imx51 squashfs as built on the armel livefs builder is broken and produces noisy kernel messages

Bug #412757 reported by Loïc Minier on 2009-08-12
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
linux-fsl-imx51 (Ubuntu)
Undecided
Canonical Kernel Team

Bug Description

Hi

The A4 imx51 babbage image doesn't find its livefs by default (due to a driver renaming its /sys data between 2.6.28 and .31) but will find it if one passes LIVEMEDIA=/dev/mmcblk0p2 on the kernel cmdline. However startup will grind to a halt after a kernel stacktrace in squashfs and I/O errors immediately thereafter.

I will attach photos of the end of the boot.

Bye

Loïc Minier (lool) wrote :
Loïc Minier (lool) wrote :

First photo is page 1/2

Now attaching page 2/2

Loïc Minier (lool) on 2009-08-13
Changed in linux-fsl-imx51 (Ubuntu):
assignee: nobody → Canonical Kernel Team (canonical-kernel-team)
milestone: none → karmic-alpha-5
Amit Kucheria (amitk) wrote :

Let's make it easy for search engines... Here is the actual bug:

Begin: Running /scripts/init-premount ...
Done.
Begin: Mounting root file system... ...
Adding 119656k swap on /dev/ramzswap0. Priority:100 extents:1 across:119656k SSD
aufs 2-standalone.tree-30-20090727
squashfs: version 4.0 (2009/01/31) Phillip Lougher
Begin: Running /scripts/casper-premount ...
Done.
Done.
SQUASHFS error: squashfs_read_data failed to read block 0x0
SQUASHFS error: Unable to read metadata cache entry [0]
SQUASHFS error: Unable to read inode 0x116b2606f7
------------[ cut here ]------------
WARNING: at /build/buildd/linux-fsl-imx51-2.6.31/fs/inode.c:699 unlock_new_inode+0x40/0x4c()
Modules linked in: squashfs aufs nls_iso8859_1 nls_cp437 dm_raid45 xor ramzswap xvmalloc lzo_decompress lzo_compress
Backtrace:
fs])
[<bf07d204>] (squashfs_get_sb+0x0/0x30 [squashfs]) from [<c00e86d4>] (vfs_kern_mount+0x70/0x130)
[<c00e8664>] (vfs_kern_mount+0x0/0x130) from [<c00e87f0>] (do_kern_mount+0x3c/0xe4)
[<c00e87b4>] (do_kern_mount+0x0/0xe4) from [<c00ff240>] (do_mount+0x208/0x230)
 r9:00000000 r8:00008001 r7:00000048 r6:d9dce000 r5:d9046000
r4:00008401
[<c00ff038>] (do_mount+0x0/0x230) from [<c00ff2f4>] (sys_mount+0x8c/0xd4)
[<c00ff268>] (sys_mount+0x0/0xd4) from [<c0033940>] (ret_fast_syscall+0x0/0x2c)
 r7:00000015 r6:00008401 r5:4001faf0 r4:00000000
---[ end trace 06d337adeed8ced6 ]---

BusyBox v1.13.3 (Ubuntu 1:1.13.3-1ubuntu5) built-in shell (ash)
Enter 'help' for a list of built-in commands.

(initramfs) usb 1-1.6: new high speed USB device using fsl-ehci and address 3
mount: mounting /dev/loop0 on //filesystem.squashfs failed: Input/output error
Can not mount /dev/loop0 (/cdrom/casper/filesystem.squashfs) on //filesystem.squashfs
usb 1-1.6: configuration #1 chosen from 1 choice
scsi0 : SCSI emulation for USB Mass Storage devices

Oliver Grawert (ogra) wrote :

the squashfs itself is broken, this is no issue with the kernel module (apart from it being noisy if it cant use the squashfs)

ogra@antimony:~$ file /srv/cdimage.ubuntu.com/scratch/ubuntu/ports_daily-live/live/armel+imx51.squashfs
/srv/cdimage.ubuntu.com/scratch/ubuntu/ports_daily-live/live/armel+imx51.squashfs: Squashfs filesystem, little endian, version 4.0, -755002881 bytes, 113901 inodes, blocksize: 41 bytes, created: Mon Mar 26 22:52:00 2103

antimony runs hardy, file and libmagic in hardy cant handle squashfs4 files so reports negative size values

manoao runs lenny with a karmic chroot to build the squashfs, trying the following on manoao results in a segfault:

mksquashfs etc foo.sq && unsquashfs -li foo.sq
Found a valid exportable SQUASHFS superblock on foo.sq.
        Inodes are compressed
        Data is compressed
        Fragments are compressed
        Fragments are present in the filesystem
        Always_use_fragments option is not specified
        Duplicates are removed
        Filesystem size 214.14 Kbytes (0.21 Mbytes)
        Block size 131072
        Number of fragments 3
        Number of inodes 648
        Number of ids 2

