity in this case.
>>>>>>>>
>>>>>>>> It's probably easiest to have an immutable branch for the driver
>>>>>>>> changes I can base the dts changes on. Or I can ack the dts changes
>>>>>>>> if they get merged with the driver.
>>>>>>>
>>>>>>> Both drivers are already using gpiod API:
>>>>>>>
>>>>>>> drivers/gpu/drm/panel/panel-sony-acx565akm.c
>>>>>>> drivers/gpu/drm/panel/panel-dsi-cm.c
>>>>>>
>>>>>> I was looking at
>>>>>>
>>>>>> drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
>>>>>> drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c
>>>>>
>>>>> Ah OK that explains :)
>>>>>
>>>>>> which are not using gpiod. Should they be retired?
>>>>>
>>>>> Yes we should just get rid of them with omapdrm working just fine.
>>>>
>>>> Will you be submitting such patches? I'd like to get rid of
>>>> of_get_named_gpio() and friends if I can...
>>>
>>> Adding Tomi to Cc, my guess is he already has such patches and knows
>>> better which ones can go :)
>>
>> To be honest, I haven't really even had a glance towards fbdev for a long
>> time.
>>
>> There is one thing that omapdrm doesn't support, which is VRFB rotation. I
>> cannot say if the users of those above-mentioned panels require VRFB.
>>
>>>>>>> So this just breaks things.
>>>>>>
>>>>>> I missed the drivers in drivers/gpu/... and I see that they essentially
>>>>>> abuse gpiod API as gpiod_set_value() operates on logical level
>>>>>> (active/inactive) and not absolute (high/low). They should either use
>>>>>> the gpiod_*_raw() variants, or they should be adjusted to do the proper
>>>>>> thing together with the accompanying DTS change.
>>>>>>
>>>>>> What are your preferences?
>>>>>
>>>>> Seems like high/low at the connected device end is what we should use,
>>>>> right? Otherwise things will misbehave if the panel is connected to
>>>>> some other SoC possibly.
>>>>
>>>> It is exactly because of this case the driver should use active/inactive
>>>> and follow polarity described in DTS. If the driver does:
>>>>
>>>> 	gpiod_set_value_cansleep(d->reset, 1);
>>>>
>>>> then if DTS is saying that the reset line is active low, under the wraps
>>>> the line will be driven to "0", but if DTS is saying that the line is
>>>> active high, then the very same call will drive the line to "1".
>>>>
>>>> This allows accommodating different designs without having to change the
>>>> driver code.
>>
>> Isn't breaking an old dts file quite a bad thing? Why not just add a comment
>> to the .dts and to the driver about the situation. I don't quite see that
>> the fixing the dts (And, if done properly, adding a boot time fixup for old
>> dtbs) and changing the drivers is worth the hassle.
>>
>> Unless we see new users for these drivers, which would require the new users
>> to write broken dts files.
> 
> Or maybe there are devices with fixed DTSes and fixed up kernels but the
> fixes have not been contributed upstream. I don't know...
> 
> My personal opinion is that we pay too much attention to DTS
> compatibility in cases when it is not totally clear if there are devices
> that use DTSes that are not bundled with the kernel and also have a
> chance to have their kernel updated (and be lucky enough for the
> upstream kernel to work on such device without extensive work).
> 
> Anyway, my goal is to stop exposing of_get_named_gpio() and its
> derivatives, so please let me know your preference. Should I:
> 
> - mirror in omapfb drivers what gpu drivers do and use inverted
>    polarity

I would just go with the above for the time being. It should be an easy 
change, and as these omapfb and drm panel drivers are kind of copies of 
each other, I think it makes sense to use the same code in both.

That said, I personally don't mind fixing the dts files and the drivers, 
and even dropping the omapfb panel drivers. However, as I don't know if 
someone needs the omapfb drivers or has to use an old dtb, I don't want 
to step on that possible mine field. If someone else wants to go there 
(without my involvement), fine for me =).

  Tomi
Tony Lindgren Oct. 13, 2022, 11:08 a.m. UTC | #10
Hi,

* Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [221013 06:13]:
> I would just go with the above for the time being. It should be an easy
> change, and as these omapfb and drm panel drivers are kind of copies of each
> other, I think it makes sense to use the same code in both.

Maybe if a fix is needed, sure let's fix things first, then drop
the unused panel drivers.

