On Wed, Sep 03, 2025 at 05:48:46PM +0200, Miklos Szeredi wrote:
> On Tue, 26 Aug 2025 at 20:52, Darrick J. Wong <djwong@xxxxxxxxxx> wrote:
>
> > Hrmm. I was thinking that it would be very nice to have
> > fuse_request_{send,end} bracket the start and end of a fuse request,
> > even if we kill it immediately.
>
> I'm fine with that, and would possibly simplify some code that checks
> for an error and calls ->end manually. But that makes it a
> non-trivial change unfortunately.
Yes, and then you have to poke the idr structure for a request id even
if that caller already knows that the connection's dead. That seems
like a waste of cycles, but OTOH maybe we just don't care?
(Though I suppose seeing more than one request id of zero in the trace
output implies very strongly that the connection is really dead)
--D
> Thanks,
> Miklos
>
