From 63551ae0feaaa23807ebea60de1901564bbef32e Mon Sep 17 00:00:00 2001 From: David Gibson Date: Tue, 21 Jun 2005 17:14:44 -0700 Subject: [PATCH] Hugepage consolidation A lot of the code in arch/*/mm/hugetlbpage.c is quite similar. This patch attempts to consolidate a lot of the code across the arch's, putting the combined version in mm/hugetlb.c. There are a couple of uglyish hacks in order to covert all the hugepage archs, but the result is a very large reduction in the total amount of code. It also means things like hugepage lazy allocation could be implemented in one place, instead of six. Tested, at least a little, on ppc64, i386 and x86_64. Notes: - this patch changes the meaning of set_huge_pte() to be more analagous to set_pte() - does SH4 need s special huge_ptep_get_and_clear()?? Acked-by: William Lee Irwin Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- arch/i386/mm/hugetlbpage.c | 170 +++------------------------------------------ 1 file changed, 8 insertions(+), 162 deletions(-) (limited to 'arch/i386/mm') diff --git a/arch/i386/mm/hugetlbpage.c b/arch/i386/mm/hugetlbpage.c index 171fc925e1e..5aa06001a4b 100644 --- a/arch/i386/mm/hugetlbpage.c +++ b/arch/i386/mm/hugetlbpage.c @@ -18,7 +18,7 @@ #include #include -static pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr) +pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr) { pgd_t *pgd; pud_t *pud; @@ -30,7 +30,7 @@ static pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr) return (pte_t *) pmd; } -static pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) +pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) { pgd_t *pgd; pud_t *pud; @@ -42,21 +42,6 @@ static pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) return (pte_t *) pmd; } -static void set_huge_pte(struct mm_struct *mm, struct vm_area_struct *vma, struct page *page, pte_t * page_table, int write_access) -{ - pte_t entry; - - add_mm_counter(mm, rss, HPAGE_SIZE / PAGE_SIZE); - if (write_access) { - entry = - pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot))); - } else - entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot)); - entry = pte_mkyoung(entry); - mk_pte_huge(entry); - set_pte(page_table, entry); -} - /* * This function checks for proper alignment of input addr and len parameters. */ @@ -69,77 +54,6 @@ int is_aligned_hugepage_range(unsigned long addr, unsigned long len) return 0; } -int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src, - struct vm_area_struct *vma) -{ - pte_t *src_pte, *dst_pte, entry; - struct page *ptepage; - unsigned long addr = vma->vm_start; - unsigned long end = vma->vm_end; - - while (addr < end) { - dst_pte = huge_pte_alloc(dst, addr); - if (!dst_pte) - goto nomem; - src_pte = huge_pte_offset(src, addr); - entry = *src_pte; - ptepage = pte_page(entry); - get_page(ptepage); - set_pte(dst_pte, entry); - add_mm_counter(dst, rss, HPAGE_SIZE / PAGE_SIZE); - addr += HPAGE_SIZE; - } - return 0; - -nomem: - return -ENOMEM; -} - -int -follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma, - struct page **pages, struct vm_area_struct **vmas, - unsigned long *position, int *length, int i) -{ - unsigned long vpfn, vaddr = *position; - int remainder = *length; - - WARN_ON(!is_vm_hugetlb_page(vma)); - - vpfn = vaddr/PAGE_SIZE; - while (vaddr < vma->vm_end && remainder) { - - if (pages) { - pte_t *pte; - struct page *page; - - pte = huge_pte_offset(mm, vaddr); - - /* hugetlb should be locked, and hence, prefaulted */ - WARN_ON(!pte || pte_none(*pte)); - - page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)]; - - WARN_ON(!PageCompound(page)); - - get_page(page); - pages[i] = page; - } - - if (vmas) - vmas[i] = vma; - - vaddr += PAGE_SIZE; - ++vpfn; - --remainder; - ++i; - } - - *length = remainder; - *position = vaddr; - - return i; -} - #if 0 /* This is just for testing */ struct page * follow_huge_addr(struct mm_struct *mm, unsigned long address, int write) @@ -204,83 +118,15 @@ follow_huge_pmd(struct mm_struct *mm, unsigned long address, } #endif -void unmap_hugepage_range(struct vm_area_struct *vma, - unsigned long start, unsigned long end) +void hugetlb_clean_stale_pgtable(pte_t *pte) { - struct mm_struct *mm = vma->vm_mm; - unsigned long address; - pte_t pte, *ptep; + pmd_t *pmd = (pmd_t *) pte; struct