3 proposed patches for next 3.8.13-bone5x update

GoogleGroups
1

I’ve attached three patches for consideration in the Robert Nelson’s 3.8.13-bone5x update, all of which are independent of each other (i.e., you can patch your kernel with any one of them, separate from the others):

0001-am335x-features

Detect AM335x-specific features and CPU version. Doesn’t do anything significant, other than accurately report the CPU and SGX, L2 cache presence in the bootup dmesg output.

0002-element14-bb-view-lcd-capes

Add Element14’s “BB-VIEW” LCD cape device trees to firmware/capes.

0003-sitara_rb_swap_workaround

Create a workaround to the TI Sitara red/blue swap erratum through device tree properties, ‘bgrx_16bpp’ and ‘bgrx_24bpp’. The ‘bgrx_16bpp’ property swaps red and blue if 16bpp color depth is used and the LCD cape itself swaps red and blue at higher color depths (i.e., the cape designer “fixed” the problem by swapping signals). The ‘bgrx_24bpp’ property swaps red and blue at the 24bpp color depth, which addresses TI’s erratum.

Also, the tilcdc LCD driver queries the panel’s panel-info/bpp device tree property to find the preferred BPP when initializing the console framebuffer. This was originally hardcoded into the driver at 16bpp. If the panel says that it wants 32bpp, the framebuffer initializes to a preferred 32bpp.

I’ve been testing these changes relative to Robert’s linux-dev 3.8.15-bone53 tag. They may apply cleanly against earlier tags, but I can’t guarantee they will.

I’m sure I have canonical kernel source formatting issues; comments, testing and suggestions are welcome (and advocacy for inclusion in Robert’s linux-dev git repo also helpful.)

-scooter

0001-am335x-features.patch (3.31 KB)

0002-element14-bb-view-lcd-capes.patch (17.6 KB)

0003-sitara_rb_swap_workaround.patch (26.3 KB)

2

Thanks! They are all queue'd up:

https://github.com/RobertCNelson/bb-kernel/commit/5e50ae5c219b52bf70193bffdef7c07c8b26b90f

Just have to run a few checks on the lcd3/lcd7 as this did revert the
custom touchscreen filtering.

I'm "pretty" sure we fixed the adc issue in:
https://github.com/RobertCNelson/bb-kernel/commit/3a66618d3bbd86a4e7655e07bc48838fb967d5ed

So it shouldn't be needed anymore now..

Regards,

3

Wow. That was fast. My e-mail is in the patches, so… I’m sure the bug reports will trickle in…

-scooter

4

The sitara and/or bb-view patches are causing problems for my LCD4 based 4D Systems LCD cape.
http://www.4dsystems.com.au/product/4DCAPE_43/

The image is squished to the left, with blank space on the right, and the colors are messed up. When I remove the patch in the most recent kernel (bone62), the problem with the screen goes away.

Do you know what the problem might be? I’ve spent two days trying to find and fix this issue. At this point, I’m planning to run our system with a kernel sans your patch, but I wouldn’t be surprised if there are others out there with the same issue.

5

I should add that the patch causes the fbdev graphics driver to fail (squished image, blue colors). The modesetting driver works fine. We have to use fbdev because modesetting doesn’t allow screen rotation.

Do I need to change the device tree property to get this screen to work? I would prefer to use your stock kernel. Any guidance would be really appreciated. It’s been humbling to see all of the work that y’all have put in to this.

Alex

6

I should add that with the patch applied, the modesetting driver works fine, but fbdev has the squished image/mangled colors. In my application, I can’t use modesetting because we need to rotate the screen.

It’s been humbling to see all of the work y’all have put in to this project. I’d appreciate any insight you might have. Do I need to modify a dts file?

Best,
Alex

7

Hi Alexander,

There was some driver level changes, i yanked the two offending
patches on v3.8.x branch. My lcd4 is working now.

