Hi,
On Thu, May 25, 2023 at 2:32 AM Cong Yang
<[email protected]> wrote:
>
> The Starry-ili9882 is a 10.51" WUXGA TFT panel. which fits in nicely with
> the existing panel-boe-tv101wum-nl6 driver. From the datasheet,MIPI need
> to keep the LP11 state before the lcm_reset pin is pulled high. So add
> lp11_before_reset flag.
>
> Signed-off-by: Cong Yang <[email protected]>
> Reviewed-by: Douglas Anderson <[email protected]>
> ---
> .../gpu/drm/panel/panel-boe-tv101wum-nl6.c | 371 ++++++++++++++++++
> 1 file changed, 371 insertions(+)
Applied to drm-misc-next:
8716a6473e6c drm/panel: Support for Starry-ili9882t TDDI MIPI-DSI panel
On Thu, Jun 1, 2023 at 5:55 PM Doug Anderson <[email protected]> wrote:
> On Thu, May 25, 2023 at 2:32 AM Cong Yang
> <[email protected]> wrote:
> >
> > The Starry-ili9882 is a 10.51" WUXGA TFT panel. which fits in nicely with
> > the existing panel-boe-tv101wum-nl6 driver. From the datasheet,MIPI need
> > to keep the LP11 state before the lcm_reset pin is pulled high. So add
> > lp11_before_reset flag.
> >
> > Signed-off-by: Cong Yang <[email protected]>
> > Reviewed-by: Douglas Anderson <[email protected]>
> > ---
> > .../gpu/drm/panel/panel-boe-tv101wum-nl6.c | 371 ++++++++++++++++++
> > 1 file changed, 371 insertions(+)
>
> Applied to drm-misc-next:
>
> 8716a6473e6c drm/panel: Support for Starry-ili9882t TDDI MIPI-DSI panel
Sorry for noticing too late and coming after the fact and complaining.
We must stop using the panel-boe-tv101wum-nl6.c driver as a
one-stop-shop for Chromium panels. The Starry panel in particular
hardware-wise has nothing in common with the other panels in this
driver and I'm suspicious about patch 3/4 as well.
Please check my patch breaking it out to a separate driver, and
if you could check internally if you have a datasheet for Ilitek
ILI9882t or can use your vendor leverage to get one to improve
on the driver (such as define the DCS commands...) that would
be great.
There are good reasons for grouping the panel drivers into
respective display controller such as fixing bugs in one place
and if we ever want to properly support things such as
gamma correction it will provide the proper per-display-controller
approach.
Yours,
Linus Walleij
Hi,
On Tue, Jul 4, 2023 at 12:47 AM Linus Walleij <[email protected]> wrote:
>
> On Thu, Jun 1, 2023 at 5:55 PM Doug Anderson <[email protected]> wrote:
> > On Thu, May 25, 2023 at 2:32 AM Cong Yang
> > <[email protected]> wrote:
> > >
> > > The Starry-ili9882 is a 10.51" WUXGA TFT panel. which fits in nicely with
> > > the existing panel-boe-tv101wum-nl6 driver. From the datasheet,MIPI need
> > > to keep the LP11 state before the lcm_reset pin is pulled high. So add
> > > lp11_before_reset flag.
> > >
> > > Signed-off-by: Cong Yang <[email protected]>
> > > Reviewed-by: Douglas Anderson <[email protected]>
> > > ---
> > > .../gpu/drm/panel/panel-boe-tv101wum-nl6.c | 371 ++++++++++++++++++
> > > 1 file changed, 371 insertions(+)
> >
> > Applied to drm-misc-next:
> >
> > 8716a6473e6c drm/panel: Support for Starry-ili9882t TDDI MIPI-DSI panel
>
> Sorry for noticing too late and coming after the fact and complaining.
>
> We must stop using the panel-boe-tv101wum-nl6.c driver as a
> one-stop-shop for Chromium panels. The Starry panel in particular
> hardware-wise has nothing in common with the other panels in this
> driver and I'm suspicious about patch 3/4 as well.
>
> Please check my patch breaking it out to a separate driver, and
> if you could check internally if you have a datasheet for Ilitek
> ILI9882t or can use your vendor leverage to get one to improve
> on the driver (such as define the DCS commands...) that would
> be great.
>
> There are good reasons for grouping the panel drivers into
> respective display controller such as fixing bugs in one place
> and if we ever want to properly support things such as
> gamma correction it will provide the proper per-display-controller
> approach.
I mentioned in response to your patch #3 also [1], but closing the
loop here as well. The original reason several panels all ended up in
one driver was in response to Sam's feedback [2]. That was even
documented when the first of the "Chromium" panels landed in commit
93ee1a2c0f08 ("drm/panel: support for BOE and INX video mode panel").
