Re: [PATCH v3 10/26] mm: Add more flags for __apply_to_page_range()

[Yosry Ahmed <yosry@kernel.org>]

On Sun, Jul 26, 2026 at 10:22:43PM +0000, Brendan Jackman wrote:
> Add two flags to make this API more generic:
> 
> 1. Separate "create" into two levels - one to allow creating new
>    mappings without allocating pagetables, and one for the current
>    behaviour that allows both of these.
> 
> 2. Create a new flag to report that the caller has taken care of
>    synchronization and no locks are required.
> 
> Both of these will serve to allow calling this API from restricted
> contexts where allocation and pagetable locking are not possible.
> 
> Signed-off-by: Brendan Jackman <jackmanb@google.com>
> ---
>  mm/internal.h | 26 +++++++++++++++++++++++++-
>  mm/memory.c   | 59 ++++++++++++++++++++++++++++++++++-------------------------
>  2 files changed, 59 insertions(+), 26 deletions(-)
> 
> diff --git a/mm/internal.h b/mm/internal.h
> index 395331a12d62d..5a237d9c5fa96 100644
> --- a/mm/internal.h
> +++ b/mm/internal.h
> @@ -1662,9 +1662,33 @@ static inline bool can_spin_trylock(void)
>  
>  /*
>   * Create a mapping if it doesn't exist. (Otherwise, skip regions with no
> - * existing mapping, and return an error for regions with no leaf pagetable).
> + * existing mapping). This doesn't allow allocating, most users will want
> + * PGRANGE_ALLOC.
> + *
> + * Do not test this bit directly as it is implied by PGRANGE_ALLOC, use
> + * pgrange_create() instead.
>   */
>  #define PGRANGE_CREATE		(1 << 0)
> +/*
> + * Allocate a pagetable if one is missing. (Otherwise, return an error for
> + * regions with no leaf pagetable). Also implies PGRANGE_CREATE.
> + *
> + * Note that __apply_to_page_range() assumes that pagetables for the area are
> + * already initialised down to PMD level, so this only affects PTEs in practice.
> + */
> +#define PGRANGE_ALLOC		(1 << 1)
> +/*
> + * Do not take any locks. This means the caller has taken care of
> + * synchronisation. This is incompatible with PGRANGE_ALLOC and also with
> + * mm=&init_mm.
> + */
> +#define PGRANGE_NOLOCK		(1 << 2)

I assume this is used by the mermap as locking is not required because
the mappings are per-CPU and migration is disabled while the mermap is
used?

Also, why is this incompatible with init_mm? It actually seems like
apply_to_pte_range() is always lockless for init_mm (uses
pte_offset_kernel()), probably callers are also synchronizing in their
own way (e.g. exclusive access to a vmap area?).

> +
> +
> +static inline bool pgrange_create(unsigned int flags)
> +{
> +	return flags & (PGRANGE_CREATE | PGRANGE_ALLOC);
> +}
>  
>  int __apply_to_page_range(struct mm_struct *mm, unsigned long addr,
>  			  unsigned long size, pte_fn_t fn,
> diff --git a/mm/memory.c b/mm/memory.c
> index c4defefea1574..d00508db1021e 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -3441,30 +3441,35 @@ static int apply_to_pte_range(struct mm_struct *mm, pmd_t *pmd,
>  				     pte_fn_t fn, void *data, unsigned int flags,
>  				     pgtbl_mod_mask *mask)
>  {
> -	bool create = flags & PGRANGE_CREATE;
>  	pte_t *pte, *mapped_pte;
>  	int err = 0;
>  	spinlock_t *ptl;
>  
> -	if (create) {
> +	if (flags & PGRANGE_ALLOC) {
> +		VM_WARN_ON(flags & PGRANGE_NOLOCK);
> +
>  		mapped_pte = pte = (mm == &init_mm) ?
>  			pte_alloc_kernel_track(pmd, addr, mask) :
>  			pte_alloc_map_lock(mm, pmd, addr, &ptl);
>  		if (!pte)
>  			return -ENOMEM;
>  	} else {
> -		mapped_pte = pte = (mm == &init_mm) ?
> -			pte_offset_kernel(pmd, addr) :
> -			pte_offset_map_lock(mm, pmd, addr, &ptl);
> +		if (mm == &init_mm)
> +			pte = pte_offset_kernel(pmd, addr);
> +		else if (flags & PGRANGE_NOLOCK)
> +			pte = pte_offset_map(pmd, addr);
> +		else
> +			pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
>  		if (!pte)
>  			return -EINVAL;
> +		mapped_pte = pte;
>  	}
>  
>  	lazy_mmu_mode_enable();
>  
>  	if (fn) {
>  		do {
> -			if (create || !pte_none(ptep_get(pte))) {
> +			if (pgrange_create(flags) || !pte_none(ptep_get(pte))) {
>  				err = fn(pte, addr, data);
>  				if (err)
>  					break;
> @@ -3475,8 +3480,13 @@ static int apply_to_pte_range(struct mm_struct *mm, pmd_t *pmd,
>  
>  	lazy_mmu_mode_disable();
>  
> -	if (mm != &init_mm)
> -		pte_unmap_unlock(mapped_pte, ptl);
> +	if (mm != &init_mm) {
> +		if (flags & PGRANGE_NOLOCK)
> +			pte_unmap(mapped_pte);
> +		else
> +			pte_unmap_unlock(mapped_pte, ptl);
> +	}
> +
>  	return err;
>  }
[..]