Regards,

8

Alexander:

I apologize for the delayed response – I’ve been on travel for a customer with more customers keeping me tied up in meetings yesterday and today.

I was going to ask which DTB you were using because your description sounded as if pixel timings were off. Sounds like Robert found the issue (were they my patches??)

-scooter

9

My suspicion, it might be a compiler bug..

0009-sitara_red_blue_swap_workaround.patch

We are dealing with:

gcc version 4.6.3 (Debian 4.6.3-14)

I just disabled them for the time being..

https://github.com/RobertCNelson/bb-kernel/commit/59ba8323a910c7c3c032ee455d534ad43fba07f8

Regards,

10

Robert:

I don’t have the 4" LCD cape, so I really have no way of testing on the smaller screen. I could break the patch into individual file patches or more atomic pieces to isolate the offender, if there were someone willing to test on my behalf.

-scooter

11

Robert:

I'll try to get my hands on the Element-14 4" LCD to see if it happens with that LCD. I have the same company's 7" LCD cape -- should be able to narrow it down to either the board DTB or a compiler bug.

-scooter

12

I can arrange to have one of the 4D Systems 4" LCD's sent to you. If
that would help. I could also do the kernel testing here to help
isolate the offender. I should be able to apply the patches one at a
time and do a quick partial recompile after each patch is applied?

To answer your question about the DTB, I was just using whatever DTB's
are used in the Debian build generated via Robert Nelson's
omap-image-builder.

Alex

13

Alexander:

Short answer: It works for me. :slight_smile:

It turns out that a colleague has an Element-14 4.3" LCD cape here at work. I just booted up after changing uEnv.txt. Hard to work on the BBB at that screen resolution, but I don’t have the image compression that you experienced.

I’m not at the most recent tag on Robert’s tree, but I could test it later this week. But given that the E-14 4.3" LCD cape works, it makes me wonder if the problem isn’t with the DTB’s pixel and dot clocking params?

Now, I’ll admit that I compiled my kernel with the 4.7.3 20130328 prerelease compiler, so that could have made a difference if there were a compiler bug:

[ 0.000000] Linux version 3.8.13-bone53 (-----@------------) (gcc version 4.7.3 20130328 (prerelease) (crosstool-NG linaro-1.13.1-4.7-2013.04-20130415 - Linaro GCC 2013.04) ) #40 SMP Fri May 30 13:18:49 PDT 2014
[ 0.000000] CPU: ARMv7 Processor [413fc082] revision 2 (ARMv7), cr=50c5387d
[ 0.000000] CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
[ 0.000000] Machine: Generic AM33XX (Flattened Device Tree), model: TI AM335x BeagleBone
[ 0.000000] Memory policy: ECC disabled, Data cache writeback
[ 0.000000] On node 0 totalpages: 130816
[ 0.000000] free_area_init_node: node 0, pgdat c0880640, node_mem_map c08fb000
[ 0.000000] Normal zone: 1024 pages used for memmap
[ 0.000000] Normal zone: 0 pages reserved
[ 0.000000] Normal zone: 129792 pages, LIFO batch:31
[ 0.000000] AM335X ES2.1 (l2cache sgx neon )
[ 0.000000] PERCPU: Embedded 9 pages/cpu @c0d0b000 s14080 r8192 d14592 u36864
[ 0.000000] pcpu-alloc: s14080 r8192 d14592 u36864 alloc=9*4096
[ 0.000000] pcpu-alloc: [0] 0
[ 0.000000] Built 1 zonelists in Zone order, mobility grouping on. Total pages: 129792
[ 0.000000] Kernel command line: console=tty0 console=ttyO0,115200n8 drm.debug=0x05 capemgr.disable_partno=BB-BONELT-HDMI,BB-BONELT-HDMIN capemgr.enable_partno=BB-VIEW-LCD4-01 root=/dev/mmcblk0p2 ro rootfstype=ext4 rootwait fixrtc quiet init=/lib/systemd/systemd

