On Wed, Apr 15, 2026 at 07:29:04PM -0400, Stefan Hajnoczi wrote:
> The SCSI Persistent Reservation code can fail if another node has interfered
> with the reservation (e.g. by sending a PERSISTENT RESERVE OUT command with a
> PREEMPT or CLEAR service action). This is rare but can happen if the
> application or clustering framework performs recovery and wants to take away
> the guest's reservation.
>
> This patch series adds an extra step when loading device state that
> synchronizes the transferred vmstate against the actual reservation state of
> the disk before moving the reservation. This way we can gracefully handle
> changes to reservations that were performed externally.
>
> Stefan Hajnoczi (2):
> scsi: change buf_size to unsigned int in scsi_SG_IO()
> scsi: handle reservation changes across migration
>
> include/hw/scsi/scsi.h | 3 +-
> hw/scsi/scsi-generic.c | 175 +++++++++++++++++++++++++++++++++++------
> 2 files changed, 152 insertions(+), 26 deletions(-)
Ping.
Stefan