75 replies to this topic

[supertx2]

Dualboot?

[android@sam]

Sulthekk, on 18 July 2012 - 10:44 AM, said:

Wait... Recovery in the source of an experimental dual boot kernel? Is it a samsung device?

Yes

[Sulthekk]

android@sam, on 18 July 2012 - 01:51 PM, said:

Yes

Galaxy S and some other samsung devices have their recovery integrated into the kernel instead of having it on a separate partition, unlike most of other phones, including Blade... Could you link the patches for me to see how they actually work?

Edited by Sulthekk, 18 July 2012 - 02:39 PM.

[Snap.IT]

Sulthekk, on 15 July 2012 - 07:11 PM, said:

That's why I was thinking on kexec... You can just place a kernel and an initrd directly in the root of the partition, and from now on you would only need to tell the loader program where the partition is, then it would find the necessary things by itself, since the paths would be the same for everytime. But I'm still doing a research on it as I'm not experienced with it, so what I say maybe isn't working... Handle it that way.

Aye, my thoughts exactly, this would also mean that just copying the files into the correct directories would work for installing any rom.

Using grub legacy code for ARM i can get it to boot grub2 but there isn't much i can do with that yet, the mappings just doesn't work at all.

[Snap.IT]

Tried to make it boot first device but i'm so tired i forgot that i don't have a zImage on first device.

I'm going to bed.

[Sulthekk]

Snap.IT, on 18 July 2012 - 09:46 PM, said:

Aye, my thoughts exactly, this would also mean that just copying the files into the correct directories would work for installing any rom.

Using grub legacy code for ARM i can get it to boot grub2 but there isn't much i can do with that yet, the mappings just doesn't work at all.

Great

I didn't know that GRUB ever had ARM support...

[android@sam]

I have cloned successfully NAND rom into sdcard by doing some fun with dev/block/mmcblk.Just needed some modifications in init.blade.rc and a script for alternating the boot.

[android@sam]

@sulthekk Any progress?

[Sulthekk]

android@sam, on 21 July 2012 - 05:21 AM, said:

@sulthekk Any progress?

Still had no time at PC, and I'm limited to android...
Until I don't resolve that problem I can't compile. Or more precisely, I don't want to compile on the Blade because it's damn long.

[android@sam]

Sulthekk, on 21 July 2012 - 07:54 AM, said:

Still had no time at PC, and I'm limited to android...
Until I don't resolve that problem I can't compile. Or more precisely, I don't want to compile on the Blade because it's damn long.

Still no progress?

[Sulthekk]

Aww... I totally forgot about this during the  vacation.

Before leaving I also forgot to mention that I found some really interesting things at https://wiki.ubuntu....MSoftbootLoader

Some of that looks promising.
For example, kboot is written in shellscript, all it needs is a kexec-able kernel (and probably kexec-tools too) with minor modifications to load the kernel from a specified path, and to attach the 'root=' parameter to he kernel cmdline.

Edited by Sulthekk, 06 August 2012 - 03:20 PM.

[android@sam]

@sulthekk  I need you  help.Could you tell me which block is used for mounting system,data and cache in /dev/block/?

[Sulthekk]

The mount command shows that. I post it there anyway:


dev:    size   erasesize  name
mtd0: 00500000 00020000 "recovery"
mtd1: 00500000 00020000 "boot"
mtd2: 00180000 00020000 "splash"
mtd3: 00080000 00020000 "misc"
mtd4: 00200000 00020000 "cache"
mtd5: 0a000000 00020000 "system"
mtd6: 115e0000 00020000 "userdata"
mtd7: 00020000 00020000 "oem"
mtd8: 00180000 00020000 "persist"


———————————————


Today I have compiled burstlam's Turies GB kernel with CONFIG_KEXEC enabled, it boots and works fine for standard use. Haven't tested kexec'ing because I fail to compile kexec-tools, which we need to use that feature. I guess something is bad around my toolchain as it compiled nicely on x86, but fails on arm. I am not at the PC at the moment, but I remember a warning about implicit declaration of if_nameindex and and muliple error's saying something like 'converting integer to pointer without a cast' or 'incomplete type' from multiple lines between 45-60. I will check on that later.

Edit: /proc/mtd seems a better way to go than mount as it shows only the partitions on the NAND.

Edited by Sulthekk, 07 August 2012 - 09:18 PM.

[android@sam]

Finally got success but partial.I have successfully booted ics from sdcard and it is fast as NAND .Soon will create thread.

[narkoman88]

look forward to....

[Sulthekk]

android@sam, on 08 August 2012 - 10:20 AM, said:

Finally got success but partial.I have successfully booted ics from sdcard and it is fast as NAND .Soon will create thread.

That's good to hear. Does partial succes mean that you have repacked the boot.img with the root and/or loop parameters pointing to the fs on the sdcard, and then placed the system manually there?