Hi Tony, Thierry, Laurent,
After the recent change of moving from omapdrm specific panel-dpi driver to the DRM simple panel, AM4 EVM/ePOS's panel is not working quite right. This series has fixes for it, but I'm not sure if these are the right ways to fix the issues, so comments welcome.
1) Panel driver is not probed. With omapdrm's panel-dpi, the match happened with "panel-dpi" compatible string. Now with panel-simple, the match should happen with the panel model compatible string, which is "osddisplays,osd057T0559-34ts" in the DT file. However, no such compatible exists in panel-simple.
Interestingly, the actual panel at least on my EVMs and ePOSes is not osd057T0559-34ts, but osd070t1718-19ts. Also, I was unable to find any information about osd057T0559-34ts. I don't know the history with this, so it is possible that the early versions of the boards did have osd057T0559-34ts, but was later changed to osd070t1718-19ts.
As osd070t1718-19ts is supported by panel-simple, changing the compatible string to osd070t1718-19ts in the DT file solves this one.
2) Timings in DT file cause a kernel warning. Omapdrm's panel-dpi used video timings from the DT file, so they are present in all the DT files. panel-simple uses timings from a table in the panel-simple driver, but gives a kernel warning if the DT file contains timings.
This can be solved by removing the timings from the DT file.
3) Sync drive edge is not right. This one might have been present also with panel-dpi, I didn't verify. The problem is that the panel-simple data for osddisplays_osd070t1718_19ts defines bus_flags for DE polarity and pixdata edge, but not for sync edge.
The datasheet for the panel does not give any hint on what the edge should be. Omapdrm defaults to driving syncs on falling edge, which caused the image to be shifted one pixel to the right.
Adding DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE bus_flag solves the problem. AM5 EVM also has the same panel with the same behavior.
====
The reason I'm not sure if the 1) and 2) fixes are correct is that they're breaking DT compatibility. Should we instead make changes to panel-simple to keep the same DT files working?
This would mean adding a new entry for the osd057T0559-34ts panel, but as we don't have datasheet for it, I think we could just append the compatible string to osd070t1718-19t's data.
It would also mean doing some change to the panel-simple code that gives the warning about timings in DT data. This might make sense, as I think we have other DT files with video timings too.
For 3), I think the patch is fine, but I'm not sure if the display controller driver should be able to deduce the sync drive edge from the pixdata drive edge. Are they usually the same? I have no idea...
Tomi
Tomi Valkeinen (3): ARM: dts: am437x-gp/epos-evm: fix panel compatible ARM: dts: am437x-gp/epos-evm: drop unused panel timings drm/panel: simple: fix osd070t1718_19ts sync drive edge
arch/arm/boot/dts/am437x-gp-evm.dts | 18 +----------------- arch/arm/boot/dts/am43x-epos-evm.dts | 18 +----------------- drivers/gpu/drm/panel/panel-simple.c | 3 ++- 3 files changed, 4 insertions(+), 35 deletions(-)
-- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
The LCD panel on AM4 GP EVMs and ePOS boards seems to be osd070t1718-19ts. The current dts files say osd057T0559-34ts. Possibly the panel has changed since the early EVMs, or there has been a mistake with the panel type.
Update the DT files accordingly.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com --- arch/arm/boot/dts/am437x-gp-evm.dts | 2 +- arch/arm/boot/dts/am43x-epos-evm.dts | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/arm/boot/dts/am437x-gp-evm.dts b/arch/arm/boot/dts/am437x-gp-evm.dts index cae4500194fe..811c8cae315b 100644 --- a/arch/arm/boot/dts/am437x-gp-evm.dts +++ b/arch/arm/boot/dts/am437x-gp-evm.dts @@ -86,7 +86,7 @@ };
lcd0: display { - compatible = "osddisplays,osd057T0559-34ts", "panel-dpi"; + compatible = "osddisplays,osd070t1718-19ts", "panel-dpi"; label = "lcd";
backlight = <&lcd_bl>; diff --git a/arch/arm/boot/dts/am43x-epos-evm.dts b/arch/arm/boot/dts/am43x-epos-evm.dts index 95314121d111..078cb473fa7d 100644 --- a/arch/arm/boot/dts/am43x-epos-evm.dts +++ b/arch/arm/boot/dts/am43x-epos-evm.dts @@ -42,7 +42,7 @@ };
lcd0: display { - compatible = "osddisplays,osd057T0559-34ts", "panel-dpi"; + compatible = "osddisplays,osd070t1718-19ts", "panel-dpi"; label = "lcd";
backlight = <&lcd_bl>;
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:48AM +0200, Tomi Valkeinen wrote:
The LCD panel on AM4 GP EVMs and ePOS boards seems to be osd070t1718-19ts. The current dts files say osd057T0559-34ts. Possibly the panel has changed since the early EVMs, or there has been a mistake with the panel type.
Update the DT files accordingly.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
Acked-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
arch/arm/boot/dts/am437x-gp-evm.dts | 2 +- arch/arm/boot/dts/am43x-epos-evm.dts | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/arm/boot/dts/am437x-gp-evm.dts b/arch/arm/boot/dts/am437x-gp-evm.dts index cae4500194fe..811c8cae315b 100644 --- a/arch/arm/boot/dts/am437x-gp-evm.dts +++ b/arch/arm/boot/dts/am437x-gp-evm.dts @@ -86,7 +86,7 @@ };
lcd0: display {
compatible = "osddisplays,osd057T0559-34ts", "panel-dpi";
compatible = "osddisplays,osd070t1718-19ts", "panel-dpi";label = "lcd";
backlight = <&lcd_bl>;
diff --git a/arch/arm/boot/dts/am43x-epos-evm.dts b/arch/arm/boot/dts/am43x-epos-evm.dts index 95314121d111..078cb473fa7d 100644 --- a/arch/arm/boot/dts/am43x-epos-evm.dts +++ b/arch/arm/boot/dts/am43x-epos-evm.dts @@ -42,7 +42,7 @@ };
lcd0: display {
compatible = "osddisplays,osd057T0559-34ts", "panel-dpi";
compatible = "osddisplays,osd070t1718-19ts", "panel-dpi";label = "lcd";
backlight = <&lcd_bl>;
* Laurent Pinchart laurent.pinchart@ideasonboard.com [191202 13:02]:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:48AM +0200, Tomi Valkeinen wrote:
The LCD panel on AM4 GP EVMs and ePOS boards seems to be osd070t1718-19ts. The current dts files say osd057T0559-34ts. Possibly the panel has changed since the early EVMs, or there has been a mistake with the panel type.
