Possible Fix for our HDMI-DVI Issues

[Arcanjo Azrael]

So... Is it compatible with our wandboard quad ?

Leonardo

Source: http://boundarydevices.com/dvi-support-on-i-mx6-boards/

At long last, DVI support is here for our i.MX6 boards.

When we built our very first SABRE Lite boards, one of the first things we tested was the HDMI video port, and at that time, most of the monitors we used for testing were natively DVI, not HDMI. All of the newer monitors were being used by our staff, and the ones available for testing were older monitors which only supported DVI.

Way back then, we weren’t booting Linux, and were running some bare metal code, and we found that the hot-plug detect code wasn’t working quite right. If we looked at the HDMI_PHY_STAT0 register, we could see that bits 7-4 (the RX_SENSE bits) would toggle along with connection to the monitor, but bit 1 (HPD) wouldn’t.

We updated the code to look at the others and didn’t look back until very recently. Along the way, we’ve told customers to just use HDMI instead, knowing that there was a software fix, but kicked the can down the road. Sorry about that.

We finally had a chance to investigate more fully, and to compare the operation of Freescale’s SABRE SDB board with ours. With a little nudge from RooT, we figured out that the core of the issue is a voltage dividing resistor we placed on the HPD signal to prevent 5V from hitting the CPU.

With the latest specs from Freescale, we found that this resistor could be bypassed (the pad is 5V-tolerant), but we also added some patches to allow the RX_SENSE bits to be used instead, providing a software solution.

The patches come in three parts:

1. U-Boot patch to change the hdmidet command and internal HDMI detection to use the RX_SENSE bits, and
2. A non-Android Linux patch to do the same, and
3. A patch for the Android (Jellybean) kernel

Note that the U-Boot patch is only in our staging branch at the moment, and will be released to production in the next couple of weeks.

[Otavio Salvador]

On Thu, Jul 25, 2013 at 3:58 PM, Arcanjo Azrael
<arcanjo...@gmail.com> wrote:
> So... Is it compatible with our wandboard quad ?

...

> With the latest specs from Freescale, we found that this resistor could be
> bypassed (the pad is 5V-tolerant), but we also added some patches to allow
> the RX_SENSE bits to be used instead, providing a software solution.
>
> The patches come in three parts:
>
> U-Boot patch to change the hdmidet command and internal HDMI detection to
> use the RX_SENSE bits, and
> A non-Android Linux patch to do the same, and
> A patch for the Android (Jellybean) kernel

...

I am cooking a test image with the U-Boot and Linux Kernel patches for
testing in Wandboard Quad (as it is the most problematic one) for
testing; my monitors work fine so I cannot test.

--
Otavio Salvador O.S. Systems
http://www.ossystems.com.br http://projetos.ossystems.com.br
Mobile: +55 (53) 9981-7854 Mobile: +1 (347) 903-9750

[jth...@gmail.com]

Hi Leonardo,

I pulled the mxc_hdmi.[ch] file changes into Robert Nelson's 3.0.35.wand6.3 source, rebuilt and deployed.
This did not resolve the display trouble I experience.  Looking forward to seeing how it works for you.

All the best,   Joe

[John Weber]

> I am cooking a test image with the U-Boot and Linux Kernel patches for
> testing in Wandboard Quad (as it is the most problematic one) for
> testing; my monitors work fine so I cannot test.
>

Is there an opinion that this should fix the EDID issues we've been seeing?
Those have been difficult to pin down.

Otavio - you going to put this is Yocto master-next?

[Otavio Salvador]

I plan to make this in an image for testing and ask for feedback, if
this improves it than we can add it.

[Otavio Salvador]

Hello folks,

On Thu, Jul 25, 2013 at 5:51 PM, Otavio Salvador
<ota...@ossystems.com.br> wrote:
> On Thu, Jul 25, 2013 at 3:58 PM, Arcanjo Azrael
> <arcanjo...@gmail.com> wrote:
>> So... Is it compatible with our wandboard quad ?
> ...
>> With the latest specs from Freescale, we found that this resistor could be
>> bypassed (the pad is 5V-tolerant), but we also added some patches to allow
>> the RX_SENSE bits to be used instead, providing a software solution.
>>
>> The patches come in three parts:
>>
>> U-Boot patch to change the hdmidet command and internal HDMI detection to
>> use the RX_SENSE bits, and
>> A non-Android Linux patch to do the same, and
>> A patch for the Android (Jellybean) kernel
> ...
>
> I am cooking a test image with the U-Boot and Linux Kernel patches for
> testing in Wandboard Quad (as it is the most problematic one) for
> testing; my monitors work fine so I cannot test.

