Re: [RESEND PATCH V10 06/15] monitor: Update tracking iothread users with holder

On Fri, Jul 31, 2026, 3:23 PM Markus Armbruster <[email protected]> wrote:

> Zhang Chen <[email protected]> writes:
>
> > On Tue, Jul 21, 2026 at 8:33 PM Markus Armbruster <[email protected]>
> wrote:
> >>
> >> Zhang Chen <[email protected]> writes:
> >>
> >> > Based on monitor QOM path tracking iothread users with holder.
> >> > Introduce the AioContext in the Monitor struct to avoid repeated
> calls to
> >> > iothread_get_aio_context() and ensure symmetrical ref/unref during
> >> > monitor lifecycle.
> >> >
> >> > Signed-off-by: Zhang Chen <[email protected]>
> >>
> >> Why are repeated calls bad?
> >>
> >> How would symmetry be lost?
> >>
> >
> > OK, will add detailed comments like this in the next version:
> >
> > Avoid repeated calls: Under the holder tracking model, fetching the
> > AioContext increments the IOThread refcount and registers holder
> metadata.
> > Dynamically querying it in helper functions (e.g., suspend, BH
> scheduling)
> > leads to redundant refcount churn.
>
> Right here, we still have
>
>     AioContext *iothread_get_aio_context(IOThread *iothread)
>     {
>         return iothread->ctx;
>     }
>
> and complicating code to avoid additional calls feels dubious.
>
> PATCH 14 changes it to
>
>     AioContext *iothread_get_aio_context(IOThread *iothread,
>                                          const IOThreadHolder *holder)
>     {
>         /* Add IOThreadHolder to the list */
>         iothread_ref(iothread, holder);
>
>         return iothread->ctx;
>      }
>
> Are you making this change in preparation of PATCH 14?
>


Yes, following Stefan's earlier suggestion allows for lifecycle management
to be completed with minimal disruption.

Thanks
Chen


> > Ensure lifecycle symmetry: Storing AioContext in mon->ctx establishes a
> > strict 1-to-1 lifecycle pairing: a single
> iothread_ref_and_get_aio_context()
> > in monitor_complete() matched by a single iothread_put_aio_context()
> > in monitor_cleanup().
> >
> > Thanks
> > Chen
>
>