Update the DT files accordingly.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
Acked-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
Applying this one into fixes thanks.
Regards,
Tony
panel-simple now handled panel osd070t1718-19ts, and we no longer need the panel timings in the DT file. So remove them.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com --- arch/arm/boot/dts/am437x-gp-evm.dts | 16 ---------------- arch/arm/boot/dts/am43x-epos-evm.dts | 16 ---------------- 2 files changed, 32 deletions(-)
diff --git a/arch/arm/boot/dts/am437x-gp-evm.dts b/arch/arm/boot/dts/am437x-gp-evm.dts index 811c8cae315b..040403049cea 100644 --- a/arch/arm/boot/dts/am437x-gp-evm.dts +++ b/arch/arm/boot/dts/am437x-gp-evm.dts @@ -91,22 +91,6 @@
backlight = <&lcd_bl>;
- panel-timing { - clock-frequency = <33000000>; - hactive = <800>; - vactive = <480>; - hfront-porch = <210>; - hback-porch = <16>; - hsync-len = <30>; - vback-porch = <10>; - vfront-porch = <22>; - vsync-len = <13>; - hsync-active = <0>; - vsync-active = <0>; - de-active = <1>; - pixelclk-active = <1>; - }; - port { lcd_in: endpoint { remote-endpoint = <&dpi_out>; diff --git a/arch/arm/boot/dts/am43x-epos-evm.dts b/arch/arm/boot/dts/am43x-epos-evm.dts index 078cb473fa7d..330fce060e8b 100644 --- a/arch/arm/boot/dts/am43x-epos-evm.dts +++ b/arch/arm/boot/dts/am43x-epos-evm.dts @@ -47,22 +47,6 @@
backlight = <&lcd_bl>;
- panel-timing { - clock-frequency = <33000000>; - hactive = <800>; - vactive = <480>; - hfront-porch = <210>; - hback-porch = <16>; - hsync-len = <30>; - vback-porch = <10>; - vfront-porch = <22>; - vsync-len = <13>; - hsync-active = <0>; - vsync-active = <0>; - de-active = <1>; - pixelclk-active = <1>; - }; - port { lcd_in: endpoint { remote-endpoint = <&dpi_out>;
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:49AM +0200, Tomi Valkeinen wrote:
panel-simple now handled panel osd070t1718-19ts, and we no longer need the panel timings in the DT file. So remove them.
Should you in that case drop the panel-dpi compatible string too, as the panel-dpi bindings require panel timings in DT ?
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
arch/arm/boot/dts/am437x-gp-evm.dts | 16 ---------------- arch/arm/boot/dts/am43x-epos-evm.dts | 16 ---------------- 2 files changed, 32 deletions(-)
diff --git a/arch/arm/boot/dts/am437x-gp-evm.dts b/arch/arm/boot/dts/am437x-gp-evm.dts index 811c8cae315b..040403049cea 100644 --- a/arch/arm/boot/dts/am437x-gp-evm.dts +++ b/arch/arm/boot/dts/am437x-gp-evm.dts @@ -91,22 +91,6 @@
backlight = <&lcd_bl>;
panel-timing {clock-frequency = <33000000>;hactive = <800>;vactive = <480>;hfront-porch = <210>;hback-porch = <16>;hsync-len = <30>;vback-porch = <10>;vfront-porch = <22>;vsync-len = <13>;hsync-active = <0>;vsync-active = <0>;de-active = <1>;pixelclk-active = <1>;};- port { lcd_in: endpoint { remote-endpoint = <&dpi_out>;
diff --git a/arch/arm/boot/dts/am43x-epos-evm.dts b/arch/arm/boot/dts/am43x-epos-evm.dts index 078cb473fa7d..330fce060e8b 100644 --- a/arch/arm/boot/dts/am43x-epos-evm.dts +++ b/arch/arm/boot/dts/am43x-epos-evm.dts @@ -47,22 +47,6 @@
backlight = <&lcd_bl>;
panel-timing {clock-frequency = <33000000>;hactive = <800>;vactive = <480>;hfront-porch = <210>;hback-porch = <16>;hsync-len = <30>;vback-porch = <10>;vfront-porch = <22>;vsync-len = <13>;hsync-active = <0>;vsync-active = <0>;de-active = <1>;pixelclk-active = <1>;};- port { lcd_in: endpoint { remote-endpoint = <&dpi_out>;
* Laurent Pinchart laurent.pinchart@ideasonboard.com [191202 13:05]:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:49AM +0200, Tomi Valkeinen wrote:
panel-simple now handled panel osd070t1718-19ts, and we no longer need the panel timings in the DT file. So remove them.
Should you in that case drop the panel-dpi compatible string too, as the panel-dpi bindings require panel timings in DT ?
Yeah sounds like if panel-dpi is no longer usable for this device it should be dropped from the compatible list.
Regards,
Tony
On 11/12/2019 18:53, Tony Lindgren wrote:
- Laurent Pinchart laurent.pinchart@ideasonboard.com [191202 13:05]:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:49AM +0200, Tomi Valkeinen wrote:
panel-simple now handled panel osd070t1718-19ts, and we no longer need the panel timings in the DT file. So remove them.
Should you in that case drop the panel-dpi compatible string too, as the panel-dpi bindings require panel timings in DT ?
Yeah sounds like if panel-dpi is no longer usable for this device it should be dropped from the compatible list.
Ok, I agree.
Looking at the dts files, panel-dpi is used in a bunch of boards. But we even have 3 dts files with panel-dpi, without the detailed panel model in compatible...
Fixing those will break the compatibility with old dtbs and new kernel, unless we add timings-from-dt to a panel driver that handles panel-dpi.
Tomi
Hi Tomi,
On Thu, Dec 12, 2019 at 11:37:51AM +0200, Tomi Valkeinen wrote:
On 11/12/2019 18:53, Tony Lindgren wrote:
- Laurent Pinchart laurent.pinchart@ideasonboard.com [191202 13:05]:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:49AM +0200, Tomi Valkeinen wrote:
panel-simple now handled panel osd070t1718-19ts, and we no longer need the panel timings in the DT file. So remove them.
