On Tue, Sep 08, 2026 at 12:53:52PM -0400, Mathieu Desnoyers wrote:
> On 2026-09-08 12:22, Paul E. McKenney wrote:
> > On Tue, Sep 08, 2026 at 11:29:31AM -0400, Mathieu Desnoyers wrote:
> > > On 2026-09-08 11:24, Bradley Morgan wrote:
> > > > On 8 September 2026 16:22:14 BST, Mathieu Desnoyers
> > > > <[email protected]> wrote:
> > > > > When hazptr_acquire loads a NULL pointer, it sets:
> > > > >
> > > > > - slot_item->slot.addr = NULL,
> > > > > - slot_item->ctx.ctx = ctx
> > > > > - ctx->slot = slot
> > > > >
> > > > > And it returns NULL.
> > > > >
> > > > > Then hazptr_detach is called on this ctx, it will act on the ctx as if
> > > > > needed to be promoted to backup slot, even though it has a NULL addr.
> > > > >
> > > > > Looking at what hazptr_note_context_switch() does before promoting
> > > > > to backup slot, it checks for a NULL slot->addr, which is exactly
> > > > > what is missing from hazptr_detach.
> > > > >
> > > > > With this in place there would be no need to explicitly check the
> > > > > hazptr_acquire() return value before calling hazptr_detach().
> > > > >
> > > > > hazptr_release() has a early return check for NULL addr as well, so it
> > > > > makes sense that detach does an early return (no-op) similarly.
> > > > >
> > > >
> > > > You shall kill me for this!!
> > > >
> > > > Could you perhaps do a splat in ze commit description pls?
> > >
> > > The splat is available at the "Closes" URL below. I'm not sure whether
> > > we should duplicate this verbose information ?
> > >
> > > Paul ?
> >
> > I am fine either way, as long as the information is reasonably easily
> > accessible. Which is the case either way. ;-)
>
> Then I favor a concise commit message, leaving the splat details to the
> "Closes" link. Feel free to pick up this patch for testing.
>
> Thank you!
Done, thank you! With this applied, the CPU-hotplug-related failures
are gone. Or at least much lower probability. ;-)
Thanx, Paul