I have made the image with the kernel patch applied. Please give it a try:

http://bit.ly/1bSkthR

This is a very small image just for test. It should be enough to get
the HDMI tested. I hope it does work.

[Arcanjo Azrael]

On 25/07/2013 18:37, Otavio Salvador wrote:
> Hello folks,
>
> On Thu, Jul 25, 2013 at 5:51 PM, Otavio Salvador
> <ota...@ossystems.com.br> wrote:
>> On Thu, Jul 25, 2013 at 3:58 PM, Arcanjo Azrael
>> <arcanjo...@gmail.com> wrote:
>>> So... Is it compatible with our wandboard quad ?
>> ...
>>> With the latest specs from Freescale, we found that this resistor could be
>>> bypassed (the pad is 5V-tolerant), but we also added some patches to allow
>>> the RX_SENSE bits to be used instead, providing a software solution.
>>>
>>> The patches come in three parts:
>>>
>>> U-Boot patch to change the hdmidet command and internal HDMI detection to
>>> use the RX_SENSE bits, and
>>> A non-Android Linux patch to do the same, and
>>> A patch for the Android (Jellybean) kernel
>> ...
>>
>> I am cooking a test image with the U-Boot and Linux Kernel patches for
>> testing in Wandboard Quad (as it is the most problematic one) for
>> testing; my monitors work fine so I cannot test.
> I have made the image with the kernel patch applied. Please give it a try:
>
> http://bit.ly/1bSkthR
>
> This is a very small image just for test. It should be enough to get
> the HDMI tested. I hope it does work.
>

I saw your logo, at 1024x768@56.* (not sure about 56.*hz, too fast to
get my monitor settings) . But the display remains in standby.

[ 3.238537] mxc_hdmi mxc_hdmi: No modes read from edid
[ 3.238545] mxc_hdmi mxc_hdmi: create default modelist

I compiled a kernel and applied the patches and I got this in dmesg:

fbcvt: 1680x1050@60: CVT Name - 1.764MA

with uEnv.txt settings, set to 1680x1050.

and cat /sys/class/graphics/fb0/mode returns:
U:1920x1080p-60

Yet, no fixed yet for wandboard.

Leonardo

[Robert Nelson]

On Thu, Jul 25, 2013 at 4:37 PM, Otavio Salvador

<ota...@ossystems.com.br> wrote:
> Hello folks,
>
> On Thu, Jul 25, 2013 at 5:51 PM, Otavio Salvador
> <ota...@ossystems.com.br> wrote:
>> On Thu, Jul 25, 2013 at 3:58 PM, Arcanjo Azrael
>> <arcanjo...@gmail.com> wrote:
>>> So... Is it compatible with our wandboard quad ?
>> ...
>>> With the latest specs from Freescale, we found that this resistor could be
>>> bypassed (the pad is 5V-tolerant), but we also added some patches to allow
>>> the RX_SENSE bits to be used instead, providing a software solution.
>>>
>>> The patches come in three parts:
>>>
>>> U-Boot patch to change the hdmidet command and internal HDMI detection to
>>> use the RX_SENSE bits, and
>>> A non-Android Linux patch to do the same, and
>>> A patch for the Android (Jellybean) kernel
>> ...
>>
>> I am cooking a test image with the U-Boot and Linux Kernel patches for
>> testing in Wandboard Quad (as it is the most problematic one) for
>> testing; my monitors work fine so I cannot test.
>
> I have made the image with the kernel patch applied. Please give it a try:
>
> http://bit.ly/1bSkthR
>
> This is a very small image just for test. It should be enough to get
> the HDMI tested. I hope it does work.

It looks closer, still no dice on mine.. (Note: u-boot displays it's
framebuffer just fine.. )

http://pastebin.com/QpKVXSjk

Regards,

--
Robert Nelson
http://www.rcn-ee.com/

[Alberto F]

Hi there, and ... OMG....

I'm THAT root from Boundary Devices, this... this is so embarrassing!

I adquired recently -this last week- a wandboard ( quad ), I was looking for common problems before turning it on and .. tachan!

