Re: [PATCH v4 08/36] arm: Implement the new page table range API
From: Mike Rapoport
Date: Wed Mar 15 2023 - 05:50:40 EST
On Wed, Mar 15, 2023 at 05:14:16AM +0000, Matthew Wilcox (Oracle) wrote:
> Add set_ptes(), update_mmu_cache_range(), flush_dcache_folio() and
> flush_icache_pages(). Change the PG_dcache_clear flag from being per-page
> to per-folio which makes __dma_page_dev_to_cpu() a bit more exciting.
> Also add flush_cache_pages(), even though this isn't used by generic code
> (yet?)
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
> Cc: Russell King <linux@xxxxxxxxxxxxxxx>
> Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
Acked-by: Mike Rapoport (IBM) <rppt@xxxxxxxxxx>
> ---
> arch/arm/include/asm/cacheflush.h | 24 +++++---
> arch/arm/include/asm/pgtable.h | 5 +-
> arch/arm/include/asm/tlbflush.h | 13 ++--
> arch/arm/mm/copypage-v4mc.c | 5 +-
> arch/arm/mm/copypage-v6.c | 5 +-
> arch/arm/mm/copypage-xscale.c | 5 +-
> arch/arm/mm/dma-mapping.c | 24 ++++----
> arch/arm/mm/fault-armv.c | 14 ++---
> arch/arm/mm/flush.c | 99 +++++++++++++++++++------------
> arch/arm/mm/mm.h | 2 +-
> arch/arm/mm/mmu.c | 14 +++--
> 11 files changed, 125 insertions(+), 85 deletions(-)
>
> diff --git a/arch/arm/include/asm/cacheflush.h b/arch/arm/include/asm/cacheflush.h
> index a094f964c869..841e268d2374 100644
> --- a/arch/arm/include/asm/cacheflush.h
> +++ b/arch/arm/include/asm/cacheflush.h
> @@ -231,14 +231,15 @@ vivt_flush_cache_range(struct vm_area_struct *vma, unsigned long start, unsigned
> vma->vm_flags);
> }
>
> -static inline void
> -vivt_flush_cache_page(struct vm_area_struct *vma, unsigned long user_addr, unsigned long pfn)
> +static inline void vivt_flush_cache_pages(struct vm_area_struct *vma,
> + unsigned long user_addr, unsigned long pfn, unsigned int nr)
> {
> struct mm_struct *mm = vma->vm_mm;
>
> if (!mm || cpumask_test_cpu(smp_processor_id(), mm_cpumask(mm))) {
> unsigned long addr = user_addr & PAGE_MASK;
> - __cpuc_flush_user_range(addr, addr + PAGE_SIZE, vma->vm_flags);
> + __cpuc_flush_user_range(addr, addr + nr * PAGE_SIZE,
> + vma->vm_flags);
> }
> }
>
> @@ -247,15 +248,17 @@ vivt_flush_cache_page(struct vm_area_struct *vma, unsigned long user_addr, unsig
> vivt_flush_cache_mm(mm)
> #define flush_cache_range(vma,start,end) \
> vivt_flush_cache_range(vma,start,end)
> -#define flush_cache_page(vma,addr,pfn) \
> - vivt_flush_cache_page(vma,addr,pfn)
> +#define flush_cache_pages(vma, addr, pfn, nr) \
> + vivt_flush_cache_pages(vma, addr, pfn, nr)
> #else
> -extern void flush_cache_mm(struct mm_struct *mm);
> -extern void flush_cache_range(struct vm_area_struct *vma, unsigned long start, unsigned long end);
> -extern void flush_cache_page(struct vm_area_struct *vma, unsigned long user_addr, unsigned long pfn);
> +void flush_cache_mm(struct mm_struct *mm);
> +void flush_cache_range(struct vm_area_struct *vma, unsigned long start, unsigned long end);
> +void flush_cache_pages(struct vm_area_struct *vma, unsigned long user_addr,
> + unsigned long pfn, unsigned int nr);
> #endif
>
> #define flush_cache_dup_mm(mm) flush_cache_mm(mm)
> +#define flush_cache_page(vma, addr, pfn) flush_cache_pages(vma, addr, pfn, 1)
>
> /*
> * flush_icache_user_range is used when we want to ensure that the
> @@ -289,7 +292,9 @@ extern void flush_cache_page(struct vm_area_struct *vma, unsigned long user_addr
> * See update_mmu_cache for the user space part.
