On 09/04/16 16:13, Aneesh Kumar K.V wrote:
> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar at linux.vnet.ibm.com>
> ---
> arch/powerpc/include/asm/book3s/64/pgalloc.h | 34 ++++++++++++++++++++++++----
> arch/powerpc/include/asm/book3s/64/pgtable.h | 10 ++++++--
> arch/powerpc/mm/hash_utils_64.c | 7 ++++++
> arch/powerpc/mm/pgtable-radix.c | 5 +++-
> arch/powerpc/mm/pgtable_64.c | 6 +++++
> 5 files changed, 55 insertions(+), 7 deletions(-)
>
> diff --git a/arch/powerpc/include/asm/book3s/64/pgalloc.h b/arch/powerpc/include/asm/book3s/64/pgalloc.h
> index faad1319ba26..a282674b2378 100644
> --- a/arch/powerpc/include/asm/book3s/64/pgalloc.h
> +++ b/arch/powerpc/include/asm/book3s/64/pgalloc.h
> @@ -50,19 +50,45 @@ extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, int shift);
> extern void __tlb_remove_table(void *_table);
> #endif
>
> +static inline pgd_t *rpgd_alloc(struct mm_struct *mm)
> +{
> +#ifdef CONFIG_PPC_64K_PAGES
> + return (pgd_t *)__get_free_page(PGALLOC_GFP);
> +#else
> + struct page *page;
> + page = alloc_pages(PGALLOC_GFP, 4);
We need a 2^4*4k = 64k page for PGD even for 4k pages? Why? One concern is that PGALLOC_GFP does not use CMA if I read the PGALLOC_GFP flags correct
> + if (!page)
> + return NULL;
> + return (pgd_t *) page_address(page);
> +#endif
> +}
> +
> +static inline void rpgd_free(struct mm_struct *mm, pgd_t *pgd)
> +{
> +#ifdef CONFIG_PPC_64K_PAGES
> + free_page((unsigned long)pgd);
> +#else
> + free_pages((unsigned long)pgd, 4);
> +#endif
> +}
> +
> static inline pgd_t *pgd_alloc(struct mm_struct *mm)
> {
> + if (radix_enabled())
> + return rpgd_alloc(mm);
> return kmem_cache_alloc(PGT_CACHE(PGD_INDEX_SIZE), GFP_KERNEL);
> }
>
> static inline void pgd_free(struct mm_struct *mm, pgd_t *pgd)
> {
> + if (radix_enabled())
> + return rpgd_free(mm, pgd);
> kmem_cache_free(PGT_CACHE(PGD_INDEX_SIZE), pgd);
> }
>
> static inline void pgd_populate(struct mm_struct *mm, pgd_t *pgd, pud_t *pud)
> {
> - pgd_set(pgd, __pgtable_ptr_val(pud));
> + pgd_set(pgd, __pgtable_ptr_val(pud) | PGD_VAL_BITS);
> }
>
> static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
> @@ -78,7 +104,7 @@ static inline void pud_free(struct mm_struct *mm, pud_t *pud)
>
> static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
> {
> - pud_set(pud, __pgtable_ptr_val(pmd));
> + pud_set(pud, __pgtable_ptr_val(pmd) | PUD_VAL_BITS);
> }
>
> static inline void __pud_free_tlb(struct mmu_gather *tlb, pud_t *pud,
> @@ -107,13 +133,13 @@ static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd,
> static inline void pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd,
> pte_t *pte)
> {
> - pmd_set(pmd, __pgtable_ptr_val(pte));
> + pmd_set(pmd, __pgtable_ptr_val(pte) | PMD_VAL_BITS);
> }
>
> static inline void pmd_populate(struct mm_struct *mm, pmd_t *pmd,
> pgtable_t pte_page)
> {
> - pmd_set(pmd, __pgtable_ptr_val(pte_page));
> + pmd_set(pmd, __pgtable_ptr_val(pte_page) | PMD_VAL_BITS);
> }
>
> static inline pgtable_t pmd_pgtable(pmd_t pmd)