I'm deepend a little more in this annoying problem, and found a hardware solution - I just realized that ericn answer my comment -

https://dl.dropboxusercontent.com/u/40319604/HDMI_HDP_issues.PNG

With relation to the problem of EDID, I found that some monitors ( DVI always, sorry HDMI people! ) don't send the complete parameters -just partial list of capabilities- until HPD ( pin 16 of DVI ) is put on high level. Some strange behavior.

Good luck!

[abr...@gmail.com]

I can't quite follow that diagram but I understand you are jumping a voltage divider to effectively take it out, yes? Any chance you could post a picture of the pads you are using for the bypass? I have had no end of problems with my quad on every display in my house. I'm about to quit trying and go headless while this is sorted out.

[Alberto F]

Hi there,

THAT solution was implemented on a custom board that is being designed right now. It's one of the multiple solutions for implement the HDMI/DVI hotplug pins -high input impedance ( aka very low current ), low output impedance, safe levels on i.MX6 pin, etc.-

The currently solution for nitrogen6x boards comes from a solder jumper located on the bottom part of the board, this force the hotplug event always.

The wandboard has not had the luck, yet...  but the workaround with the alternative register... looks elegant and simple.

[朱含]

Today, I brought a WandBoard Dual board, and tried the Ubuntu Images and Yocto Images.

Ops, connecting to a DVI monitor SyncMaster 940bw, and consolo:
[    3.012378] mxc_hdmi mxc_hdmi: No modes read from edid
[    3.012386] mxc_hdmi mxc_hdmi: create default modelist

My board version is A0.
So does this patch could be used to solve this issue ?



在 2013年7月26日星期五UTC+8上午2时58分21秒，ArchAngel Azrael写道：

[John Weber]

Hi,

On 4/18/14, 10:03 AM, 朱含 wrote:

Today, I brought a WandBoard Dual board, and tried the Ubuntu Images and Yocto Images.

Ops, connecting to a DVI monitor SyncMaster 940bw, and consolo:
[    3.012378] mxc_hdmi mxc_hdmi: No modes read from edid
[    3.012386] mxc_hdmi mxc_hdmi: create default modelist

My board version is A0.
So does this patch could be used to solve this issue ?

This issue is not related to the DVI patches described below.  However, in Yocto we have U-boot 2014.1 working and this has the first patch applied already.

I've applied these patches to the kernel and this does not solve the EDID issue, which has been prevalent in Wandboard from the beginning.

John

在 2013年7月26日星期五UTC+8上午2时58分21秒，ArchAngel Azrael写道：

So... Is it compatible with our wandboard quad ?

Leonardo

Source: http://boundarydevices.com/dvi-support-on-i-mx6-boards/

At long last, DVI support is here for our i.MX6 boards.

When we built our very first SABRE Lite boards, one of the first things we tested was the HDMI video port, and at that time, most of the monitors we used for testing were natively DVI, not HDMI. All of the newer monitors were being used by our staff, and the ones available for testing were older monitors which only supported DVI.

Way back then, we weren’t booting Linux, and were running some bare metal code, and we found that the hot-plug detect code wasn’t working quite right. If we looked at the HDMI_PHY_STAT0 register, we could see that bits 7-4 (the RX_SENSE bits) would toggle along with connection to the monitor, but bit 1 (HPD) wouldn’t.

We updated the code to look at the others and didn’t look back until very recently. Along the way, we’ve told customers to just use HDMI instead, knowing that there was a software fix, but kicked the can down the road. Sorry about that.

We finally had a chance to investigate more fully, and to compare the operation of Freescale’s SABRE SDB board with ours. With a little nudge from RooT, we figured out that the core of the issue is a voltage dividing resistor we placed on the HPD signal to prevent 5V from hitting the CPU.

With the latest specs from Freescale, we found that this resistor could be bypassed (the pad is 5V-tolerant), but we also added some patches to allow the RX_SENSE bits to be used instead, providing a software solution.

The patches come in three parts:

1. U-Boot patch to change the hdmidet command and internal HDMI detection to use the RX_SENSE bits, and
2. A non-Android Linux patch to do the same, and
3. A patch for the Android (Jellybean) kernel

Note that the U-Boot patch is only in our staging branch at the moment, and will be released to production in the next couple of weeks.