> */
> #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1
> -extern void flush_dcache_page(struct page *);
> +void flush_dcache_page(struct page *);
> +void flush_dcache_folio(struct folio *folio);
> +#define flush_dcache_folio flush_dcache_folio
>
> #define ARCH_IMPLEMENTS_FLUSH_KERNEL_VMAP_RANGE 1
> static inline void flush_kernel_vmap_range(void *addr, int size)
> @@ -321,6 +326,7 @@ static inline void flush_anon_page(struct vm_area_struct *vma,
> * duplicate cache flushing elsewhere performed by flush_dcache_page().
> */
> #define flush_icache_page(vma,page) do { } while (0)
> +#define flush_icache_pages(vma, page, nr) do { } while (0)
>
> /*
> * flush_cache_vmap() is used when creating mappings (eg, via vmap,
> diff --git a/arch/arm/include/asm/pgtable.h b/arch/arm/include/asm/pgtable.h
> index a58ccbb406ad..841001ab495c 100644
> --- a/arch/arm/include/asm/pgtable.h
> +++ b/arch/arm/include/asm/pgtable.h
> @@ -207,8 +207,9 @@ static inline void __sync_icache_dcache(pte_t pteval)
> extern void __sync_icache_dcache(pte_t pteval);
> #endif
>
> -void set_pte_at(struct mm_struct *mm, unsigned long addr,
> - pte_t *ptep, pte_t pteval);
> +void set_ptes(struct mm_struct *mm, unsigned long addr,
> + pte_t *ptep, pte_t pteval, unsigned int nr);
> +#define set_ptes set_ptes
>
> static inline pte_t clear_pte_bit(pte_t pte, pgprot_t prot)
> {
> diff --git a/arch/arm/include/asm/tlbflush.h b/arch/arm/include/asm/tlbflush.h
> index 0ccc985b90af..7d792e485f4f 100644
> --- a/arch/arm/include/asm/tlbflush.h
> +++ b/arch/arm/include/asm/tlbflush.h
> @@ -619,18 +619,21 @@ extern void flush_bp_all(void);
> * If PG_dcache_clean is not set for the page, we need to ensure that any
> * cache entries for the kernels virtual memory range are written
> * back to the page. On ARMv6 and later, the cache coherency is handled via
> - * the set_pte_at() function.
> + * the set_ptes() function.
> */
> #if __LINUX_ARM_ARCH__ < 6
> -extern void update_mmu_cache(struct vm_area_struct *vma, unsigned long addr,
> - pte_t *ptep);
> +void update_mmu_cache_range(struct vm_area_struct *vma, unsigned long addr,
> + pte_t *ptep, unsigned int nr);
> #else
> -static inline void update_mmu_cache(struct vm_area_struct *vma,
> - unsigned long addr, pte_t *ptep)
> +static inline void update_mmu_cache_range(struct vm_area_struct *vma,
> + unsigned long addr, pte_t *ptep, unsigned int nr)
> {
> }
> #endif
>
> +#define update_mmu_cache(vma, addr, ptep) \
> + update_mmu_cache_range(vma, addr, ptep, 1)
> +
> #define update_mmu_cache_pmd(vma, address, pmd) do { } while (0)
>
> #endif
> diff --git a/arch/arm/mm/copypage-v4mc.c b/arch/arm/mm/copypage-v4mc.c
> index f1da3b439b96..7ddd82b9fe8b 100644
> --- a/arch/arm/mm/copypage-v4mc.c
> +++ b/arch/arm/mm/copypage-v4mc.c
> @@ -64,10 +64,11 @@ static void mc_copy_user_page(void *from, void *to)
> void v4_mc_copy_user_highpage(struct page *to, struct page *from,
> unsigned long vaddr, struct vm_area_struct *vma)
> {
> + struct folio *src = page_folio(from);
> void *kto = kmap_atomic(to);
>
> - if (!test_and_set_bit(PG_dcache_clean, &from->flags))
> - __flush_dcache_page(page_mapping_file(from), from);
> + if (!test_and_set_bit(PG_dcache_clean, &src->flags))
> + __flush_dcache_folio(folio_flush_mapping(src), src);
>