I’m also using fbdev for the X server:

[ 11.250] X Protocol Version 11, Revision 0
[ 11.251] Build Operating System: Linux 3.2.0-4-mx5 armv7l Debian
[ 11.251] Current Operating System: Linux beaglebone 3.8.13-bone53 #40 SMP Fri May 30 13:18:49 PDT 2014 armv7l
[ 11.253] Kernel command line: console=tty0 console=ttyO0,115200n8 drm.debug=0x05 capemgr.disable_partno=BB-BONELT-HDMI,BB-BONELT-HDMIN capemgr.enable_partno=BB-VIEW-LCD4-01 root=/dev/mmcblk0p2 ro rootfstype=ext4 rootwait fixrtc quiet init=/lib/systemd/systemd
[ 11.256] Build Date: 17 December 2013 08:51:06PM
[ 11.256] xorg-server 2:1.12.4-6+deb7u2 (Julien Cristau <jcristau@debian.org>)
[ 11.256] Current version of pixman: 0.26.0
[ 11.258] Before reporting problems, check http://wiki.x.org
to make sure that you have the latest version.
[ 11.258] Markers: (–) probed, () from config file, (==) default setting,
(++) from command line, (!!) notice, (II) informational,
(WW) warning, (EE) error, (NI) not implemented, (??) unknown.
[ 11.264] (==) Log file: “/var/log/Xorg.0.log”, Time: Wed Apr 23 13:20:01 2014
[ 11.292] (==) Using config file: “/etc/X11/xorg.conf”
[ 11.293] (==) Using system config directory “/usr/share/X11/xorg.conf.d”
[ 11.312] (==) ServerLayout “Builtin Default Layout”
[ 11.313] () |–>Screen “Builtin Default fbdev Screen 0” (0)
[ 11.313] () | |–>Monitor “Builtin Default Monitor”
[ 11.317] () | |–>Device “Builtin Default fbdev Device 0”
[ 11.318] (==) Automatically adding devices

-scooter

14

I am using Robert’s bb-kernel git repo to build the kernel. It looks like we are using the exact same compiler.
Linux version 3.8.13-bone62 (root@debian) (gcc version 4.7.3 20130328 (prerelease) (crosstool-NG linaro-1.13.1-4.7-2013.04-20130415 - Linaro GCC 2013.04) )

I noticed that you are passing specific commands to the capemgr. “capemgr.enable_partno=BB-VIEW-LCD4-01” This I am not doing. Should I be doing this? If I force it to capemgr.enable_partno=BB-BONE-LCD4-01, then maybe it will work even if your patch is installed?

This is what the detection of the screen looks like on my kernel without the patches. Maybe the kernel is getting confused and using BB-VIEW dtb for my LCD instead of BB-BONE dtb?