--
You received this message because you are subscribed to the Google Groups "Wandboard" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wandboard+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[hans]

Thanks, John

 

  Now I try to find u-boot 2014.1 for wandboard, but all files exit on wiki.wandboard.org has not such u-boot version.

 

  Now what I’m using is U-Boot 2013.10, used in dora image.

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber
发送时间: 2014年4月19日 星期六 0:29
收件人: wand...@googlegroups.com
主题: Re: Possible Fix for our HDMI-DVI Issues

--
You received this message because you are subscribed to a topic in the Google Groups "Wandboard" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/wandboard/KgFACCiAEPk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to wandboard+...@googlegroups.com.

[John Weber]

Hi Hans,

On 4/18/14, 6:07 PM, hans wrote:

Thanks, John

 

  Now I try to find u-boot 2014.1 for wandboard, but all files exit on wiki.wandboard.org has not such u-boot version.

 

  Now what I’m using is U-Boot 2013.10, used in dora image.

Here is what is used in Yocto:
https://github.com/Freescale/u-boot-imx/tree/patches-2014.01

Keep in mind that this is mainline u-boot with a few patches applied, so it does NOT include SPL support.  It needs to be built for the specific Wandboard you have (e.g. solo, dual, quad).  Once SPL is mainlined we'll use that for all boards in Yocto, including Wandboard.

John

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber

发 送时间: 2014年4月19日 星期六 0:29

收件人: wand...@googlegroups.com
主题: Re: Possible Fix for our HDMI-DVI Issues

 

Hi,

On 4/18/14, 10:03 AM, 朱含 wrote:

Today, I brought a WandBoard Dual board, and tried the Ubuntu Images and Yocto Images.

Ops, connecting to a DVI monitor SyncMaster 940bw, and consolo:
[    3.012378] mxc_hdmi mxc_hdmi: No modes read from edid
[    3.012386] mxc_hdmi mxc_hdmi: create default modelist

My board version is A0.
So does this patch could be used to solve this issue ?

This issue is not related to the DVI patches described below.  However, in Yocto we have U-boot 2014.1 working and this has the first patch applied already.

I've applied these patches to the kernel and this does not solve the EDID issue, which has been prevalent in Wandboard from the beginning.

John

在 2013年7月26日星期五UTC+8上 午2时58分21秒，ArchAngel Azrael写道：

[hans]

Hi John

 

 

   I just download U-Boot 2014.1 and compile it for Wandboard, HDMI in Uboot worked, but kernel still

cannot.

 

   I’m trying to update the kernel image to see what happened?

 

   So why not merge the change to Dora ?

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber

发送时间: 2014年4月19日 星期六 10:44
收件人: wand...@googlegroups.com
主题: Re: 答复: Possible Fix for our HDMI-DVI Issues

[John Weber]

Hi Hans,

On 4/19/14, 12:14 AM, hans wrote:

Hi John

 

 

   I just download U-Boot 2014.1 and compile it for Wandboard, HDMI in Uboot worked, but kernel still

cannot.

The kernel doesn't boot?  We made a change in 2014.01 to use zImages so this is probably what you are seeing.  Actually, you might want to check 2013.10 of u-boot, because that might have had the hdmidet patch already added.

 

   I’m trying to update the kernel image to see what happened?

 

   So why not merge the change to Dora ?

This is too big of a change to merge into a stable branch.  It will be in Daisy.

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber

发 送时间: 2014年4月19日 星期六 10:44

收件人: wand...@googlegroups.com
主题: Re: 答复: Possible Fix for our HDMI-DVI Issues

 

Hi Hans,

On 4/18/14, 6:07 PM, hans wrote:

Thanks, John

 

  Now I try to find u-boot 2014.1 for wandboard, but all files exit on wiki.wandboard.org has not such u-boot version.

 

  Now what I’m using is U-Boot 2013.10, used in dora image.

Here is what is used in Yocto:
https://github.com/Freescale/u-boot-imx/tree/patches-2014.01

Keep in mind that this is mainline u-boot with a few patches applied, so it does NOT include SPL support.  It needs to be built for the specific Wandboard you have (e.g. solo, dual, quad).  Once SPL is mainlined we'll use that for all boards in Yocto, including Wandboard.

John

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代 表 John Weber

[hans]

Hi John

 

 Chnaged to 2013.10, this time, kernel start, but there is no HDMI output to DVI in kernel.