page *page; - BUG_ON(start & (HPAGE_SIZE - 1)); - BUG_ON(end & (HPAGE_SIZE - 1)); - - for (address = start; address < end; address += HPAGE_SIZE) { - ptep = huge_pte_offset(mm, address); - if (!ptep) - continue; - pte = ptep_get_and_clear(mm, address, ptep); - if (pte_none(pte)) - continue; - page = pte_page(pte); - put_page(page); - } - add_mm_counter(mm ,rss, -((end - start) >> PAGE_SHIFT)); - flush_tlb_range(vma, start, end); -} - -int hugetlb_prefault(struct address_space *mapping, struct vm_area_struct *vma) -{ - struct mm_struct *mm = current->mm; - unsigned long addr; - int ret = 0; - - BUG_ON(vma->vm_start & ~HPAGE_MASK); - BUG_ON(vma->vm_end & ~HPAGE_MASK); - - spin_lock(&mm->page_table_lock); - for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) { - unsigned long idx; - pte_t *pte = huge_pte_alloc(mm, addr); - struct page *page; - - if (!pte) { - ret = -ENOMEM; - goto out; - } - - if (!pte_none(*pte)) - continue; - - idx = ((addr - vma->vm_start) >> HPAGE_SHIFT) - + (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT)); - page = find_get_page(mapping, idx); - if (!page) { - /* charge the fs quota first */ - if (hugetlb_get_quota(mapping)) { - ret = -ENOMEM; - goto out; - } - page = alloc_huge_page(); - if (!page) { - hugetlb_put_quota(mapping); - ret = -ENOMEM; - goto out; - } - ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC); - if (! ret) { - unlock_page(page); - } else { - hugetlb_put_quota(mapping); - free_huge_page(page); - goto out; - } - } - set_huge_pte(mm, vma, page, pte, vma->vm_flags & VM_WRITE); - } -out: - spin_unlock(&mm->page_table_lock); - return ret; + page = pmd_page(*pmd); + pmd_clear(pmd); + dec_page_state(nr_page_table_pages); + page_cache_release(page); } /* x86_64 also uses this file */ -- cgit v1.2.3 From 1363c3cd8603a913a27e2995dccbd70d5312d8e6 Mon Sep 17 00:00:00 2001 From: Wolfgang Wander Date: Tue, 21 Jun 2005 17:14:49 -0700 Subject: [PATCH] Avoiding mmap fragmentation Ingo recently introduced a great speedup for allocating new mmaps using the free_area_cache pointer which boosts the specweb SSL benchmark by 4-5% and causes huge performance increases in thread creation. The downside of this patch is that it does lead to fragmentation in the mmap-ed areas (visible via /proc/self/maps), such that some applications that work fine under 2.4 kernels quickly run out of memory on any 2.6 kernel. The problem is twofold: 1) the free_area_cache is used to continue a search for memory where the last search ended. Before the change new areas were always searched from the base address on. So now new small areas are cluttering holes of all sizes throughout the whole mmap-able region whereas before small holes tended to close holes near the base leaving holes far from the base large and available for larger requests. 2) the free_area_cache also is set to the location of the last munmap-ed area so in scenarios where we allocate e.g. five regions of 1K each, then free regions 4 2 3 in this order the next request for 1K will be placed in the position of the old region 3, whereas before we appended it to the still active region 1, placing it at the location of the old region 2. Before we had 1 free region of 2K, now we only get two free regions of 1K -> fragmentation. The patch addresses thes issues by introducing yet another cache descriptor cached_hole_size that contains the largest known hole size below the current free_area_cache. If a new request comes in the size is compared against the cached_hole_size and if the request can be filled with a hole below free_area_cache the search is started from the base instead. The results look promising: Whereas 2.6.12-rc4 fragments quickly and my (earlier posted) leakme.c test program terminates after 50000+ iterations with 96 distinct and fragmented maps in /proc/self/maps it performs nicely (as expected) with thread creation, Ingo's test_str02 with 20000 threads requires 0.7s system time. Taking out Ingo's patch (un-patch available per request) by basically deleting all mentions of free_area_cache from the kernel and starting the search for new memory always at the respective bases we observe: leakme terminates successfully with 11 distinctive hardly fragmented areas in /proc/self/maps but thread creating is gringdingly slow: 