Re: [PATCH v3 2/2] mm: disable CONFIG_PER_VMA_LOCK until its fixed

[Suren Baghdasaryan]

On Wed, Jul 5, 2023 at 5:44 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> On Wed, 5 Jul 2023 17:32:09 -0700 Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote:
>
> > On Wed, Jul 5, 2023 at 5:30 PM Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote:
> > >
> > > On Wed, Jul 5, 2023 at 5:24 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> > > >
> > > > On Wed, 5 Jul 2023 13:33:26 -0700 Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote:
> > > >
> > > > > I was hoping we could re-enable VMA locks in 6.4 once we get more
> > > > > confirmations that the problem is gone. Is that not possible once the
> > > > > BROKEN dependency is merged?
> > > >
> > > > I think "no".  By doing this we're effectively backporting a minor
> > > > performance optimization, which isn't a thing we'd normally do.
> > >
> > > In that case, maybe for 6.4 we send the fix and only disable it by
> > > default without marking BROKEN? That way we still have a way to enable
> > > it if desired?
> >
> > I'm preparing the next version with Liam's corrections. If the above
> > option I suggested is acceptable I can send a modified second patch
> > which would not have BROKEN dependency.
>
> I think just mark it broken and move on.  At some later time we can
> consider backporting the fixes into 6.4.x and reenabling, but I don't
> think it's likely that we'll do this.

Uh, ok. I'll send the next version shortly with the patch fixing the
issue and another one marking it BROKEN. Hopefully in the next version
we can roll it our more carefully, removing BROKEN dependency but
keeping it disabled by default?

>