Tools boot kindle from kernel image on USB drive

[geekmaster]

Originally Posted by wsxhwyy

can i use the same method to boot main?

As you can read in the Select Boot thread, the MfgTool you can download there contains bootmode profiles for main, diags, and fastboot. Just select the bootmode you want from the menu.

Lets get back on topic. Just because I am trying to be helpful here is not a reason to continue this debricking conversation in this thread.

Please ask future debricking questions in a debricking thread. Thanks...

[wsxhwyy]

Originally Posted by geekmaster

As you can read in the Select Boot thread, the MfgTool you can download there contains bootmode profiles for main, diags, and fastboot. Just select the bootmode you want from the menu.

Lets get back on topic. Just because I am trying to be helpful here is not a reason to continue this debricking conversation in this thread.

Please ask future debricking questions in a debricking thread. Thanks...

Report:
I used this method to boot main and succeeded. You only have to change mmcblk0p2 to mmcblk0p1.

[seaniko7]

Does anybody know which toolchain is appropriate for KT kernel cross-compiling ? I used code sourcery and eldk without success..

[hawhill]

Hm, CodeSourcery should work, I think, they have one of the most versatile ARM targeted gcc suites. In what way it doesn't work? But then, I don't have a Touch...

[seaniko7]

I can compile kernel without issues, but after flashing it with kexec KT just freezes.

[eureka]

Originally Posted by seaniko7

I can compile kernel without issues, but after flashing it with kexec KT just freezes.

Did you tried kexec'ing with KT connected to power source (wall outlet)?

[seaniko7]

Of course. I can still use stock kernels, which kexec flash even without additional console arguments.

EDIT:

Executing on KT :

ldd --version

Gave me this output :

ldd (Sourcery G++ Lite 2009q3-67) 2.10.1

Which led me to downloading correct version of toolchain :

arm-2009q3-67-arm-none-linux-gnueabi

This particural one seems to work !

[geekmaster]

Originally Posted by seaniko7

Of course. I can still use stock kernels, which kexec flash even without additional console arguments.

A little friendly advice: using the word "flash" in this context is misleading. You should use the term "execute" instead (as correctly implied by the name of the "kexec" command). "Flash" generally means writing to the MMC using a tool like "fastboot", when used in this forum.

@seaniko7: In general, your command of the English language is excellent.

Originally Posted by seaniko7

arm-2009q3-67-arm-none-linux-gnueabi

This particural one seems to work !

Great news! That should save some of us (like me) a LOT of wasted effort trying to learn this on our own. Thanks!


P.S. The first post now contains a link to the "getkernels" tool, which extracts kernel images from mmcblk0.img or from /dev/mmcblk0, an all eink kindles:

http://www.mobileread.com/forums/sho...d.php?t=174674

[geekmaster]

I tried launching the Paperwhite diags kernel from the simple debricking pastebin using kexec, but I get this output:

[root@kindle us]# ./kexec.sh
Cannot open /proc/atags: No such file or directory
kexec_load failed: Function not implemented
entry       = 0x70008000 flags = 280000
nr_segments = 2
segment[0].buf   = 0x2a068
segment[0].bufsz = a0
segment[0].mem   = 0x70001000
segment[0].memsz = 1000
segment[1].buf   = 0x4012d048
segment[1].bufsz = 4653c0
segment[1].mem   = 0x70008000
segment[1].memsz = 466000

Bummer. I was hoping to boot the Paperwhite diags partition this way, because ENABLE_DIAGS does not work.

EDIT: Also, kexec -e drops USBNet so you cannot start a shell, but the kindle still shows up as an RNDIS device. The kindle framework and menu system still works though. But the search bar ";un" command does nothing, and I need to restart the kindle before I can use USBNet again...

[eureka]

@geekmaster, could you please post output of Paperwhite's:

zcat /proc/config.gz |grep -i "kexec\|atags"

[geekmaster]

Originally Posted by eureka

@geekmaster, could you please post output of Paperwhite's:

zcat /proc/config.gz |grep -i "kexec\|atags"

[root@kindle root]# zcat /proc/config.gz |grep -i "kexec\|atags"
zcat: /proc/config.gz: No such file or directory
[root@kindle root]# 

[NiLuJe]

Awesome. -_-".