> diff --git a/arch/powerpc/include/asm/book3s/64/pgtable.h b/arch/powerpc/include/asm/book3s/64/pgtable.h
> index c16037116625..b8ee70458bae 100644
> --- a/arch/powerpc/include/asm/book3s/64/pgtable.h
> +++ b/arch/powerpc/include/asm/book3s/64/pgtable.h
> @@ -170,11 +170,17 @@ extern unsigned long __pgd_table_size;
> #define PMD_TABLE_SIZE __pmd_table_size
> #define PUD_TABLE_SIZE __pud_table_size
> #define PGD_TABLE_SIZE __pgd_table_size
> +
> +extern unsigned long __pmd_val_bits;
> +extern unsigned long __pud_val_bits;
> +extern unsigned long __pgd_val_bits;
> +#define PMD_VAL_BITS __pmd_val_bits
> +#define PUD_VAL_BITS __pud_val_bits
> +#define PGD_VAL_BITS __pgd_val_bits
> /*
> * Pgtable size used by swapper, init in asm code
> - * We will switch this later to radix PGD
> */
> -#define MAX_PGD_TABLE_SIZE (sizeof(pgd_t) << H_PGD_INDEX_SIZE)
> +#define MAX_PGD_TABLE_SIZE (sizeof(pgd_t) << R_PGD_INDEX_SIZE)
>
> #define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
> #define PTRS_PER_PMD (1 << PMD_INDEX_SIZE)
> diff --git a/arch/powerpc/mm/hash_utils_64.c b/arch/powerpc/mm/hash_utils_64.c
> index be5d123b3f61..aef691b75784 100644
> --- a/arch/powerpc/mm/hash_utils_64.c
> +++ b/arch/powerpc/mm/hash_utils_64.c
> @@ -878,6 +878,13 @@ void __init hlearly_init_mmu(void)
> __pmd_table_size = H_PMD_TABLE_SIZE;
> __pud_table_size = H_PUD_TABLE_SIZE;
> __pgd_table_size = H_PGD_TABLE_SIZE;
> + /*
> + * 4k use hugepd format, so for hash set then to
^ them
> + * zero
> + */
The comment is not very clear
> + __pmd_val_bits = 0;
> + __pud_val_bits = 0;
> + __pgd_val_bits = 0;
> /* Initialize the MMU Hash table and create the linear mapping
> * of memory. Has to be done before SLB initialization as this is
> * currently where the page size encoding is obtained.
> diff --git a/arch/powerpc/mm/pgtable-radix.c b/arch/powerpc/mm/pgtable-radix.c
> index 52ed820a386a..5a0400fb5d71 100644
> --- a/arch/powerpc/mm/pgtable-radix.c
> +++ b/arch/powerpc/mm/pgtable-radix.c
> @@ -318,8 +318,11 @@ void __init rearly_init_mmu(void)
> __pud_table_size = R_PUD_TABLE_SIZE;
> __pgd_table_size = R_PGD_TABLE_SIZE;
>
> - radix_init_page_sizes();
> + __pmd_val_bits = R_PMD_VAL_BITS;
> + __pud_val_bits = R_PUD_VAL_BITS;
> + __pgd_val_bits = R_PGD_VAL_BITS;
>
> + radix_init_page_sizes();
> if (!firmware_has_feature(FW_FEATURE_LPAR))
> radix_init_partition_table();
>
> diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c
> index 78910035cbca..a58259793198 100644
> --- a/arch/powerpc/mm/pgtable_64.c
> +++ b/arch/powerpc/mm/pgtable_64.c
> @@ -91,6 +91,12 @@ unsigned long __pud_table_size;
> EXPORT_SYMBOL(__pud_table_size);
> unsigned long __pgd_table_size;
> EXPORT_SYMBOL(__pgd_table_size);
> +unsigned long __pmd_val_bits;
> +EXPORT_SYMBOL(__pmd_val_bits);
> +unsigned long __pud_val_bits;
> +EXPORT_SYMBOL(__pud_val_bits);
> +unsigned long __pgd_val_bits;
> +EXPORT_SYMBOL(__pgd_val_bits);
>
> #endif
> unsigned long ioremap_bot = IOREMAP_BASE;
>