l-m-c fails when image and hwpack combine to give >1 kernel

[Tom Gall]

See http://pastebin.ubuntu.com/498743/ for the full log

In a nutshell

./linaro-media-create --rootfs ext3 --mmc /dev/sdc --dev beagle --hwpack /home/tgall/hwpacks/hwpack_linaro-omap3_20100917-16_armel_supported.tar.gz --binary ~tgall/linaro-alip-0922.tar.gz

(that's a linaro-alip image I built 9/22 via live helper)

l-m-c is the latest as is the hwpack.

Last messages from the log:

-----------------------------

Installing Boot Loader

`binary/usr/lib/x-loader-omap/MLO' -> `/home/tgall/bzr/projects/linaro-image-tools/linaro-image-tools/disk/MLO'
`binary/usr/lib/u-boot/omap3_beagle/u-boot.bin' -> `/home/tgall/bzr/projects/linaro-image-tools/linaro-image-tools/disk/u-boot.bin'
done
Usage: mkimage -l image
          -l ==> list image header information
       mkimage [-x] -A arch -O os -T type -C comp -a addr -e ep -n name -d data_file[:data_file...] image
          -A ==> set architecture to 'arch'
          -O ==> set operating system to 'os'
          -T ==> set image type to 'type'
          -C ==> set compression type 'comp'
          -a ==> set load address to 'addr' (hex)
          -e ==> set entry point to 'ep' (hex)
          -n ==> set image name to 'name'
          -d ==> use image data from 'datafile'
          -x ==> set XIP (execute in place)
umount: /tmp/tmp.UcaQ5PIqTA/binary/proc: not mounted

[Tom Gall (tom-gall)]

rerunning with -x ... more shortly.

[Tom Gall (tom-gall)]

+ sudo mkimage -A arm -O linux -T kernel -C none -a 0x80008000 -e 0x80008000 -n Linux -d /home/tgall/bzr/projects/linaro-image-tools/linaro-image-tools/binary/boot/vmlinuz-2.6.35-1005-linaro-omap /home/tgall/bzr/projects/linaro-image-tools/linaro-image-tools/binary/boot/vmlinuz-2.6.35-1006-linaro-omap /home/tgall/bzr/projects/linaro-image-tools/linaro-image-tools/disk/uImage
Usage: mkimage -l image
          -l ==> list image header information
       mkimage [-x] -A arch -O os -T type -C comp -a addr -e ep -n name -d data_file[:data_file...] image
          -A ==> set architecture to 'arch'
          -O ==> set operating system to 'os'
          -T ==> set image type to 'type'
          -C ==> set compression type 'comp'
          -a ==> set load address to 'addr' (hex)
          -e ==> set entry point to 'ep' (hex)
          -n ==> set image name to 'name'
          -d ==> use image data from 'datafile'
          -x ==> set XIP (execute in place)
+ sudo umount /tmp/tmp.8IsfAUYsL9/binary/proc
umount: /tmp/tmp.8IsfAUYsL9/binary/proc: not mounted
+ true
+ sudo rm -rf /tmp/tmp.8IsfAUYsL9
root@aku:/home/tgall/bzr/projects/linaro-image-tools/linaro-image-tools#

[Tom Gall (tom-gall)]

new hwpack hwpack_linaro-omap3_20100922-23_armel_supported.tar.gz fixes this issue.

closing.

[James Westby (james-w)]

Hi,

I think we should still make the code safe against two kernels ending up in the image.

Thanks,

James

[Dave Martin (dave-martin-arm)]

+1

...but shouldn't we also try hard to prevent the image ending up with two kernels in it? ... That's a huge waste of space.

This bug still hits me with hwpack_linaro-omap3_20101004-35_armel_supported.tar.gz + releases.linaro.org/platform/linaro-m/headless/beta/linaro-m-headless-tar-20100830-2.tar.gz.

(linaro-image-tools r132)

Maybe it's wrong to combine those two images, but if so it would be good if linaro-image-create could flag it up.

[Steve Langasek (vorlon)]

On Tue, Oct 05, 2010 at 09:41:04AM -0000, Dave Martin wrote:
> ...but shouldn't we also try hard to prevent the image ending up with
> two kernels in it? ... That's a huge waste of space.

The current daily headless images no longer include a kernel.

> Maybe it's wrong to combine those two images, but if so it would be good
> if linaro-image-create could flag it up.

As this is a transitional issue between pre-hwpack images and hwpacks, I
don't think it's going to get attention. We should still solve the more
general problem of l-m-c failing when more than one kernel is present,
however.

--
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
Ubuntu Developer http://www.debian.org/
<email address hidden> <email address hidden>

[Tom Gall (tom-gall)]

with the 1006 and 1007 API change in the midst of the 10/14 daily snapshot tests. With this merge proposal https://code.launchpad.net/~james-w/linaro-image-tools/fix-install-latest-logic/+merge/38426 l-m-c works correctly.

[Jamie Bennett (jamiebennett)]

Marking 'Fixed Released' (branch has been merged) based on Tom's comment above.

[James Westby (james-w)]

No, the problem is still there lurking, ready to bite us next time
someone somehow ends up with two kernels.

I just fixed the immediate cause this time, we should make the code
more robust against this issue.

Thanks,

James

[Loïc Minier (lool)]

New hardware pack format should help here.

[Loïc Minier (lool)]

bug #744857 discusses a quicker solution for this than hardware pack v2 (which would solve this too)

[Mattias Backman (mabac)]

This can't be an issue anymore right? The change for kernel_flavors should have fixed this. Also the hwpack v2 format is available now to play around with,

Can we close this bug?

[James Westby (james-w)]

Hi,

I think that we still use globs for the vmlinuz file and fail if multiple files are found,
so I think it could still happen.

hwpacksv2 will make the problem disappear though, so I don't think we need to do anything
on this beyond move to v2.

Thanks,

James