20 replies to this topic

[christophecvr]

Hello as saw that there was an new kernel avbl 2.6.30 for dm8000. I just tried to compile the new openpli 2.1 dm8000 for kernel 2.6.30

I did changed file dreambox-mipsel.inc located
openplitest/openembedded/conf/machine/include

PREFERRED_VERSION_linux-${MACHINE} = "2.6.30" before it was 2.6.18
PREFERRED_VERSION_linux-libc-headers = "2.6.30" before it was 2.6.18

apart from a small compile error very classic by migration from .c files to kernel 2.6.30 about declaration DMX_REMOVE_PID (first use in this function not declared) and the same for DMX_ADD_PID)
in streamproxy.c and usbtunerhelper.c which I solved by declaring them as
unsigned long DMX_ADD_PID;/*added by cvr*/
unsigned long DMX_REMOVE_PID; /*added by cvr*/

all fine. I do have a succesfull compilation and nfi file.

I installed the file to the box but it hangs during boot. Do I have to change also boot stage loader or so ?

[littlesat]

I do not think this will work... as the drivers are nog 2.6.30 compatible.

[christophecvr]

Yes perhaps not everything will work . But the boot should. Also the basic drivers from which source is not released by dreambox are in the openembedded and ready and compiled. all the dvb modelu drivers. The other's did compile nicely (at least it seems like they are) I obatianed full ready compilation.

But I think that the second stage boot loader needs to be changed as well in the nand. and if i'm not wrong it can be done by dreamload 1.5.1 1.6 doe not seems to be able to do it anymore.

And if i'm not wrong
openpli-enigma2-2.1-dm8000.boot.jffs2

is the second stage boot loader ?

[MiLo]

As I recall it, Dream attempted to build a 2.6.30 based system (and various other attempts for the 7025 too), but they never got it to work propely and abandoned it. The "leftovers" are still visible in the source.

[christophecvr]

Hello yes it seems indeed like that. cause the very last attempts seems to be from 2010 . However it was just a try out. just seeing now if i can try to fix something at the boot but even there lucky it will perhaps bug later on on other drivers.

Somehow it's compiling very well .

[MiLo]

It's the closed-source drivers where the problem is. The "open" stuff, like the kernel itself, runs fine. So you can only "fix" it if you manage to fix those drivers. Which you can basically only do if you happen to work for DMM.

[christophecvr]

Yes it's a problem. As far as I could see they seems to start working on a kernel 2.6.32. Perhaps they abandoned the work with kernel and will come out with 2.6.32 . With closed sources for the dvb stuff and so. Also I think the licensing problem on some points is what causes delay. Implementing some work arounds for licences they can't obtain and stay legal is not so easy I gues .

[malakudi]

Xtrend has released drivers for 2.6.31 kernel. Has anyone checked them?

http://www.et-view.c...31-20110625.zip

[littlesat]

What are the advantages to use the newer kernel?

If the ET9000/5000 boxes have drivers for a new kernel we also should build special new kernal images only for these two type of boxes when DMM (and VU) do not create new kernel drivers.

[christophecvr]

Yes that's a good question. I'm just experimenting some things. And hit the problem that some stuff do require a kernel minimum 2.6.30 However i'm not really convinced with those explanations. First I wan't to try if my stuff is working on 2.6.30 yes or no. If well the explanations is ok. If not ....

For this test working dm 2.6.30 kernel on dm 8000 is needed. That's all. The things i'm testing may well or not beeing a big help in future developpement. Now i'm just performing some tests.

In the mean time I do have a console output off minicom follow up from dm8000 start. I tried to clean it up as much I could but its' still a mess.

What can be seen: it's a couple of missing stuf for booting. I'm shure that these can be corrected in order to let boot the dream at least. If it boots ... Well the further things which happen or do not happen are a big surprisse /images/smiley/shock.gif /images/smiley/smile.gif

added an console log from boot process dream.

Attached Files

•  bootprob.txt.tar.gz   9.48KB   2 downloads

[christophecvr]

Still trying to get 2.6.30 booting.
In the mean time I added the patches for cdfs to compile on 2.6.30 , nand patch, cpu patch and cross compile patches are made as well installed and compiling ok also the result in work dir are ok.

