what linux distro's still support the pandaboard

[Kristoff]

Hi all,



I have a pandaboard (non-ES) laying on my desk for ... well ... very
long. Althou I have moved into microcontrollers (and now also FPGAs), I
do would like the thing to do something usefull, be it a NAS,
mediaplayer, or an android box or whatever.


Are there still linux distro's that officially support the pandaboard?


(I did try archlinux about a year ago which worked, ... untill you
pulled a lot of traffic over it and then the ethernet-interface froze. :-( )



Any advice on still using the pandaboard beside a (very expensive) paper
holder?



Cheerio! Kr. Bonne.

[H. Nikolaus Schaller]

Hi,

we use Debian Jessie and this U-Boot / kernel (and maintain on low priority):
but on PandaBoard ES. If you can find out for us what has to be changed (especially
for booting) we can add Pandaboard (non-ES).

http://download.goldelico.com/letux-kernel/letux-4.7-rc7/
http://download.goldelico.com/letux-u-boot/PandaBoard-ES/
http://download.goldelico.com/letux-debian-rootfs/

A bootable SD card can be pulled together by http://projects.goldelico.com/p/gta04-makesd/

As soon as it works you can extend or replace Debian with something different.

-- hns

[Han Hartgers]

Dear Kristoff,

Happy new year!
Did you find a good working distribution for the Pandaboard?  I have a Pandaboard A1 and I am also looking  for a good (modern) distribution for my board to use it as scan and print server.

The things Robert Nelson is doing at:  https://eewiki.net/display/linuxonarm/PandaBoard is quite good. However mainline u-boot (16.11) and kernel (4.9x) are not configuring "things" correctly any more. Therefore is at the on-board Wifi interface not working at this moment. I am currently wondering if I should try to get the things or Robert Nelson working or if I should look for a different distribution.
Therefore I wanted to know if you have find a good working distribution except for the old ubuntu 12.04.

Greetings,

Han
 

[Robert Nelson]

Other then testing the display in v4.9.x. I haven't done really much
with the panda's anymore (they aren't the backbone of my build farm
anymore, wandboard's have replaced them).

For wifi on the A1, i thing we are missing a 32khz clock somewhere, as
back in the v3.7.x era (pre-device tree's), my A1's needed this:

https://github.com/RobertCNelson/stable-kernel/commit/2b19f1fd3b53f3a3519a2809cff59735361bbd80

Regards,

--
Robert Nelson
https://rcn-ee.com/

[Han Hartgers]

Dear Robert,

Thanks for the pointer to the 3.7.x kernels. I have compiled and tried but that was also not a big success.
(I did "get clone https://github.com/RobertCNelson/stable-kernel ; git checkout origin origin/v3.7.x ; needed to adapt tools/host_det.sh to support ubuntu 16.04 and ./build_kernel.sh)

The kernel boots but cannot find the root file system, it stops with "[    2.194671] Waiting for root device /dev/mmcblk0p1..."
It can be that I look for the wrong root device (did that change?) but I see quite some errors relating with "mmc" on the serial terminal so I think that this kernel cannot start the mmc system correctly. 
[    1.731323] omap_hsmmc 4809c000.mmc: unable to select pin group
[    1.737640] omap_hsmmc 4809c000.mmc: Failed to get debounce clk
[    1.743896] /ocp/mmc@4809c000: can't find DMA channel
[    1.749206] omap_hsmmc 4809c000.mmc: unable to obtain RX DMA engine channel 62
[    1.756866] omap_hsmmc 480d5000.mmc: unable to select pin group
[    1.763153] omap_hsmmc 480d5000.mmc: Failed to get debounce clk
[    1.769378] /ocp/mmc@480d5000: can't find DMA channel
[    1.774719] omap_hsmmc 480d5000.mmc: unable to obtain RX DMA engine channel 60
Giving other errors in the attached log of the serial console it nearly looks if the board is not set-up as described in the device tree. Did there change something that I should "communicate" through u-boot?
Probably I have to look for an u-boot out of the same era and not using the 2016.11 one. Do you have in your git a pointer for that as well?

An other route, i am looking into, is to use the patches mentioned here https://bugzilla.kernel.org/show_bug.cgi?id=119331 with some effort I get this compiled for kernel 4.9 but no real success. I have the suspicion that I have to adapt the dts file as well or some hook in the patch is not valid and the init/activation of the CLK32GK  is therefore not done.