Should you in that case drop the panel-dpi compatible string too, as the panel-dpi bindings require panel timings in DT ?
Yeah sounds like if panel-dpi is no longer usable for this device it should be dropped from the compatible list.
Ok, I agree.
Looking at the dts files, panel-dpi is used in a bunch of boards. But we even have 3 dts files with panel-dpi, without the detailed panel model in compatible...
Fixing those will break the compatibility with old dtbs and new kernel, unless we add timings-from-dt to a panel driver that handles panel-dpi.
I know, and I don't have a perfect answer for this :-( I don't see a third option, it's either breaking DT backward compatibility or adding timings parsing to a panel driver (either a new panel-dpi driver or to panel-simple). What's your preferred option ?
On 12/12/2019 22:35, Laurent Pinchart wrote:
Hi Tomi,
On Thu, Dec 12, 2019 at 11:37:51AM +0200, Tomi Valkeinen wrote:
On 11/12/2019 18:53, Tony Lindgren wrote:
- Laurent Pinchart laurent.pinchart@ideasonboard.com [191202 13:05]:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:49AM +0200, Tomi Valkeinen wrote:
panel-simple now handled panel osd070t1718-19ts, and we no longer need the panel timings in the DT file. So remove them.
Should you in that case drop the panel-dpi compatible string too, as the panel-dpi bindings require panel timings in DT ?
Yeah sounds like if panel-dpi is no longer usable for this device it should be dropped from the compatible list.
Ok, I agree.
Looking at the dts files, panel-dpi is used in a bunch of boards. But we even have 3 dts files with panel-dpi, without the detailed panel model in compatible...
Fixing those will break the compatibility with old dtbs and new kernel, unless we add timings-from-dt to a panel driver that handles panel-dpi.
I know, and I don't have a perfect answer for this :-( I don't see a third option, it's either breaking DT backward compatibility or adding timings parsing to a panel driver (either a new panel-dpi driver or to panel-simple). What's your preferred option ?
Hmm, I just realized that changing these will break omapfb. It relies on panel-dpi and timings from DT...
Tomi
On 13/01/2020 14:01, Tomi Valkeinen wrote:
On 12/12/2019 22:35, Laurent Pinchart wrote:
Hi Tomi,
On Thu, Dec 12, 2019 at 11:37:51AM +0200, Tomi Valkeinen wrote:
On 11/12/2019 18:53, Tony Lindgren wrote:
- Laurent Pinchart laurent.pinchart@ideasonboard.com [191202 13:05]:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:49AM +0200, Tomi Valkeinen wrote:
panel-simple now handled panel osd070t1718-19ts, and we no longer need the panel timings in the DT file. So remove them.
Should you in that case drop the panel-dpi compatible string too, as the panel-dpi bindings require panel timings in DT ?
Yeah sounds like if panel-dpi is no longer usable for this device it should be dropped from the compatible list.
Ok, I agree.
Looking at the dts files, panel-dpi is used in a bunch of boards. But we even have 3 dts files with panel-dpi, without the detailed panel model in compatible...
Fixing those will break the compatibility with old dtbs and new kernel, unless we add timings-from-dt to a panel driver that handles panel-dpi.
I know, and I don't have a perfect answer for this :-( I don't see a third option, it's either breaking DT backward compatibility or adding timings parsing to a panel driver (either a new panel-dpi driver or to panel-simple). What's your preferred option ?
Hmm, I just realized that changing these will break omapfb. It relies on panel-dpi and timings from DT...
If no one objects, I think we should just drop the timings from the .dts, and say that these boards are no longer supported with omapfb. I don't think there's much point in trying to keep omapfb working fine for boards that are fully supported by omapdrm.
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
Tomi
Hi Tomi,
On Tue, Feb 11, 2020 at 12:01:31PM +0200, Tomi Valkeinen wrote:
On 13/01/2020 14:01, Tomi Valkeinen wrote:
On 12/12/2019 22:35, Laurent Pinchart wrote:
On Thu, Dec 12, 2019 at 11:37:51AM +0200, Tomi Valkeinen wrote:
On 11/12/2019 18:53, Tony Lindgren wrote:
- Laurent Pinchart laurent.pinchart@ideasonboard.com [191202 13:05]:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:49AM +0200, Tomi Valkeinen wrote: > panel-simple now handled panel osd070t1718-19ts, and we no longer need > the panel timings in the DT file. So remove them.
Should you in that case drop the panel-dpi compatible string too, as the panel-dpi bindings require panel timings in DT ?
Yeah sounds like if panel-dpi is no longer usable for this device it should be dropped from the compatible list.
Ok, I agree.
Looking at the dts files, panel-dpi is used in a bunch of boards. But we even have 3 dts files with panel-dpi, without the detailed panel model in compatible...
Fixing those will break the compatibility with old dtbs and new kernel, unless we add timings-from-dt to a panel driver that handles panel-dpi.
I know, and I don't have a perfect answer for this :-( I don't see a third option, it's either breaking DT backward compatibility or adding timings parsing to a panel driver (either a new panel-dpi driver or to panel-simple). What's your preferred option ?
Hmm, I just realized that changing these will break omapfb. It relies on panel-dpi and timings from DT...
If no one objects, I think we should just drop the timings from the .dts, and say that these boards are no longer supported with omapfb. I don't think there's much point in trying to keep omapfb working fine for boards that are fully supported by omapdrm.
No objection from me.
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
On 11/02/2020 13:07, Laurent Pinchart wrote:
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
Not before we add DSI support to omapdrm...
Tomi
On Tue, Feb 11, 2020 at 01:08:12PM +0200, Tomi Valkeinen wrote:
On 11/02/2020 13:07, Laurent Pinchart wrote:
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
Not before we add DSI support to omapdrm...
Details, details ;-)
Seriously speaking, Sebastian's patches are on my todo list.
Hi,
On Tue, Feb 11, 2020 at 01:10:07PM +0200, Laurent Pinchart wrote:
On Tue, Feb 11, 2020 at 01:08:12PM +0200, Tomi Valkeinen wrote:
On 11/02/2020 13:07, Laurent Pinchart wrote:
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
Not before we add DSI support to omapdrm...
Details, details ;-)
Seriously speaking, Sebastian's patches are on my todo list.
