On 04/03/2023 00:29, Bart Van Assche wrote:
Access the qla2xxx_driver_template data structure directly instead of via
the host pointer. This patch prepares for declaring the 'hostt' pointer
const.
Cc: Nilesh Javali <njavali@xxxxxxxxxxx>
Signed-off-by: Bart Van Assche <bvanassche@xxxxxxx>
---
drivers/scsi/qla2xxx/qla_target.c | 4 ++--
Hi Bart,
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c
index aa0cf5ca6c1c..8d9a6aa3ea61 100644
--- a/drivers/scsi/qla2xxx/qla_target.c
+++ b/drivers/scsi/qla2xxx/qla_target.c
@@ -6395,8 +6395,8 @@ int qlt_add_target(struct qla_hw_data *ha, struct scsi_qla_host *base_vha)
return -ENOMEM;
}
- if (!(base_vha->host->hostt->supported_mode & MODE_TARGET))
- base_vha->host->hostt->supported_mode |= MODE_TARGET;
+ if (!(qla2xxx_driver_template.supported_mode & MODE_TARGET))
+ qla2xxx_driver_template.supported_mode |= MODE_TARGET;
So we're saying if that MODE_TARGET bit is not set, then set it. It
would be neater to just always set it, right?
Apart from that, I will say that I haven't studied the driver in detail,
but my impression is that we should just set this flag per-shost in
base_vha->host.active_mode, and not the host template supported_mode
member. Indeed, we don't even seem to be making this driver
scsi_host_template as const in this series, which I thought was the aim
(and I assume because of this).
Thanks,
John
rc = btree_init64(&tgt->lun_qpair_map);
if (rc) {
