Panda How to kernel 3 3 rcx
- 1 Introduction
- 2 rc1
- 3 rc2
- 4 rc3
- 5 rc4
- 6 rc5
- 7 rc6
- 8 rc7
- 9 PandaBoard ES
- 10 wlan12xx
- 11 wlan12xx patch
- 12 I2C Patch
- 13 Building
- 14 Testing
The merge window for 3.3 is over and we are currently at v3.3-rc4. Lots of work on the OMAP platforms and on Pandaboard as well as the wl12xx wlan driver . Let's see what the various release candidates have provided.
You can download a tarball of the mainline kernel at http://kernel.org/
or you can clone a copy of mainline kernel with:
git clone http://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git cd linux git checkout v3.3-rcx
where x is the release candidate version.
Important Note There has been a lot of work done on the HDMI interface and its related driver omapdrm. Between kernel 3.1 and 3.2-rc1 enough changed so that the Testing section on the HDMI interface is no longer correct. The hdmi init functions are no longer in arch/arm/mach_omap2/board-omap4panda.c. When this situation stabilises, the HDMI testing section below will be updated.
The following seems to be only true for the PandaBoard ES. Tests with a PandaBoard EA3 confirm that the patch is still needed
The WLAN no longer requires a patch! The WL12xx driver needs current firmware. When the driver isn't happy, the error messages are somewhat less than useful, however the drivers/firmware are being constantly improved and it would not be a good idea to have the driver support anything but the latest firmware. Still a work in progress.
No PandaBoard changes, and with no patches required this is an easy release to build.
Skipped due to the HDMI changes which cropped up in 3.2-rc1 Lots of changes working towards getting HDMI functional
resolve nebulous 'Error setting wl12xx data' fix to PandaBoard. "This should be fixed properly for the next merge window so we don't issue error messages merely because a driver is not configured."
Devicetree-enabled kernels crash during boot due to the UART driver (http://www.spinics.net/lists/linux-omap/msg64921.html). A patch has been proposed and should be merged before release.
The WL12xx driver still needs the same patch as used for 3.1 and 3.2. when the driver isn't happy, the error messages are somewhat less than useful, however the drivers/firmware are being constantly improved and it would not be a good idea to have the driver support anything but the latest firmware. The race issue that required building as a module has returned and with the modules built in, operation is not consistant.
There is now a PandaBoard ES http://pandaboard.org/content/pandaboard-es which includes an OMAP 4460 at up to 1.2GHz. Several important differences make it important (at the present time) that the MLO/u-boot be specifically crafted for the 4460. The thermal management is not in the mainline 4430 code as yet and therefore the max clock frequency when running the OMAP4460 on the PandaBoard ES with the mainline kernel is 920MHz(same as the OMAP4430).
This patch is still necessary to resolve the issue noted in 3.0, 3.1 and 3.2, the code was moved to twl_common.c.
Apply it like so: (from inside the kernel sources directory)
patch -p1 < 0001a-omap4-pandaboard-wlan-fix.patch
The latest wlan firmware is available from git: git://git.kernel.org/pub/scm/linux/kernel/git/dwmw2/linux-firmware.git just copy the contents of the ti-connectivity folder to /lib/firmware/ti-connectivity.
The patch is no longer necessary for the PandaBoard ES but does not seem to hurt if applied. In addition, the race issue that required building as a module has returned as of -rc2. As part of the code cleanup, the wl12xx and wl12xx_sdio drivers no longer depend on each other. This creates an issue with systems that do not use udev or mdev (with as somewhat fiddly & slow script) to load MODALIAS drivers. The quick solution is to modprobe both, the order no longer matters. Just modprobing wl12xx_sdio will no longer automatically load wl12xx. It is suggested that these 2 drivers be built as modules.
i2c character device driver patch that has been necessary is no longer required, and the issues from kernel 3.2 have been resolved.
Building 3.3-rc1 or -rc2 is fairly straight forward.
The 3.3-rc1 .config enables Sound builtin and wl12xx as modules and the 3.3-rc2 enables Sound and the wl12xx drivers builtin, so the resulting kernel will be a bit bigger than previously. The builtin sound does not presently work, but the enabled configuration allows USB sound devices, which function properly.
Then compile like so:
make ARCH=arm CROSS_COMPILE=Path_to_your/arm-2010q1/bin/arm-none-linux-gnueabi- uImage
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 --> Currently not functional <--
Make sure that a monitor is plugged into the HDMI port before doing the following:
# 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: http://omappedia.org/wiki/Bootargs_for_enabling_display for lots of useful info on the display subsystem. Be aware that the display, manager and overlay numbers don't match the panda configuration.
Make sure that a monitor is plugged into the DVI port before doing the following:
# 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 --> Currently not functional <--
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: -- -- -- -- -- -- -- --
Run the following commands after the PandaBoard is booted:
modprobe wl12xx ** only if you built the wl12xx drivers as modules modprobe wl12xx_sdio ** only if you built the wl12xx drivers as module ifconfig wlan0 up iwconfig wlan0 essid "Your AccessPoint Name" udhcpc -i wlan0
If your network is set up to provide DHCP services, the PandaBoard will get all the "right stuff(tm)" and you will be able to access the Internet.
# ping www.google.com PING www.google.com (126.96.36.199): 56 data bytes 64 bytes from 188.8.131.52: seq=0 ttl=43 time=62.683 ms 64 bytes from 184.108.40.206: seq=1 ttl=43 time=54.077 ms 64 bytes from 220.127.116.11: seq=2 ttl=43 time=51.484 ms 64 bytes from 18.104.22.168: seq=3 ttl=43 time=54.108 ms
Insert a USB memory stick into one of the usb ports
Run dmesg to see what sdx the stick was recognised as, then:
hdparam -tT /dev/sdx
If you did this on a kernel without the patch you will see some improvement, if you run the same command on a desktop Linux system, with the same USB memory stick, the PandaBoard's speed should roughly be the same.