Re: [Qemu-discuss] How to debug blocked QEMU run

[Peter Maydell]

On 6 January 2016 at 04:55, Ilya Katsnelson <address@hidden> wrote:
> I'm trying to test my ARM build on QEMU from Yocto script.
>
> The command line is this:
>
>
> runqemu qemuarm qemuparams="-D /tmp/a.log -smp cpus=1 -nographic -serial
> mon:stdio -append 'console=ttyS0'"
> ./tmp/deploy/images/vcm4-qemu/zImage-vcm4-qemu.bin
> ./tmp/deploy/images/vcm4-qemu/core-image-minimal-vcm4-qemu.ext4
>
>
> The core rootfs is correct because I've tested it with a different kernel.
> With the different kernel the simulator comes up fine, but with this one it
> just tits there at this output:

> Is there any way to know what is really going on in QEMU? To see what it is
> really unhappy about? Otherwise I'm flying blind and can't really debug the
> problem.

If kernel A works and kernel B doesn't with the same command line
then the problem is probably in the kernel configuration. If you
get the kernel config wrong then booting it on QEMU will behave
exactly as if you booted a busted kernel on real hardware: it will
crash before it gets round to printing anything on the serial port.
I would try checking the difference between the two configs, and
in particular making sure that the kernel is built for the right
machine type (looks like Versatile PB926). If the kernel needs a
device tree blob (modern ones often do) then you should also supply
that to QEMU. Finally, try enabling the earlyprintk support in
the kernel and turning it on via the kernel command line -- this
may result in some output useful to debugging if the kernel
crashed after producing some output but before the console driver
got itself set up to actually send it anywhere.

thanks
-- PMM