30+s(!) system time for Ingo's test_str02 with 20000 threads. Now - drumroll ;-) the appended patch works fine with leakme: it ends with only 7 distinct areas in /proc/self/maps and also thread creation seems sufficiently fast with 0.71s for 20000 threads. Signed-off-by: Wolfgang Wander Credit-to: "Richard Purdie" Signed-off-by: Ken Chen Acked-by: Ingo Molnar (partly) Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- arch/i386/mm/hugetlbpage.c | 34 +++++++++++++++++++++++++++++----- 1 file changed, 29 insertions(+), 5 deletions(-) (limited to 'arch/i386/mm') diff --git a/arch/i386/mm/hugetlbpage.c b/arch/i386/mm/hugetlbpage.c index 5aa06001a4b..3b099f32b94 100644 --- a/arch/i386/mm/hugetlbpage.c +++ b/arch/i386/mm/hugetlbpage.c @@ -140,7 +140,12 @@ static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *file, struct vm_area_struct *vma; unsigned long start_addr; - start_addr = mm->free_area_cache; + if (len > mm->cached_hole_size) { + start_addr = mm->free_area_cache; + } else { + start_addr = TASK_UNMAPPED_BASE; + mm->cached_hole_size = 0; + } full_search: addr = ALIGN(start_addr, HPAGE_SIZE); @@ -154,6 +159,7 @@ full_search: */ if (start_addr != TASK_UNMAPPED_BASE) { start_addr = TASK_UNMAPPED_BASE; + mm->cached_hole_size = 0; goto full_search; } return -ENOMEM; @@ -162,6 +168,8 @@ full_search: mm->free_area_cache = addr + len; return addr; } + if (addr + mm->cached_hole_size < vma->vm_start) + mm->cached_hole_size = vma->vm_start - addr; addr = ALIGN(vma->vm_end, HPAGE_SIZE); } } @@ -173,12 +181,17 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file, struct mm_struct *mm = current->mm; struct vm_area_struct *vma, *prev_vma; unsigned long base = mm->mmap_base, addr = addr0; + unsigned long largest_hole = mm->cached_hole_size; int first_time = 1; /* don't allow allocations above current base */ if (mm->free_area_cache > base) mm->free_area_cache = base; + if (len <= largest_hole) { + largest_hole = 0; + mm->free_area_cache = base; + } try_again: /* make sure it can fit in the remaining address space */ if (mm->free_area_cache < len) @@ -199,13 +212,21 @@ try_again: * vma->vm_start, use it: */ if (addr + len <= vma->vm_start && - (!prev_vma || (addr >= prev_vma->vm_end))) + (!prev_vma || (addr >= prev_vma->vm_end))) { /* remember the address as a hint for next time */ - return (mm->free_area_cache = addr); - else + mm->cached_hole_size = largest_hole; + return (mm->free_area_cache = addr); + } else { /* pull free_area_cache down to the first hole */ - if (mm->free_area_cache == vma->vm_end) + if (mm->free_area_cache == vma->vm_end) { mm->free_area_cache = vma->vm_start; + mm->cached_hole_size = largest_hole; + } + } + + /* remember the largest hole we saw so far */ + if (addr + largest_hole < vma->vm_start) + largest_hole = vma->vm_start - addr; /* try just below the current vma->vm_start */ addr = (vma->vm_start - len) & HPAGE_MASK; @@ -218,6 +239,7 @@ fail: */ if (first_time) { mm->free_area_cache = base; + largest_hole = 0; first_time = 0; goto try_again; } @@ -228,6 +250,7 @@ fail: * allocations. */ mm->free_area_cache = TASK_UNMAPPED_BASE; + mm->cached_hole_size = ~0UL; addr = hugetlb_get_unmapped_area_bottomup(file, addr0, len, pgoff, flags); @@ -235,6 +258,7 @@ fail: * Restore the topdown base: */ mm->free_area_cache = base; + mm->cached_hole_size = ~0UL; return addr; } -- cgit v1.2.3 From cbe37d093707762fc0abb280781e6a82a9d8d568 Mon Sep 17 00:00:00 2001 From: Badari Pulavarty Date: Tue, 21 Jun 2005 17:14:52 -0700 Subject: [PATCH] mm: remove PG_highmem Remove PG_highmem, to save a page flag. Use is_highmem() instead. It'll generate a little more code, but we don't use PageHigheMem() in many places. Signed-off-by: Badari Pulavarty Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- arch/i386/mm/init.c | 1 - 1 file changed, 1 deletion(-) (limited to 'arch/i386/mm') diff --git a/arch/i386/mm/init.c b/arch/i386/mm/init.c index 7a7ea373726..8766c771bb4 100644 --- a/arch/i386/mm/init.c +++ b/arch/i386/mm/init.c @@ -269,7 +269,6 @@ void __init one_highpage_init(struct page *page, int pfn, int bad_ppro) { if (page_is_ram(pfn) && !(bad_ppro && page_kills_ppro(pfn))) { ClearPageReserved(page); - set_bit(PG_highmem, &page->flags); set_page_count(page, 1); __free_page(page); totalhigh_pages++; -- cgit v1.2.3