Do I need a right version of kernel?

 

  Besides, I’m downloading Yocto master also, hope to be successful.

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber

发送时间: 2014年4月19日 星期六 13:29
收件人: wand...@googlegroups.com
主题: Re: 答复: 答复: Possible Fix for our HDMI-DVI Issues

[hans]

Hi John

 

U-Boot, 2013.10, 2014.1, 2014.4, all u-boot versions work all right for the HDMI.

 

Linux kernel

3.10.17-1.0.0-wandboard 

    1) once a DVI monitor was connected, system crashed, no matter hot plug or always exist.

    mxc_hdmi 20e0000.hdmi_video: No modes read from edid

mxc_hdmi 20e0000.hdmi_video: create default modelist

fbcvt: 1920x1080@60: CVT Name - 2.073M9

   

3.0.35

    1) EDID cannot be read, but the system can still function.

   

      So until now, how could I find a solution about HDMI display ?

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber

发送时间: 2014年4月19日 星期六 13:29
收件人: wand...@googlegroups.com
主题: Re: 答复: 答复: Possible Fix for our HDMI-DVI Issues

[John Weber]

Hi Hans,

On 4/19/14, 11:09 PM, hans wrote:

Hi John

 

U-Boot, 2013.10, 2014.1, 2014.4, all u-boot versions work all right for the HDMI.

 

Linux kernel

3.10.17-1.0.0-wandboard 

    1) once a DVI monitor was connected, system crashed, no matter hot plug or always exist.

    mxc_hdmi 20e0000.hdmi_video: No modes read from edid

mxc_hdmi 20e0000.hdmi_video: create default modelist

fbcvt: 1920x1080@60: CVT Name - 2.073M9

OK.  I have seen this happen with HDMI (not DVI) on one Wandboard Solo I have here.  Since it doesn't happen on the Duallite or Quad, I had assumed that it was due to some sort of electrical damage to the port (probably affecting the I2C/DDC bus) and haven't attempted to debug it further.  I suspect that it is due to a dependency in the EDID.  This is also more confusing to me because I know that at one point I tested the same Wandboard Solo and it worked, so this failure has happened over time.

Can you share some information on your system configuration?  For example, are you using Yocto?  If so what rev?  What machine are you building (core-image-x11)?  Kernel configuration and any DTS changes?

[hite...@gmail.com]

all source code are the original code from.theweb, I changed nothing

I think the possible reason may be
1. HDMI and DVI.have difference during vblank period, so DVI monitors cannot recognize HDMI signals.

2. edid couldn't be read, so nothing can be output.

I also test the android image from the website, edid couldbe read, but still no image could be output to the monitors

anyway, i would test a HDMI monitor next week

John Weber <rjohn...@gmail.com>编写：

[John Weber]

Hans,

On 4/21/14, 10:16 AM, hite...@gmail.com wrote:

all source code are the original code from.theweb, I changed nothing

OK.  There is a lot of source code out there.  What are you using?  Yocto?

[Joshua Kurland]

Hey John, I just wanted to throw my two cents in to this too.  I am seeing the exact same crash on the wandboard-quad when using HDMI.  At the moment I am on the latest commit of wandboard-3.10.17 and uboot-2014.01.  However it does not happen every time for me.  Pressing the reset button does nothing, but I did notice that unplugging the power cable for an extended period (~1min) will clear the issue, allowing the board to boot.  

On Monday, April 21, 2014 11:17:42 AM UTC-4, John Weber wrote:

Hans,

On 4/21/14, 10:16 AM, hite...@gmail.com wrote:

all source code are the original code from.theweb, I changed nothing

OK.  There is a lot of source code out there.  What are you using?  Yocto?

I think the possible reason may be
1. HDMI and DVI.have difference during vblank period, so DVI monitors cannot recognize HDMI signals.

2. edid couldn't be read, so nothing can be output.

I also test the android image from the website, edid couldbe read, but still no image could be output to the monitors

anyway, i would test a HDMI monitor next week

John Weber <rjoh...@gmail.com>编写：

[hans]

Yocto  Dora Image

 

Ubunt 12.04

 

Android 4.3

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber
发送时间: 2014年4月21日 星期一 23:18
收件人: wand...@googlegroups.com
主题: Re: 答复: 答复: 答复: Possible Fix for our HDMI-DVI Issues

--
You received this message because you are subscribed to a topic in the Google Groups "Wandboard" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/wandboard/KgFACCiAEPk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to wandboard+...@googlegroups.com.

