Original "Simple" debricking methods for K5-Touch, K4-Mini, & K3-Keyboard

[knc1]

Originally Posted by kirokko

Maybe I missed something but is there fastboot tool for linux? Or should I use Windows?

https://www.mobileread.com/forums/sho...d.php?t=174638

[kirokko]

So, after trying to flash kernels and images with fastboot I got the same problem as it was described early with K4. I can load fastboot mode, I can load diags mode, but I can't load main mode. MgfTool also doesn't work. Fastboot also can't help me to get into main mode using this commands:

fastboot setvar bootmode main
fastboot reboot

I checked everything before flashing, and I'm sure that I choose correct files.
May be we have the wrong kernel image?

I had 4.1.0 before. And I had succesfull experience when I flashed main image using dd command through ssh. I didn't flash kernel and diags images and at that time everything was fine.

[kirokko]

I can't edit my previous post

Size of main kernel in bytes: 4724736
Size of diags kernel in bytes: 5123072

MD5 of both kernels:

99c31fa3ad2618e577194a912f139550 *diags_kernel.img
2e880fdfd48c3d460151c5051ca8161b *kernel-4.0.1.img

[Azrhei]

Ok, I posted somewhere else but it looks like this is the place to be for me. My kindle wasn't bricked, I just wanted to update. After messing around I bricked it. Then I followed this guide and ....

K4NT is my toy 4.0.1

MFGTool never worked, never detected my Kindle, but I was able to instal drivers and get fastboot to work. I flashed the K4 main img and diag img and that seemed to work. I SSH'd into the device and that worked. I dd'd the imgs, and that worked (it seemed) and then I zero'd the partition and that... well, then I disabled DIAG and rebooted and now I'm bricked again. I can't get my device to show up on MFGtool. I can't even get my computer to realized it's plugged in when I do the power/home trick.

I'm stuck on the USBnet screen, if it matters.
I did this on Windows 7 AMD x64

The only thing that I may have done wrong. I did the flashing on one computer, but then switched computers to do the SSH part (because I had that already setup on my other computer)

[Azrhei]

Sorry for the double post:

I was able to fastboot again, followed all steps again... When I try to disable diag mode, it bricks... I've tried using both the diag menu to do this and fastboot and either way it bricks. I'm wondering if I used the wrong images? I grabbed the ones marked K4 and had no errors along the way with them.

[kirokko]

Originally Posted by Azrhei

Sorry for the double post:

I was able to fastboot again, followed all steps again... When I try to disable diag mode, it bricks... I've tried using both the diag menu to do this and fastboot and either way it bricks. I'm wondering if I used the wrong images? I grabbed the ones marked K4 and had no errors along the way with them.

It seem like you have the same problem as I describe above. Everything works fine when flashing diags and kernels until it stucks and the same point at load to normal mode.

[Azrhei]

Originally Posted by kirokko

It seem like you have the same problem as I describe above. Everything works fine when flashing diags and kernels until it stucks and the same point at load to normal mode.

I noticed your post. Only difference is that I was at 4.0.1 otherwise I think we are at the same point.

I can repeat all the steps as much as I want. I can fastboot/flash. I can SSH in and re-dd the imgs etc.

[kirokko]

As I've said before, I thought we have wrong kernel image. I flashed the new one by hostar form this post and kindle works fine. I think it should be replaced in pastebin page. I flashed both diags_kernel and main kernel, so I'm not sure which was caused of such problems.

Sorry for my English, it's not perfect

[kirokko]

Sorry for double post, I just want that all this should be clear.

This is the MD5 hash for all kernels which I found:

99c31fa3ad2618e577194a912f139550  diags_kernel_forensic.img
3566d91f123e609b6a1aa8e679d65bfc  diags_kernel_hostar.img
2e880fdfd48c3d460151c5051ca8161b  main_kernel_forensic.img
f8ee194c84bbc824537b469265785691  main_kernel_hostar.img
2e880fdfd48c3d460151c5051ca8161b  main_kernel_mediafire.img

As you can see, main kernel form forensic server and from mediafire are the same. The main kernel from hostar differ from main kernel form forensic.

Also I attach info using file utility. This is for main kernel from forensic:

main_kernel_forensic.img: u-boot legacy uImage, Linux-2.6.31-rt11-lab126, Linux/ARM, OS Kernel Image (Not compressed), 4760896 bytes, Fri Sep 23 11:34:55 2011, Load Address: 0x70008000, Entry Point: 0x70008000, Header CRC: 0xEE89786D, Data CRC: 0x04F95A54

And this is for main kernel by hostar:

main_kernel_hostar.img: u-boot legacy uImage, Linux-2.6.31-rt11-lab126, Linux/ARM, OS Kernel Image (Not compressed), 4760896 bytes, Sat Sep  3 12:46:54 2011, Load Address: 0x70008000, Entry Point: 0x70008000, Header CRC: 0x3E5D2E75, Data CRC: 0x8C67E408

We can see that Header CRC and Data CRC differ.