In my mind it's really a tradeoff and I'm happy to go with whatever
consensus that others agree with. What I'm never super happy with,
though, is changing the bikeshed color too often, so I'd be really
curious to hear Sam's thoughts on the issue and whether he'd like to
see this driver broken out into a dozen drivers.
[1] https://lore.kernel.org/r/CAD=FV=Xkr3Qpd8m_6Xta_2jL_ezbxsmMyarbKXTXL+UJLG9xNw@mail.gmail.com
[2] https://lore.kernel.org/all/[email protected]/
On Thu, Jul 6, 2023 at 11:25 PM Doug Anderson <[email protected]> wrote:
> In my mind it's really a tradeoff and I'm happy to go with whatever
> consensus that others agree with. What I'm never super happy with,
> though, is changing the bikeshed color too often, so I'd be really
> curious to hear Sam's thoughts on the issue and whether he'd like to
> see this driver broken out into a dozen drivers.
This is not question about a dozen drivers, to be clear.
I just want to break out the drivers that have an identifiable
display controller that differs from the others, especially this one.
The rest of the drivers inside of this boe driver I can't really tell,
they seem related? We don't know?
So the Ilitek ILI9882t is an obvious break-out.
For the rest, I guess I would be happier if the Chromium people
could use their leverage with vendors to muscle out the details
about display controller vendors and provide #defines for all
magic commands, we all dislike these opaque firmware-looking
jam tables.
Cong already stated that he indeed has the datasheet for the
ILI9882t controller at hand, had I come in earlier I would have
asked for all of those sequences to be provided with proper
#defines instead of 0xff 0x98 0x82 0x01... but I'm late to the
show.
Yours,
Linus Walleij
Hi,
On Thu, Jul 6, 2023 at 2:36 PM Linus Walleij <[email protected]> wrote:
>
> On Thu, Jul 6, 2023 at 11:25 PM Doug Anderson <[email protected]> wrote:
>
> > In my mind it's really a tradeoff and I'm happy to go with whatever
> > consensus that others agree with. What I'm never super happy with,
> > though, is changing the bikeshed color too often, so I'd be really
> > curious to hear Sam's thoughts on the issue and whether he'd like to
> > see this driver broken out into a dozen drivers.
>
> This is not question about a dozen drivers, to be clear.
>
> I just want to break out the drivers that have an identifiable
> display controller that differs from the others, especially this one.
>
> The rest of the drivers inside of this boe driver I can't really tell,
> they seem related? We don't know?
>
> So the Ilitek ILI9882t is an obvious break-out.
I guess. To me it feels like the concept of breaking the driver into
multiple sub-drivers and the idea of supporting ILI9882t more cleanly
are orthogonal. You could still do your patch #4 and break out the
page switching function without breaking up the driver.
It feels to me fairly likely that many of the panels here are just as
different from each other as the ILI9882t is from them. I guess it's
not a dozen, but it feels like using the same "how different are they
from each other" metric we'd end up with at least 5-6 new drivers. It
seems clear to me that the panel that Sam first commented on is as
different from the others in the BOE driver as the ILI9882t is.
Certainly it has a pretty darn big unique command sequence for init...
> For the rest, I guess I would be happier if the Chromium people
> could use their leverage with vendors to muscle out the details
> about display controller vendors and provide #defines for all
> magic commands, we all dislike these opaque firmware-looking
> jam tables.
Yeah, I've grumbled about this with each new blob added. For instance:
https://lore.kernel.org/r/CAD=FV=Uo-7rFWGiJG0oJ0ydosA4DxhFqiWGrab1zoZyxyPsOBg@mail.gmail.com/
Where I said:
> I'm not really an expert on
> MIPI panels but the convention of a big stream of binary commands
> seems to match what other panels in this driver do, even if their
> table of binary data isn't quite as long as yours (are all of yours
> actually needed?)
The problem is that it's hard for me to make a strong argument here
when there is prior art of panels being supported with blob-sequences.
In this case, I think you as an upstream developer have more leverage.
I can help put pressure to make sure that upstream concerns are
addressed, but I think it's on upstream to put their foot down and say
that these blob sequences are not OK for new panels. In each case I
landed a patch with a new blob sequence I tried to give the community
time to respond and I tried to telegraph what I was going to do to
make sure nobody was surprised...
-Doug
On Thu, Jul 6, 2023 at 11:58 PM Doug Anderson <[email protected]> wrote:
> > So the Ilitek ILI9882t is an obvious break-out.
>
> I guess. To me it feels like the concept of breaking the driver into
> multiple sub-drivers and the idea of supporting ILI9882t more cleanly
> are orthogonal. You could still do your patch #4 and break out the
> page switching function without breaking up the driver.
