On Mon, Jun 5, 2023 at 7:06 AM Will Deacon <will@xxxxxxxxxx> wrote:
>
> Hi Peter,
>
> On Mon, May 22, 2023 at 05:43:08PM -0700, Peter Collingbourne wrote:
> > Commit c145e0b47c77 ("mm: streamline COW logic in do_swap_page()") moved
> > the call to swap_free() before the call to set_pte_at(), which meant that
> > the MTE tags could end up being freed before set_pte_at() had a chance
> > to restore them. Fix it by adding a call to the arch_swap_restore() hook
> > before the call to swap_free().
> >
> > Signed-off-by: Peter Collingbourne <pcc@xxxxxxxxxx>
> > Link: https://linux-review.googlesource.com/id/I6470efa669e8bd2f841049b8c61020c510678965
> > Cc: <stable@xxxxxxxxxxxxxxx> # 6.1
> > Fixes: c145e0b47c77 ("mm: streamline COW logic in do_swap_page()")
> > Reported-by: Qun-wei Lin (林群崴) <Qun-wei.Lin@xxxxxxxxxxxx>
> > Closes: https://lore.kernel.org/all/5050805753ac469e8d727c797c2218a9d780d434.camel@xxxxxxxxxxxx/
> > Acked-by: David Hildenbrand <david@xxxxxxxxxx>
> > Acked-by: "Huang, Ying" <ying.huang@xxxxxxxxx>
> > Reviewed-by: Steven Price <steven.price@xxxxxxx>
> > Acked-by: Catalin Marinas <catalin.marinas@xxxxxxx>
> > ---
> > v2:
> > - Call arch_swap_restore() directly instead of via arch_do_swap_page()
> >
> > mm/memory.c | 7 +++++++
> > 1 file changed, 7 insertions(+)
> >
> > diff --git a/mm/memory.c b/mm/memory.c
> > index f69fbc251198..fc25764016b3 100644
> > --- a/mm/memory.c
> > +++ b/mm/memory.c
> > @@ -3932,6 +3932,13 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> > }
> > }
> >
> > + /*
> > + * Some architectures may have to restore extra metadata to the page
> > + * when reading from swap. This metadata may be indexed by swap entry
> > + * so this must be called before swap_free().
> > + */
> > + arch_swap_restore(entry, folio);
> > +
> > /*
> > * Remove the swap entry and conditionally try to free up the swapcache.
> > * We're already holding a reference on the page but haven't mapped it
>
> It looks like the intention is for this patch to land in 6.4, whereas the
> other two in the series could go in later, right? If so, I was expecting
> Andrew to pick this one up but he's not actually on CC. I've added him now,
> but you may want to send this as a separate fix so it's obvious what needs
> picking up for this cycle.
I was expecting that this whole series could be picked up in mm. There
was a previous attempt to apply v3 of this series to mm, but that
failed because a dependent patch (commit c4c597f1b367 ("arm64: mte: Do
not set PG_mte_tagged if tags were not initialized")) hadn't been
merged into Linus's master branch yet. The series should be good to go
in now that that patch has been merged.
Peter