[John Weber]

Hi Joshua and Hans -

On 4/21/14, 2:44 PM, Joshua Kurland wrote:
> Hey John, I just wanted to throw my two cents in to this too. I am seeing the
> exact same crash on the wandboard-quad when using HDMI. At the moment I am on
> the latest commit of wandboard-3.10.17 and uboot-2014.01. However it does not
> happen every time for me. Pressing the reset button does nothing, but I did
> notice that unplugging the power cable for an extended period (~1min) will
> clear the issue, allowing the board to boot.
>

Thanks. It looks like we have a little work cut out for us here to debug this.
Since 3.0.35 works, I think it's safe to assume that this is a software problem
in 3.10.17 beta.

Tonight, I'm working on a set of Yocto recipes to turn the Wandboard into a
wireless access point.

John

[Fabio Estevam]

Hi,

In case someone is interested in trying HDMI on wandboard running
3.15-rc1, here is a patch I submitted today:
http://www.spinics.net/lists/arm-kernel/msg324380.html

It worked fine here on a full HDMI monitor and also tested via
HDMI-DVI adapter on another monitor at 1680x1050.

Regards,

Fabio Estevam

[John Weber]

Thanks Fabio.

I've seen evidence of working HDMI in mainline, but it would not use the FSL
mxcfb devices, correct? Does EDID work in mainline?
>
> Regards,
>
> Fabio Estevam

[Fabio Estevam]

Hi John,

On Tue, Apr 22, 2014 at 11:58 AM, John Weber <rjohn...@gmail.com> wrote:

> I've seen evidence of working HDMI in mainline, but it would not use the FSL
> mxcfb devices, correct? Does EDID work in mainline?

Correct, mainline does not use the FSL mxcfb driver and EDID is
working fine in mainline.

Regards,

Fabio Estevam

[sels...@gmail.com]

On 22/04/14 15:33, Fabio Estevam wrote:

> In case someone is interested in trying HDMI on wandboard running
> 3.15-rc1, here is a patch I submitted today:

Fabio,

Is that with a pure 3.15-rc1, or do you have other patches as well?

I'm still having the problem RMK posted about where EDID is fine but
the wrong clock gets configured:
http://article.gmane.org/gmane.linux.ports.arm.kernel/314767

I didn't think whatever fix for that was eventually chosen had made
it into 3.15-rc1, seems to be several sets of patches for imx-drm
floating around on the ml just now and I've still to work out which
ones I need.

[Fabio Estevam]

On Tue, Apr 22, 2014 at 1:09 PM, <sels...@gmail.com> wrote:
> On 22/04/14 15:33, Fabio Estevam wrote:
>
>> In case someone is interested in trying HDMI on wandboard running
>> 3.15-rc1, here is a patch I submitted today:
>
> Fabio,
>
> Is that with a pure 3.15-rc1, or do you have other patches as well?

I applied the HDMI dts patch against this 3.15-rc1 based tree:
https://git.kernel.org/cgit/linux/kernel/git/shawnguo/linux.git/log/?h=for-next

Let me know if it works for you.

[John Weber]

Hans, Joshua, and others that might be interested.

I've pulled in some changes to the kernel in a test branch.  It has solved the problem that I was seeing on the Wandboard Solo.  Would one of you mind giving it a try?  The branch is 'hdmi-fixes'

https://github.com/wandboard-org/linux/tree/hdmi-fixes

FYI - I'm seeing the following behavior:
- HDMI mode defaults to 1024x768 resolution regardless of the video= bootarg.  On 3.0.35 it defaulted to 1920x1080 if I remember correctly.
- DVI mode defaults to 640x480.

Thanks,
John 

On 4/21/14, 8:13 PM, hans wrote:

Yocto  Dora Image

 

Ubunt 12.04

 

Android 4.3

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber

发 送时间: 2014年4月21日 星期一 23:18
收件人: wand...@googlegroups.com
主题: Re: 答复: 答 复: 答复: Possible Fix for our HDMI-DVI Issues

[John Weber]

Hans, Joshua, and others

Good news - Freescale just released their 3.10.17 GA kernel, rev 1.0.0.  Or, at least they pushed it to their repository.