Yeah that's true. But with Ilitek in particular we have these nice
precedents:
drivers/gpu/drm/panel/panel-ilitek-ili9322.c
drivers/gpu/drm/panel/panel-ilitek-ili9341.c
drivers/gpu/drm/panel/panel-ilitek-ili9881c.c
So it looks disorganized to me if this one Ilitek panel controller
now goes inside another driver with other completely unrelated
drivers.
> It feels to me fairly likely that many of the panels here are just as
> different from each other as the ILI9882t is from them. I guess it's
> not a dozen, but it feels like using the same "how different are they
> from each other" metric we'd end up with at least 5-6 new drivers. It
> seems clear to me that the panel that Sam first commented on is as
> different from the others in the BOE driver as the ILI9882t is.
> Certainly it has a pretty darn big unique command sequence for init...
It doesn't really matter until we can say certainly what display controller
each of them is. It seems we can't, but for this one we can.
> The problem is that it's hard for me to make a strong argument here
> when there is prior art of panels being supported with blob-sequences.
> In this case, I think you as an upstream developer have more leverage.
> I can help put pressure to make sure that upstream concerns are
> addressed, but I think it's on upstream to put their foot down and say
> that these blob sequences are not OK for new panels. In each case I
> landed a patch with a new blob sequence I tried to give the community
> time to respond and I tried to telegraph what I was going to do to
> make sure nobody was surprised...
I would say it is not fair to block driver coming from hobbyists or minor
vendors just trying to make something work. In general I think a working
something is better than nothing so I wouldn't block anything.
But with big companies who actually talk to Ilitek, Novotek and the other
companies ending with -tek that make these display controllers I would
certainly like to send the message that datasheets and proper
defines would be appreciated, and say it is also for their best, because
I mentioned proper gamma correction is possible if the driver author
just invest time and works with the DRM community and that should
be in their best interest. Feel free to pass this along the supply
chain if you can.
Yours,
Linus Walleij
Hi all,
On Thu, Jul 06, 2023 at 02:25:16PM -0700, Doug Anderson wrote:
> Hi,
>
> On Tue, Jul 4, 2023 at 12:47 AM Linus Walleij <[email protected]> wrote:
> >
> > On Thu, Jun 1, 2023 at 5:55 PM Doug Anderson <[email protected]> wrote:
> > > On Thu, May 25, 2023 at 2:32 AM Cong Yang
> > > <[email protected]> wrote:
> > > >
> > > > The Starry-ili9882 is a 10.51" WUXGA TFT panel. which fits in nicely with
> > > > the existing panel-boe-tv101wum-nl6 driver. From the datasheet,MIPI need
> > > > to keep the LP11 state before the lcm_reset pin is pulled high. So add
> > > > lp11_before_reset flag.
> > > >
> > > > Signed-off-by: Cong Yang <[email protected]>
> > > > Reviewed-by: Douglas Anderson <[email protected]>
> > > > ---
> > > > .../gpu/drm/panel/panel-boe-tv101wum-nl6.c | 371 ++++++++++++++++++
> > > > 1 file changed, 371 insertions(+)
> > >
> > > Applied to drm-misc-next:
> > >
> > > 8716a6473e6c drm/panel: Support for Starry-ili9882t TDDI MIPI-DSI panel
> >
> > Sorry for noticing too late and coming after the fact and complaining.
> >
> > We must stop using the panel-boe-tv101wum-nl6.c driver as a
> > one-stop-shop for Chromium panels. The Starry panel in particular
> > hardware-wise has nothing in common with the other panels in this
> > driver and I'm suspicious about patch 3/4 as well.
> >
> > Please check my patch breaking it out to a separate driver, and
> > if you could check internally if you have a datasheet for Ilitek
> > ILI9882t or can use your vendor leverage to get one to improve
> > on the driver (such as define the DCS commands...) that would
> > be great.
> >
> > There are good reasons for grouping the panel drivers into
> > respective display controller such as fixing bugs in one place
> > and if we ever want to properly support things such as
> > gamma correction it will provide the proper per-display-controller
> > approach.
>
> I mentioned in response to your patch #3 also [1], but closing the
> loop here as well. The original reason several panels all ended up in
> one driver was in response to Sam's feedback [2]. That was even
> documented when the first of the "Chromium" panels landed in commit
> 93ee1a2c0f08 ("drm/panel: support for BOE and INX video mode panel").
If we should go with any sort of guideline then one-driver-per-controller.
So we do not mix display controllers in one driver, but we can have
different panels in one driver.
Then there may be two almost identical controllers that can share the
same driver, or there can be controllers used in two different ways so
they warrant independent drivers. In other words this should be used
with common sense.
And if someone can help naming all the magic constant that would be
super.
Sam