Re: Problem with creating a root device

[Roger James]

Peter,

Thank you for your prompt response. It does look like the failure in the scsi driver is what is causing the root device not to be created. I can see that there has been an issue raised on this at the qemu-rpi-kernel repo on github.

I am puzzled about why the kernel thinks there is a scsi device present. Could this be an issue with the device tree that I am using?

I have now  tried yet another config, still no success.

-drive "file=2020-05-27-raspios-buster-lite-armhf.img,if=none,index=0,media=disk,format=raw,id=disk0" \
-device "virtio-blk-pci,drive=disk0,disable-modern=on,disable-legacy=off" \

I think I need to either disable the scsi driver in the kernel, or stop qemu emulating one on the pci bus.

Any ideas/comments anyone.

Roger

On 20 September 2020 21:15:46 Peter Maydell <peter.maydell@linaro.org> wrote:

On Sun, 20 Sep 2020 at 19:37, roger <roger@beardandsandals.co.uk> wrote:

The only other thing I see at the start of the boot process is the scsi device driver timing out, and giving warning stack traces.

sym53c8xx 0000:00:0c.0: enabling device (0100 -> 0103)

sym0: <895a> rev 0x0 at pci 0000:00:0c.0 irq 66

sym0: No NVRAM, ID 7, Fast-40, ??, parity checking

random: fast init done

CACHE TEST FAILED: timeout.

sym0: CACHE INCORRECTLY CONFIGURED.

sym0: giving up ...

------------[ cut here ]------------

This is your problem, I think -- the new kernel fails to correctly

init the SCSI adaptor, and so it never sees the disk, and then

later the kernel panics becasue you told it to use a disk that's

not there. It's identified the adaptor, so your command line is

correct. This looks like it's going to either be a bug in QEMU's

emulation of the scsi card or in the kernel's driver for it.

Googling for the error message I find an email from Paolo

from 2015 (about a similar but not identical failure) advising

that this particular SCSI controller is not a good one to use

because its emulation is incomplete and it's slow:

https://lists.cs.columbia.edu/pipermail/kvmarm/2015-April/014380.html

So my suggestions:

 (1) if you're not using a current QEMU you could try using

     the latest version to see if this particular bug has been

     fixed

 (2) you could try changing your commandline to explicitly

     plug in a better-supported SCSI controller rather than

     relying on the lsi53c8xx that the versatilepb board

     will give you by default. Using a virtio disk device

     of some form is usually the best choice, assuming the

     drivers have been compiled into your guest kernel.

thanks

-- PMM