Ubuntu 11.04 - the installer start again after reboot

52 views
Skip to first unread message

David Hautbois

unread,
Apr 19, 2011, 1:51:27 PM4/19/11
to pandaboard
Hi
I try to install ubuntu natty daily build.
I had the partition corruption because 8Gb SD card...
After fsck, the installer starts.
At the end of the install process, the board reboots, and the
installer start again....
I don't know if it is a "normal" reboot.

Does anyone have the same issue ?
fsck does not repair all errors ?
What log file do I have to analyze ?

Thanks.

David.

Dechesne, Nicolas

unread,
Apr 19, 2011, 2:09:07 PM4/19/11
to panda...@googlegroups.com, David Hautbois
On Tue, Apr 19, 2011 at 7:51 PM, David Hautbois <david.h...@free.fr> wrote:
Hi
I try to install ubuntu natty daily build.
I had the partition corruption because 8Gb SD card...
After fsck, the installer starts.
At the end of the install process, the board reboots, and the
installer start again....
I don't know if it is a "normal" reboot.


the normal installation process is something like this:

1- flash the .img file on your SD. the .img contains partition table, so you don't need to worry about fdisk and creating the partition
2- the .img will create 2 partitions (FAT and EXT3) and it will leave unused space at the end of your SD, so that the same .img can be used with any size >2G
3- at first boot ubuntu installer will expand the EXT3 partition so that its size matches the size of your SD. it ends up removing the unused space
4- then it reboots
5- at 2nd boot you should get up to the ubuntu installer. if you use a headless image you don't need a monitor, all installation happens on the serial console. if you use a desktop image, you need a monitor plugged on the HDMI port since installation is graphical and nothing is displayed on the serial console

if you have problems, you can report them here, but also it might be a good idea to report them on Launchpad. For every issue before the first reboot, you may report them here: https://bugs.launchpad.net/ubuntu/+source/jasper and for issues after the first reboot here: https://bugs.launchpad.net/ubuntu/+source/ubiquity


 

Dechesne, Nicolas

unread,
Apr 19, 2011, 2:10:58 PM4/19/11
to panda...@googlegroups.com, David Hautbois


On Tue, Apr 19, 2011 at 8:09 PM, Dechesne, Nicolas <n-dec...@ti.com> wrote:
For every issue before the first reboot, you may report them here: https://bugs.launchpad.net/ubuntu/+source/jasper

correction: the right place is here: https://bugs.launchpad.net/ubuntu/+source/jasper-initramfs


David Hautbois

unread,
Apr 19, 2011, 2:58:21 PM4/19/11
to panda...@googlegroups.com
The board rebooted 4 or 5 times.
The first time, I entered all information during the installer.
The next times, the installer remembered information : language, location, hostname. but not the username
The last reboot (5th?, I don't counted...) was ok.

Now, the system is up and ready
I have some work : the resolution of my hdmi display is not correctly detected...

Thanks.

David.

Frantisek Dufka

unread,
Apr 19, 2011, 7:04:03 PM4/19/11
to panda...@googlegroups.com, Dechesne, Nicolas, David Hautbois
Hi,

I have similar issue. I have tried to install beta1 and beta2 image to
Transcend class6 8GB sd card several times (like 3 times the beta1, 2
times beta 2) and had no luck. It went always like this:

- extracted image to card via zcat | dd
- booted in pandaboard
- first boot the screen on hdmi monitor said it is resizing image
- then it rebooted and showed some progress dots going indefinitely
- after waiting hour or more I powered off the board

There is no kernel console output either on serial console or any
virtual console (it can be switched via alt Fx keys) so I am not sure
there are I/O errors while it runs but it is very likely because after
trying beta 1 and removing it from pandaboard the card became corrupted
to the point that I could no longer rewrite the image via dd. Also
panasonic sd formatter failed repeatedly when formatting via full
format. After few hours of formatting it repeatedly it always failed at
some sector and card became dead, after reinserting card the format got
a bit further - the first bad sector was few megabytes later. After
several formats (5-10?) it healed itself and now works fine again. I
used this card quite a lot with linux filesystems before trying it in
pandaboard and it appeared to be good and had no issues with it before.

