On 2017年01月09日 16:17, Mike Christie wrote:
On 01/06/2017 05:26 PM, Andy Grover wrote:
On 01/05/2017 06:30 PM, Xiubo Li wrote:
What is the point of the target_core_user timer though? The initiators
already do the same thing, so it seems like one more thing to configure.
For example, for tape commands we will want to increase it, so why
configure it twice. Is it just a hack around target_core_tmr.c not being
able to call into the backend modules to perform TMFs?
If there is a good use for the timer, should it be in userspace? It
seems like it would be easier for tcmu-runner to drive it. We could have
a tcmu-runner timer in userspace, call into the handlers and when they
kill a command we can fail it with a task aborted status or something
like that.
Sounds good.
Is it just supposed to detect when tcmu-runner has died?
Yep, that's why it's there. No doubt a misunderstanding of the
requirements on my part.
What I'm hearing is: we can eliminate the timeouts in TCMU now, but then
we need to add a way to get initiator-generated task management
functions (TMFs) like ABORT_TASK to backstores, and once we have that,
we can add an opcode to pass the TMF from the backstore to userspace.
Yeah, that sounds good.
That's fine.
We will try to fix this.
Thanks,
BRs
Xiubo Li
--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