The patches need to be rebased. My hack for supporting panel un/rebind stole connector and drm_dev pointer from drm_panel, which is no longer possible. I hoped this could be cleaned up once your and my omapdrm patches landed...
I do have a compile-tested-only WIP branch here, but I doubt I got this working on the first try:
https://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-n900.git/log/?h=om...
-- Sebastian
Hi,
On 11/02/2020 12:08, Tomi Valkeinen wrote:
On 11/02/2020 13:07, Laurent Pinchart wrote:
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
Not before we add DSI support to omapdrm...
This is probably known, but for devices that would like use PowerVR SGX, there sometimes is only userspace available that works with omap(l)fb, and not with DRM. The Nokia N900 is such an example.
There might be a newer release of (closed) userspace coming for the aging device(s), but as it stands, I don't think it's possible to do 3D with PowerVR SGX on omapdrm currently.
But I might be wrong...
Cheers, Merlijn
* Merlijn Wajer merlijn@wizzup.org [200211 12:54]:
Hi,
On 11/02/2020 12:08, Tomi Valkeinen wrote:
On 11/02/2020 13:07, Laurent Pinchart wrote:
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
Not before we add DSI support to omapdrm...
Hmm do your mean RFBI? The DSI support we already have :)
This is probably known, but for devices that would like use PowerVR SGX, there sometimes is only userspace available that works with omap(l)fb, and not with DRM. The Nokia N900 is such an example.
There might be a newer release of (closed) userspace coming for the aging device(s), but as it stands, I don't think it's possible to do 3D with PowerVR SGX on omapdrm currently.
But I might be wrong...
Yes SGX is a bottleneck currently for omap3 users. And I think RFBI is blocking n8x0 from moving to omapdrm and finally leaving drivers/video/fbdev/omap2 behind.
With those blockers fixed, I'd also be happy to just get rid of drivers/video/fbdev/omap2 as there should be no reason to keep using it.
Regards,
Tony
On 11/02/2020 18:05, Tony Lindgren wrote:
- Merlijn Wajer merlijn@wizzup.org [200211 12:54]:
Hi,
On 11/02/2020 12:08, Tomi Valkeinen wrote:
On 11/02/2020 13:07, Laurent Pinchart wrote:
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
Not before we add DSI support to omapdrm...
Hmm do your mean RFBI? The DSI support we already have :)
Oh, I didn't even remember RFBI... But it is not supported in omapfb either, so it's not blocking removal of omapfb.
We are still missing DSI command mode support, and moving it to the common DRM model.
This is probably known, but for devices that would like use PowerVR SGX, there sometimes is only userspace available that works with omap(l)fb, and not with DRM. The Nokia N900 is such an example.
There might be a newer release of (closed) userspace coming for the aging device(s), but as it stands, I don't think it's possible to do 3D with PowerVR SGX on omapdrm currently.
But I might be wrong...
Yes SGX is a bottleneck currently for omap3 users. And I think RFBI is blocking n8x0 from moving to omapdrm and finally leaving drivers/video/fbdev/omap2 behind.
With those blockers fixed, I'd also be happy to just get rid of drivers/video/fbdev/omap2 as there should be no reason to keep using it.
I don't know what to do about SGX, but if someone sends patches to omapdrm to allow using SGX, I'll review =).
Tomi
* Tomi Valkeinen tomi.valkeinen@ti.com [200211 16:14]:
On 11/02/2020 18:05, Tony Lindgren wrote:
- Merlijn Wajer merlijn@wizzup.org [200211 12:54]:
Hi,
On 11/02/2020 12:08, Tomi Valkeinen wrote:
On 11/02/2020 13:07, Laurent Pinchart wrote:
Hopefully soon (in five years? =) we can say that omapdrm supports all the boards, and we can deprecate omapfb.
I'd love to send a patch to remove omapfb, but I'll let you do the honours :-)
Not before we add DSI support to omapdrm...
Hmm do your mean RFBI? The DSI support we already have :)
Oh, I didn't even remember RFBI... But it is not supported in omapfb either, so it's not blocking removal of omapfb.
Oh OK.
We are still missing DSI command mode support, and moving it to the common DRM model.
Nope, DSI command mode support has been working just fine for a while now :) And Sebastian has a WIP git tree of the common DRM model changes for it. I don't think we have devices with DSI command mode working for omapfb but not for omapdrm?
What got missed for v5.6-rc1 is the LCD backlight patch though, I think the only issue there is default-brightness vs more common default-brightness-value usage if you have any input to that.
This is probably known, but for devices that would like use PowerVR SGX, there sometimes is only userspace available that works with omap(l)fb, and not with DRM. The Nokia N900 is such an example.
There might be a newer release of (closed) userspace coming for the aging device(s), but as it stands, I don't think it's possible to do 3D with PowerVR SGX on omapdrm currently.
But I might be wrong...
Yes SGX is a bottleneck currently for omap3 users. And I think RFBI is blocking n8x0 from moving to omapdrm and finally leaving drivers/video/fbdev/omap2 behind.
With those blockers fixed, I'd also be happy to just get rid of drivers/video/fbdev/omap2 as there should be no reason to keep using it.
I don't know what to do about SGX, but if someone sends patches to omapdrm to allow using SGX, I'll review =).
OK so then the only blocker seems to be the lack of SGX blobs for omap3 devices which seems to be in process of getting fixed.
Regards,
Tony
On 11/02/2020 18:27, Tony Lindgren wrote:
We are still missing DSI command mode support, and moving it to the common DRM model.
Nope, DSI command mode support has been working just fine for a while now :) And Sebastian has a WIP git tree of the common DRM
Indeed... It had been going on for so long that now my mind is stuck at dsi-command-mode-not-yet-in =).
model changes for it. I don't think we have devices with DSI command mode working for omapfb but not for omapdrm?
Yes, I think that is true.
What got missed for v5.6-rc1 is the LCD backlight patch though, I think the only issue there is default-brightness vs more common default-brightness-value usage if you have any input to that.
At least for some boards a power supply is needed, and I think there was no conclusion on who should enable that. It didn't seem to fit in anywhere...
But need to check on the latest status. I wasn't following that work closely, as JJ was working on it.
Tomi
Hi,
On Tue, Feb 11, 2020 at 07:22:14PM +0200, Tomi Valkeinen wrote:
On 11/02/2020 18:27, Tony Lindgren wrote:
We are still missing DSI command mode support, and moving it to the common DRM model.