And the 5.2.0 source packages aren't out, so no way to check the defconfig. :/

[geekmaster]

[root@kindle /proc]# ls -al|grep -v ^d
-r--r--r--    1 root     root             0 Oct  6 22:31 board_id
-r--r--r--    1 root     root             0 Oct  6 22:31 bootmode
-r--r--r--    1 root     root             0 Oct  6 22:31 buddyinfo
-r--r--r--    1 root     root             0 Oct  6 22:31 cmdline
-r--r--r--    1 root     root             0 Oct  6 22:31 cpuinfo
-r--r--r--    1 root     root             0 Oct  6 22:31 crypto
-r--r--r--    1 root     root             0 Oct  6 22:31 devices
-r--r--r--    1 root     root             0 Oct  6 22:31 diskstats
-r--r--r--    1 root     root             0 Oct  6 22:31 dma-engine
-r--r--r--    1 root     root             0 Oct  6 22:31 execdomains
-r--r--r--    1 root     root             0 Oct  6 22:31 fb
-r--r--r--    1 root     root             0 Oct  6 22:31 filesystems
-r--r--r--    1 root     root             0 Oct  6 22:31 interrupts
-r--r--r--    1 root     root             0 Oct  6 22:31 iomem
-r--r--r--    1 root     root             0 Oct  6 22:31 ioports
-r--r--r--    1 root     root             0 Oct  6 22:31 kallsyms
-r--------    1 root     root             0 Oct  6 22:30 kmsg
-r--------    1 root     root             0 Oct  6 22:31 kpagecount
-r--------    1 root     root             0 Oct  6 22:31 kpageflags
-rw-r--r--    1 root     root             0 Oct  6 22:31 latency_stats
-r--r--r--    1 root     root             0 Oct  6 22:31 loadavg
-r--r--r--    1 root     root             0 Oct  6 22:31 locks
-r--r--r--    1 root     root             0 Oct  6 22:31 mac_addr
-r--r--r--    1 root     root             0 Oct  6 22:31 mac_sec
-r--r--r--    1 root     root             0 Oct  6 22:31 meminfo
-r--r--r--    1 root     root             0 Oct  6 22:31 misc
-r--r--r--    1 root     root             0 Oct  6 22:31 modules
lrwxrwxrwx    1 root     root            11 Oct  6 22:31 mounts -> self/mounts
lrwxrwxrwx    1 root     root             8 Oct  6 22:31 net -> self/net
-r--r--r--    1 root     root             0 Oct  6 22:31 pagetypeinfo
-r--r--r--    1 root     root             0 Oct  6 22:31 panel_id
-r--r--r--    1 root     root             0 Oct  6 22:31 partitions
-r--r--r--    1 root     root             0 Oct  6 22:31 postmode
-r--r--r--    1 root     root             0 Oct  6 22:31 sched_debug
-r--r--r--    1 root     root             0 Oct  6 22:31 schedstat
lrwxrwxrwx    1 root     root            64 Oct  6 21:46 self -> 7004
-rw-r--r--    1 root     root             0 Oct  6 22:31 slabinfo
-r--r--r--    1 root     root             0 Oct  6 22:31 softirqs
-r--r--r--    1 root     root             0 Oct  6 22:31 stat
-r--r--r--    1 root     root             0 Oct  6 22:31 swaps
--w--w--w-    1 root     root             0 Oct  6 22:31 tcpreset
-r--r--r--    1 root     root             0 Oct  6 22:31 timer_list
-rw-r--r--    1 root     root             0 Oct  6 22:31 touch
-r--r--r--    1 root     root             0 Oct  6 22:31 uptime
-r--r--r--    1 root     root             0 Oct  6 22:31 usid
-r--r--r--    1 root     root             0 Oct  6 22:31 version
-r--------    1 root     root             0 Oct  6 22:31 vmallocinfo
-r--r--r--    1 root     root             0 Oct  6 22:31 vmstat
-r--r--r--    1 root     root             0 Oct  6 22:31 zoneinfo
[root@kindle /proc]#

[geekmaster]

This happens on other ARM devices too, and there is a kernel patch to fix it:
http://docs.openmoko.org/trac/attach...2.6.22.5.patch

So we need a patched kernel to support kexec, so we can load a custom kernel? Catch-22?

I could see a multi-stage leveraged approach, where we tether boot a custom u-boot that does not need atags, and use that to run kexec to load a kernel without the missing atags problem (if I am beginning to understand how that works). At least it is a place to start looking. I hope we find an untethered boot approach instead, but this is a start until then...

[knc1]

Originally Posted by geekmaster

This happens on other ARM devices too, and there is a kernel patch to fix it:
http://docs.openmoko.org/trac/attach...2.6.22.5.patch

Did I not read somewhere in one of these information post dumps that the K5/Kpw was running 2.6.31?

That patch from back in the 2.6.22 days should have made it into the code, although it may not have been optioned in the build configuration.

Hint:
If you have read access to the u-boot environment area, dump the script strings and see how it is setup for network booting.

Those defaults may be the same as the K4/K5 - but might not.