Re: [PATCH 5/5] powerpc: Don't use SLAB/SLUB for PTE pages
From: Hugh Dickins <hidden>
Date: 2007-05-08 20:41:28
On Tue, 8 May 2007, Benjamin Herrenschmidt wrote:
From: Hugh Dickins <redacted> The SLUB allocator relies on struct page fields first_page and slab, overwritten by ptl when SPLIT_PTLOCK: so the SLUB allocator cannot then be used for the lowest level of pagetable pages. This was obstructing SLUB on PowerPC, which uses kmem_caches for its pagetables. So convert its pte level to use normal gfp pages (whereas pmd, pud and 64k-page pgd want partpages, so continue to use kmem_caches for pmd, pud and pgd). Signed-off-by: Hugh Dickins <redacted> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> ---
Looks and runs fine for me (though would have seemed nicer to use __GFP_ZERO than clear_page - or does that subvert your fast zeroing?). But you also need to undo the ARCH_USES_SLAB_PAGE_STRUCT from arch/powerpc/Kconfig now that's gone into Linus' tree: patch I added in testing below (either append to this one or add a 6/5). For some people that interim disablement was working, for other people it wasn't: mysterious, and the sooner it's got rid of the better. I'm sorry to find that oldconfig doesn't actually offer the SLAB or SLUB menu (to me anyway) when moving from SLUB disabled to SLUB enabled: so again, the sooner we get rid of that interim disablement the better for SLUB exposure. Thanks, Hugh
quoted hunk ↗ jump to hunk
arch/powerpc/mm/init_64.c | 17 ++++++----------- include/asm-powerpc/pgalloc.h | 34 ++++++++++++++++------------------ 2 files changed, 22 insertions(+), 29 deletions(-) Index: linux-cell/arch/powerpc/mm/init_64.c ===================================================================--- linux-cell.orig/arch/powerpc/mm/init_64.c 2007-05-08 11:46:50.000000000 +1000 +++ linux-cell/arch/powerpc/mm/init_64.c 2007-05-08 15:45:46.000000000 +1000@@ -146,21 +146,16 @@ static void zero_ctor(void *addr, struct memset(addr, 0, kmem_cache_size(cache)); } -#ifdef CONFIG_PPC_64K_PAGES -static const unsigned int pgtable_cache_size[3] = { - PTE_TABLE_SIZE, PMD_TABLE_SIZE, PGD_TABLE_SIZE -}; -static const char *pgtable_cache_name[ARRAY_SIZE(pgtable_cache_size)] = { - "pte_pmd_cache", "pmd_cache", "pgd_cache", -}; -#else static const unsigned int pgtable_cache_size[2] = { - PTE_TABLE_SIZE, PMD_TABLE_SIZE + PGD_TABLE_SIZE, PMD_TABLE_SIZE }; static const char *pgtable_cache_name[ARRAY_SIZE(pgtable_cache_size)] = { - "pgd_pte_cache", "pud_pmd_cache", -}; +#ifdef CONFIG_PPC_64K_PAGES + "pgd_cache", "pmd_cache", +#else + "pgd_cache", "pud_pmd_cache", #endif /* CONFIG_PPC_64K_PAGES */ +}; #ifdef CONFIG_HUGETLB_PAGE /* Hugepages need one extra cache, initialized in hugetlbpage.c. WeIndex: linux-cell/include/asm-powerpc/pgalloc.h ===================================================================--- linux-cell.orig/include/asm-powerpc/pgalloc.h 2007-04-27 14:13:24.000000000 +1000 +++ linux-cell/include/asm-powerpc/pgalloc.h 2007-05-08 16:01:00.000000000 +1000@@ -13,18 +13,11 @@ extern struct kmem_cache *pgtable_cache[]; -#ifdef CONFIG_PPC_64K_PAGES -#define PTE_CACHE_NUM 0 -#define PMD_CACHE_NUM 1 -#define PGD_CACHE_NUM 2 -#define HUGEPTE_CACHE_NUM 3 -#else -#define PTE_CACHE_NUM 0 -#define PMD_CACHE_NUM 1 -#define PUD_CACHE_NUM 1 -#define PGD_CACHE_NUM 0 -#define HUGEPTE_CACHE_NUM 2 -#endif +#define PGD_CACHE_NUM 0 +#define PUD_CACHE_NUM 1 +#define PMD_CACHE_NUM 1 +#define HUGEPTE_CACHE_NUM 2 +#define PTE_NONCACHE_NUM 3 /* from GFP rather than kmem_cache */ /* * This program is free software; you can redistribute it and/or@@ -97,8 +90,10 @@ static inline void pmd_free(pmd_t *pmd) static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address) { - return kmem_cache_alloc(pgtable_cache[PTE_CACHE_NUM], - GFP_KERNEL|__GFP_REPEAT); + pte_t *ptepage = (pte_t *)__get_free_page(GFP_KERNEL | __GFP_REPEAT); + if (ptepage) + clear_page(ptepage); + return ptepage; } static inline struct page *pte_alloc_one(struct mm_struct *mm,@@ -109,12 +104,12 @@ static inline struct page *pte_alloc_one static inline void pte_free_kernel(pte_t *pte) { - kmem_cache_free(pgtable_cache[PTE_CACHE_NUM], pte); + free_page((unsigned long)pte); } static inline void pte_free(struct page *ptepage) { - pte_free_kernel(page_address(ptepage)); + __free_page(ptepage); } #define PGF_CACHENUM_MASK 0x3@@ -136,14 +131,17 @@ static inline void pgtable_free(pgtable_ void *p = (void *)(pgf.val & ~PGF_CACHENUM_MASK); int cachenum = pgf.val & PGF_CACHENUM_MASK; - kmem_cache_free(pgtable_cache[cachenum], p); + if (cachenum == PTE_NONCACHE_NUM) + free_page((unsigned long)p); + else + kmem_cache_free(pgtable_cache[cachenum], p); } extern void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf); #define __pte_free_tlb(tlb, ptepage) \ pgtable_free_tlb(tlb, pgtable_free_cache(page_address(ptepage), \ - PTE_CACHE_NUM, PTE_TABLE_SIZE-1)) + PTE_NONCACHE_NUM, PTE_TABLE_SIZE-1)) #define __pmd_free_tlb(tlb, pmd) \ pgtable_free_tlb(tlb, pgtable_free_cache(pmd, \ PMD_CACHE_NUM, PMD_TABLE_SIZE-1))
--- 2.6.21-git9/arch/powerpc/Kconfig 2007-05-08 12:32:33.000000000 +0100
+++ linux/arch/powerpc/Kconfig 2007-05-08 13:30:51.000000000 +0100@@ -117,19 +117,6 @@ config GENERIC_BUG default y depends on BUG -# -# Powerpc uses the slab allocator to manage its ptes and the -# page structs of ptes are used for splitting the page table -# lock for configurations supporting more than SPLIT_PTLOCK_CPUS. -# -# In that special configuration the page structs of slabs are modified. -# This setting disables the selection of SLUB as a slab allocator. -# -config ARCH_USES_SLAB_PAGE_STRUCT - bool - default y - depends on SPLIT_PTLOCK_CPUS <= NR_CPUS - config DEFAULT_UIMAGE bool help