And info about diags kernel. As you can see, MD5 also differ.
This info for diags kernel from mediafire (there is no such kernel on forensic server):

diags_kernel_forensic.img: u-boot legacy uImage, Linux-2.6.31-rt11-lab126, Linux/ARM, OS Kernel Image (Not compressed), 5122432 bytes, Sat Sep  3 09:01:08 2011, Load Address: 0x70008000, Entry Point: 0x70008000, Header CRC: 0x2A4A2E2C, Data CRC: 0xAB6660C5

And info for diags kernel by hostar:

diags_kernel_hostar.img: u-boot legacy uImage, Linux-2.6.31-rt11-lab126, Linux/ARM, OS Kernel Image (Not compressed), 5122432 bytes, Sat Sep  3 09:01:08 2011, Load Address: 0x70008000, Entry Point: 0x70008000, Header CRC: 0x2A4A2E2C, Data CRC: 0xAB6660C5

As we can see, Header CRC and Data CRC are the same.

Summarize all, the reason is wrong main kernel from forensic and mediafire. I think it would be useful to correct all links.

[knc1]

About the md5sum's of the image files . . . .

• Each image file contains (in /etc/shadow) a machine serial number based password for the user 'root'. The md5sum's will differ if the images are from machines with different serial numbers.
• Each image file contains files with three timestamps, one of which is the "last accessed" (atime) field. Even if the images are from the same machine with the same serial number, if the last time any of the files where accessed are different, the md5sums will be different.
• The non-windows file system (the ones listed here) have meta-data super blocks, each with a "last mount time" - which gets written to even when the media is mounted "read-only". This will also cause the md5sums to differ.

Instead of comparing image md5sums -
Mount each of the images;
Copy out ("cat") the contents of <mount-point>/etc/version.txt and <mount-point>/etc/prettyversion.txt and post those contents.
It those exact file names are not present in an image, browse around in <mount-point>/etc until you find something similar.

The <mount-point>/etc/version.txt will have a software "build code" in it and a software build date.
It would be nice to see those.

[kirokko]

But this is kernel images, how to mount it? Also I don't think that this kernels have /etc/version.txt and so on.

Also you can watch on CRC header and CRC data. This is important, not md5sum. It was created to show that forensic and mediafire kernels are the same. And you miss something. Are you sure that last access date will change md5? I tried with simple text file filled with random characters. Last access date doesn't change, but if I save file without add or remove or change something inside of it date of modification si changing but md5 are the same.

Even the date is different. In hostar files all dates are 3 Sep (for main and for diags), but for forensic main kernel date is 23 Sep and mediafire diags are 3 Sep.

[knc1]

Originally Posted by kirokko

But this is kernel images, how to mount it? Also I don't think that this kernels have /etc/version.txt and so on.

My bad.
I will leave the post above for the information that it contains.

Originally Posted by kirokko

And you miss something. Are you sure that last access date will change md5? I tried with simple text file filled with random characters. Last access date doesn't change, but if I save file without add or remove or change something inside of it date of modification si changing but md5 are the same.

IF we here talking about file system images, rather than kernel images: Yes.
Depending on your mount options. Check the mount options on the file you are testing with - many distro's default to "noatime".

[geekmaster]

Originally Posted by kirokko

As I've said before, I thought we have wrong kernel image. I flashed the new one by hostar form this post and kindle works fine. I think it should be replaced in pastebin page. I flashed both diags_kernel and main kernel, so I'm not sure which was caused of such problems.

Sorry for my English, it's not perfect

I cannot access pastebin.com from my current location. I get a redirect to a "pastebin.com Blocked by Government Act (translated)" page. I cannot update it now. I will add Hostar's post to the first post for now.

EDIT: Done! Hostar's images are 4.0.0, according to the post immediately before that linked post. It looks like we need a mirror for pastebin.com now that it is banned and blocked in some countries (at least by the ISP I am now using).

Request: Can somebody please COPY the pastebin page to another site, and I will add that mirror link to the first post in this thread?

[knc1]

Originally Posted by kirokko

But this is kernel images, how to mount it?

The short answer: It is possible.
The long answer: It isn't fun and simple, I will have to write that one up.

One way to compare the binary kernels is to dump its symbol table.
If two kernels support the same symbols, then they may be functionally the same (or may not be).

(Kernels have a symbol table, even when "stripped", to make loadable modules work.)

In general, u-boot/ARM uses a zImage with a perpended u-boot header.
And like any other kernel, that image may also have an appended initramfs file (which may in turn, have other files appended to it).

The u-boot header will not be compressed, the kernel image may well be, after the leading "piggy" block (the relocation and decompression code).

There seems to be a lot of information on the 'net about taking apart the omap (TI, ARM) images. I guess we need to re-write that info for the Freescale/Amazon images.

(I have done several write-up on taking apart the MIPS images, which is not relevant here. Those can be found by clicking through to my "home website" in the left hand info blocks.)

[kirokko]

Originally Posted by geekmaster

[COLOR="DarkRed"]EDIT: Done! Hostar's images are 4.0.0, according to the post immediately before that linked post. It looks like we need a mirror for pastebin.com now that it is banned and blocked in Turkey (at least by the ISP I am now using).

So, mmcblk0p1.img 4.0.1 works with kernel 4.0.0 and mmcblk0p1.img 4.0.1 doesn't work with kernel 4.0.1.
I didn't try to flash mmcblk0p1.img 4.0.0 with dd's command, 4.0.1 works fine.

Does the diags partition and diags kernel also update with new firmware 4.1.0? How to check main kernel and diags kernel version in my Kindle? It looks like 4.0.1 fw update didn't change diags_kernel.