I've rebased the Wandboard 3.0.17_1.0.0_beta kernel against the GA kernel and it went cleanly.  I was able to do a quick build of core-image-x11 and check it out.  The good news is that it seems to have resolved the resolution and stability problems I've seen on Wandboard Quad.  DVI now defaults to 1920x1080 and HDMI does as well. EDID is still a problem, unfortunately.  MIPI CSI capture works well, as do the GLX benchmarks.

I haven't had a chance to check this on Dual and Solo yet.

I wanted to get this out to you guys because it probably does not make sense to waste much time testing the hdmi-fixes branch I just sent out.

https://github.com/wandboard-org/linux/tree/wandboard_imx_3.10.17_1.0.0_ga_test

John

On 4/21/14, 8:13 PM, hans wrote:

Yocto  Dora Image

 

Ubunt 12.04

 

Android 4.3

 

 

Thanks and Best Regards !

 

发件人: wand...@googlegroups.com [mailto:wand...@googlegroups.com] 代表 John Weber

发 送时间: 2014年4月21日 星期一 23:18
收件人: wand...@googlegroups.com
主题: Re: 答复: 答 复: 答复: Possible Fix for our HDMI-DVI Issues

[hite...@gmail.com]

since edid could not be read, how could video be output to hdmi?

John Weber <rjohn...@gmail.com>编写：

[John Weber]

It picks a default resolution and fills the modelist with a set of standard resolutions. 

—
Sent from Mailbox

[sels...@gmail.com]

Ah, when did Shawn switch his tree to kernel.org ? I've been following the
one he has at linaro.

Certainly works better, as long as I only enable HDMI it's working fine.
However unplugging one display and plugging in a different display after
boot doesn't cause a change in available modes in
/sys/class/drm/card0-HDMI-A-1/modes even though
/sys/class/drm/card0-HDMI-A-1/edid shows the change
A reboot gets the mode updated correctly though.

On some other i.MX6 based boards I have LVDS based displays in addition
to HDMI, but as soon as I enable the ldb node in the dts it seems to
cause problems with HDMI.

Now maybe I'm just doing something wrong enabling it, but I'd really
like a single dts that could support the three combinations:
1. HDMI only
2. LVDS only
3. both simultaneously

Seems I can do either HDMI only or LVDS only but need different dts files
to do it.
I need to investigate further, but it seems just having the ldb node with
status=disabled and a timings section in there is a problem.

Are there any good example dts files where both are enabled?

[Fabio Estevam]

On Wed, Apr 23, 2014 at 2:45 PM, <sels...@gmail.com> wrote:

> Certainly works better, as long as I only enable HDMI it's working fine.

Hopefully this will help Wandboard users using HDMI only display.

> However unplugging one display and plugging in a different display after
> boot doesn't cause a change in available modes in
> /sys/class/drm/card0-HDMI-A-1/modes even though
> /sys/class/drm/card0-HDMI-A-1/edid shows the change
> A reboot gets the mode updated correctly though.
>
> On some other i.MX6 based boards I have LVDS based displays in addition
> to HDMI, but as soon as I enable the ldb node in the dts it seems to
> cause problems with HDMI.

This is a known issue currently:
http://www.spinics.net/lists/arm-kernel/msg324710.html

[sels...@gmail.com]

On 23/04/14 18:53, Fabio Estevam wrote:
>
> Hopefully this will help Wandboard users using HDMI only display.

I've tried it on several boards with solo, dual and quad and it works
on them all.

> This is a known issue currently:
> http://www.spinics.net/lists/arm-kernel/msg324710.html

Hmm. ok, well maybe someone will get the itch and think it's worth fixing
at some point.

Thanks!

[Fabio Estevam]

On Fri, Apr 25, 2014 at 2:53 PM, <sels...@gmail.com> wrote:
> On 23/04/14 18:53, Fabio Estevam wrote:
>>
>> Hopefully this will help Wandboard users using HDMI only display.
>
> I've tried it on several boards with solo, dual and quad and it works
> on them all.

Excellent! ;-) Thanks for testing.

>> This is a known issue currently:
>> http://www.spinics.net/lists/arm-kernel/msg324710.html
>
> Hmm. ok, well maybe someone will get the itch and think it's worth fixing
> at some point.

In the initial version of the imx-hdmi driver I submitted to mainline
I was able to get LVDS and HDMI working fine.

I will try to look at this issue when time permits.