Nope, DSI command mode support has been working just fine for a while now :) And Sebastian has a WIP git tree of the common DRM
Indeed... It had been going on for so long that now my mind is stuck at dsi-command-mode-not-yet-in =).
Welcome in the future :)
model changes for it. I don't think we have devices with DSI command mode working for omapfb but not for omapdrm?
Yes, I think that is true.
Note, that OMAP3 quirk is missing (IDK if its supported in omapfb, haven't used it for ages). I planned to have a look at OMAP3 once the patchset moving omapdrm DSI to common DRM is merged, which needs a non-trivial rebase.
What got missed for v5.6-rc1 is the LCD backlight patch though, I think the only issue there is default-brightness vs more common default-brightness-value usage if you have any input to that.
At least for some boards a power supply is needed, and I think there was no conclusion on who should enable that. It didn't seem to fit in anywhere...
But need to check on the latest status. I wasn't following that work closely, as JJ was working on it.
FWIW omapdrm's DSI driver is ready for that and omapfb is not :P
-- Sebastian
The panel datasheet says that the panel samples at falling edge, but does not say anything about h/v sync signals. Testing shows that if the sync signals are driven on falling edge, the picture on the panel will be slightly shifted right.
Setting sync drive edge to the same as data drive edge fixes this issue.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com --- drivers/gpu/drm/panel/panel-simple.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 5d487686d25c..0784536ae6af 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -2397,7 +2397,8 @@ static const struct panel_desc osddisplays_osd070t1718_19ts = { .height = 91, }, .bus_format = MEDIA_BUS_FMT_RGB888_1X24, - .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE, + .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE | + DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE, .connector_type = DRM_MODE_CONNECTOR_DPI, };
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:50AM +0200, Tomi Valkeinen wrote:
The panel datasheet says that the panel samples at falling edge, but does not say anything about h/v sync signals. Testing shows that if the sync signals are driven on falling edge, the picture on the panel will be slightly shifted right.
Setting sync drive edge to the same as data drive edge fixes this issue.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
I don't have access to the documentation, but this makes sense, so
Acked-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
drivers/gpu/drm/panel/panel-simple.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 5d487686d25c..0784536ae6af 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -2397,7 +2397,8 @@ static const struct panel_desc osddisplays_osd070t1718_19ts = { .height = 91, }, .bus_format = MEDIA_BUS_FMT_RGB888_1X24,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE |
DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE,};
Hi Thierry,
On 02/12/2019 15:07, Laurent Pinchart wrote:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:50AM +0200, Tomi Valkeinen wrote:
The panel datasheet says that the panel samples at falling edge, but does not say anything about h/v sync signals. Testing shows that if the sync signals are driven on falling edge, the picture on the panel will be slightly shifted right.
Setting sync drive edge to the same as data drive edge fixes this issue.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
I don't have access to the documentation, but this makes sense, so
Acked-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
drivers/gpu/drm/panel/panel-simple.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 5d487686d25c..0784536ae6af 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -2397,7 +2397,8 @@ static const struct panel_desc osddisplays_osd070t1718_19ts = { .height = 91, }, .bus_format = MEDIA_BUS_FMT_RGB888_1X24,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE |
DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE,
Can you pick this one up?
Tomi
Hi Thierry,
On 02/12/2019 15:07, Laurent Pinchart wrote:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:50AM +0200, Tomi Valkeinen wrote:
The panel datasheet says that the panel samples at falling edge, but does not say anything about h/v sync signals. Testing shows that if the sync signals are driven on falling edge, the picture on the panel will be slightly shifted right.
Setting sync drive edge to the same as data drive edge fixes this issue.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
I don't have access to the documentation, but this makes sense, so
Acked-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
drivers/gpu/drm/panel/panel-simple.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 5d487686d25c..0784536ae6af 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -2397,7 +2397,8 @@ static const struct panel_desc osddisplays_osd070t1718_19ts = { .height = 91, }, .bus_format = MEDIA_BUS_FMT_RGB888_1X24,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE |
DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE,
Can this be merged?
Tomi
Hi Tomi.
On Mon, Feb 10, 2020 at 10:15:33AM +0200, Tomi Valkeinen wrote:
Hi Thierry,
On 02/12/2019 15:07, Laurent Pinchart wrote:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:50AM +0200, Tomi Valkeinen wrote:
The panel datasheet says that the panel samples at falling edge, but does not say anything about h/v sync signals. Testing shows that if the sync signals are driven on falling edge, the picture on the panel will be slightly shifted right.
Setting sync drive edge to the same as data drive edge fixes this issue.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
I don't have access to the documentation, but this makes sense, so
Acked-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
drivers/gpu/drm/panel/panel-simple.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 5d487686d25c..0784536ae6af 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -2397,7 +2397,8 @@ static const struct panel_desc osddisplays_osd070t1718_19ts = { .height = 91, }, .bus_format = MEDIA_BUS_FMT_RGB888_1X24,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE |
DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE,Can this be merged?
I have lost the original mail. Can you re-send or provide a patchwork pointer or similar. Then I will apply.
PS. Mail had been stuck in my spam quarantine so did not get it until now.
Sam
Hi Sam,
On 14/02/2020 23:39, Sam Ravnborg wrote:
Hi Tomi.
On Mon, Feb 10, 2020 at 10:15:33AM +0200, Tomi Valkeinen wrote:
Hi Thierry,
On 02/12/2019 15:07, Laurent Pinchart wrote:
Hi Tomi,
Thank you for the patch.
On Thu, Nov 14, 2019 at 11:39:50AM +0200, Tomi Valkeinen wrote:
The panel datasheet says that the panel samples at falling edge, but does not say anything about h/v sync signals. Testing shows that if the sync signals are driven on falling edge, the picture on the panel will be slightly shifted right.
Setting sync drive edge to the same as data drive edge fixes this issue.
Signed-off-by: Tomi Valkeinen tomi.valkeinen@ti.com
I don't have access to the documentation, but this makes sense, so
Acked-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
drivers/gpu/drm/panel/panel-simple.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 5d487686d25c..0784536ae6af 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -2397,7 +2397,8 @@ static const struct panel_desc osddisplays_osd070t1718_19ts = { .height = 91, }, .bus_format = MEDIA_BUS_FMT_RGB888_1X24,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE |
DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE,Can this be merged?
