Re: [PATCH] mm/madvise: reclaim isolated folios if PTE restart fails

[Gregory Price]

On Tue, Sep 15, 2026 at 04:55:47PM +0100, Lorenzo Stoakes (ARM) wrote:
> > diff --git a/mm/madvise.c b/mm/madvise.c
> > index 574aa2bb7c7e..ba5a3d77241a 100644
> > --- a/mm/madvise.c
> > +++ b/mm/madvise.c
> > @@ -464,7 +464,7 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *pmd,
> 
> This function is truly some of the absolute worst code I've read in
> mm. Shocking.
>

relevant meme: https://i.imgur.com/W0N40BK.png

No argument here. This function in particular is really horrendous.

I think i'm now 4 or 5 stacks deep on the "Thanks for fixing a bug, but
while reviewing your fix sashiko found yet another CVE-worthy bug".

I'm convinced (like you below) madvise needs some attention.

> >  restart:
> >  	start_pte = pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
> >  	if (!start_pte)
> > -		return 0;
> > +		goto out;
> >  	flush_tlb_batched_pending(mm);
> >  	lazy_mmu_mode_enable();
> >  	for (; addr < end; pte += nr, addr += nr * PAGE_SIZE) {
> > @@ -568,6 +568,7 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *pmd,
> >  			folio_deactivate(folio);
> >  	}
> >
> > +out:
> 
> Seems odd to put this here for a condition that explicitly guarantees
> !start_pte? It should be before the if (pageout) reclaim_pages(...);
> surely?
> 

I wanted to prevent any future use of the label from skipping the
start_pte check.  The function is so bad I was err'ing on the side
of being a bit defensive without making it worse.

I suppose reasonable to say the goto is worse.

> I honestly wonder whether, rather than adding yet another label/goto into this
> absolute bloody mess, whether we should just live with a bit of duplication and do:
> 
> 	if (!start_pte) {
> 		if (pageout)
> 			reclaim_pages(&folio_list);
> 		return 0;
> 	}
> 
> I actually think that'd be clearer at this point than throwing in some more
> indirection.
> 

I'll spin a v2.

> That's for the backport but somebody needs to rework this entire bloody
> function going forwards...

The longer I look at madvise, the more I think it is a wart on mm/
as-written.  A userland-driven page table walker that requires all
the special-casing of Yooj Pages, zone device, swap, soft leafs, a
bunch of tlb flushing 5-tabs deep...

Scary surface.

I've been trying to think about how we might make this more reasonable,
but generalizing the different walkers across mm is pretty hard.

~Gregory