It keeps me busy when the children are in bed.... (and I only wanted to make the copy button on the scanner working for my wife. ;-))

It is a pity that TI had abandoned this OMAP4.... it was a very promising board. The beagle-family and panda made the arm boards more accessible for hobby developers like me.  
I like the wandboard as well, but does your kernels support the VPU? 

Kind regards,

Han

[Robert Nelson]

> Dear Robert,
>
> Thanks for the pointer to the 3.7.x kernels. I have compiled and tried but
> that was also not a big success.
> (I did "get clone https://github.com/RobertCNelson/stable-kernel ; git
> checkout origin origin/v3.7.x ; needed to adapt tools/host_det.sh to support
> ubuntu 16.04 and ./build_kernel.sh)

yeah, it's been a few years for the branch, hence why i just linked to
the specific patch ;)

> The kernel boots but cannot find the root file system, it stops with "[
> 2.194671] Waiting for root device /dev/mmcblk0p1..."
> It can be that I look for the wrong root device (did that change?) but I see
> quite some errors relating with "mmc" on the serial terminal so I think that
> this kernel cannot start the mmc system correctly.
> [ 1.731323] omap_hsmmc 4809c000.mmc: unable to select pin group
> [ 1.737640] omap_hsmmc 4809c000.mmc: Failed to get debounce clk
> [ 1.743896] /ocp/mmc@4809c000: can't find DMA channel
> [ 1.749206] omap_hsmmc 4809c000.mmc: unable to obtain RX DMA engine
> channel 62
> [ 1.756866] omap_hsmmc 480d5000.mmc: unable to select pin group
> [ 1.763153] omap_hsmmc 480d5000.mmc: Failed to get debounce clk
> [ 1.769378] /ocp/mmc@480d5000: can't find DMA channel
> [ 1.774719] omap_hsmmc 480d5000.mmc: unable to obtain RX DMA engine
> channel 60

well v3.7.x is "pre-dt" for omap4, so boot the board file... It was
really till 3.10/v3.11 did the omap4 becomes "useable" with dt..

> Giving other errors in the attached log of the serial console it nearly
> looks if the board is not set-up as described in the device tree. Did there
> change something that I should "communicate" through u-boot?
> Probably I have to look for an u-boot out of the same era and not using the
> 2016.11 one. Do you have in your git a pointer for that as well?
>
> An other route, i am looking into, is to use the patches mentioned here
> https://bugzilla.kernel.org/show_bug.cgi?id=119331 with some effort I get
> this compiled for kernel 4.9 but no real success. I have the suspicion that
> I have to adapt the dts file as well or some hook in the patch is not valid
> and the init/activation of the CLK32GK is therefore not done.
>
> It keeps me busy when the children are in bed.... (and I only wanted to make
> the copy button on the scanner working for my wife. ;-))
>
> It is a pity that TI had abandoned this OMAP4.... it was a very promising
> board.

Well it was abandoned in late 2012

http://www.eetimes.com/document.asp?doc_id=1262782

4 years later, most of it still works.. ;)

> The beagle-family and panda made the arm boards more accessible for
> hobby developers like me.
> I like the wandboard as well, but does your kernels support the VPU?

wandboard has real opensource 2d/3d acceleration with etnaviv

Just follow:
https://eewiki.net/display/linuxonarm/Wandboard

and use v4.9.x+

Regards.

[Han Hartgers]

On Thursday, January 5, 2017 at 10:49:56 PM UTC+1, Robert Nelson wrote:

> Dear Robert,
>
> Thanks for the pointer to the 3.7.x kernels. I have compiled and tried but
> that was also not a big success.
> (I did "get clone https://github.com/RobertCNelson/stable-kernel ; git
> checkout origin origin/v3.7.x ; needed to adapt tools/host_det.sh to support
> ubuntu 16.04 and ./build_kernel.sh)

yeah, it's been a few years for the branch, hence why i just linked to
the specific patch ;)

