Re: [PATCH for-9.2 00/10] s390: Convert virtio-ccw, cpu to three-phase reset, and followup cleanup

[Peter Maydell]

On Wed, 28 Aug 2024 at 09:13, Nico Boehr <nrb@linux.ibm.com> wrote:
>
> Quoting Nico Boehr (2024-08-26 14:08:20)
> > There was a little hickup without the fixup to patch 2, but after Nina
> > pushed the fixup, we did not observe any failures related to your
> > changes in our CI. Thanks!
>
> Peter, after a few CI runs, we unfortunately did find some issues with your
> patch :-(
>
> Rebooting a guest in a loop sometimes fails. Michael was able to bisect it
> to your series.
>
> The problem is intermittent. The guest is unable to load its initramfs:
>
>   [    0.560674] rootfs image is not initramfs (no cpio magic); looks like an 
> initrd
>   [    0.588605] Freeing initrd memory: 95680K
>   [    0.593143] md: Waiting for all devices to be available before autodetect
>   [    0.593144] md: If you don't use raid, use raid=noautodetect
>   [    0.593145] md: Autodetecting RAID arrays.
>   [    0.593146] md: autorun ...
>   [    0.593147] md: ... autorun DONE.
>   [    0.593156] RAMDISK: gzip image found at block 0
>   [    0.609110] RAMDISK: incomplete write (29120 != 32768)
>   [    0.609113] write error
>
> ...and then a panic because the kernel doesn't find a rootfs.
>
> It seems like the compressed initramfs is corrupted somehow, since "rootfs
> image is not initramfs" doesn't appear on a successful boot.
>
> initramfs and kernel are loaded via direct kernel boot. Running under KVM.
>
> Some vhost error messages do appear before the guest panics, but it is not
> entirely clear to me whether they are related:
>
>   [...]
>   2024-08-28T06:56:29.765324Z qemu-system-s390x: vhost vring error in 
> virtqueue 0: Invalid argument (22)
>   2024-08-28T06:56:32.210982Z qemu-system-s390x: vhost vring error in 
> virtqueue 0: Invalid argument (22)
>   2024-08-28 06:56:35.430+0000: panic s390: core='0' 
> psw-mask='0x0002000180000000' psw-addr='0x00000387b028c67e' 
> reason='disabled-wait'
>
> Any idea?

Well, the series is *supposed* to be just a refactoring, not a change of
behaviour, so I'm not sure. I don't suppose you have a reproduce case
that I can run? (I do have access to an s390 machine if that helps.)

thanks
-- PMM