Re: [PATCH v2 2/3] ksm: perform a range-walk in break_ksm

[Pedro Demarchi Gomes]

On Mon, Nov 03, 2025 at 06:06:26PM +0100, David Hildenbrand (Red Hat) wrote:
> On 31.10.25 18:46, Pedro Demarchi Gomes wrote:
> > Make break_ksm() receive an address range and change
> > break_ksm_pmd_entry() to perform a range-walk and return the address of
> > the first ksm page found.
> > 
> > This change allows break_ksm() to skip unmapped regions instead of
> > iterating every page address. When unmerging large sparse VMAs, this
> > significantly reduces runtime.
> > 
> > In a benchmark unmerging a 32 TiB sparse virtual address space where
> > only one page was populated, the runtime dropped from 9 minutes to less
> > then 5 seconds.
> > 
> > Suggested-by: David Hildenbrand <david@xxxxxxxxxx>
> > Signed-off-by: Pedro Demarchi Gomes <pedrodemargomes@xxxxxxxxx>
> > ---
> >   mm/ksm.c | 88 ++++++++++++++++++++++++++++++--------------------------
> >   1 file changed, 48 insertions(+), 40 deletions(-)
> > 
> > diff --git a/mm/ksm.c b/mm/ksm.c
> > index 922d2936e206..64d66699133d 100644
> > --- a/mm/ksm.c
> > +++ b/mm/ksm.c
> > @@ -607,35 +607,55 @@ static inline bool ksm_test_exit(struct mm_struct *mm)
> >   	return atomic_read(&mm->mm_users) == 0;
> >   }
> > -static int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next,
> > +struct break_ksm_arg {
> > +	unsigned long addr;
> > +};
> 
> Leftover? :)
> 

Yes, I am sorry.
I will remove it in the v3.

> > +
> > +static int break_ksm_pmd_entry(pmd_t *pmdp, unsigned long addr, unsigned long end,
> >   			struct mm_walk *walk)
> >   {
> > -	struct folio *folio = NULL;
> > +	unsigned long *found_addr = (unsigned long *) walk->private;
> > +	struct mm_struct *mm = walk->mm;
> > +	pte_t *start_ptep, *ptep;
> >   	spinlock_t *ptl;
> > -	pte_t *pte;
> > -	pte_t ptent;
> > -	int ret;
> > +	int found = 0;
> 
> Best to perform the ret -> found rename already in patch #1.
>

Ok

> With both things
> 
> Acked-by: David Hildenbrand (Red Hat) <david@xxxxxxxxxx>

Thanks!

> -- 
> Cheers
> 
> David
>