> raw_spin_lock(&minicache_lock);
>
> diff --git a/arch/arm/mm/copypage-v6.c b/arch/arm/mm/copypage-v6.c
> index d8a115de5507..a1a71f36d850 100644
> --- a/arch/arm/mm/copypage-v6.c
> +++ b/arch/arm/mm/copypage-v6.c
> @@ -69,11 +69,12 @@ static void discard_old_kernel_data(void *kto)
> static void v6_copy_user_highpage_aliasing(struct page *to,
> struct page *from, unsigned long vaddr, struct vm_area_struct *vma)
> {
> + struct folio *src = page_folio(from);
> unsigned int offset = CACHE_COLOUR(vaddr);
> unsigned long kfrom, kto;
>
> - if (!test_and_set_bit(PG_dcache_clean, &from->flags))
> - __flush_dcache_page(page_mapping_file(from), from);
> + if (!test_and_set_bit(PG_dcache_clean, &src->flags))
> + __flush_dcache_folio(folio_flush_mapping(src), src);
>
> /* FIXME: not highmem safe */
> discard_old_kernel_data(page_address(to));
> diff --git a/arch/arm/mm/copypage-xscale.c b/arch/arm/mm/copypage-xscale.c
> index bcb485620a05..f1e29d3e8193 100644
> --- a/arch/arm/mm/copypage-xscale.c
> +++ b/arch/arm/mm/copypage-xscale.c
> @@ -84,10 +84,11 @@ static void mc_copy_user_page(void *from, void *to)
> void xscale_mc_copy_user_highpage(struct page *to, struct page *from,
> unsigned long vaddr, struct vm_area_struct *vma)
> {
> + struct folio *src = page_folio(from);
> void *kto = kmap_atomic(to);
>
> - if (!test_and_set_bit(PG_dcache_clean, &from->flags))
> - __flush_dcache_page(page_mapping_file(from), from);
> + if (!test_and_set_bit(PG_dcache_clean, &src->flags))
> + __flush_dcache_folio(folio_flush_mapping(src), src);
>
> raw_spin_lock(&minicache_lock);
>
> diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
> index 8bc01071474a..5ecfde41d70a 100644
> --- a/arch/arm/mm/dma-mapping.c
> +++ b/arch/arm/mm/dma-mapping.c
> @@ -693,6 +693,7 @@ static void __dma_page_cpu_to_dev(struct page *page, unsigned long off,
> static void __dma_page_dev_to_cpu(struct page *page, unsigned long off,
> size_t size, enum dma_data_direction dir)
> {
> + struct folio *folio = page_folio(page);
> phys_addr_t paddr = page_to_phys(page) + off;
>
> /* FIXME: non-speculating: not required */
> @@ -707,19 +708,18 @@ static void __dma_page_dev_to_cpu(struct page *page, unsigned long off,
> * Mark the D-cache clean for these pages to avoid extra flushing.
> */
> if (dir != DMA_TO_DEVICE && size >= PAGE_SIZE) {
> - unsigned long pfn;
> - size_t left = size;
> -
> - pfn = page_to_pfn(page) + off / PAGE_SIZE;
> - off %= PAGE_SIZE;
> - if (off) {
> - pfn++;
> - left -= PAGE_SIZE - off;
> + ssize_t left = size;
> + size_t offset = offset_in_folio(folio, paddr);
> +
> + if (offset) {
> + left -= folio_size(folio) - offset;
> + folio = folio_next(folio);
> }
> - while (left >= PAGE_SIZE) {
> - page = pfn_to_page(pfn++);
> - set_bit(PG_dcache_clean, &page->flags);
> - left -= PAGE_SIZE;
> +
> + while (left >= (ssize_t)folio_size(folio)) {
> + set_bit(PG_dcache_clean, &folio->flags);
> + left -= folio_size(folio);
> + folio = folio_next(folio);
> }
> }
> }
> diff --git a/arch/arm/mm/fault-armv.c b/arch/arm/mm/fault-armv.c
> index 0e49154454a6..e2c869b8f012 100644
> --- a/arch/arm/mm/fault-armv.c
> +++ b/arch/arm/mm/fault-armv.c
> @@ -178,8 +178,8 @@ make_coherent(struct address_space *mapping, struct vm_area_struct *vma,
> *
> * Note that the pte lock will be held.
