On 2023/9/6 23:56, Mike Christie wrote:
On 9/6/23 6:15 AM, haowenchao (C) wrote:
If the driver supports performing multiple TMFs/resets in parallel then why
not always enable it?
Not all hardware/drivers support performing multiple TMFs/resets in parallel,
so I think it is necessary to call scsi_device_setup_eh/scsi_device_clear_eh
in specific drivers.
Ah shoot sorry. I edited my email before I sent it and dropped part of it.
For the scsi_device_setup_eh/scsi_device_clear_eh comment I just meant it could
be a scsi_host_template field. scsi-ml would then see it and do the
scsi_device_setup_eh/scsi_device_clear_eh calls for the driver. The drivers
then don't have to deal with doing slave callouts and handling errors.
Also for the error handling case I think we want to still proceed if
scsi_device_setup_eh fails. Just use the old EH in that case.
It looks better to add setup_eh/clear_eh in scsi_host_template, I would update in
next version.
