On Thu, Mar 19, 2026 at 10:19:00PM -0400, Martin K. Petersen wrote: > > This patch avoids a kernel warning that may occur if a virtio_scsi > > controller is detached immediately following a disk detach. See the > > commit message for details. The following are instructions to > > produce the warning (without the proposed patch). > > A few issues were flagged. Please review: > > > https://sashiko.dev/#/patchset/20260316153341.2062278-1-jdaley%40linux.ibm.com
Hi Joshua, I am responding to the following sashiko review comment (haven't figured out a way to reply in the web UI or via direct email to sashiko). I feel responsible for this one since I suggested the change that sashiko is questioning. I haven't looked at the other review comments, please triage them yourself. From Sashiko: > Does this code violate the virtio-scsi specification? > > The specification mandates that a driver must not place buffers into the > event virtqueue if neither VIRTIO_SCSI_F_HOTPLUG nor VIRTIO_SCSI_F_CHANGE > has been negotiated. > > By completely removing the VIRTIO_SCSI_F_HOTPLUG check without expanding it > to check for VIRTIO_SCSI_F_CHANGE, could this unconditionally populate the > event queue and cause strict implementations to reject the buffers or > transition the device into a broken state? No, this is a hallucination. The spec does not mandate that a driver must not place buffers into the event virtqueue when neither VIRTIO_SCSI_F_HOTPLUG nor VIRTIO_SCSI_F_CHANGE has been negotiated: https://docs.oasis-open.org/virtio/virtio/v1.4/virtio-v1.4.html#x1-4510006 The event virtqueue still serves a purpose when both VIRTIO_SCSI_F_HOTPLUG and VIRTIO_SCSI_F_CHANGE are not negotiated. For example, see "Asynchronous notification subscription" and the VIRTIO_SCSI_T_ASYNC_NOTIFY event type. Stefan
signature.asc
Description: PGP signature