> */
> -void update_mmu_cache(struct vm_area_struct *vma, unsigned long addr,
> - pte_t *ptep)
> +void update_mmu_cache_range(struct vm_area_struct *vma, unsigned long addr,
> + pte_t *ptep, unsigned int nr)
> {
> unsigned long pfn = pte_pfn(*ptep);
> struct address_space *mapping;
> @@ -192,13 +192,13 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long addr,
> * The zero page is never written to, so never has any dirty
> * cache lines, and therefore never needs to be flushed.
> */
> - page = pfn_to_page(pfn);
> - if (page == ZERO_PAGE(0))
> + if (is_zero_pfn(pfn))
> return;
>
> - mapping = page_mapping_file(page);
> - if (!test_and_set_bit(PG_dcache_clean, &page->flags))
> - __flush_dcache_page(mapping, page);
> + folio = page_folio(pfn_to_page(pfn));
> + mapping = folio_flush_mapping(page);
> + if (!test_and_set_bit(PG_dcache_clean, &folio->flags))
> + __flush_dcache_folio(mapping, folio);
> if (mapping) {
> if (cache_is_vivt())
> make_coherent(mapping, vma, addr, ptep, pfn);
> diff --git a/arch/arm/mm/flush.c b/arch/arm/mm/flush.c
> index 7ff9feea13a6..07ea0ab51099 100644
> --- a/arch/arm/mm/flush.c
> +++ b/arch/arm/mm/flush.c
> @@ -95,10 +95,10 @@ void flush_cache_range(struct vm_area_struct *vma, unsigned long start, unsigned
> __flush_icache_all();
> }
>
> -void flush_cache_page(struct vm_area_struct *vma, unsigned long user_addr, unsigned long pfn)
> +void flush_cache_pages(struct vm_area_struct *vma, unsigned long user_addr, unsigned long pfn, unsigned int nr)
> {
> if (cache_is_vivt()) {
> - vivt_flush_cache_page(vma, user_addr, pfn);
> + vivt_flush_cache_pages(vma, user_addr, pfn, nr);
> return;
> }
>
> @@ -196,29 +196,31 @@ void copy_to_user_page(struct vm_area_struct *vma, struct page *page,
> #endif
> }
>
> -void __flush_dcache_page(struct address_space *mapping, struct page *page)
> +void __flush_dcache_folio(struct address_space *mapping, struct folio *folio)
> {
> /*
> * Writeback any data associated with the kernel mapping of this
> * page. This ensures that data in the physical page is mutually
> * coherent with the kernels mapping.
> */
> - if (!PageHighMem(page)) {
> - __cpuc_flush_dcache_area(page_address(page), page_size(page));
> + if (!folio_test_highmem(folio)) {
> + __cpuc_flush_dcache_area(folio_address(folio),
> + folio_size(folio));
> } else {
> unsigned long i;
> if (cache_is_vipt_nonaliasing()) {
> - for (i = 0; i < compound_nr(page); i++) {
> - void *addr = kmap_atomic(page + i);
> + for (i = 0; i < folio_nr_pages(folio); i++) {
> + void *addr = kmap_local_folio(folio,
> + i * PAGE_SIZE);
> __cpuc_flush_dcache_area(addr, PAGE_SIZE);
> - kunmap_atomic(addr);
> + kunmap_local(addr);
> }
> } else {
> - for (i = 0; i < compound_nr(page); i++) {
> - void *addr = kmap_high_get(page + i);
> + for (i = 0; i < folio_nr_pages(folio); i++) {
> + void *addr = kmap_high_get(folio_page(folio, i));
> if (addr) {
> __cpuc_flush_dcache_area(addr, PAGE_SIZE);
> - kunmap_high(page + i);
> + kunmap_high(folio_page(folio, i));
> }
> }
> }
> @@ -230,15 +232,14 @@ void __flush_dcache_page(struct address_space *mapping, struct page *page)
> * userspace colour, which is congruent with page->index.
