On 12/16/24 13:53, Klaus Jensen wrote:
> Apart from some random small fixes here and there, the major thing here
> is cleaning up how we handle command sets. Prior to this series, the
> controller would not correctly validate namespace command sets against
> CC.CSS. This is fixed here.
>
> The most clean way of doing this (as far as I could tell) was to make
> sure an nvme-subsys device exists (creating it if necessary). This
> allows us to "store" the namespaces in the subsystem, using existing
> functionality, and attach supported namespaces when the device is
> started (instead of when the device is created/realized).
>
> Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
> ---
> Klaus Jensen (9):
> hw/nvme: always initialize a subsystem
> hw/nvme: make oacs dynamic
> hw/nvme: add knob for doorbell buffer config support
> nvme: fix iocs status code values
> hw/nvme: be compliant wrt. dsm processing limits
> hw/nvme: rework csi handling
> hw/nvme: only set command abort requested when cancelled due to Abort
> hw/nvme: set error status code explicitly for misc commands
> hw/nvme: remove nvme_aio_err()
>
> hw/nvme/ctrl.c | 430 ++++++++++++++++++++++++++++-----------------------
> hw/nvme/ns.c | 62 ++------
> hw/nvme/nvme.h | 10 +-
> include/block/nvme.h | 22 ++-
> 4 files changed, 276 insertions(+), 248 deletions(-)
> ---
> base-commit: ca80a5d026a280762e0772615f1988db542b3ade
> change-id: 20241216-nvme-queue-f4151c5d7507
>
> Best regards,
Read, looks good to me.
Reviewed-by: Jesper Wendel Devantier <foss@defmacro.it>