But it still not boot, bootlog when connecting to serial console always state that :

Scanning JFFS2 FS: done.
find_inode failed for name=secondstage.conf
load: Failed to find inode
- failed to open /boot/secondstage.conf
- failed to open /cf/autorun.bat

I think there must be somewhere a patch to aply especially for open pli image to boot. But I can not find which one.
somehow the settings are set now for opendreambox image, and they work a bit different whit nand then open pli. Also this can be seen
in different ways of mounting the boot partition in fstab.

Would be nice if somebody can tell were I can find the boot patch on 2.6.18. then I can try to adapt that one for the 2.6.30 kernel.

[pieterg]

there is no such thing as a 'boot patch'.
But from your log, it looks like the kernel is not even started by the bootloader, all you see are secondstage bootloader messages.

[christophecvr]

yes , it can't find /boot/secondstage.conf

by opendreambox image we find this normally in

fstab from working opendream experimental 2.6.18-4
rootfs / auto defaults 1 1
proc /proc proc defaults 0 0
devpts /dev/pts devpts mode=0620,gid=5 0 0
usbfs /proc/bus/usb usbfs defaults 0 0
/dev/mtdblock2 /boot jffs2 ro 0 0
tmpfs /var tmpfs defaults 0 0
tmpfs /tmp tmpfs defaults 0 0
/dev/sda1 /media/hdd auto defaults 0 0

fstab from working openpli 2.1 kernel 2.6.18
rootfs / auto defaults 1 1
proc /proc proc defaults 0 0
devpts /dev/pts devpts mode=0620,gid=5 0 0
usbfs /proc/bus/usb usbfs defaults 0 0
tmpfs /var/volatile tmpfs defaults 0 0
tmpfs /dev/shm tmpfs mode=0777 0 0

I think there is a difference, the 2.6.30 kernel was configured for opendreambox. Somewhere there should be a difference
on how the flash mtd is mounted between the two.

With 2.6.30 , I boots first stage, and the start openpli enigma 2. The second nice screen does show up and then somewhere it hangs. On lcd can be seen that it well already loads openpli.

[pieterg]

You seem to be sure the kernel is started? If so, you should enable the serial console, and see what it does.
Add 'console=ttyS0,115200' to the kernel arguments (second stage loader menu).

[christophecvr]

Yes somehow I think kernel started. At last I had on lcd dreambox multimedia. This comes with kernel. And the video screen showed the open pli image flash screen. The second good working one, only can come when the kernel-module dvb-video is running i think . I just did an boot up with 2.6.18 openpli. I 've got the same message in beginning that /boot..... could not be loaded but it boot very well and work's fine. With this message serial console output from boot process.

I as well Put an file where all my changes to the basic git in openembedded are done for openpli 2.1 to compile with kernel 2.6.30

Attached Files

•  bootconsole2.6.18openpliok.tar.gz   1.84KB   3 downloads
•  openpli2.1cvrkernelpatches2.6.30.tar.gz   22.39KB   7 downloads

[pieterg]

Yes somehow I think kernel started. At last I had on lcd dreambox multimedia. This comes with kernel.

no, that is still done by the secondstage loader.
The kernel does not have access to the lcd/oled (till the lcd driver has been loaded)

And the video screen showed the open pli image flash screen.

that's the jpg image, handled by bootlogo.elf (started by the seconstage loader, before starting the kernel)

So I think the kernel is not started. It might have the wrong format. I believe you have to use vmlinuz (gzipped vmlinux) format?

[christophecvr]

Hello, Indeed it's even not seems to start the kernel.

But it request vmlinux.gz, so that file is ok.

I reinstalled an image compiled from scratch 2.6.30 with all patches in but still not ok oeeps just saw that all my patches where erased.

Just try again with new patches and will try to use to vmlinux.gz from working 2.6.18 instead of that just compiled

[christophecvr]

I tried with different vmlinux.gz. But it does not seems to bring a solution. I'm out of ideas what it could be. But indeed it hangs just on the point where i should start the kernel itself.
Just the last point . Could it be the dvb api patch which are not applied the three mentionned below as in kernel 2.618 . If so go and look to eventually try to make such one for 2.6.30
they are in file linux-opendreambox-2.6.18.inc located openembedded/recipes/linux

file://dvb-api-2.6.18-5.3.patch \
file://linux-2.6.18-dvb-core-headers-20100904.patch \
file://linux-2.6.18-dvb-frontends-headers-20100904.patch \

[pieterg]

no, dvbapi is not used till the dmm drivers are loaded, kernel has way booted by then.

[christophecvr]

Yes that's what I thaught. It would caused prob's later on that well but at least it should have booted