> The kernel boots but cannot find the root file system, it stops with "[
> 2.194671] Waiting for root device /dev/mmcblk0p1..."
> It can be that I look for the wrong root device (did that change?) but I see
> quite some errors relating with "mmc" on the serial terminal so I think that
> this kernel cannot start the mmc system correctly.
> [    1.731323] omap_hsmmc 4809c000.mmc: unable to select pin group
> [    1.737640] omap_hsmmc 4809c000.mmc: Failed to get debounce clk
> [    1.743896] /ocp/mmc@4809c000: can't find DMA channel
> [    1.749206] omap_hsmmc 4809c000.mmc: unable to obtain RX DMA engine
> channel 62
> [    1.756866] omap_hsmmc 480d5000.mmc: unable to select pin group
> [    1.763153] omap_hsmmc 480d5000.mmc: Failed to get debounce clk
> [    1.769378] /ocp/mmc@480d5000: can't find DMA channel
> [    1.774719] omap_hsmmc 480d5000.mmc: unable to obtain RX DMA engine
> channel 60

It is a kind of a mix to me. I do have a omap4-panda.dtb file with this kernel. Probably is the contents not so "complete" or is only u-boot using this? (Loading the omap4-pandaES.dtb did not change the behaviour of my kernel booting...)
Any thought on what this waiting for root device /dev/mmcblk0p1 can be?

 

well v3.7.x is "pre-dt" for omap4, so boot the board file... It was
really till 3.10/v3.11 did the omap4 becomes "useable" with dt..

> Giving other errors in the attached log of the serial console it nearly
> looks if the board is not set-up as described in the device tree. Did there
> change something that I should "communicate" through u-boot?
> Probably I have to look for an u-boot out of the same era and not using the
> 2016.11 one. Do you have in your git a pointer for that as well?
>
> An other route, i am looking into, is to use the patches mentioned here
> https://bugzilla.kernel.org/show_bug.cgi?id=119331 with some effort I get
> this compiled for kernel 4.9 but no real success. I have the suspicion that
> I have to adapt the dts file as well or some hook in the patch is not valid
> and the init/activation of the CLK32GK  is therefore not done.
>
> It keeps me busy when the children are in bed.... (and I only wanted to make
> the copy button on the scanner working for my wife. ;-))
>
> It is a pity that TI had abandoned this OMAP4.... it was a very promising
> board.

Well it was abandoned in late 2012

http://www.eetimes.com/document.asp?doc_id=1262782

4 years later, most of it still works.. ;)

> The beagle-family and panda made the arm boards more accessible for
> hobby developers like me.
> I like the wandboard as well, but does your kernels support the VPU?

 

I don't need the 2d/3d acceleration but I need the hardware h264 encoder/decoder as can used with gstreamer, that was a few years back a bit of a hit or miss with which kernel I used.

[Robert Nelson]

> It is a kind of a mix to me. I do have a omap4-panda.dtb file with this
> kernel. Probably is the contents not so "complete" or is only u-boot using
> this? (Loading the omap4-pandaES.dtb did not change the behaviour of my
> kernel booting...)
> Any thought on what this waiting for root device /dev/mmcblk0p1 can be?

Yeah, like i said previously, v3.7.x is "pre-dt" for omap4, so STOP
using "omap4-panda.dtb/omap4-pandaES.dtb" and just use the classic
board file..

> I don't need the 2d/3d acceleration but I need the hardware h264
> encoder/decoder as can used with gstreamer, that was a few years back a bit
> of a hit or miss with which kernel I used.

Well someone needs to reverse-engineer that first..

Regards,

[Han Hartgers]

On Thursday, January 5, 2017 at 11:23:11 PM UTC+1, Robert Nelson wrote:

> It is a kind of a mix to me. I do have a omap4-panda.dtb file with this
> kernel. Probably is the contents not so "complete" or is only u-boot using
> this? (Loading the omap4-pandaES.dtb did not change the behaviour of my
> kernel booting...)
> Any thought on what this waiting for root device /dev/mmcblk0p1 can be?

Yeah, like i said previously, v3.7.x is "pre-dt" for omap4, so STOP
using "omap4-panda.dtb/omap4-pandaES.dtb" and just use the classic
board file..

I do not fully understand what you mean here.
I (or actually your build script) is doing a  make ARCH=arm CROSS_COMPILE=omap2plus_defconfig  is that the classic board file way-of-work you point to?
To my surprise I do get as well: am335x-bone.dtb    am335x-evm.dtb   omap3-beagle-xm.dtb  omap3-tobi.dtb   omap4-pandaES.dtb  omap4-var_som.dtb
am335x-bonelt.dtb  omap2420-h4.dtb  omap3-evm.dtb        omap4-panda.dtb  omap4-sdp.dtb      omap5-evm.dtb (of which I am interested in only hte omap4-pandaboard.dtb)
I had not expected these and I assumed that they are ignored by the kernel??? Or is the kernel ignoring the board file data when it gets a (wrong) device tree data as well?

If the last part is true then I have to look into u-boot not giving device tree data to the 3.7.x kernel


Greetings,

Han


 

[Robert Nelson]

No, you don't get it.. Device Tree's where not created over night.
There was a long transition period, as each and every driver had to be
converted.

board file boot -> device tree boot.

v3.7.x is in the early/middle section of that..

Looking at your boot log:

******************
Running uname_boot ...
loading /boot/vmlinuz-3.7.10-x13 ...
3419792 bytes read in 209 ms (15.6 MiB/s)
loading /boot/dtbs/3.7.10-x13/omap4-panda.dtb ...
11810 bytes read in 38 ms (302.7 KiB/s)
debug: [console=ttyO2,115200n8 root=/dev/mmcblk0p1 ro rootfstype=ext4
rootwait] ...
debug: [bootz 0x82000000 - 0x88000000] ...
## Flattened Device Tree blob at 88000000
Booting using the fdt blob at 0x88000000
reserving fdt memory region: addr=9d000000 size=3000000
Loading Device Tree to 8fffa000, end 8ffffe21 ... OK

Starting kernel ...
******************

You want it to look like:

******************
Running uname_boot ...
loading /boot/vmlinuz-3.7.10-x13 ...
3419792 bytes read in 209 ms (15.6 MiB/s)
debug: [console=ttyO2,115200n8 root=/dev/mmcblk0p1 ro rootfstype=ext4
rootwait] ...
debug: [bootz 0x82000000] ...

Starting kernel ...
******************

See how the device tree is NOT used? That's the old board-file boot.

[stevesan...@gmail.com]

I'm using OpenWrt - Linux 4.4.14:

root@OpenWrt:/# uname -a
Linux OpenWrt 4.4.14 #84 SMP Sun Dec 18 19:29:30 UTC 2016 armv7l GNU/Linux

PandaBoard ES rev. B1

The only thing that's not working is audio...

Steve

 

 

[Robert Nelson]

> I'm using OpenWrt - Linux 4.4.14:
>
> root@OpenWrt:/# uname -a
> Linux OpenWrt 4.4.14 #84 SMP Sun Dec 18 19:29:30 UTC 2016 armv7l GNU/Linux
>
> PandaBoard ES rev. B1

Wrong board, wifi works fine on the ES's on mainline..

It's the older non-ES original panda that had wifi clock issues..

[Han Hartgers]

OK now I understand. With the device tree available are the board-file settings overwritten with a incomplete device tree data. I had expected that they would be more additive construction so that there is more smooth transition.

I deleted the device tree data from the file system and kernel 3.7.10 is booting.

Wifi works also a sort of; CLK32KG is there and the WL12xx driver loads. I can connect to an AP but I have problems with a "beacon loss" and that makes the connection not so reliable.  This was also what I now start to remember from my experiences in  2011/2012...

I am probably better off by staying at kernel 4.9.x and  soldering an additional USB connector to add a USB wlan-stick. It is unsatisfying; I will have to leave the on board module for what it is. (I can give it a try if I understand why the old 2014 patches do not enable the CLK32KG signal at a later time.)
By the way can it be that the cpu govenor is not working properly in kernel 4.9.x?
I see the following in my logs:
[   15.748077] cpu cpu0: dev_pm_opp_set_rate: Regulator not registered with OPP core
[   15.755767] cpufreq: __target_index: Failed to change cpu frequency: -22
[ 1092.638519] cpu cpu0: dev_pm_opp_set_rate: Regulator not registered with OPP core
[ 1092.646636] cpufreq: __target_index: Failed to change cpu frequency: -22

And when I remove the heatsink from the processor it crashes with a thermal issue...
(I am not 100% certain about the crash because I was also measuring the power rails and clks to the wifi module; so maybe I have loaded some signal too much.)
 
Thanks for the support and much respect the overall very good work you are presenting at the eewiki site!   May I ask if you are employee of Digikey so that you can do this or is this just your hobby?