[ 0.725548] bone-capemgr bone_capemgr.9: Baseboard: ‘A335BNLT,00A5,4110BBBK0000’
[ 0.725572] bone-capemgr bone_capemgr.9: compatible-baseboard=ti,beaglebone-black
[ 0.755349] bone-capemgr bone_capemgr.9: slot #0: No cape found
[ 0.792456] bone-capemgr bone_capemgr.9: slot #1: No cape found
[ 0.829564] bone-capemgr bone_capemgr.9: slot #2: No cape found
[ 0.859739] bone-capemgr bone_capemgr.9: slot #3: ‘4D 4.3 LCD CAPE- 4DCAPE-43T ,00A1,4D SYSTEMS ,BB-BONE-LCD4-01’
[ 0.859839] bone-capemgr bone_capemgr.9: slot #4: specific override
[ 0.859859] bone-capemgr bone_capemgr.9: bone: Using override eeprom data at slot 4
[ 0.859873] bone-capemgr bone_capemgr.9: slot #4: ‘Bone-LT-eMMC-2G,00A0,Texas Instrument,BB-BONE-EMMC-2G’
[ 0.859945] bone-capemgr bone_capemgr.9: slot #5: specific override
[ 0.859963] bone-capemgr bone_capemgr.9: bone: Using override eeprom data at slot 5
[ 0.859977] bone-capemgr bone_capemgr.9: slot #5: ‘Bone-Black-HDMI,00A0,Texas Instrument,BB-BONELT-HDMI’
[ 0.860046] bone-capemgr bone_capemgr.9: slot #6: specific override
[ 0.860063] bone-capemgr bone_capemgr.9: bone: Using override eeprom data at slot 6
[ 0.860077] bone-capemgr bone_capemgr.9: slot #6: ‘Bone-Black-HDMIN,00A0,Texas Instrument,BB-BONELT-HDMIN’
[ 0.860405] bone-capemgr bone_capemgr.9: loader: before slot-3 BB-BONE-LCD4-01:00A1 (prio 0)
[ 0.860420] bone-capemgr bone_capemgr.9: loader: check slot-3 BB-BONE-LCD4-01:00A1 (prio 0)
[ 0.860500] bone-capemgr bone_capemgr.9: loader: before slot-4 BB-BONE-EMMC-2G:00A0 (prio 1)
[ 0.860512] bone-capemgr bone_capemgr.9: loader: check slot-4 BB-BONE-EMMC-2G:00A0 (prio 1)
[ 0.860585] bone-capemgr bone_capemgr.9: loader: before slot-5 BB-BONELT-HDMI:00A0 (prio 1)
[ 0.860598] bone-capemgr bone_capemgr.9: loader: check slot-5 BB-BONELT-HDMI:00A0 (prio 1)
[ 0.860632] bone-capemgr bone_capemgr.9: initialized OK.
[ 0.861126] bone-capemgr bone_capemgr.9: loader: after slot-3 BB-BONE-LCD4-01:00A1 (prio 0)
[ 0.861145] bone-capemgr bone_capemgr.9: slot #3: Requesting part number/version based 'BB-BONE-LCD4-01-00A1.dtbo
[ 0.861167] bone-capemgr bone_capemgr.9: slot #3: Requesting firmware ‘BB-BONE-LCD4-01-00A1.dtbo’ for board-name '4D 4.3 LCD CAPE- 4DCAPE-43T ', version ‘00A1’
[ 0.861186] bone-capemgr bone_capemgr.9: slot #3: dtbo ‘BB-BONE-LCD4-01-00A1.dtbo’ loaded; converting to live tree
[ 0.861686] bone-capemgr bone_capemgr.9: slot #3: #4 overlays

15

I checked the difference between the dts’s. It looks like the only difference that matters is the panel-info { bpp } value. BB-BONE has a bpp of 24. BB-VIEW has a bpp of 32. The display-timings information is all identical.

Would it then make sense that the image is scaled and the colors are corrupted if the 4D Systems screen is using the BB-VIEW dtb file with the slightly higher bpp value?

Alex

16

Alex:

This conversation seemed to have gotten lost in the myriad of e-mails I get per day. Sorry for the delayed reply.

Disable the HDMI interface and enable the 4" LCD, by putting this line into /boot/uboot/uEnv.txt:

optargs=capemgr.disable_partno=BB-BONELT-HDMI,BB-BONELT-HDMIN capemgr.enable_partno=BB-BONE-LCD4-01

If you have an existing “optargs=” line, comment it out with a hash mark, e.g., “#optargs =”.

I don’t think the bpp parameter makes much difference to the frame buffer device or the X server.

Let me know how this work out!

-scooter

17

I don’t currently have time to check whether these patches will work. I’m happy to give you the benefit of the doubt. :slight_smile:

Alex

18

Okay, noticed what may be going on. I've been porting my collection
of lcd's to v3.14.x (which include a 24bit lcd).. On the 24bit
(bbb-exp-c), we need to make sure to:

sudo sed -i -e 's:16:24:g' /etc/X11/xorg.conf

Otherwise the graphics look shifted..