Parallel mksquashfs: Using 1 processor
Scanning existing filesystem...
Segmentation fault

debugging it further showed the following:

mksquashfs test foo.sq ; find test -print0 | xargs -0 ls -ld
Found a valid exportable SQUASHFS superblock on foo.sq.
        Inodes are compressed
        Data is compressed
        Fragments are compressed
        Fragments are present in the filesystem
        Always_use_fragments option is not specified
        Duplicates are removed
        Filesystem size 225.53 Kbytes (0.22 Mbytes)
        Block size 131072
        Number of fragments 3
        Number of inodes 1215
        Number of ids 2

Parallel mksquashfs: Using 1 processor
Scanning existing filesystem...
Segmentation fault
drwxr-xr-x 4 root root 4096 Aug 14 13:06 test
drwxr-xr-x 2 root root 4096 Aug 14 13:05 test/init.d
-rw-r--r-- 1 root root 0 Aug 14 13:05 test/init.d/bootlogs.sh
-rw-r--r-- 1 root root 0 Aug 14 13:05 test/init.d/dbus
-rw-r--r-- 1 root root 0 Aug 14 13:05 test/init.d/killprocs
-rw-r--r-- 1 root root 0 Aug 14 13:05 test/init.d/rsync
-rw-r--r-- 1 root root 0 Aug 14 13:05 test/init.d/single
drwxr-xr-x 2 root root 4096 Aug 14 13:06 test/rc1.d
lrwxrwxrwx 1 root root 15 May 14 07:55 test/rc1.d/K20rsync -> ../init.d/rsync
lrwxrwxrwx 1 root root 14 Jul 15 01:16 test/rc1.d/K88dbus -> ../init.d/dbus
-rw-r--r-- 1 root root 369 Aug 7 10:31 test/rc1.d/README
lrwxrwxrwx 1 root root 19 May 14 07:53 test/rc1.d/S30killprocs -> ../init.d/killprocs
lrwxrwxrwx 1 root root 21 May 14 07:53 test/rc1.d/S70bootlogs.sh -> ../init.d/bootlogs.sh
lrwxrwxrwx 1 root root 16 May 14 07:53 test/rc1.d/S90single -> ../init.d/single

so the lenny+karmic combination makes mksquashfs incapable of handling symlinks it seems.

summary: - I/O error and stacktrace on squashfs on MMC on Babbage 2.5
+ imx51 squashfs unusable
Oliver Grawert (ogra) wrote :

oh, i forgot, the squashfs produced by manoao is neither loop mountable nor unsquashable anywhere

trying to reproduce the segfaults and breakage on a babbage board running jaunty with karmic chroot shows that this combination executes the commands fine and produces a usable squashfs

summary: - imx51 squashfs unusable
+ imx51 squashfs as built on the armel livefs builder is broken
Oliver Grawert (ogra) wrote :

the build server side is being taken care of, i adjusted the description to point out the relevance for the asignee.

while it is nice to have a warning from squashfs in dmesg, it would be nice if it would not look like an oops with stacktrace, can the warning message that is show be quietened down and not contain the backtrace by default ?

summary: - imx51 squashfs as built on the armel livefs builder is broken
+ imx51 squashfs as built on the armel livefs builder is broken and
+ produces noisy kernel messages
Amit Kucheria (amitk) wrote :

No, unfortunately it can't be quietened down. The messages are due to the way the BUG and WARN macros in the kernel are defined. And they serve a useful function in that they allow kerneloops.org to track what messages/oops/warnings are being triggered most and what is the callpath.

Loïc Minier (lool) wrote :

This was an unidentified kernel issue on the v5 livefs buildd; we switched to more recent buildds + kernels and the issue is gone

Changed in linux-fsl-imx51 (Ubuntu):
status: New → Invalid
tags: added: iso-testing
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers