diff options
author | Linus Torvalds <torvalds@g5.osdl.org> | 2006-03-22 10:56:57 -0800 |
---|---|---|
committer | Linus Torvalds <torvalds@g5.osdl.org> | 2006-03-22 10:56:57 -0800 |
commit | d04ef3a795b3b7b376a02713ed5e211e9ae1f917 (patch) | |
tree | 837da034751a2fc1be0fc5a105c218d41a498eb6 /arch/sparc64/mm | |
parent | 36177ba655c238e33400cc2837a28720b62784bd (diff) | |
parent | dcc1e8dd88d4bc55e32a26dad7633d20ffe606d2 (diff) |
Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6
* master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
[SPARC64]: Add a secondary TSB for hugepage mappings.
[SPARC]: Respect vm_page_prot in io_remap_page_range().
Diffstat (limited to 'arch/sparc64/mm')
-rw-r--r-- | arch/sparc64/mm/fault.c | 15 | ||||
-rw-r--r-- | arch/sparc64/mm/generic.c | 1 | ||||
-rw-r--r-- | arch/sparc64/mm/hugetlbpage.c | 28 | ||||
-rw-r--r-- | arch/sparc64/mm/init.c | 21 | ||||
-rw-r--r-- | arch/sparc64/mm/tsb.c | 234 |
5 files changed, 197 insertions, 102 deletions
diff --git a/arch/sparc64/mm/fault.c b/arch/sparc64/mm/fault.c index 63b6cc0cd5d5..d21ff3230c02 100644 --- a/arch/sparc64/mm/fault.c +++ b/arch/sparc64/mm/fault.c @@ -410,9 +410,18 @@ good_area: up_read(&mm->mmap_sem); mm_rss = get_mm_rss(mm); - if (unlikely(mm_rss >= mm->context.tsb_rss_limit)) - tsb_grow(mm, mm_rss); - +#ifdef CONFIG_HUGETLB_PAGE + mm_rss -= (mm->context.huge_pte_count * (HPAGE_SIZE / PAGE_SIZE)); +#endif + if (unlikely(mm_rss >= + mm->context.tsb_block[MM_TSB_BASE].tsb_rss_limit)) + tsb_grow(mm, MM_TSB_BASE, mm_rss); +#ifdef CONFIG_HUGETLB_PAGE + mm_rss = mm->context.huge_pte_count; + if (unlikely(mm_rss >= + mm->context.tsb_block[MM_TSB_HUGE].tsb_rss_limit)) + tsb_grow(mm, MM_TSB_HUGE, mm_rss); +#endif return; /* diff --git a/arch/sparc64/mm/generic.c b/arch/sparc64/mm/generic.c index 5fc5c579e35e..8cb06205d265 100644 --- a/arch/sparc64/mm/generic.c +++ b/arch/sparc64/mm/generic.c @@ -140,7 +140,6 @@ int io_remap_pfn_range(struct vm_area_struct *vma, unsigned long from, vma->vm_flags |= VM_IO | VM_RESERVED | VM_PFNMAP; vma->vm_pgoff = phys_base >> PAGE_SHIFT; - prot = __pgprot(pg_iobits); offset -= from; dir = pgd_offset(mm, from); flush_cache_range(vma, beg, end); diff --git a/arch/sparc64/mm/hugetlbpage.c b/arch/sparc64/mm/hugetlbpage.c index 280dc7958a13..074620d413d4 100644 --- a/arch/sparc64/mm/hugetlbpage.c +++ b/arch/sparc64/mm/hugetlbpage.c @@ -199,13 +199,11 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr) pte_t *pte = NULL; pgd = pgd_offset(mm, addr); - if (pgd) { - pud = pud_offset(pgd, addr); - if (pud) { - pmd = pmd_alloc(mm, pud, addr); - if (pmd) - pte = pte_alloc_map(mm, pmd, addr); - } + pud = pud_alloc(mm, pgd, addr); + if (pud) { + pmd = pmd_alloc(mm, pud, addr); + if (pmd) + pte = pte_alloc_map(mm, pmd, addr); } return pte; } @@ -231,13 +229,14 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) return pte; } -#define mk_pte_huge(entry) do { pte_val(entry) |= _PAGE_SZHUGE; } while (0) - void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t entry) { int i; + if (!pte_present(*ptep) && pte_present(entry)) + mm->context.huge_pte_count++; + for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) { set_pte_at(mm, addr, ptep, entry); ptep++; @@ -253,6 +252,8 @@ pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, int i; entry = *ptep; + if (pte_present(entry)) + mm->context.huge_pte_count--; for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) { pte_clear(mm, addr, ptep); @@ -290,6 +291,15 @@ static void context_reload(void *__data) void hugetlb_prefault_arch_hook(struct mm_struct *mm) { + struct tsb_config *tp = &mm->context.tsb_block[MM_TSB_HUGE]; + + if (likely(tp->tsb != NULL)) + return; + + tsb_grow(mm, MM_TSB_HUGE, 0); + tsb_context_switch(mm); + smp_tsb_sync(mm); + /* On UltraSPARC-III+ and later, configure the second half of * the Data-TLB for huge pages. */ diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c index 2ae143ba50d8..ded63ee9c4fd 100644 --- a/arch/sparc64/mm/init.c +++ b/arch/sparc64/mm/init.c @@ -283,6 +283,7 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t p struct mm_struct *mm; struct tsb *tsb; unsigned long tag, flags; + unsigned long tsb_index, tsb_hash_shift; if (tlb_type != hypervisor) { unsigned long pfn = pte_pfn(pte); @@ -312,10 +313,26 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t p mm = vma->vm_mm; + tsb_index = MM_TSB_BASE; + tsb_hash_shift = PAGE_SHIFT; + spin_lock_irqsave(&mm->context.lock, flags); - tsb = &mm->context.tsb[(address >> PAGE_SHIFT) & - (mm->context.tsb_nentries - 1UL)]; +#ifdef CONFIG_HUGETLB_PAGE + if (mm->context.tsb_block[MM_TSB_HUGE].tsb != NULL) { + if ((tlb_type == hypervisor && + (pte_val(pte) & _PAGE_SZALL_4V) == _PAGE_SZHUGE_4V) || + (tlb_type != hypervisor && + (pte_val(pte) & _PAGE_SZALL_4U) == _PAGE_SZHUGE_4U)) { + tsb_index = MM_TSB_HUGE; + tsb_hash_shift = HPAGE_SHIFT; + } + } +#endif + + tsb = mm->context.tsb_block[tsb_index].tsb; + tsb += ((address >> tsb_hash_shift) & + (mm->context.tsb_block[tsb_index].tsb_nentries - 1UL)); tag = (address >> 22UL); tsb_insert(tsb, tag, pte_val(pte)); diff --git a/arch/sparc64/mm/tsb.c b/arch/sparc64/mm/tsb.c index b2064e2a44d6..beaa02810f0e 100644 --- a/arch/sparc64/mm/tsb.c +++ b/arch/sparc64/mm/tsb.c @@ -15,9 +15,9 @@ extern struct tsb swapper_tsb[KERNEL_TSB_NENTRIES]; -static inline unsigned long tsb_hash(unsigned long vaddr, unsigned long nentries) +static inline unsigned long tsb_hash(unsigned long vaddr, unsigned long hash_shift, unsigned long nentries) { - vaddr >>= PAGE_SHIFT; + vaddr >>= hash_shift; return vaddr & (nentries - 1); } @@ -36,7 +36,8 @@ void flush_tsb_kernel_range(unsigned long start, unsigned long end) unsigned long v; for (v = start; v < end; v += PAGE_SIZE) { - unsigned long hash = tsb_hash(v, KERNEL_TSB_NENTRIES); + unsigned long hash = tsb_hash(v, PAGE_SHIFT, + KERNEL_TSB_NENTRIES); struct tsb *ent = &swapper_tsb[hash]; if (tag_compare(ent->tag, v)) { @@ -46,49 +47,91 @@ void flush_tsb_kernel_range(unsigned long start, unsigned long end) } } -void flush_tsb_user(struct mmu_gather *mp) +static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, unsigned long tsb, unsigned long nentries) { - struct mm_struct *mm = mp->mm; - unsigned long nentries, base, flags; - struct tsb *tsb; - int i; - - spin_lock_irqsave(&mm->context.lock, flags); - - tsb = mm->context.tsb; - nentries = mm->context.tsb_nentries; + unsigned long i; - if (tlb_type == cheetah_plus || tlb_type == hypervisor) - base = __pa(tsb); - else - base = (unsigned long) tsb; - for (i = 0; i < mp->tlb_nr; i++) { unsigned long v = mp->vaddrs[i]; unsigned long tag, ent, hash; v &= ~0x1UL; - hash = tsb_hash(v, nentries); - ent = base + (hash * sizeof(struct tsb)); + hash = tsb_hash(v, hash_shift, nentries); + ent = tsb + (hash * sizeof(struct tsb)); tag = (v >> 22UL); tsb_flush(ent, tag); } +} + +void flush_tsb_user(struct mmu_gather *mp) +{ + struct mm_struct *mm = mp->mm; + unsigned long nentries, base, flags; + + spin_lock_irqsave(&mm->context.lock, flags); + base = (unsigned long) mm->context.tsb_block[MM_TSB_BASE].tsb; + nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries; + if (tlb_type == cheetah_plus || tlb_type == hypervisor) + base = __pa(base); + __flush_tsb_one(mp, PAGE_SHIFT, base, nentries); + +#ifdef CONFIG_HUGETLB_PAGE + if (mm->context.tsb_block[MM_TSB_HUGE].tsb) { + base = (unsigned long) mm->context.tsb_block[MM_TSB_HUGE].tsb; + nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries; + if (tlb_type == cheetah_plus || tlb_type == hypervisor) + base = __pa(base); + __flush_tsb_one(mp, HPAGE_SHIFT, base, nentries); + } +#endif spin_unlock_irqrestore(&mm->context.lock, flags); } -static void setup_tsb_params(struct mm_struct *mm, unsigned long tsb_bytes) +#if defined(CONFIG_SPARC64_PAGE_SIZE_8KB) +#define HV_PGSZ_IDX_BASE HV_PGSZ_IDX_8K +#define HV_PGSZ_MASK_BASE HV_PGSZ_MASK_8K +#elif defined(CONFIG_SPARC64_PAGE_SIZE_64KB) +#define HV_PGSZ_IDX_BASE HV_PGSZ_IDX_64K +#define HV_PGSZ_MASK_BASE HV_PGSZ_MASK_64K +#elif defined(CONFIG_SPARC64_PAGE_SIZE_512KB) +#define HV_PGSZ_IDX_BASE HV_PGSZ_IDX_512K +#define HV_PGSZ_MASK_BASE HV_PGSZ_MASK_512K +#elif defined(CONFIG_SPARC64_PAGE_SIZE_4MB) +#define HV_PGSZ_IDX_BASE HV_PGSZ_IDX_4MB +#define HV_PGSZ_MASK_BASE HV_PGSZ_MASK_4MB +#else +#error Broken base page size setting... +#endif + +#ifdef CONFIG_HUGETLB_PAGE +#if defined(CONFIG_HUGETLB_PAGE_SIZE_64K) +#define HV_PGSZ_IDX_HUGE HV_PGSZ_IDX_64K +#define HV_PGSZ_MASK_HUGE HV_PGSZ_MASK_64K +#elif defined(CONFIG_HUGETLB_PAGE_SIZE_512K) +#define HV_PGSZ_IDX_HUGE HV_PGSZ_IDX_512K +#define HV_PGSZ_MASK_HUGE HV_PGSZ_MASK_512K +#elif defined(CONFIG_HUGETLB_PAGE_SIZE_4MB) +#define HV_PGSZ_IDX_HUGE HV_PGSZ_IDX_4MB +#define HV_PGSZ_MASK_HUGE HV_PGSZ_MASK_4MB +#else +#error Broken huge page size setting... +#endif +#endif + +static void setup_tsb_params(struct mm_struct *mm, unsigned long tsb_idx, unsigned long tsb_bytes) { unsigned long tsb_reg, base, tsb_paddr; unsigned long page_sz, tte; - mm->context.tsb_nentries = tsb_bytes / sizeof(struct tsb); + mm->context.tsb_block[tsb_idx].tsb_nentries = + tsb_bytes / sizeof(struct tsb); base = TSBMAP_BASE; tte = pgprot_val(PAGE_KERNEL_LOCKED); - tsb_paddr = __pa(mm->context.tsb); + tsb_paddr = __pa(mm->context.tsb_block[tsb_idx].tsb); BUG_ON(tsb_paddr & (tsb_bytes - 1UL)); /* Use the smallest page size that can map the whole TSB @@ -147,61 +190,49 @@ static void setup_tsb_params(struct mm_struct *mm, unsigned long tsb_bytes) /* Physical mapping, no locked TLB entry for TSB. */ tsb_reg |= tsb_paddr; - mm->context.tsb_reg_val = tsb_reg; - mm->context.tsb_map_vaddr = 0; - mm->context.tsb_map_pte = 0; + mm->context.tsb_block[tsb_idx].tsb_reg_val = tsb_reg; + mm->context.tsb_block[tsb_idx].tsb_map_vaddr = 0; + mm->context.tsb_block[tsb_idx].tsb_map_pte = 0; } else { tsb_reg |= base; tsb_reg |= (tsb_paddr & (page_sz - 1UL)); tte |= (tsb_paddr & ~(page_sz - 1UL)); - mm->context.tsb_reg_val = tsb_reg; - mm->context.tsb_map_vaddr = base; - mm->context.tsb_map_pte = tte; + mm->context.tsb_block[tsb_idx].tsb_reg_val = tsb_reg; + mm->context.tsb_block[tsb_idx].tsb_map_vaddr = base; + mm->context.tsb_block[tsb_idx].tsb_map_pte = tte; } /* Setup the Hypervisor TSB descriptor. */ if (tlb_type == hypervisor) { - struct hv_tsb_descr *hp = &mm->context.tsb_descr; + struct hv_tsb_descr *hp = &mm->context.tsb_descr[tsb_idx]; - switch (PAGE_SIZE) { - case 8192: - default: - hp->pgsz_idx = HV_PGSZ_IDX_8K; + switch (tsb_idx) { + case MM_TSB_BASE: + hp->pgsz_idx = HV_PGSZ_IDX_BASE; break; - - case 64 * 1024: - hp->pgsz_idx = HV_PGSZ_IDX_64K; - break; - - case 512 * 1024: - hp->pgsz_idx = HV_PGSZ_IDX_512K; - break; - - case 4 * 1024 * 1024: - hp->pgsz_idx = HV_PGSZ_IDX_4MB; +#ifdef CONFIG_HUGETLB_PAGE + case MM_TSB_HUGE: + hp->pgsz_idx = HV_PGSZ_IDX_HUGE; break; +#endif + default: + BUG(); }; hp->assoc = 1; hp->num_ttes = tsb_bytes / 16; hp->ctx_idx = 0; - switch (PAGE_SIZE) { - case 8192: - default: - hp->pgsz_mask = HV_PGSZ_MASK_8K; - break; - - case 64 * 1024: - hp->pgsz_mask = HV_PGSZ_MASK_64K; - break; - - case 512 * 1024: - hp->pgsz_mask = HV_PGSZ_MASK_512K; + switch (tsb_idx) { + case MM_TSB_BASE: + hp->pgsz_mask = HV_PGSZ_MASK_BASE; break; - - case 4 * 1024 * 1024: - hp->pgsz_mask = HV_PGSZ_MASK_4MB; +#ifdef CONFIG_HUGETLB_PAGE + case MM_TSB_HUGE: + hp->pgsz_mask = HV_PGSZ_MASK_HUGE; break; +#endif + default: + BUG(); }; hp->tsb_base = tsb_paddr; hp->resv = 0; @@ -241,11 +272,11 @@ void __init tsb_cache_init(void) } } -/* When the RSS of an address space exceeds mm->context.tsb_rss_limit, - * do_sparc64_fault() invokes this routine to try and grow the TSB. +/* When the RSS of an address space exceeds tsb_rss_limit for a TSB, + * do_sparc64_fault() invokes this routine to try and grow it. * * When we reach the maximum TSB size supported, we stick ~0UL into - * mm->context.tsb_rss_limit so the grow checks in update_mmu_cache() + * tsb_rss_limit for that TSB so the grow checks in do_sparc64_fault() * will not trigger any longer. * * The TSB can be anywhere from 8K to 1MB in size, in increasing powers @@ -257,7 +288,7 @@ void __init tsb_cache_init(void) * the number of entries that the current TSB can hold at once. Currently, * we trigger when the RSS hits 3/4 of the TSB capacity. */ -void tsb_grow(struct mm_struct *mm, unsigned long rss) +void tsb_grow(struct mm_struct *mm, unsigned long tsb_index, unsigned long rss) { unsigned long max_tsb_size = 1 * 1024 * 1024; unsigned long new_size, old_size, flags; @@ -297,7 +328,8 @@ retry_tsb_alloc: * down to a 0-order allocation and force no TSB * growing for this address space. */ - if (mm->context.tsb == NULL && new_cache_index > 0) { + if (mm->context.tsb_block[tsb_index].tsb == NULL && + new_cache_index > 0) { new_cache_index = 0; new_size = 8192; new_rss_limit = ~0UL; @@ -307,8 +339,8 @@ retry_tsb_alloc: /* If we failed on a TSB grow, we are under serious * memory pressure so don't try to grow any more. */ - if (mm->context.tsb != NULL) - mm->context.tsb_rss_limit = ~0UL; + if (mm->context.tsb_block[tsb_index].tsb != NULL) + mm->context.tsb_block[tsb_index].tsb_rss_limit = ~0UL; return; } @@ -339,23 +371,26 @@ retry_tsb_alloc: */ spin_lock_irqsave(&mm->context.lock, flags); - old_tsb = mm->context.tsb; - old_cache_index = (mm->context.tsb_reg_val & 0x7UL); - old_size = mm->context.tsb_nentries * sizeof(struct tsb); + old_tsb = mm->context.tsb_block[tsb_index].tsb; + old_cache_index = + (mm->context.tsb_block[tsb_index].tsb_reg_val & 0x7UL); + old_size = (mm->context.tsb_block[tsb_index].tsb_nentries * + sizeof(struct tsb)); /* Handle multiple threads trying to grow the TSB at the same time. * One will get in here first, and bump the size and the RSS limit. * The others will get in here next and hit this check. */ - if (unlikely(old_tsb && (rss < mm->context.tsb_rss_limit))) { + if (unlikely(old_tsb && + (rss < mm->context.tsb_block[tsb_index].tsb_rss_limit))) { spin_unlock_irqrestore(&mm->context.lock, flags); kmem_cache_free(tsb_caches[new_cache_index], new_tsb); return; } - mm->context.tsb_rss_limit = new_rss_limit; + mm->context.tsb_block[tsb_index].tsb_rss_limit = new_rss_limit; if (old_tsb) { extern void copy_tsb(unsigned long old_tsb_base, @@ -372,8 +407,8 @@ retry_tsb_alloc: copy_tsb(old_tsb_base, old_size, new_tsb_base, new_size); } - mm->context.tsb = new_tsb; - setup_tsb_params(mm, new_size); + mm->context.tsb_block[tsb_index].tsb = new_tsb; + setup_tsb_params(mm, tsb_index, new_size); spin_unlock_irqrestore(&mm->context.lock, flags); @@ -394,40 +429,65 @@ retry_tsb_alloc: int init_new_context(struct task_struct *tsk, struct mm_struct *mm) { +#ifdef CONFIG_HUGETLB_PAGE + unsigned long huge_pte_count; +#endif + unsigned int i; + spin_lock_init(&mm->context.lock); mm->context.sparc64_ctx_val = 0UL; +#ifdef CONFIG_HUGETLB_PAGE + /* We reset it to zero because the fork() page copying + * will re-increment the counters as the parent PTEs are + * copied into the child address space. + */ + huge_pte_count = mm->context.huge_pte_count; + mm->context.huge_pte_count = 0; +#endif + /* copy_mm() copies over the parent's mm_struct before calling * us, so we need to zero out the TSB pointer or else tsb_grow() * will be confused and think there is an older TSB to free up. */ - mm->context.tsb = NULL; + for (i = 0; i < MM_NUM_TSBS; i++) + mm->context.tsb_block[i].tsb = NULL; /* If this is fork, inherit the parent's TSB size. We would * grow it to that size on the first page fault anyways. */ - tsb_grow(mm, get_mm_rss(mm)); + tsb_grow(mm, MM_TSB_BASE, get_mm_rss(mm)); - if (unlikely(!mm->context.tsb)) +#ifdef CONFIG_HUGETLB_PAGE + if (unlikely(huge_pte_count)) + tsb_grow(mm, MM_TSB_HUGE, huge_pte_count); +#endif + + if (unlikely(!mm->context.tsb_block[MM_TSB_BASE].tsb)) return -ENOMEM; return 0; } -void destroy_context(struct mm_struct *mm) +static void tsb_destroy_one(struct tsb_config *tp) { - unsigned long flags, cache_index; + unsigned long cache_index; - cache_index = (mm->context.tsb_reg_val & 0x7UL); - kmem_cache_free(tsb_caches[cache_index], mm->context.tsb); + if (!tp->tsb) + return; + cache_index = tp->tsb_reg_val & 0x7UL; + kmem_cache_free(tsb_caches[cache_index], tp->tsb); + tp->tsb = NULL; + tp->tsb_reg_val = 0UL; +} - /* We can remove these later, but for now it's useful - * to catch any bogus post-destroy_context() references - * to the TSB. - */ - mm->context.tsb = NULL; - mm->context.tsb_reg_val = 0UL; +void destroy_context(struct mm_struct *mm) +{ + unsigned long flags, i; + + for (i = 0; i < MM_NUM_TSBS; i++) + tsb_destroy_one(&mm->context.tsb_block[i]); spin_lock_irqsave(&ctx_alloc_lock, flags); |