On 3/17/21 5:46 AM, Dmitry Bogdanov wrote: > All target transport drivers have its own 'tpg/enable' attribute > implementation. This produces unnecessary code duplication I don't think that is correct. Some drivers don't have en enable attr: - vhost-scsi, xen and loop have a nexus attr that sort of leaves it in the equivalent of enabled. - usb has a nexus file like above, but after you write to it you still have to write to the enable attr. - tcm_fc does not have an enable and has it's own initialization strategy. For drivers that have an enable file your patches missed usb, srpt and ibm_vscsi. > Also it makes difficult to control that attribute and to depend on that > attribute inside of target core module. I agree with this :) It's a bit of mess, but I think it's sometimes due to how the driver is implemented and how userspace has to set it up, so it's not as simple as in this patchset due to having to support existing tools.
