From patchwork Thu Dec 11 01:55:41 2008 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ilya Yanok X-Patchwork-Id: 13389 X-Patchwork-Delegate: benh@kernel.crashing.org Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from ozlabs.org (localhost [127.0.0.1]) by ozlabs.org (Postfix) with ESMTP id 3CBFDDE0DC for ; Thu, 11 Dec 2008 12:56:51 +1100 (EST) X-Original-To: linuxppc-dev@ozlabs.org Delivered-To: linuxppc-dev@ozlabs.org Received: from ocean.emcraft.com (ocean.emcraft.com [213.221.7.182]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTPS id DCDB4DDF7E for ; Thu, 11 Dec 2008 12:56:22 +1100 (EST) Received: from [172.17.0.9] (helo=localhost.localdomain) by ocean.emcraft.com with esmtp (Exim 4.43) id 1LAama-0008V0-KJ; Thu, 11 Dec 2008 04:56:05 +0300 From: Ilya Yanok To: linuxppc-dev@ozlabs.org Subject: [PATCH] powerpc: add 16K/64K pages support for the 44x PPC32 architectures. Date: Thu, 11 Dec 2008 04:55:41 +0300 Message-Id: <1228960541-11407-1-git-send-email-yanok@emcraft.com> X-Mailer: git-send-email 1.5.6.5 In-Reply-To: <49406F61.2000004@emcraft.com> References: <49406F61.2000004@emcraft.com> Cc: wd@denx.de, dzu@denx.de, Vladimir Panfilov , Ilya Yanok X-BeenThere: linuxppc-dev@ozlabs.org X-Mailman-Version: 2.1.11 Precedence: list List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , MIME-Version: 1.0 Sender: linuxppc-dev-bounces+patchwork-incoming=ozlabs.org@ozlabs.org Errors-To: linuxppc-dev-bounces+patchwork-incoming=ozlabs.org@ozlabs.org This patch adds support for page sizes bigger than 4K (16K/64K) on PPC 44x. PGDIR table is much smaller than page in case of 16K/64K pages (512 and 32 bytes resp.) so we allocate PGDIR with kzalloc() instead of __get_free_pages(). PTE table covers rather big memory area in case of 16K/64K pages (32MB and 512MB resp.) so we can easily put FIXMAP and PKMAP in area covered by one PTE table. Signed-off-by: Yuri Tikhonov Signed-off-by: Vladimir Panfilov Signed-off-by: Ilya Yanok Acked-by: Josh Boyer --- arch/powerpc/Kconfig | 58 ++++++++++++++++++++++++-------- arch/powerpc/include/asm/highmem.h | 19 +++++++++- arch/powerpc/include/asm/mmu-44x.h | 17 +++++++++ arch/powerpc/include/asm/page.h | 13 ++++--- arch/powerpc/include/asm/page_32.h | 7 +++- arch/powerpc/kernel/asm-offsets.c | 4 ++ arch/powerpc/kernel/head_44x.S | 23 ++++++++----- arch/powerpc/kernel/misc_32.S | 12 +++--- arch/powerpc/mm/pgtable_32.c | 23 ++++++++----- arch/powerpc/platforms/Kconfig.cputype | 2 +- 10 files changed, 130 insertions(+), 48 deletions(-) diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig index 525c13a..cd8ff7c 100644 --- a/arch/powerpc/Kconfig +++ b/arch/powerpc/Kconfig @@ -401,23 +401,53 @@ config PPC_HAS_HASH_64K depends on PPC64 default n -config PPC_64K_PAGES - bool "64k page size" - depends on PPC64 - select PPC_HAS_HASH_64K +choice + prompt "Page size" + default PPC_4K_PAGES help - This option changes the kernel logical page size to 64k. On machines - without processor support for 64k pages, the kernel will simulate - them by loading each individual 4k page on demand transparently, - while on hardware with such support, it will be used to map - normal application pages. + Select the kernel logical page size. Increasing the page