Re: [PATCH v4 0/3] Only free healthy pages in high-order has_hwpoisoned folio

[Lorenzo Stoakes]

On Wed, May 27, 2026 at 09:29:19PM -0700, Jiaqi Yan wrote:
> On Tue, May 26, 2026 at 11:36 PM Lorenzo Stoakes <ljs@xxxxxxxxxx> wrote:
> >
> > On Tue, May 26, 2026 at 10:00:06PM -0700, Jiaqi Yan wrote:
> > > Thanks Lorenzo and Oscar.
> > >
> > > On Tue, May 26, 2026 at 10:17 AM Oscar Salvador (SUSE)
> > > <osalvador@xxxxxxxxxx> wrote:
> > > >
> > > > On Tue, May 26, 2026 at 05:09:28PM +0100, Lorenzo Stoakes wrote:
> > > > > On Mon, Feb 23, 2026 at 03:17:57PM -0800, Jiaqi Yan wrote:
> > > > > > Hi Vlastimil,
> > > > > >
> > > > > > Could you and other page_alloc.c reviewers share your thoughts on this
> > > > > > patchset? Thanks!
> > > > >
> > > > > Hi Jiaqi,
> > > > >
> > > > > I guess you are blocked on review here, did you intend to return to this?
> > >
> > > Yes, as long as this patchset is useful to the community.
> >
> > Yes it is, thanks!
> >
> > >
> > > > >
> > > > > I'd suggest attempted a respin to get some movement here as clearly this slipped
> > > > > through the gaps last cycle.
> > > >
> > > > This fell off my radar, I plan to have a look this week/early next week.
> >
> > Thanks Oscar!
> >
> > > >
> > > >
> > >
> > > By "respin" do you mean I should re-work the patches on top of the
> > > latest akpm/mm-stable or linus/master, for whatever the changes in
> > > mm/page_alloc.c and mm/memory-failure.c since Feb?
> >
> > Rebase on to mm-unstable would be easiest, we're late in the cycle now (rc5) so
> > this will probably have to wait for next cycle, but doing this early makes
> > reviewing easier :)
> >
> > You can either do a v4 RESEND or it may be simplier to just send a v5 where the
> > latest version is just the rebase to avoid confusion.
>
> Ack. I am preparing v5 on top of mm-unstable.

Great thanks!

>
> Interestingly, I notice [1] introduced free_prepared_contig_range()
> which does almost exactly what free_contiguous_pages() in my patch 1/3
> does. I will try to re-use free_prepared_contig_range() during the
> rebase.
>
> [1] https://lore.kernel.org/all/20260401101634.2868165-2-usama.anjum@xxxxxxx

Nice :)

>
> >
> > >
> > > > --
> > > > Oscar Salvador
> > > > SUSE Labs
> >
> > Cheers, Lorenzo

Cheers, Lorenzo