Init dies in qemu-system-arm with SIGILL with default instructions

Bug #437629 reported by Loïc Minier
10
This bug affects 2 people
Affects Status Importance Assigned to Milestone
RootStock
Invalid
Undecided
Unassigned

Bug Description

Hi

Booting a freshly created image, qemu hangs with:
[...]
input: ImExPS/2 Generic Explorer Mouse as /class/input/input1
md: Waiting for all devices to be available before autodetect
md: If you don't use raid, use raid=noautodetect
md: Autodetecting RAID arrays.
md: Scanned 0 and added 0 devices.
md: autorun ...
md: ... autorun DONE.
EXT2-fs warning: mounting unchecked fs, running e2fsck is recommended
VFS: Mounted root (ext2 filesystem).
Freeing init memory: 136K
Kernel panic - not syncing: Attempted to kill init!

I couldn't find out what's happening; the rootfs is ext2, mouting ro or rw doesn't change anything; I tried passing init=/bin/sh /bin/bash /bin/dash and init=/usr/lib/initramfs-tools/bin/busybox; I tried passing a large rootdelay instead of rootwait, but nothing helps. The rootfs does have /dev/console and /dev/tty0 and 1, so not sure what's wrong. I'm not using an initrd, just http://people.canonical.com/~ogra/arm/qemu/kernel/vmlinuz-2.6.28-versatile

Created rootfs with:
sudo ./rootstock -f localhost -l ubuntu -p "ubuntu" --notarball

Running with:
qemu-system-arm -M versatilepb -kernel vmlinuz-2.6.28-versatile -hda qemu-armel-*.img -m 256 -append "root=/dev/sda mem=256M rootwait"

Bye

Revision history for this message
Loïc Minier (lool) wrote :

This is on amd64; not sure whether it's relevant

Revision history for this message
Loïc Minier (lool) wrote : Re: Init dies in qemu-system-arm with SIGILL

So it's due to the choice of kernel: README recommends an older v5 kernel and usage of a v5 CPU, but userspace is v7 nowadays.

Using qemu-system-arm -M versatilepb -cpu cortex-a8 -kernel vmlinuz-2.6.31-rc3versatile1-cortex-a8 with the kernel from http://people.canonical.com/~ogra//arm/qemu/vmlinuz-2.6.31-rc3versatile1-cortex-a8 I could login to /bin/sh.

(However I get numerous IO errors if I try starting init normally)

I think the doc should be updated to point at the current kernel + command line which rootstock uses; or better, add a mode to launch the latest image with rootstock (e.g. "rootstock run-img" or "rootstock-run-img").

summary: - Init dies in qemu-system-arm for an unknown reason
+ Init dies in qemu-system-arm with SIGILL
summary: - Init dies in qemu-system-arm with SIGILL
+ Init dies in qemu-system-arm with SIGILL with default instructions
Revision history for this message
Oliver Grawert (ogra) wrote :

i noticed this before but didnt have enough info to debug it Keybuk told me we should try to just run:
exec /sbin/init
from a system thats been started with init=/bin/bash and configure upstart to spit out debug messages

i will take a look after the karmic beta release ....

Revision history for this message
jaap (jaap-dejong) wrote :

my target arm system crashes too...
target kernel = 2.6.31
target = karmic ubuntu-minimal
host = jaunty

//
[ 6.460000] Freeing init memory: 120K
[ 6.520000] Kernel panic - not syncing: Attempted to kill init!
[ 6.540000] [<c002cb64>] (unwind_backtrace+0x0/0xdc) from [<c0318568>] (panic+0x48/0x118)
[ 6.570000] [<c0318568>] (panic+0x48/0x118) from [<c0040d6c>] (do_exit+0x70/0x6cc)
[ 6.600000] [<c0040d6c>] (do_exit+0x70/0x6cc) from [<c0041480>] (do_group_exit+0xb8/0xec)
[ 6.630000] [<c0041480>] (do_group_exit+0xb8/0xec) from [<c004cd88>] (get_signal_to_deliver+0x3b0/0x40c)
[ 6.660000] [<c004cd88>] (get_signal_to_deliver+0x3b0/0x40c) from [<c0029558>] (do_signal+0x6c/0x544)
[ 6.690000] [<c0029558>] (do_signal+0x6c/0x544) from [<c0026e4c>] (work_pending+0x1c/0x20)

Revision history for this message
Loïc Minier (lool) wrote :

Did you try launching qemu with the kernel and command line options I indicated in comment #2?

Revision history for this message
Oliver Grawert (ogra) wrote :

@jaap, i assume you use your own kernel ? please note that by default versatile the kernels have no v6+ support, karmic userspace binaries require v6 at least

Revision history for this message
jaap (jaap-dejong) wrote :

@oliver, v6 = IPv6? allowed as module?

Revision history for this message
Oliver Grawert (ogra) wrote :

no, the ARMv6 architecture

Revision history for this message
jaap (jaap-dejong) wrote :

ahhh! yep it is arm9 (at91sam9263)

Revision history for this message
Oliver Grawert (ogra) wrote :

right, that wont work with karmic binaries

Revision history for this message
jaap (jaap-dejong) wrote :

are you sure?
A few weeks ago, when I first started working with your tool, I was able to create a karmic rootfs which worked without major problems on this same processor...

Revision history for this message
Oliver Grawert (ogra) wrote :

everything built past august 28th (past the gcc upload below) requires at least ARMv6 CPUs:

gcc-4.4 (4.4.1-3ubuntu2) karmic; urgency=low

  * On armel, configure --with-arch=armv6 --with-tune=cortex-a8
    --with-float=softfp --with-fpu=vfp.
...

Revision history for this message
jaap (jaap-dejong) wrote :

very bad news :-(((

Oliver Grawert (ogra)
Changed in project-rootstock:
status: New → Invalid
Revision history for this message
Alain SAURAT (maisondouf) wrote :

 """ On armel, configure --with-arch=armv6 --with-tune=cortex-a8
    --with-float=softfp --with-fpu=vfp. """

When I see this line, I ask myself if I can fix the architecture to arm v5 to create packages for my plug computer and install lucid on it ...

Is this only a dream ?

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.