I have lost the original mail. Can you re-send or provide a patchwork pointer or similar. Then I will apply.
Here are lore and patchwork links. Note that the first two patches in the series have already been handled.
https://lore.kernel.org/dri-devel/20191114093950.rT5m86Xh0-a9FZGq7JipQ3O3vld...
https://patchwork.freedesktop.org/series/69459/
Tomi
Hi Tomi.
+++ b/drivers/gpu/drm/panel/panel-simple.c @@ -2397,7 +2397,8 @@ static const struct panel_desc osddisplays_osd070t1718_19ts = { .height = 91, }, .bus_format = MEDIA_BUS_FMT_RGB888_1X24,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE,
- .bus_flags = DRM_BUS_FLAG_DE_HIGH | DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE |
DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE,Can this be merged?
I have lost the original mail. Can you re-send or provide a patchwork pointer or similar. Then I will apply.
Here are lore and patchwork links. Note that the first two patches in the series have already been handled.
https://lore.kernel.org/dri-devel/20191114093950.rT5m86Xh0-a9FZGq7JipQ3O3vld...
Thanks for the pointers. It is now applied to drm-misc-next.
Sam
Hi Tony, Thierry, Laurent,
Any thoughts on the below points?
I think yet another option is to write some omap boot time quirks code, which looks at the DT data, and changes the panel compatible string (for 1), and removes the timings node (for 2).
Tomi
On 14/11/2019 11:39, Tomi Valkeinen wrote:
Hi Tony, Thierry, Laurent,
After the recent change of moving from omapdrm specific panel-dpi driver to the DRM simple panel, AM4 EVM/ePOS's panel is not working quite right. This series has fixes for it, but I'm not sure if these are the right ways to fix the issues, so comments welcome.
- Panel driver is not probed. With omapdrm's panel-dpi, the match
happened with "panel-dpi" compatible string. Now with panel-simple, the match should happen with the panel model compatible string, which is "osddisplays,osd057T0559-34ts" in the DT file. However, no such compatible exists in panel-simple.
Interestingly, the actual panel at least on my EVMs and ePOSes is not osd057T0559-34ts, but osd070t1718-19ts. Also, I was unable to find any information about osd057T0559-34ts. I don't know the history with this, so it is possible that the early versions of the boards did have osd057T0559-34ts, but was later changed to osd070t1718-19ts.
As osd070t1718-19ts is supported by panel-simple, changing the compatible string to osd070t1718-19ts in the DT file solves this one.
- Timings in DT file cause a kernel warning. Omapdrm's panel-dpi used
video timings from the DT file, so they are present in all the DT files. panel-simple uses timings from a table in the panel-simple driver, but gives a kernel warning if the DT file contains timings.
This can be solved by removing the timings from the DT file.
- Sync drive edge is not right. This one might have been present also
with panel-dpi, I didn't verify. The problem is that the panel-simple data for osddisplays_osd070t1718_19ts defines bus_flags for DE polarity and pixdata edge, but not for sync edge.
The datasheet for the panel does not give any hint on what the edge should be. Omapdrm defaults to driving syncs on falling edge, which caused the image to be shifted one pixel to the right.
Adding DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE bus_flag solves the problem. AM5 EVM also has the same panel with the same behavior.
====
The reason I'm not sure if the 1) and 2) fixes are correct is that they're breaking DT compatibility. Should we instead make changes to panel-simple to keep the same DT files working?
This would mean adding a new entry for the osd057T0559-34ts panel, but as we don't have datasheet for it, I think we could just append the compatible string to osd070t1718-19t's data.
It would also mean doing some change to the panel-simple code that gives the warning about timings in DT data. This might make sense, as I think we have other DT files with video timings too.
For 3), I think the patch is fine, but I'm not sure if the display controller driver should be able to deduce the sync drive edge from the pixdata drive edge. Are they usually the same? I have no idea...
Tomi
Tomi Valkeinen (3): ARM: dts: am437x-gp/epos-evm: fix panel compatible ARM: dts: am437x-gp/epos-evm: drop unused panel timings drm/panel: simple: fix osd070t1718_19ts sync drive edge
arch/arm/boot/dts/am437x-gp-evm.dts | 18 +----------------- arch/arm/boot/dts/am43x-epos-evm.dts | 18 +----------------- drivers/gpu/drm/panel/panel-simple.c | 3 ++- 3 files changed, 4 insertions(+), 35 deletions(-)
-- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
Hi
* Tomi Valkeinen tomi.valkeinen@ti.com [191127 12:59]:
Hi Tony, Thierry, Laurent,
Any thoughts on the below points?
I think yet another option is to write some omap boot time quirks code, which looks at the DT data, and changes the panel compatible string (for 1), and removes the timings node (for 2).
Nah, seems we can just update the compatible.
On 14/11/2019 11:39, Tomi Valkeinen wrote:
- Panel driver is not probed. With omapdrm's panel-dpi, the match
happened with "panel-dpi" compatible string. Now with panel-simple, the match should happen with the panel model compatible string, which is "osddisplays,osd057T0559-34ts" in the DT file. However, no such compatible exists in panel-simple.
Interestingly, the actual panel at least on my EVMs and ePOSes is not osd057T0559-34ts, but osd070t1718-19ts. Also, I was unable to find any information about osd057T0559-34ts. I don't know the history with this, so it is possible that the early versions of the boards did have osd057T0559-34ts, but was later changed to osd070t1718-19ts.
I guess you could keep the old compatible there too if really needed. But then again if the old compatible is known to be incorrect, it should be just updated.
So it looks good to me for the dts changes. Do you want me to pick them into fixes as it seems that the panel driver fix is a separate issue?
Regards,
Tony
Hi Tony,
On 27/11/2019 17:45, Tony Lindgren wrote:
Interestingly, the actual panel at least on my EVMs and ePOSes is not osd057T0559-34ts, but osd070t1718-19ts. Also, I was unable to find any information about osd057T0559-34ts. I don't know the history with this, so it is possible that the early versions of the boards did have osd057T0559-34ts, but was later changed to osd070t1718-19ts.
I guess you could keep the old compatible there too if really needed. But then again if the old compatible is known to be incorrect, it should be just updated.
So it looks good to me for the dts changes. Do you want me to pick them into fixes as it seems that the panel driver fix is a separate issue?
Yes, the third patch can be handled separately, so please pick the first two ones.
Tomi
Hi Tomi,
On Thu, Nov 14, 2019 at 11:39:47AM +0200, Tomi Valkeinen wrote:
Hi Tony, Thierry, Laurent,
After the recent change of moving from omapdrm specific panel-dpi driver to the DRM simple panel, AM4 EVM/ePOS's panel is not working quite right. This series has fixes for it, but I'm not sure if these are the right ways to fix the issues, so comments welcome.
- Panel driver is not probed. With omapdrm's panel-dpi, the match
happened with "panel-dpi" compatible string. Now with panel-simple, the match should happen with the panel model compatible string, which is "osddisplays,osd057T0559-34ts" in the DT file. However, no such compatible exists in panel-simple.
I've also noticed that we have a few other omap-based platforms that got broken, for the same reason :-( We're missing driver support for innolux,at070tn83, samsung,lte430wq-f0c and startek,startek-kd050c, and we have a few nodes that use panel-dpi without any more precise compatible string.
Interestingly, the actual panel at least on my EVMs and ePOSes is not osd057T0559-34ts, but osd070t1718-19ts. Also, I was unable to find any information about osd057T0559-34ts. I don't know the history with this, so it is possible that the early versions of the boards did have osd057T0559-34ts, but was later changed to osd070t1718-19ts.
As osd070t1718-19ts is supported by panel-simple, changing the compatible string to osd070t1718-19ts in the DT file solves this one.
- Timings in DT file cause a kernel warning. Omapdrm's panel-dpi used
video timings from the DT file, so they are present in all the DT files. panel-simple uses timings from a table in the panel-simple driver, but gives a kernel warning if the DT file contains timings.
This can be solved by removing the timings from the DT file.
- Sync drive edge is not right. This one might have been present also
with panel-dpi, I didn't verify. The problem is that the panel-simple data for osddisplays_osd070t1718_19ts defines bus_flags for DE polarity and pixdata edge, but not for sync edge.
The datasheet for the panel does not give any hint on what the edge should be. Omapdrm defaults to driving syncs on falling edge, which caused the image to be shifted one pixel to the right.
Adding DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE bus_flag solves the problem. AM5 EVM also has the same panel with the same behavior.
====
The reason I'm not sure if the 1) and 2) fixes are correct is that they're breaking DT compatibility. Should we instead make changes to panel-simple to keep the same DT files working?
That's tempting, as breaking DT is quite bad, but that would require adding a match on panel-dpi, and parsing timings in the panel-simple driver. Thierry has always been opposed to that as far as I can tell, and even if I don't share his point of view, I don't want to move in this direction without a consensus.
Your series is fine in my opinion, as even if we decide to handle backward compatibility in this case, changing the DT files in mainline is still the right way to go (if only to avoid giving wrong examples).
This would mean adding a new entry for the osd057T0559-34ts panel, but as we don't have datasheet for it, I think we could just append the compatible string to osd070t1718-19t's data.
It would also mean doing some change to the panel-simple code that gives the warning about timings in DT data. This might make sense, as I think we have other DT files with video timings too.
For 3), I think the patch is fine, but I'm not sure if the display controller driver should be able to deduce the sync drive edge from the pixdata drive edge. Are they usually the same? I have no idea...
Tomi Valkeinen (3): ARM: dts: am437x-gp/epos-evm: fix panel compatible ARM: dts: am437x-gp/epos-evm: drop unused panel timings drm/panel: simple: fix osd070t1718_19ts sync drive edge
arch/arm/boot/dts/am437x-gp-evm.dts | 18 +----------------- arch/arm/boot/dts/am43x-epos-evm.dts | 18 +----------------- drivers/gpu/drm/panel/panel-simple.c | 3 ++- 3 files changed, 4 insertions(+), 35 deletions(-)
On Mon, Dec 2, 2019 at 7:13 AM Laurent Pinchart laurent.pinchart@ideasonboard.com wrote:
Hi Tomi,
On Thu, Nov 14, 2019 at 11:39:47AM +0200, Tomi Valkeinen wrote:
Hi Tony, Thierry, Laurent,
After the recent change of moving from omapdrm specific panel-dpi driver to the DRM simple panel, AM4 EVM/ePOS's panel is not working quite right. This series has fixes for it, but I'm not sure if these are the right ways to fix the issues, so comments welcome.
- Panel driver is not probed. With omapdrm's panel-dpi, the match
happened with "panel-dpi" compatible string. Now with panel-simple, the match should happen with the panel model compatible string, which is "osddisplays,osd057T0559-34ts" in the DT file. However, no such compatible exists in panel-simple.
I've also noticed that we have a few other omap-based platforms that got broken, for the same reason :-( We're missing driver support for innolux,at070tn83, samsung,lte430wq-f0c and startek,startek-kd050c, and we have a few nodes that use panel-dpi without any more precise compatible string.
Interestingly, the actual panel at least on my EVMs and ePOSes is not osd057T0559-34ts, but osd070t1718-19ts. Also, I was unable to find any information about osd057T0559-34ts. I don't know the history with this, so it is possible that the early versions of the boards did have osd057T0559-34ts, but was later changed to osd070t1718-19ts.
As osd070t1718-19ts is supported by panel-simple, changing the compatible string to osd070t1718-19ts in the DT file solves this one.
- Timings in DT file cause a kernel warning. Omapdrm's panel-dpi used
video timings from the DT file, so they are present in all the DT files. panel-simple uses timings from a table in the panel-simple driver, but gives a kernel warning if the DT file contains timings.
This can be solved by removing the timings from the DT file.
- Sync drive edge is not right. This one might have been present also
with panel-dpi, I didn't verify. The problem is that the panel-simple data for osddisplays_osd070t1718_19ts defines bus_flags for DE polarity and pixdata edge, but not for sync edge.
The datasheet for the panel does not give any hint on what the edge should be. Omapdrm defaults to driving syncs on falling edge, which caused the image to be shifted one pixel to the right.
Adding DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE bus_flag solves the problem. AM5 EVM also has the same panel with the same behavior.
====
The reason I'm not sure if the 1) and 2) fixes are correct is that they're breaking DT compatibility. Should we instead make changes to panel-simple to keep the same DT files working?
That's tempting, as breaking DT is quite bad, but that would require adding a match on panel-dpi, and parsing timings in the panel-simple driver. Thierry has always been opposed to that as far as I can tell, and even if I don't share his point of view, I don't want to move in this direction without a consensus.
What about a generic driver separate from the simple panel driver that does just the panel-dpi parsing?
If we exported the necessary functions from simple-panel, we could call them from the panel-dpi parser and not have to re-invent the functions to enable power, gpio or back light and/or fix them each time they get updated.
adam
Your series is fine in my opinion, as even if we decide to handle backward compatibility in this case, changing the DT files in mainline is still the right way to go (if only to avoid giving wrong examples).
This would mean adding a new entry for the osd057T0559-34ts panel, but as we don't have datasheet for it, I think we could just append the compatible string to osd070t1718-19t's data.
It would also mean doing some change to the panel-simple code that gives the warning about timings in DT data. This might make sense, as I think we have other DT files with video timings too.
For 3), I think the patch is fine, but I'm not sure if the display controller driver should be able to deduce the sync drive edge from the pixdata drive edge. Are they usually the same? I have no idea...
Tomi Valkeinen (3): ARM: dts: am437x-gp/epos-evm: fix panel compatible ARM: dts: am437x-gp/epos-evm: drop unused panel timings drm/panel: simple: fix osd070t1718_19ts sync drive edge
arch/arm/boot/dts/am437x-gp-evm.dts | 18 +----------------- arch/arm/boot/dts/am43x-epos-evm.dts | 18 +----------------- drivers/gpu/drm/panel/panel-simple.c | 3 ++- 3 files changed, 4 insertions(+), 35 deletions(-)
-- Regards,
Laurent Pinchart _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
Hi Adam,
On Mon, Dec 02, 2019 at 07:24:09AM -0600, Adam Ford wrote:
On Mon, Dec 2, 2019 at 7:13 AM Laurent Pinchart wrote:
On Thu, Nov 14, 2019 at 11:39:47AM +0200, Tomi Valkeinen wrote:
Hi Tony, Thierry, Laurent,
After the recent change of moving from omapdrm specific panel-dpi driver to the DRM simple panel, AM4 EVM/ePOS's panel is not working quite right. This series has fixes for it, but I'm not sure if these are the right ways to fix the issues, so comments welcome.
- Panel driver is not probed. With omapdrm's panel-dpi, the match
happened with "panel-dpi" compatible string. Now with panel-simple, the match should happen with the panel model compatible string, which is "osddisplays,osd057T0559-34ts" in the DT file. However, no such compatible exists in panel-simple.
I've also noticed that we have a few other omap-based platforms that got broken, for the same reason :-( We're missing driver support for innolux,at070tn83, samsung,lte430wq-f0c and startek,startek-kd050c, and we have a few nodes that use panel-dpi without any more precise compatible string.
Interestingly, the actual panel at least on my EVMs and ePOSes is not osd057T0559-34ts, but osd070t1718-19ts. Also, I was unable to find any information about osd057T0559-34ts. I don't know the history with this, so it is possible that the early versions of the boards did have osd057T0559-34ts, but was later changed to osd070t1718-19ts.
As osd070t1718-19ts is supported by panel-simple, changing the compatible string to osd070t1718-19ts in the DT file solves this one.
- Timings in DT file cause a kernel warning. Omapdrm's panel-dpi used
video timings from the DT file, so they are present in all the DT files. panel-simple uses timings from a table in the panel-simple driver, but gives a kernel warning if the DT file contains timings.
This can be solved by removing the timings from the DT file.
- Sync drive edge is not right. This one might have been present also
with panel-dpi, I didn't verify. The problem is that the panel-simple data for osddisplays_osd070t1718_19ts defines bus_flags for DE polarity and pixdata edge, but not for sync edge.
The datasheet for the panel does not give any hint on what the edge should be. Omapdrm defaults to driving syncs on falling edge, which caused the image to be shifted one pixel to the right.
Adding DRM_BUS_FLAG_SYNC_DRIVE_POSEDGE bus_flag solves the problem. AM5 EVM also has the same panel with the same behavior.
====
The reason I'm not sure if the 1) and 2) fixes are correct is that they're breaking DT compatibility. Should we instead make changes to panel-simple to keep the same DT files working?
That's tempting, as breaking DT is quite bad, but that would require adding a match on panel-dpi, and parsing timings in the panel-simple driver. Thierry has always been opposed to that as far as I can tell, and even if I don't share his point of view, I don't want to move in this direction without a consensus.
What about a generic driver separate from the simple panel driver that does just the panel-dpi parsing?
A separate driver wouldn't make too much of a difference. Thierry's point was that drivers should hardcode the timings, as the compatible string gives enough information and encoding timings in DT would duplicate that information. I have a different point of view, and there are pros and cons for both options, so we haven't been able to reach an agreement. That's why I haven't pushed too hard for timings parsing from DT, regardless of whether it would be in panel-simple or in a separate driver.
If we exported the necessary functions from simple-panel, we could call them from the panel-dpi parser and not have to re-invent the functions to enable power, gpio or back light and/or fix them each time they get updated.
Your series is fine in my opinion, as even if we decide to handle backward compatibility in this case, changing the DT files in mainline is still the right way to go (if only to avoid giving wrong examples).
This would mean adding a new entry for the osd057T0559-34ts panel, but as we don't have datasheet for it, I think we could just append the compatible string to osd070t1718-19t's data.
It would also mean doing some change to the panel-simple code that gives the warning about timings in DT data. This might make sense, as I think we have other DT files with video timings too.
For 3), I think the patch is fine, but I'm not sure if the display controller driver should be able to deduce the sync drive edge from the pixdata drive edge. Are they usually the same? I have no idea...
Tomi Valkeinen (3): ARM: dts: am437x-gp/epos-evm: fix panel compatible ARM: dts: am437x-gp/epos-evm: drop unused panel timings drm/panel: simple: fix osd070t1718_19ts sync drive edge
arch/arm/boot/dts/am437x-gp-evm.dts | 18 +----------------- arch/arm/boot/dts/am43x-epos-evm.dts | 18 +----------------- drivers/gpu/drm/panel/panel-simple.c | 3 ++- 3 files changed, 4 insertions(+), 35 deletions(-)
dri-devel@lists.freedesktop.org
-
Adam Ford -
Laurent Pinchart -
Merlijn Wajer -
Sam Ravnborg -
Sebastian Reichel -
Tomi Valkeinen -
Tony Lindgren