Help New Kernel For Pdaxrom

[pelrun]

It's unlikely that they got in; it's almost certainly changes that conflict with the patches.

As for the touchscreen stuff, X has to interpret the data it gets from the ts device, and the case of the working 2.4 can be one of two things:

1) Kernel ts device outputs sensible data (whatever that means), X reads it sensibly
2) Kernel ts device outputs weird data, X expects the weirdness and reconstructs the correct data.

If the 2.6 ts device has gone from weird to sensible, or sensible to weird, or weird to a different kind of weird, then X is getting something unexpected and isn't interpreting it correctly.

Obviously, you want both sides behaving sensibly, so if the 2.6 ts device is now sensible, the patch should be made to the X code and vice versa.

There was an earlier thread regarding the X ts code in beta3 where there was a somewhat similar problem (and patches for X's tslib.c); that might be a starting point:

https://www.oesf.org/forums/index.php?showtopic=16557&st=0

[louigi600]

Thanks pelrun .... I'll see if I've time for that too although I was hoping not to go into X stuff myself ;-)

Where has this code moved to in 2.6.22:
drivers/input/power.c ?

Apparently it's drivers/input/power.c has been removed altogether.

[Hrw]

2.6.22 is in Poky

[louigi600]

Ok ... lets leave 2.6.22 for a moment.
With the use of 2.6.21.5 (and for that matter this also happened on 2.6.20.6) I've lost the ability to boot from SD/MMC.
Looking at the kernel messages while booting I notice this difference:
2.6.16 initializes SD/MM right after setting the RTC to unixtime + a few seconds ... detects partitions and mounts root.
With 2.6.21.5  i see the kernel panic right after the RTC stuff ... so i looks like it's attempting to access SD/MMC before having initialized it.

As anyone any ideas ?

This is a chunk of dmesg while booting from SD/MMC with 2.6.16

Code: [Select]

sa1100-rtc sa1100-rtc: setting the system clock to 1970-01-01 00:00:09 (9)
mmcblk0: mmc0:a3f6 SD256 249088KiB
 mmcblk0: p1
kjournald starting.  Commit interval 5 seconds
EXT3 FS on mmcblk0p1, internal journal
EXT3-fs: mounted filesystem with ordered data mode.
VFS: Mounted root (ext3 filesystem).
And this is a chunk of dmesg while booting from nand with 2.6.21.5 (notice when mmc is initialized):

Code: [Select]

sa1100-rtc sa1100-rtc: setting the system clock to 1970-01-01 00:00:09 (9)
VFS: Mounted root (jffs2 filesystem).
Freeing init memory: 76K
mmcblk0: mmc0:a3f6 SD256 249088KiB
 mmcblk0: p1
Here's the panic while attempting to boot from SD:

Code: [Select]

sa1100-rtc sa1100-rtc: setting the system clock to 1970-01-01 00:00:09 (9)
VFS: Cannot open root device "mmcblk0p1" or unknown-block(0,0)
Please append a correct "root=" boot option
Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0)
I've attached the full dmesg for all the 3 cases above.

[Civil]

I've tried to apply CFS patches (Compleatly Fair Scheduler - new CPU shceduler, that will be included in kernel 2.6.23)... It acts very-very strange... Some times it says "Oops" on startup, but if it doesn't a boot to login prompt took a lot of time - for me it was around 12h(!) to boot...

[InSearchOf]

I've tried to apply CFS patches (Compleatly Fair Scheduler - new CPU shceduler, that will be included in kernel 2.6.23)... It acts very-very strange... Some times it says "Oops" on startup, but if it doesn't a boot to login prompt took a lot of time - for me it was around 12h(!) to boot...
[div align=\"right\"][a href=\"index.php?act=findpost&pid=164772\"][{POST_SNAPBACK}][/a][/div]

12hr boot is better then No boot  

Late

[louigi600]

I've got 2.6.22 patched and booting with my husky but:
none of the issues I had before were resolved,
sd/mmc has severe problems across suspend/resume,
leaving the device suspended for some time will drain the battery (I need the external power to boot),
reflashing the original kernel after the 2.6.22 was used has given me a corrupt root filesystem (I guess some incomatibilities between nandlogical and 2.6.22).

[louigi600]

My work on kernel can be gotten from here
username=anonymous and blank password.
If you want to use the buildscript you will also need master config file and common functions.

