scsi: target: vhost-scsi: Convert to new submission API
authorMike Christie <michael.christie@oracle.com>
Sat, 27 Feb 2021 16:59:52 +0000 (10:59 -0600)
committerMartin K. Petersen <martin.petersen@oracle.com>
Thu, 4 Mar 2021 22:37:01 +0000 (17:37 -0500)
commiteb929804db7c3525bc302aa043369193edd36542
treec72d233d9bce6714cf7f10f64d40ac53c0cc33ce
parent12340930a3e10a3bec110f808be9eeb301969abf
scsi: target: vhost-scsi: Convert to new submission API

target_submit_cmd_map_sgls() is being removed, so convert vhost-scsi to the
new submission API. This has it use target_init_cmd(),
target_submit_prep(), target_submit() because we need to have LIO core map
sgls which is now done in target_submit_prep(), and in the next patches we
will do the target_submit step from the LIO workqueue.

Note: vhost-scsi never calls target_stop_session() so
target_submit_cmd_map_sgls() never failed (in the new API target_init_cmd()
handles target_stop_session() being called when cmds are being
submitted). If it were to have used target_stop_session() and got an error,
we would have hit a refcount bug like xen and usb, because it does:

if (rc < 0) {
transport_send_check_condition_and_sense(se_cmd,
TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE, 0);
transport_generic_free_cmd(se_cmd, 0);
}

transport_send_check_condition_and_sense() calls queue_status which does
transport_generic_free_cmd(), and then we do an extra
transport_generic_free_cmd() call above which would have dropped the
refcount to -1 and the refcount code would spit out errors.

Link: https://lore.kernel.org/r/20210227170006.5077-12-michael.christie@oracle.com
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/vhost/scsi.c