Panda How to kernel 3 0 rcx

Revision as of 19:06, 13 June 2011 by KenMcGuire (talk | contribs)
Jump to: navigation, search


The 3.0-rc1 kernel has been released.


There have been regressions since 2.6.39.

The kernel hangs waiting for the sd/mmc card.

Enabling USB results in very ugly kernel oops.

Something is amuck with the wl21xx support.


Looks like Tony Lindgren's pull request to Linus missed the release of -rc2 by a couple of hours, so no fixes got into -rc2. We will have to wait a week or so for -rc3, an expanded mmc fix for the issue is included.


Well, -rc3 is out and the mmc issue is fixed, but the revert of commit 7e6502d577106fb5b202bbaac64c5f1b065e6daa is still necessary to let the kernel boot with ehci enabled, however something else has broken as the USB ethernet is not functional and DVI doesn't work either.

more to come......

A fix is available for the kernel oops by reverting commit 7e6502d577106fb5b202bbaac64c5f1b065e6daa. See for details.

git revert 7e6502d577106fb5b202bbaac64c5f1b065e6daa

Or use the parch from this post: This will be required until the hwmod code is fixed.

the following messages are still present during bootup:

[    0.243896] machine_constraints_voltage: VUSB: failed to apply 3300000uV constraint
[    0.245208] twl_reg twl_reg.46: can't register VUSB, -22
[    0.245239] twl_reg: probe of twl_reg.46 failed with error -22

and happens whether the EHCI driver is selected in the .config or not.

The mmc patch is still required, as are the dvi patch (if you want anything other than 640x480) and the i2c patch.


Building 3.0-rc1, or -rc2 is basically the same as How to build 2.6.38 kernel, but of course you need the -rc1/2 sources. The omap2plus_defconfig works once the mmc patch is applied. config.3.0-rc2.1 works for the reverted commit, mmc patch, dvi patch and i2c patch.

MMC Patch

A patch is available which addresses the mmc issue noted above:


Apply it like so: (from inside the kernel sources directory)

patch -p1 < 0002-omap4-pandaboard-fix-mmc-support.patch

The mmc issue has been fixed in linux-omap-2.6.git, and should be in -rc3 but the oopses are still present when EHCI is enabled in .config, unless the above commit is reverted..

There is also an issue with the setting up of the wl12xx in -rc1/2:

[    0.181457] error setting wl12xx data

Which is a problem new to 3.0-rc1/2, still present in linux-omap-2.6.git, and although the error is not present in the reverted -rc2, the wlan does not appear to work.

DVI Patch

This patch is no longer strictly necessary, but without it the dvi output is 640x480. Applying it doesn't seem to hurt though.

Apply it like so: (from inside the kernel sources directory)

patch -p1 < 0001-omap4-pandaboard-fix-dvi-support.patch

Then use either of the following configs

make ARCH=arm omap2plus_defconfig
make ARCH=arm panda_dvi_defconfig

Then compile with

make ARCH=arm CROSS_COMPILE=Path_to_your/arm-2009q3/bin/arm-none-linux-gnueabi- uImage
make ARCH=arm CROSS_COMPILE=Path_to_your/arm-2010q1/bin/arm-none-linux-gnueabi- uImage

i2c character device driver

Ever since 2.6.38, the i2c character device driver support from user space has been broken for OMAP44xx.
* A fix has been submitted for the linux-omap-2.6 branch, but it may be a while till it gets into an -rcx.
See;a=commit;h=39fe1a6fafe1e85c183379af9f3ceda7cd24bd65 for the commit.
* A quick fix for this issue is available File:Omap44xx-i2c-fix.patch

I2C Patch

Apply it like so: (from inside the kernel sources directory)

patch -p1 < Omap44xx-i2c-fix.patch

It applies with some fuzzes and offsets, but does apply successfully. Then compile as above.


fbtest on DVI Port

After booting run fbtest to see a nice test pattern from the dvi port.


Switching primary display to the HDMI port

# Enable HDMI
echo "1" > /sys/devices/platform/omapdss/display1/enabled

# Disable overlay0 (an overlay must be disabled before changing its properties)
echo "0" > /sys/devices/platform/omapdss/overlay0/enabled

# Set the manager of overlay0 to display1 which is HDMI
echo "tv" > /sys/devices/platform/omapdss/overlay0/manager

# Enable overlay0
echo "1" > /sys/devices/platform/omapdss/overlay0/enabled

And content on overlay 0 of primary lcd would be transferred to HDMI. It works similarly for all other overlay's.

Switching primary display to the DVI port

See: for lots of useful info on the display subsystem. Be aware that the display, manager and overlay numbers don't match the panda configuration.

# Disable HDMI
echo "0" > /sys/devices/platform/omapdss/display1/enabled

# Disable overlay0 (an overlay must be disabled before changing its properties)
echo "0" > /sys/devices/platform/omapdss/overlay0/enabled

# Set the manager of overlay0 to display0 which is DVI
echo "lcd2" > /sys/devices/platform/omapdss/overlay0/manager

# Enable overlay0
echo "1" > /sys/devices/platform/omapdss/overlay0/enabled

The above commands should be run from the serial console and the cable should be in the destination port before running the commands.

fbtest on HDMI Port

Run fbtest to see a nice test pattern from the HDMI port.



You can run i2cdetect and the results should look like this:

# i2cdetect -y -r 1
    0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          -- -- -- -- -- -- -- -- -- -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- -- -- -- -- UU UU UU UU -- -- -- -- 
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
70: -- -- -- -- -- -- -- --

Stay tuned as the 3.0-rcx series evolves, where there needs to be some serious bug squashing.