We already have drivers/gpu/drm/panel driver for both of these two
omapfb panels:

drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c

The compatible strings used translate to these dts files:

arch/arm/boot/dts/motorola-mapphone-common.dtsi
arch/arm/boot/dts/omap3-n900.dts
arch/arm/boot/dts/omap3-n950.dts
arch/arm/boot/dts/omap4-sdp.dts

These devices work with omapdrm and there should not be any need to
stick with the omapfb driver. We can just drop the omapfb panel
drivers for panel-sony-acx565akm.c and panel-dsi-cm.c. Let's put
the limited effort where there is activity instead :)

The vrfb rotation work has been discussed on the lists, so seems
like we will eventually have that for omapdrm. Meanwhile, software
rotation is being used for postmarketos and leste with omapdrm
AFAIK.

> That said, I personally don't mind fixing the dts files and the drivers, and
> even dropping the omapfb panel drivers. However, as I don't know if someone
> needs the omapfb drivers or has to use an old dtb, I don't want to step on
> that possible mine field. If someone else wants to go there (without my
> involvement), fine for me =).

I belive the only valid use case for omap2 omapfb is the n8x0 rfbi
driver that has no omapdrm driver.

Regards,

Tony
Sebastian Reichel Oct. 13, 2022, 1:17 p.m. UTC | #11
Hi,

On Thu, Oct 13, 2022 at 02:08:54PM +0300, Tony Lindgren wrote:
> Hi,
> 
> * Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [221013 06:13]:
> > I would just go with the above for the time being. It should be an easy
> > change, and as these omapfb and drm panel drivers are kind of copies of each
> > other, I think it makes sense to use the same code in both.
> 
> Maybe if a fix is needed, sure let's fix things first, then drop
> the unused panel drivers.
> 
> We already have drivers/gpu/drm/panel driver for both of these two
> omapfb panels:
> 
> drivers/video/fbdev/omap2/omapfb/displays/panel-sony-acx565akm.c
> drivers/video/fbdev/omap2/omapfb/displays/panel-dsi-cm.c
> 
> The compatible strings used translate to these dts files:
> 
> arch/arm/boot/dts/motorola-mapphone-common.dtsi
> arch/arm/boot/dts/omap3-n900.dts
> arch/arm/boot/dts/omap3-n950.dts
> arch/arm/boot/dts/omap4-sdp.dts
> 
> These devices work with omapdrm and there should not be any need to
> stick with the omapfb driver. We can just drop the omapfb panel
> drivers for panel-sony-acx565akm.c and panel-dsi-cm.c. Let's put
> the limited effort where there is activity instead :)

FWIW

Acked-by: Sebastian Reichel <sre@kernel.org>

for removal of those two omapfb panel drivers.

> The vrfb rotation work has been discussed on the lists, so seems
> like we will eventually have that for omapdrm. Meanwhile, software
> rotation is being used for postmarketos and leste with omapdrm
> AFAIK.
> 
> > That said, I personally don't mind fixing the dts files and the drivers, and
> > even dropping the omapfb panel drivers. However, as I don't know if someone
> > needs the omapfb drivers or has to use an old dtb, I don't want to step on
> > that possible mine field. If someone else wants to go there (without my
> > involvement), fine for me =).
> 
> I belive the only valid use case for omap2 omapfb is the n8x0 rfbi
> driver that has no omapdrm driver.

Is that upstream? omapfb (and omapdrm) both have this:

	/*
	 * HACK
	 * We don't have a working driver for rfbi, so skip it here always.
	 * Otherwise dss will never get probed successfully, as it will wait
	 * for rfbi to get probed.
	 */
	if (strstr(dev_name(dev), "rfbi"))
		return 0;

I've seen a few old drivers being removed by marking them as BROKEN
(and updating Kconfig help text to explain the situation). Then the
code is dropped 1-2 cycles later assuming nobody complained.

-- Sebastian
diff mbox series

Patch

diff --git a/arch/arm/boot/dts/omap3-n900.dts b/arch/arm/boot/dts/omap3-n900.dts
index dd7971556449..c2e5bde19452 100644
--- a/arch/arm/boot/dts/omap3-n900.dts
+++ b/arch/arm/boot/dts/omap3-n900.dts
@@ -1038,7 +1038,7 @@  lcd: acx565akm@2 {
 		pinctrl-0 = <&acx565akm_pins>;
 
 		label = "lcd";
-		reset-gpios = <&gpio3 26 GPIO_ACTIVE_HIGH>; /* 90 */
+		reset-gpios = <&gpio3 26 GPIO_ACTIVE_LOW>; /* 90 */
 
 		port {
 			lcd_in: endpoint {