size + will reduce software overhead at each page boundary, allow + hardware prefetch mechanisms to be more effective, and allow + larger dma transfers increasing IO efficiency and reducing + overhead. However the utilization of memory will increase. + For example, each cached file will using a multiple of the + page size to hold its contents and the difference between the + end of file and the end of page is wasted. + + Some dedicated systems, such as software raid serving with + accelerated calculations, have shown significant increases. + + If you configure a 64 bit kernel for 64k pages but the + processor does not support them, then the kernel will simulate + them with 4k pages, loading them on demand, but with the + reduced software overhead and larger internal fragmentation. + For the 32 bit kernel, a large page option will not be offered + unless it is supported by the configured processor. + + If unsure, choose 4K_PAGES. + +config PPC_4K_PAGES + bool "4k page size" + +config PPC_16K_PAGES + bool "16k page size" if 44x + +config PPC_64K_PAGES + bool "64k page size" if 44x || PPC_STD_MMU_64 + select PPC_HAS_HASH_64K if PPC_STD_MMU_64 + +endchoice config FORCE_MAX_ZONEORDER int "Maximum zone order" - range 9 64 if PPC_64K_PAGES - default "9" if PPC_64K_PAGES - range 13 64 if PPC64 && !PPC_64K_PAGES - default "13" if PPC64 && !PPC_64K_PAGES + range 9 64 if PPC_STD_MMU_64 && PPC_64K_PAGES + default "9" if PPC_STD_MMU_64 && PPC_64K_PAGES + range 13 64 if PPC_STD_MMU_64 && !PPC_64K_PAGES + default "13" if PPC_STD_MMU_64 && !PPC_64K_PAGES + range 9 64 if PPC_STD_MMU_32 && PPC_16K_PAGES + default "9" if PPC_STD_MMU_32 && PPC_16K_PAGES + range 7 64 if PPC_STD_MMU_32 && PPC_64K_PAGES + default "7" if PPC_STD_MMU_32 && PPC_64K_PAGES range 11 64 default "11" help @@ -437,7 +467,7 @@ config FORCE_MAX_ZONEORDER config PPC_SUBPAGE_PROT bool "Support setting protections for 4k subpages" - depends on PPC_64K_PAGES + depends on PPC_STD_MMU_64 && PPC_64K_PAGES help This option adds support for a system call to allow user programs to set access permissions (read/write, readonly, or no access) diff --git a/arch/powerpc/include/asm/highmem.h b/arch/powerpc/include/asm/highmem.h index 91c5895..7d6bb37 100644 --- a/arch/powerpc/include/asm/highmem.h +++ b/arch/powerpc/include/asm/highmem.h @@ -38,9 +38,24 @@ extern pte_t *pkmap_page_table; * easily, subsequent pte tables have to be allocated in one physical * chunk of RAM. */ -#define LAST_PKMAP (1 << PTE_SHIFT) -#define LAST_PKMAP_MASK (LAST_PKMAP-1) +/* + * We use one full pte table with 4K pages. And with 16K/64K pages pte + * table covers enough memory (32MB and 512MB resp.) that both FIXMAP + * and PKMAP can be placed in single pte table. We use 1024 pages for + * PKMAP in case of 16K/64K pages. + */ +#ifdef CONFIG_PPC_4K_PAGES +#define PKMAP_ORDER PTE_SHIFT +#else +#define PKMAP_ORDER 10 +#endif +#define LAST_PKMAP (1 << PKMAP_ORDER) +#ifndef CONFIG_PPC_4K_PAGES +#define PKMAP_BASE (FIXADDR_START - PAGE_SIZE*(LAST_PKMAP + 1)) +#else #define PKMAP_BASE ((FIXADDR_START - PAGE_SIZE*(LAST_PKMAP + 1)) & PMD_MASK) +#endif +#define LAST_PKMAP_MASK (LAST_PKMAP-1) #define PKMAP_NR(virt) ((virt-PKMAP_BASE) >> PAGE_SHIFT) #define PKMAP_ADDR(nr) (PKMAP_BASE + ((nr) << PAGE_SHIFT)) diff --git a/arch/powerpc/include/asm/mmu-44x.h b/arch/powerpc/include/asm/mmu-44x.h index a825524..73e1909 100644 --- a/arch/powerpc/include/asm/mmu-44x.h +++ b/arch/powerpc/include/asm/mmu-44x.h @@ -4,6 +4,8 @@ * PPC440 support */ +#include + #define PPC44x_MMUCR_TID 0x000000ff #define PPC44x_MMUCR_STS 0x00010000 @@ -73,4 +75,19 @@ typedef struct { /* Size of the TLBs used for pinning in lowmem */ #define PPC_PIN_SIZE (1 << 28) /* 256M */ +#if (PAGE_SHIFT == 12) +#define PPC44x_TLBE_SIZE PPC44x_TLB_4K +#elif (PAGE_SHIFT == 14) +#define PPC44x_TLBE_SIZE PPC44x_TLB_16K +#elif (PAGE_SHIFT == 16) +#define PPC44x_TLBE_SIZE PPC44x_TLB_64K +#else +#error "Unsupported PAGE_SIZE" +#endif + +#define PPC44x_PGD_OFF_SHIFT (32 - PGDIR_SHIFT + PGD_T_LOG2) +#define PPC44x_PGD_OFF_MASK_BIT (PGDIR_SHIFT - PGD_T_LOG2) +#define PPC44x_PTE_ADD_SHIFT (32 - PGDIR_SHIFT + PTE_SHIFT + PTE_T_LOG2) +#define PPC44x_PTE_ADD_MASK_BIT (32 - PTE_T_LOG2 - PTE_SHIFT) + #endif /* _ASM_POWERPC_MMU_44X_H_ */ diff --git a/arch/powerpc/include/asm/page.h b/arch/powerpc/include/asm/page.h index c0b8d4a..197d569 100644 --- a/arch/powerpc/include/asm/page.h +++ b/arch/powerpc/include/asm/page.h @@ -19,12 +19,15 @@ #include /* - * On PPC32 page size is 4K. For PPC64 we support either 4K or 64K software + * On regular PPC32 page size is 4K (but we support 4K/16K/64K pages + * on PPC44x). For PPC64 we support either 4K or 64K software * page size. When using 64K pages however, whether we are really supporting * 64K pages in HW or not is irrelevant to those definitions. */ -#ifdef CONFIG_PPC_64K_PAGES +#if defined(CONFIG_PPC_64K_PAGES) #define PAGE_SHIFT 16 +#elif defined(CONFIG_PPC_16K_PAGES) +#define PAGE_SHIFT 14 #else #define PAGE_SHIFT 12 #endif @@ -151,7 +154,7 @@ typedef struct { pte_basic_t pte; } pte_t; /* 64k pages additionally define a bigger "real PTE" type that gathers * the "second half" part of the PTE for pseudo 64k pages */ -#ifdef CONFIG_PPC_64K_PAGES +#if defined(CONFIG_PPC_64K_PAGES) && defined(CONFIG_PPC_STD_MMU_64) typedef struct { pte_t pte; unsigned long hidx; } real_pte_t; #else typedef struct { pte_t pte; } real_pte_t; @@ -191,10 +194,10 @@ typedef pte_basic_t pte_t; #define pte_val(x) (x) #define __pte(x) (x) -#ifdef CONFIG_PPC_64K_PAGES +#if defined(CONFIG_PPC_64K_PAGES) && defined(CONFIG_PPC_STD_MMU_64) typedef struct { pte_t pte; unsigned long hidx; } real_pte_t; #else -typedef unsigned long real_pte_t; +typedef pte_t real_pte_t; #endif diff --git a/arch/powerpc/include/asm/page_32.h b/arch/powerpc/include/asm/page_32.h index d77072a..1458d95 100644 --- a/arch/powerpc/include/asm/page_32.h +++ b/arch/powerpc/include/asm/page_32.h @@ -19,6 +19,8 @@ #define PTE_FLAGS_OFFSET 0 #endif +#define PTE_SHIFT (PAGE_SHIFT - PTE_T_LOG2) /* full page */ + #ifndef __ASSEMBLY__ /* * The basic type of a PTE - 64 bits for those CPUs with > 32 bit @@ -26,10 +28,8 @@ */ #ifdef CONFIG_PTE_64BIT typedef unsigned long long pte_basic_t; -#define PTE_SHIFT (PAGE_SHIFT - 3) /* 512 ptes per page */ #else typedef unsigned long pte_basic_t; -#define PTE_SHIFT (PAGE_SHIFT - 2) /* 1024 ptes per page */ #endif struct page; @@ -39,6 +39,9 @@ extern void copy_page(void *to, void *from); #include +#define PGD_T_LOG2 (__builtin_ffs(sizeof(pgd_t)) - 1) +#define PTE_T_LOG2 (__builtin_ffs(sizeof(pte_t)) - 1) + #endif /* __ASSEMBLY__ */ #endif /* _ASM_POWERPC_PAGE_32_H */ diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c index 75c5dd0..0142318 100644 --- a/arch/powerpc/kernel/asm-offsets.c +++ b/arch/powerpc/kernel/asm-offsets.c @@ -378,6 +378,10 @@ int main(void) DEFINE(VCPU_FAULT_DEAR, offsetof(struct kvm_vcpu, arch.fault_dear)); DEFINE(VCPU_FAULT_ESR, offsetof(struct kvm_vcpu, arch.fault_esr)); #endif +#ifdef CONFIG_44x + DEFINE(PGD_T_LOG2, PGD_T_LOG2); + DEFINE(PTE_T_LOG2, PTE_T_LOG2); +#endif return 0; } diff --git a/arch/powerpc/kernel/head_44x.S b/arch/powerpc/kernel/head_44x.S index f3a1ea9..3bf6bd1 100644 --- a/arch/powerpc/kernel/head_44x.S +++ b/arch/powerpc/kernel/head_44x.S @@ -391,12 +391,14 @@ interrupt_base: rlwimi r13,r12,10,30,30 /* Load the PTE */ - rlwinm r12, r10, 13, 19, 29 /* Compute pgdir/pmd offset */ + /* Compute pgdir/pmd offset */ + rlwinm r12, r10, PPC44x_PGD_OFF_SHIFT, PPC44x_PGD_OFF_MASK_BIT, 29 lwzx r11, r12, r11 /* Get pgd/pmd entry */ rlwinm. r12, r11, 0, 0, 20 /* Extract pt base address */ beq 2f /* Bail if no table */ - rlwimi r12, r10, 23, 20, 28 /* Compute pte address */ + /* Compute pte address */ + rlwimi r12, r10, PPC44x_PTE_ADD_SHIFT, PPC44x_PTE_ADD_MASK_BIT, 28 lwz r11, 0(r12) /* Get high word of pte entry */ lwz r12, 4(r12) /* Get low word of pte entry */ @@ -485,12 +487,14 @@ tlb_44x_patch_hwater_D: /* Make up the required permissions */ li r13,_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_HWEXEC - rlwinm r12, r10, 13, 19, 29 /* Compute pgdir/pmd offset */ + /* Compute pgdir/pmd offset */ + rlwinm r12, r10, PPC44x_PGD_OFF_SHIFT, PPC44x_PGD_OFF_MASK_BIT, 29 lwzx r11, r12, r11 /* Get pgd/pmd entry */ rlwinm. r12, r11, 0, 0, 20 /* Extract pt base address */ beq 2f /* Bail if no table */ - rlwimi r12, r10, 23, 20, 28 /* Compute pte address */ + /* Compute pte address */ + rlwimi r12, r10, PPC44x_PTE_ADD_SHIFT, PPC44x_PTE_ADD_MASK_BIT, 28 lwz r11, 0(r12) /* Get high word of pte entry */ lwz r12, 4(r12) /* Get low word of pte entry */ @@ -554,15 +558,16 @@ tlb_44x_patch_hwater_I: */ finish_tlb_load: /* Combine RPN & ERPN an write WS 0 */ - rlwimi r11,r12,0,0,19 + rlwimi r11,r12,0,0,31-PAGE_SHIFT tlbwe r11,r13,PPC44x_TLB_XLAT /* * Create WS1. This is the faulting address (EPN), * page size, and valid flag. */ - li r11,PPC44x_TLB_VALID | PPC44x_TLB_4K - rlwimi r10,r11,0,20,31 /* Insert valid and page size*/ + li r11,PPC44x_TLB_VALID | PPC44x_TLBE_SIZE + /* Insert valid and page size */ + rlwimi r10,r11,0,PPC44x_PTE_ADD_MASK_BIT,31 tlbwe r10,r13,PPC44x_TLB_PAGEID /* Write PAGEID */ /* And WS 2 */ @@ -634,12 +639,12 @@ _GLOBAL(set_context) * goes at the beginning of the data segment, which is page-aligned. */ .data - .align 12 + .align PAGE_SHIFT .globl sdata sdata: .globl empty_zero_page empty_zero_page: - .space 4096 + .space PAGE_SIZE /* * To support >32-bit physical addresses, we use an 8KB pgdir. diff --git a/arch/powerpc/kernel/misc_32.S b/arch/powerpc/kernel/misc_32.S index bdc8b0e..950b708 100644 --- a/arch/powerpc/kernel/misc_32.S +++ b/arch/powerpc/kernel/misc_32.S @@ -647,8 +647,8 @@ _GLOBAL(__flush_dcache_icache) BEGIN_FTR_SECTION blr END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE) - rlwinm r3,r3,0,0,19 /* Get page base address */ - li r4,4096/L1_CACHE_BYTES /* Number of lines in a page */ + rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */ + li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */ mtctr r4 mr r6,r3 0: dcbst 0,r3 /* Write line to ram */ @@ -688,8 +688,8 @@ END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE) rlwinm r0,r10,0,28,26 /* clear DR */ mtmsr r0 isync - rlwinm r3,r3,0,0,19 /* Get page base address */ - li r4,4096/L1_CACHE_BYTES /* Number of lines in a page */ + rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */ + li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */ mtctr r4 mr r6,r3 0: dcbst 0,r3 /* Write line to ram */ @@ -713,7 +713,7 @@ END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE) * void clear_pages(void *page, int order) ; */ _GLOBAL(clear_pages) - li r0,4096/L1_CACHE_BYTES + li r0,PAGE_SIZE/L1_CACHE_BYTES slw r0,r0,r4 mtctr r0 #ifdef CONFIG_8xx @@ -771,7 +771,7 @@ _GLOBAL(copy_page) dcbt r5,r4 li r11,L1_CACHE_BYTES+4 #endif /* MAX_COPY_PREFETCH */ - li r0,4096/L1_CACHE_BYTES - MAX_COPY_PREFETCH + li r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH crclr 4*cr0+eq 2: mtctr r0 diff --git a/arch/powerpc/mm/pgtable_32.c b/arch/powerpc/mm/pgtable_32.c index c31d6d2..f966a5e 100644 --- a/arch/powerpc/mm/pgtable_32.c +++ b/arch/powerpc/mm/pgtable_32.c @@ -72,24 +72,29 @@ extern unsigned long p_mapped_by_tlbcam(unsigned long pa); #define p_mapped_by_tlbcam(x) (0UL) #endif /* HAVE_TLBCAM */ -#ifdef CONFIG_PTE_64BIT -/* Some processors use an 8kB pgdir because they have 8-byte Linux PTEs. */ -#define PGDIR_ORDER 1 -#else -#define PGDIR_ORDER 0 -#endif +#define PGDIR_ORDER (32 + PGD_T_LOG2 - PGDIR_SHIFT) pgd_t *pgd_alloc(struct mm_struct *mm) { pgd_t *ret; - ret = (pgd_t *)__get_free_pages(GFP_KERNEL|__GFP_ZERO, PGDIR_ORDER); + /* pgdir take page or two with 4K pages and a page fraction otherwise */ +#ifndef CONFIG_PPC_4K_PAGES + ret = (pgd_t *)kzalloc(1 << PGDIR_ORDER, GFP_KERNEL); +#else + ret = (pgd_t *)__get_free_pages(GFP_KERNEL|__GFP_ZERO, + PGDIR_ORDER - PAGE_SHIFT); +#endif return ret; } void pgd_free(struct mm_struct *mm, pgd_t *pgd) { - free_pages((unsigned long)pgd, PGDIR_ORDER); +#ifndef CONFIG_PPC_4K_PAGES + kfree((void *)pgd); +#else + free_pages((unsigned long)pgd, PGDIR_ORDER - PAGE_SHIFT); +#endif } __init_refok pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address) @@ -400,7 +405,7 @@ void kernel_map_pages(struct page *page, int numpages, int enable) #endif /* CONFIG_DEBUG_PAGEALLOC */ static int fixmaps; -unsigned long FIXADDR_TOP = 0xfffff000; +unsigned long FIXADDR_TOP = (-PAGE_SIZE); EXPORT_SYMBOL(FIXADDR_TOP); void __set_fixmap (enum fixed_addresses idx, phys_addr_t phys, pgprot_t flags) diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype index 548efa5..51098bc 100644 --- a/arch/powerpc/platforms/Kconfig.cputype +++ b/arch/powerpc/platforms/Kconfig.cputype @@ -204,7 +204,7 @@ config PPC_STD_MMU_32 config PPC_MM_SLICES bool - default y if HUGETLB_PAGE || PPC_64K_PAGES + default y if HUGETLB_PAGE || (PPC_STD_MMU_64 && PPC_64K_PAGES) default n config VIRT_CPU_ACCOUNTING