I had no time to investigate further. It would be nice if Natty beta
images had kernel console output or some sort or verbose output enabled
or at least had some debug version of boot.scr ready on the card for
those without u-boot utils ready. Also root shell on some console would
be nice when the install runs.

Some ideas about i/o errors:
Are SD card partitions (at least the second linux ext3 or ext4? one)
properly aligned to 1 or 4MB boundary ( before or after fs resize) as per
https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey ?
If not then it could explain lot of i/o error issues with larger cards.

Would a larger ext3/ext4 filesystem block size (like 16/32/64KB instead
of 4KB) help with forcing the kernel to write in bigger blocks?

Frantisek

GrueMaster

unread,
Apr 20, 2011, 12:32:14 PM4/20/11
to pandaboard
I have followed this thread, and it mirrors what I have seen with the
daily panda netbook images from cdimage.ubuntu.com. I seem to get
this more frequently when piping the decompressed images through dd
(i.e. zcat <image>.gz | dd of=<SD card device>). Could you try
gunzipping the image first, then using "dd if=<image> of=<SD card dev>
bs=4M" and see if you still have issues? Post the results here so I
can add the data to my list.

Tobin Davis
Ubuntu-arm QA Tech.

Dechesne, Nicolas

unread,
Apr 21, 2011, 3:30:30 AM4/21/11
to panda...@googlegroups.com


On Wed, Apr 20, 2011 at 6:32 PM, GrueMaster <gruem...@gmail.com> wrote:
 I seem to get
this more frequently when piping the decompressed images through dd
(i.e. zcat <image>.gz | dd of=<SD card device>).  Could you try
gunzipping the image first, then using "dd if=<image> of=<SD card dev>
bs=4M"  and see if you still have issues?

well... I would be surprised. are you suggesting there is a bug in zcat?

Frantisek Dufka

unread,
Apr 21, 2011, 5:24:44 AM4/21/11
to panda...@googlegroups.com
That would surprise me too. I can further add that I did in on PC with
Ubuntu via USB card reader (so I was writing to /dev/sdb) and I used
bs=4M (or 1M, not sure). I also checked kernel log for I/O errors after
writing the image. That's how I found the card was later corrupted when
trying to write the image again for second test. So it definitely
happened several times while the image was correctly written.

I will try again (hopefully today evening) and use cmp to compare the
card data back after writing the image (and reinserting the card reader)
and also check partition alignment both before and after resizing to see
whether this may be an issue.

David Hautbois

unread,
Apr 21, 2011, 1:40:00 PM4/21/11
to panda...@googlegroups.com
Hello
You're right !

gunzip natty-preinstalled-netbook-armel+omap4.img.gz
sudo dd if=natty-preinstalled-netbook-armel+omap4.img of=/dev/sdb bs=4M

This works fine.

I modify the wiki ?

Thanks.

David.

Le 20/04/2011 18:32, GrueMaster a �crit :

Ricardo Salveti

unread,
Apr 22, 2011, 3:58:28 AM4/22/11
to panda...@googlegroups.com
On Tue, Apr 19, 2011 at 3:58 PM, David Hautbois <david.h...@free.fr> wrote:
> Le 19/04/2011 20:09, Dechesne, Nicolas a écrit :
>
> On Tue, Apr 19, 2011 at 7:51 PM, David Hautbois <david.h...@free.fr>
>>
>> Does anyone have the same issue ?
>> fsck does not repair all errors ?
>> What log file do I have to analyze ?
>
> The board rebooted 4 or 5 times.
> The first time, I entered all information during the installer.
> The next times, the installer remembered information : language, location,
> hostname. but not the username
> The last reboot (5th?, I don't counted...) was ok.

This is probably because the installer failed to recreate the initrd
with the proper content, that would avoid loading oem-config again at
the next reboot. As gunzip and then dd worked for you, it could be the
case that previous method failed to properly write the data at your SD
card.

Cheers,
--
Ricardo Salveti de Araujo

Reply all
Reply to author
Forward
0 new messages