> */
> if (mapping && cache_is_vipt_aliasing())
> - flush_pfn_alias(page_to_pfn(page),
> - page->index << PAGE_SHIFT);
> + flush_pfn_alias(folio_pfn(folio), folio_pos(folio));
> }
>
> -static void __flush_dcache_aliases(struct address_space *mapping, struct page *page)
> +static void __flush_dcache_aliases(struct address_space *mapping, struct folio *folio)
> {
> struct mm_struct *mm = current->active_mm;
> - struct vm_area_struct *mpnt;
> - pgoff_t pgoff;
> + struct vm_area_struct *vma;
> + pgoff_t pgoff, pgoff_end;
>
> /*
> * There are possible user space mappings of this page:
> @@ -246,21 +247,36 @@ static void __flush_dcache_aliases(struct address_space *mapping, struct page *p
> * data in the current VM view associated with this page.
> * - aliasing VIPT: we only need to find one mapping of this page.
> */
> - pgoff = page->index;
> + pgoff = folio->index;
> + pgoff_end = pgoff + folio_nr_pages(folio) - 1;
>
> flush_dcache_mmap_lock(mapping);
> - vma_interval_tree_foreach(mpnt, &mapping->i_mmap, pgoff, pgoff) {
> - unsigned long offset;
> + vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff_end) {
> + unsigned long start, offset, pfn;
> + unsigned int nr;
>
> /*
> * If this VMA is not in our MM, we can ignore it.
> */
> - if (mpnt->vm_mm != mm)
> + if (vma->vm_mm != mm)
> continue;
> - if (!(mpnt->vm_flags & VM_MAYSHARE))
> + if (!(vma->vm_flags & VM_MAYSHARE))
> continue;
> - offset = (pgoff - mpnt->vm_pgoff) << PAGE_SHIFT;
> - flush_cache_page(mpnt, mpnt->vm_start + offset, page_to_pfn(page));
> +
> + start = vma->vm_start;
> + pfn = folio_pfn(folio);
> + nr = folio_nr_pages(folio);
> + offset = pgoff - vma->vm_pgoff;
> + if (offset > -nr) {
> + pfn -= offset;
> + nr += offset;
> + } else {
> + start += offset * PAGE_SIZE;
> + }
> + if (start + nr * PAGE_SIZE > vma->vm_end)
> + nr = (vma->vm_end - start) / PAGE_SIZE;
> +
> + flush_cache_pages(vma, start, pfn, nr);
> }
> flush_dcache_mmap_unlock(mapping);
> }
> @@ -269,7 +285,7 @@ static void __flush_dcache_aliases(struct address_space *mapping, struct page *p
> void __sync_icache_dcache(pte_t pteval)
> {
> unsigned long pfn;
> - struct page *page;
> + struct folio *folio;
> struct address_space *mapping;
>
> if (cache_is_vipt_nonaliasing() && !pte_exec(pteval))
> @@ -279,14 +295,14 @@ void __sync_icache_dcache(pte_t pteval)
> if (!pfn_valid(pfn))
> return;
>
> - page = pfn_to_page(pfn);
> + folio = page_folio(pfn_to_page(pfn));
> if (cache_is_vipt_aliasing())
> - mapping = page_mapping_file(page);
> + mapping = folio_flush_mapping(folio);
> else
> mapping = NULL;
>
> - if (!test_and_set_bit(PG_dcache_clean, &page->flags))
> - __flush_dcache_page(mapping, page);
> + if (!test_and_set_bit(PG_dcache_clean, &folio->flags))
> + __flush_dcache_folio(mapping, folio);
>
> if (pte_exec(pteval))
> __flush_icache_all();
> @@ -312,7 +328,7 @@ void __sync_icache_dcache(pte_t pteval)
> * Note that we disable the lazy flush for SMP configurations where
> * the cache maintenance operations are not automatically broadcasted.