I added the bb-view-43 couldn't get your tilcd_fbdev.c to port.. (i'm
going to add the bb-view-7 tomorrow)

http://elinux.org/Beagleboard:Capes_3.8_to_3.14#bb-view-43

sudo apt-get update
sudo apt-get install linux-image-3.14.19-ti-r23

the lcd4 should also just work with this kernel..

Regards,

19

Thanks Robert. I’ve started using the 3.14 kernel (although the less bleeding edge bone8 version), and I built a custom dtb to enable both the 4Dcape-43t and a secondary non-pin-conflicting uart using your bb-kernel repo. It’s worked beautifully. Apologies to Scott for not being able to follow up on his 3.8 kernel patch.

I ran into issues when using apt to install the ti-rXX kernels. The apt-get install command fails due to single quotes in an sed command in a zz-uenv file. We tried to get it to run by replacing the single quotes around older_kernel and latest_kernel with ’ " ’ " ', but then something broke on reset, and I forget now what errors we were seeing. It’s possible these issues have been resolved recently. I’ll try installing the latest kernel sometime when it isn’t past 3am. :slight_smile:

We’ve been using omap-image-builder, and setup-sdcard.sh exits when a -ti- kernel is specified. This is another reason, albeit possibly misguided, that we shied away from using the -ti- kernels and stuck to the bone kernels. I’m not sure what the difference is?

One more comment about our adventures with omap-image-builder. We added usbmount to the package list in omap-image-builder, and usbmount uses /media/usbX as mount points, so it creates those directories.

The init-eMMC-flasher-v3 excludes /media/* from being rsync’ed so we added the necessary mkdir -p’s after the relevant rsync to fix this.

Thanks again for all of your hard work.

Alex

20

Thanks Robert. I've started using the 3.14 kernel (although the less
bleeding edge bone8 version), and I built a custom dtb to enable both the
4Dcape-43t and a secondary non-pin-conflicting uart using your bb-kernel
repo. It's worked beautifully. Apologies to Scott for not being able to
follow up on his 3.8 kernel patch.

I ran into issues when using apt to install the ti-rXX kernels. The apt-get
install command fails due to single quotes in an sed command in a zz-uenv
file. We tried to get it to run by replacing the single quotes around
older_kernel and latest_kernel with ' " ' " ', but then something broke on
reset, and I forget now what errors we were seeing. It's possible these
issues have been resolved recently. I'll try installing the latest kernel
sometime when it isn't past 3am. :slight_smile:

it took a couple release to get all the bugs out, can you retest with
the git version?

https://raw.githubusercontent.com/RobertCNelson/omap-image-builder/master/target/other/zz-uenv_txt

We've been using omap-image-builder, and setup-sdcard.sh exits when a -ti-
kernel is specified. This is another reason, albeit possibly misguided,
that we shied away from using the -ti- kernels and stuck to the bone
kernels. I'm not sure what the difference is?

I'll hack on this today, i have a bunch of logic for the bone's to
make sure they got the "-bone" kernel at all cost, i'm going to change
that to a "-bone" || "-ti".

One more comment about our adventures with omap-image-builder. We added
usbmount to the package list in omap-image-builder, and usbmount uses
/media/usbX as mount points, so it creates those directories.
The init-eMMC-flasher-v3 excludes /media/* from being rsync'ed so we added
the necessary mkdir -p's after the relevant rsync to fix this.

I think we can safely enable this now. Since the flasher is running at
boot in a single user environment, usbmount shouldn't mount an
external device, so only the base /media/<dir> should be present while
it's flashing. The whole point of disabling /media/ from rsync, when
we ran after the board fully booted, there was a risk the user had a
big full usb drive plugged in and was mounted to /media/, this would
have overfilled the eMMC.

Regards,