Re: [PATCH v1 2/2] virtio-blk: mark disk dead on ERS permanent failure

[Xixin Liu]

Hi,

On Thu, 30 Jul 2026, Michael S. Tsirkin wrote:
> Can you tell me was this written with use of AI? I ask because not many
> people go out of their way to put in unicode dashes. Next step is what
> emojis? Let's not go there pls.
> And if yes pls disclose that as per Documentation.

Yes. I will add Assisted-by for the LLM help in v2.

> Do we need this? virtqueue add will fail anyway.

For a broken vq, yes, virtqueue_add already fails. The useful part is
the !vblk->vqs check: after reset_prepare/freeze_priv the vqs pointer
can be NULL, and virtio_queue_rq would deref it before add. v2 drops
disk_live() and keeps only the NULL guards.

> why not break first?

Agreed. v2 calls virtio_break_device() before flush_work().

Thanks,
Xixin