> */
> -void flush_dcache_page(struct page *page)
> +void flush_dcache_folio(struct folio *folio)
> {
> struct address_space *mapping;
>
> @@ -320,31 +336,36 @@ void flush_dcache_page(struct page *page)
> * The zero page is never written to, so never has any dirty
> * cache lines, and therefore never needs to be flushed.
> */
> - if (page == ZERO_PAGE(0))
> + if (is_zero_pfn(folio_pfn(folio)))
> return;
>
> if (!cache_ops_need_broadcast() && cache_is_vipt_nonaliasing()) {
> - if (test_bit(PG_dcache_clean, &page->flags))
> - clear_bit(PG_dcache_clean, &page->flags);
> + if (test_bit(PG_dcache_clean, &folio->flags))
> + clear_bit(PG_dcache_clean, &folio->flags);
> return;
> }
>
> - mapping = page_mapping_file(page);
> + mapping = folio_flush_mapping(folio);
>
> if (!cache_ops_need_broadcast() &&
> - mapping && !page_mapcount(page))
> - clear_bit(PG_dcache_clean, &page->flags);
> + mapping && !folio_mapped(folio))
> + clear_bit(PG_dcache_clean, &folio->flags);
> else {
> - __flush_dcache_page(mapping, page);
> + __flush_dcache_folio(mapping, folio);
> if (mapping && cache_is_vivt())
> - __flush_dcache_aliases(mapping, page);
> + __flush_dcache_aliases(mapping, folio);
> else if (mapping)
> __flush_icache_all();
> - set_bit(PG_dcache_clean, &page->flags);
> + set_bit(PG_dcache_clean, &folio->flags);
> }
> }
> -EXPORT_SYMBOL(flush_dcache_page);
> +EXPORT_SYMBOL(flush_dcache_folio);
>
> +void flush_dcache_page(struct page *page)
> +{
> + flush_dcache_folio(page_folio(page));
> +}
> +EXPORT_SYMBOL(flush_dcache_page);
> /*
> * Flush an anonymous page so that users of get_user_pages()
> * can safely access the data. The expected sequence is:
> diff --git a/arch/arm/mm/mm.h b/arch/arm/mm/mm.h
> index d7ffccb7fea7..419316316711 100644
> --- a/arch/arm/mm/mm.h
> +++ b/arch/arm/mm/mm.h
> @@ -45,7 +45,7 @@ struct mem_type {
>
> const struct mem_type *get_mem_type(unsigned int type);
>
> -extern void __flush_dcache_page(struct address_space *mapping, struct page *page);
> +void __flush_dcache_folio(struct address_space *mapping, struct folio *folio);
>
> /*
> * ARM specific vm_struct->flags bits.
> diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
> index 463fc2a8448f..9947bbc32b04 100644
> --- a/arch/arm/mm/mmu.c
> +++ b/arch/arm/mm/mmu.c
> @@ -1788,7 +1788,7 @@ void __init paging_init(const struct machine_desc *mdesc)
> bootmem_init();
>
> empty_zero_page = virt_to_page(zero_page);
> - __flush_dcache_page(NULL, empty_zero_page);
> + __flush_dcache_folio(NULL, page_folio(empty_zero_page));
> }
>
> void __init early_mm_init(const struct machine_desc *mdesc)
> @@ -1797,8 +1797,8 @@ void __init early_mm_init(const struct machine_desc *mdesc)
> early_paging_init(mdesc);
> }
>
> -void set_pte_at(struct mm_struct *mm, unsigned long addr,
> - pte_t *ptep, pte_t pteval)
> +void set_ptes(struct mm_struct *mm, unsigned long addr,
> + pte_t *ptep, pte_t pteval, unsigned int nr)
> {
> unsigned long ext = 0;
>
> @@ -1808,5 +1808,11 @@ void set_pte_at(struct mm_struct *mm, unsigned long addr,
> ext |= PTE_EXT_NG;
> }
>
> - set_pte_ext(ptep, pteval, ext);
> + for (;;) {
> + set_pte_ext(ptep, pteval, ext);
> + if (--nr == 0)
> + break;
> + ptep++;
> + pte_val(pteval) += PAGE_SIZE;
> + }
> }
> --
> 2.39.2
>
--
Sincerely yours,
Mike.