Re: [PATCH v3 08/17] mm/sparse-vmemmap: support section-based vmemmap optimization

[Muchun Song]

On 2026/8/6 11:04, Muchun Song wrote:
> On 2026/8/4 11:55, Muchun Song wrote:
> > Teach sparse-vmemmap population code to use the compound page order
> > when deciding whether a vmemmap page can be optimized.
> >
> > With this information, the common sparse-vmemmap population path can
> > allocate or reuse shared tail vmemmap pages directly instead of relying
> > on HugeTLB-specific handling.
> >
> > This centralizes vmemmap optimization logic in the sparse-vmemmap code,
> > based on section metadata, and prepares for sharing the same mechanism
> > across different users of vmemmap optimization, including HugeTLB and
> > DAX.
> >
> > Signed-off-by: Muchun Song <songmuchun@xxxxxxxxxxxxx>
> > ---
> > v2:
> > - Keep vmemmap accounting and population logic in sparse-vmemmap.c
> >    (suggested by Mike Rapoport)
> > - Move vmemmap_get_tail() before its first use instead of adding only a
> >    forward declaration in the previous patch (suggested by Mike 
> > Rapoport)
> > - Simplify the PMD path handling for HVO-covered sections
> > ---
> >   mm/sparse-vmemmap.c | 36 ++++++++++++++++++++++++++++++------
> >   mm/sparse.c         |  4 ++--
> >   mm/sparse.h         |  7 +++++++
> >   3 files changed, 39 insertions(+), 8 deletions(-)
> >
> > diff --git a/mm/sparse-vmemmap.c b/mm/sparse-vmemmap.c
> > index b770fe2428fd..b69a7af76858 100644
> > --- a/mm/sparse-vmemmap.c
> > +++ b/mm/sparse-vmemmap.c
> > @@ -186,6 +186,11 @@ static __meminit struct page 
> > *vmemmap_get_tail(unsigned int order, struct zone *
> >         return tail;
> >   }
> > +#else
> > +static inline struct page *vmemmap_get_tail(unsigned int order, 
> > struct zone *zone)
> > +{
> > +    return NULL;
> > +}
> >   #endif
> >     static pte_t * __meminit vmemmap_pte_populate(pmd_t *pmd, 
> > unsigned long addr, int node,
> > @@ -193,12 +198,24 @@ static pte_t * __meminit 
> > vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, in
> >                          unsigned long ptpfn, unsigned long flags)
> >   {
> >       pte_t *pte = pte_offset_kernel(pmd, addr);
> > +    unsigned long pfn = page_to_pfn((struct page *)addr);
> > +
> >       if (pte_none(ptep_get(pte))) {
> >           pte_t entry;
> > -        void *p;
> > +
> > +        if (pfn_vmemmap_optimizable(pfn) && ptpfn == (unsigned 
> > long)-1) {
> > +            unsigned int order = pfn_to_section_order(pfn);
> > +            struct zone *zone = pfn_to_zone(pfn, node);
> > +            struct page *page = vmemmap_get_tail(order, zone);
> > +
> > +            if (!page)
> > +                return NULL;
> > +            ptpfn = page_to_pfn(page);
> > +        }
> >             if (ptpfn == (unsigned long)-1) {
> > -            p = vmemmap_alloc_block_buf(PAGE_SIZE, node, altmap);
> > +            void *p = vmemmap_alloc_block_buf(PAGE_SIZE, node, altmap);
> > +
> >               if (!p)
> >                   return NULL;
> >               ptpfn = PHYS_PFN(__pa(p));
> > @@ -217,7 +234,8 @@ static pte_t * __meminit 
> > vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, in
> >           }
> >           entry = pfn_pte(ptpfn, PAGE_KERNEL);
> >           set_pte_at(&init_mm, addr, pte, entry);
> > -    }
> > +    } else if (WARN_ON_ONCE(pfn_vmemmap_optimizable(pfn)))
> > +        return NULL;
> >       return pte;
> >   }
> >   @@ -406,6 +424,9 @@ int __meminit 
> > vmemmap_populate_hugepages(unsigned long start, unsigned long end,
> >       pmd_t *pmd;
> >         for (addr = start; addr < end; addr = next) {
> > +        unsigned long pfn = page_to_pfn((struct page *)addr);
> > +        const struct mem_section *ms = __pfn_to_section(pfn);
> > +
> >           next = pmd_addr_end(addr, end);
> >             pgd = vmemmap_pgd_populate(addr, node);
> > @@ -421,7 +442,7 @@ int __meminit vmemmap_populate_hugepages(unsigned 
> > long start, unsigned long end,
> >               return -ENOMEM;
> >             pmd = pmd_offset(pud, addr);
> > -        if (pmd_none(pmdp_get(pmd))) {
> > +        if (pmd_none(pmdp_get(pmd)) && 
> > !section_vmemmap_optimizable(ms)) {
> >               void *p;
> >                 p = vmemmap_alloc_block_buf(PMD_SIZE, node, altmap);
> > @@ -439,8 +460,11 @@ int __meminit 
> > vmemmap_populate_hugepages(unsigned long start, unsigned long end,
> >                    */
> >                   return -ENOMEM;
> >               }
> > -        } else if (vmemmap_check_pmd(pmd, node, addr, next))
> > +        } else if (vmemmap_check_pmd(pmd, node, addr, next)) {
> > +            if (WARN_ON_ONCE(section_vmemmap_optimizable(ms)))
> > +                return -ENOTSUPP;
> >               continue;
> > +        }
> >           if (vmemmap_populate_basepages(addr, next, node, altmap))
> >               return -ENOMEM;
> >       }
> > @@ -648,7 +672,7 @@ void offline_mem_sections(unsigned long 
> > start_pfn, unsigned long end_pfn)
> >       }
> >   }
> >   -static int __meminit section_nr_vmemmap_pages(unsigned long pfn, 
> > unsigned long nr_pages,
> > +int __meminit section_nr_vmemmap_pages(unsigned long pfn, unsigned 
> > long nr_pages,
>
> The kernel test robot reported a compilation issue: when
> CONFIG_MEMORY_HOTPLUG = n && CONFIG_SPARSEMEM_VMEMMAP=y,
> section_nr_vmemmap_pages is undefined. This problem is easy to fix, and
> I will move the entire function outside the CONFIG_MEMORY_HOTPLUG guard
> in the next version. 

Hi,

I will wait a few more days to see if there are any reviews. If there is
no further input during this period, I will update a new version to resolve
this compilation issue.

Muchun,
Thanks