I'm still lost about the odd behavour in X ... anyone have any ideas ?
[div align=\"right\"][{POST_SNAPBACK}][/a][/div]

The 2.6.21.5 kernel I was talking about above can be built with [a href=\"http://www.codesourcery.com/gnu_toolchains/arm/download.html]Code Sourcery[/url] xross toolchain.
I've not had the time yet to test if it actuallu boots and works normally but I was able to build it.

In order to use the EABI toolchain use this script:

Code: [Select]

#!/bin/bash
. /etc/profile
CROSS=arm-none-eabi-
CROSS_COMPILE=$CROSS
ARCH=arm
PATH=/arm-2007q1/bin:$PATH

export ARCH CROSS CROSS_COMPILE CFLAGS PATH

echo "Type exit for leave eabi cross environment."
/bin/bash

[psycoman]

why note use kernel build in OE in pdaXrom distro ?

[psycoman]

just a notice:


im boot pdaXii13 with angstrom last kernel i have some problems like modules not instaled and detection system of pdaxii13


maybe with some fixex can be possible.

[louigi600]

just a notice:

im boot pdaXii13 with angstrom last kernel i have some problems like modules not instaled and detection system of pdaxii13

maybe with some fixex can be possible.
[div align=\"right\"][a href=\"index.php?act=findpost&pid=165116\"][{POST_SNAPBACK}][/a][/div]

Angstrom kernel is 2.6 while pdaxii13 is 2.4 ... I think a lot of things will brake.
If you insist you can copy the modules form angstrom's /lib/modules/2.6.* to your /lib/modules/

[psycoman]

i know louigi600, i think the better usage for this kernel is with pdaXrom 198+, i make this test just to check the possibility, i do not use it in pdaxii13.

but, why you try make a kernel with this already exist ? is a posibillity to you use angstrom kernel and why not, develop in ?

[louigi600]

i know louigi600, i think the better usage for this kernel is with pdaXrom 198+, i make this test just to check the possibility, i do not use it in pdaxii13.

;-)

but, why you try make a kernel with this already exist ?

Angstrom kernel is built with OE (and maybe some Angtrom specific patches).
A kernel thus built will brake some things in pdaXrom (like u-boot parameters and zaurushw).
I'm not targetting primarely Angstrom but pdaXrom so my goal would be to benefit from all that has been fixed since 2.6.16 + all the pdaXrom specific stuff.
I was trying EABI toolchain for personal intrest so I attempted to build an EABI capable kernel.

is a posibillity to you use angstrom kernel and why not, develop in ?

I'm not developing anything but just collecting fixtes from here and ther, adjusting them so that they apply to the kernel I'm trying to use and testing if the result is any better that the kernel that comes with r197/198.

Also whether pdaXrom kernel developement should or should not be merged with OE kernel developement is not a matter of discussion here but I'm very happy and gratefull that help has come from people involved on many distributions.

A lot more things could be said about this but in short I hope that this will contribute positively to:
newer kernel for pdaXrom (with more things functional and less problems),
a cooperative attitude between all the various distribution developers (at least for kernel developement) and
have yet another option for people wishing to dual boot different distributions.

[psycoman]

well,

   i make some testes with pdaxrom 198 and angstrom kernel, some points

1 - kernel work with minor bugs
2 - USB network work like a charm =]
2 - i extract the jffs2 form 198  and put in hda1 so i boot form hd no flash used =/
3 - my unique problem is compiler diferences: examples
    alsa utilities cannot comunicate via ioctl
ERROR: alsamixer: function snd_mixer_load failed: Inappropriate ioctl for device

maybe if pdaxrom gcc 4.x and glibc 2.3.x, can be possible do it, but need recompile many things

maybe use OE kernel as a central kernel to emebebed devices is the best choice i think...

Also whether pdaXrom kernel developement should or should not be merged with OE kernel developement is not a matter of discussion here but I'm very happy and gratefull that help has come from people involved on many distributions.

my idea is always have a updated kernel, like a 2.6.23 with CFS and other, and if a "central kernel" (not especific the OE kernel) can be used, we can focus on distro develop like important softwares..


btw,
 
   what stage are you kernel ? actually

[louigi600]

2.6.21.5 builds and works with some issues (it even builds EABI but I've not yet tested).
2.6.22 builds but my setup is currently too unsafe for any real use.
2.6.23 does not exist yet (2.6.22.1 is current stable at kernel.org).