On 21 May 2019, at 11:58, J . Bruce Fields wrote:
> On Mon, May 20, 2019 at 10:33:07AM -0400, Benjamin Coddington wrote:
>> This reverts most of commit b8eee0e90f97 ("lockd: Show pid of lockd for
>> remote locks"), which caused remote locks to not be differentiated between
>> remote processes for NLM.
>
> To make sure I understand: I assume a client resolves conflicts among
> its own processes before involving the server, so the server only needs
> to resolve conflicts among nlm_hosts. Is that right? So the only
> practical affect is the missing grant callbacks that Xuewei noticed?
For the linux client, I think that is correct.
> Or is my assumption not true in general? Or is it true only for some
> client implementations?
I don't know the answer here. I imagine a client could leave off checking
for local lock conflicts.
I have a set of patches that stuff nlm_lockowner into fl_owner on the
server. That allows us to set fl_pid of lockd and not lose the svid. I am
just testing them thoroughly at this point, and I'll probably get them
posted tomorrow.
Ben
