Re: [RFC PATCH v3 09/13] mm/damon/vaddr: Add vaddr versions of migrate_{hot,cold}

[SeongJae Park <sj@kernel.org>]

On Wed,  2 Jul 2025 16:51:38 -0700 SeongJae Park <sj@kernel.org> wrote:

> On Wed,  2 Jul 2025 15:13:32 -0500 Bijan Tabatabai <bijan311@gmail.com> wrote:
> 
> > From: Bijan Tabatabai <bijantabatab@micron.com>
> > 
> > migrate_{hot,cold} are paddr schemes that are used to migrate hot/cold
> > data to a specified node. However, these schemes are only available when
> > doing physical address monitoring. This patch adds an implementation for
> > them virtual address monitoring as well.
> > 
> > Co-developed-by: Ravi Shankar Jonnalagadda <ravis.opensrc@micron.com>
> > Signed-off-by: Ravi Shankar Jonnalagadda <ravis.opensrc@micron.com>
> > Signed-off-by: Bijan Tabatabai <bijantabatab@micron.com>
> > ---
> >  mm/damon/vaddr.c | 102 +++++++++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 102 insertions(+)
> > 
> > diff --git a/mm/damon/vaddr.c b/mm/damon/vaddr.c
> > index 46554e49a478..5cdfdc47c5ff 100644
> > --- a/mm/damon/vaddr.c
> > +++ b/mm/damon/vaddr.c
> > @@ -15,6 +15,7 @@
> >  #include <linux/pagewalk.h>
> >  #include <linux/sched/mm.h>
> >  
> > +#include "../internal.h"
> >  #include "ops-common.h"
> >  
> >  #ifdef CONFIG_DAMON_VADDR_KUNIT_TEST
> > @@ -610,6 +611,65 @@ static unsigned int damon_va_check_accesses(struct damon_ctx *ctx)
> >  	return max_nr_accesses;
> >  }
> >  
> > +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > +static int damos_va_migrate_pmd_entry(pmd_t *pmd, unsigned long addr,
> > +		unsigned long next, struct mm_walk *walk)
> 
> I'd suggest to put CONFIG_TRANSPARENT_HUGEPAGE check into the body of this
> function and handle both pmd and pte here, consistent to
> damon_young_pmd_entry().

Ah, unlike damon_young_pmd_entry() which is for a single address, this is for
walking the range of a given DAMON region, and hence should have a separate pte
entry function.  Please ignore the above comment.

[...]
> > +static int damos_va_migrate_pte_entry(pte_t *pte, unsigned long addr,
> > +		unsigned long enxt, struct mm_walk *walk)
> 
> Nit.  s/enxt/next/ ?
> 
> > +{
> > +	struct list_head *migration_list = walk->private;
> > +	struct folio *folio;
> > +	pte_t ptent;
> > +
> > +	ptent = ptep_get(pte);
> > +	if (pte_none(*pte) || !pte_present(*pte))
> > +		return 0;
> 
> Shouldn't we use cached pte value (ptent) instad of *pte?  I'd suggest merging
> this into damos_va_migrate_pmd_entry() consistent to damon_young_pmd_entry().

Again, I overlooked the fact that this is for walking not only single address
point but a range.  Please ignore the latter suggestion.


Thanks,
SJ

[...]