From a91690ddd05ab0b7fbdd37733875525ac75c20f2 Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Fri, 17 Mar 2006 14:41:03 -0800 Subject: [PATCH] [SPARC64]: Top-down address space allocation for 32-bit tasks. Currently allocations are very constrained for 32-bit processes. It grows down-up from 0x70000000 to 0xf0000000 which gives about 2GB of stack + dynamic mmap() space. So support the top-down method, and we need to override the generic helper function in order to deal with D-cache coloring. With these changes I was able to squeeze out a mmap() just over 3.6GB in size in a 32-bit process. Signed-off-by: David S. Miller --- arch/sparc64/kernel/binfmt_aout32.c | 2 + arch/sparc64/kernel/sys_sparc.c | 190 ++++++++++++++++++++++++++-- include/asm-sparc64/pgtable.h | 5 +- include/asm-sparc64/processor.h | 2 + 4 files changed, 189 insertions(+), 10 deletions(-) diff --git a/arch/sparc64/kernel/binfmt_aout32.c b/arch/sparc64/kernel/binfmt_aout32.c index cb9ecd0172..d7caa60a00 100644 --- a/arch/sparc64/kernel/binfmt_aout32.c +++ b/arch/sparc64/kernel/binfmt_aout32.c @@ -239,6 +239,8 @@ static int load_aout32_binary(struct linux_binprm * bprm, struct pt_regs * regs) (current->mm->start_data = N_DATADDR(ex)); current->mm->brk = ex.a_bss + (current->mm->start_brk = N_BSSADDR(ex)); + current->mm->free_area_cache = current->mm->mmap_base; + current->mm->cached_hole_size = 0; current->mm->mmap = NULL; compute_creds(bprm); diff --git a/arch/sparc64/kernel/sys_sparc.c b/arch/sparc64/kernel/sys_sparc.c index 8840415408..61dffb9349 100644 --- a/arch/sparc64/kernel/sys_sparc.c +++ b/arch/sparc64/kernel/sys_sparc.c @@ -82,9 +82,34 @@ static inline int straddles_64bit_va_hole(unsigned long start, unsigned long end return 1; } -#define COLOUR_ALIGN(addr,pgoff) \ - ((((addr)+SHMLBA-1)&~(SHMLBA-1)) + \ - (((pgoff)< task_size || len >= VA_EXCLUDE_START) + if (unlikely(len > task_size || len >= VA_EXCLUDE_START)) return -ENOMEM; do_color_align = 0; @@ -125,11 +150,12 @@ unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr, unsi return addr; } - if (len <= mm->cached_hole_size) { + if (len > mm->cached_hole_size) { + start_addr = addr = mm->free_area_cache; + } else { + start_addr = addr = TASK_UNMAPPED_BASE; mm->cached_hole_size = 0; - mm->free_area_cache = TASK_UNMAPPED_BASE; } - start_addr = addr = mm->free_area_cache; task_size -= len; @@ -146,7 +172,7 @@ full_search: addr = VA_EXCLUDE_END; vma = find_vma(mm, VA_EXCLUDE_END); } - if (task_size < addr) { + if (unlikely(task_size < addr)) { if (start_addr != TASK_UNMAPPED_BASE) { start_addr = addr = TASK_UNMAPPED_BASE; mm->cached_hole_size = 0; @@ -154,7 +180,7 @@ full_search: } return -ENOMEM; } - if (!vma || addr + len <= vma->vm_start) { + if (likely(!vma || addr + len <= vma->vm_start)) { /* * Remember the place where we stopped the search: */ @@ -170,6 +196,121 @@ full_search: } } +unsigned long +arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0, + const unsigned long len, const unsigned long pgoff, + const unsigned long flags) +{ + struct vm_area_struct *vma; + struct mm_struct *mm = current->mm; + unsigned long task_size = 0xf0000000UL; + unsigned long addr = addr0; + int do_color_align; + + /* This should only ever run for 32-bit processes. */ + BUG_ON(!test_thread_flag(TIF_32BIT)); + + if (flags & MAP_FIXED) { + /* We do not accept a shared mapping if it would violate + * cache aliasing constraints. + */ + if ((flags & MAP_SHARED) && + ((addr - (pgoff << PAGE_SHIFT)) & (SHMLBA - 1))) + return -EINVAL; + return addr; + } + + if (unlikely(len > task_size)) + return -ENOMEM; + + do_color_align = 0; + if (filp || (flags & MAP_SHARED)) + do_color_align = 1; + + /* requesting a specific address */ + if (addr) { + if (do_color_align) + addr = COLOUR_ALIGN(addr, pgoff); + else + addr = PAGE_ALIGN(addr); + + vma = find_vma(mm, addr); + if (task_size - len >= addr && + (!vma || addr + len <= vma->vm_start)) + return addr; + } + + /* check if free_area_cache is useful for us */ + if (len <= mm->cached_hole_size) { + mm->cached_hole_size = 0; + mm->free_area_cache = mm->mmap_base; + } + + /* either no address requested or can't fit in requested address hole */ + addr = mm->free_area_cache; + if (do_color_align) { + unsigned long base = COLOUR_ALIGN_DOWN(addr-len, pgoff); + + addr = base + len; + } + + /* make sure it can fit in the remaining address space */ + if (likely(addr > len)) { + vma = find_vma(mm, addr-len); + if (!vma || addr <= vma->vm_start) { + /* remember the address as a hint for next time */ + return (mm->free_area_cache = addr-len); + } + } + + if (unlikely(mm->mmap_base < len)) + goto bottomup; + + addr = mm->mmap_base-len; + if (do_color_align) + addr = COLOUR_ALIGN_DOWN(addr, pgoff); + + do { + /* + * Lookup failure means no vma is above this address, + * else if new region fits below vma->vm_start, + * return with success: + */ + vma = find_vma(mm, addr); + if (likely(!vma || addr+len <= vma->vm_start)) { + /* remember the address as a hint for next time */ + return (mm->free_area_cache = addr); + } + + /* remember the largest hole we saw so far */ + if (addr + mm->cached_hole_size < vma->vm_start) + mm->cached_hole_size = vma->vm_start - addr; + + /* try just below the current vma->vm_start */ + addr = vma->vm_start-len; + if (do_color_align) + addr = COLOUR_ALIGN_DOWN(addr, pgoff); + } while (likely(len < vma->vm_start)); + +bottomup: + /* + * A failed mmap() very likely causes application failure, + * so fall back to the bottom-up function here. This scenario + * can happen with large stack limits and large mmap() + * allocations. + */ + mm->cached_hole_size = ~0UL; + mm->free_area_cache = TASK_UNMAPPED_BASE; + addr = arch_get_unmapped_area(filp, addr0, len, pgoff, flags); + /* + * Restore the topdown base: + */ + mm->free_area_cache = mm->mmap_base; + mm->cached_hole_size = ~0UL; + + return addr; +} + /* Try to align mapping such that we align it as much as possible. */ unsigned long get_fb_unmapped_area(struct file *filp, unsigned long orig_addr, unsigned long len, unsigned long pgoff, unsigned long flags) { @@ -213,6 +354,37 @@ unsigned long get_fb_unmapped_area(struct file *filp, unsigned long orig_addr, u return addr; } +/* Essentially the same as PowerPC... */ +void arch_pick_mmap_layout(struct mm_struct *mm) +{ + /* + * Fall back to the standard layout if the personality + * bit is set, or if the expected stack growth is unlimited: + */ + if (!test_thread_flag(TIF_32BIT) || + (current->personality & ADDR_COMPAT_LAYOUT) || + current->signal->rlim[RLIMIT_STACK].rlim_cur == RLIM_INFINITY || + sysctl_legacy_va_layout) { + mm->mmap_base = TASK_UNMAPPED_BASE; + mm->get_unmapped_area = arch_get_unmapped_area; + mm->unmap_area = arch_unmap_area; + } else { + /* We know it's 32-bit */ + unsigned long task_size = 0xf0000000UL; + unsigned long gap; + + gap = current->signal->rlim[RLIMIT_STACK].rlim_cur; + if (gap < 128 * 1024 * 1024) + gap = 128 * 1024 * 1024; + if (gap > (task_size / 6 * 5)) + gap = (task_size / 6 * 5); + + mm->mmap_base = task_size - (gap & PAGE_MASK); + mm->get_unmapped_area = arch_get_unmapped_area_topdown; + mm->unmap_area = arch_unmap_area_topdown; + } +} + asmlinkage unsigned long sparc_brk(unsigned long brk) { /* People could try to be nasty and use ta 0x6d in 32bit programs */ diff --git a/include/asm-sparc64/pgtable.h b/include/asm-sparc64/pgtable.h index d427ce6492..ed4124edf8 100644 --- a/include/asm-sparc64/pgtable.h +++ b/include/asm-sparc64/pgtable.h @@ -752,8 +752,11 @@ extern int io_remap_pfn_range(struct vm_area_struct *vma, unsigned long from, #include -/* We provide our own get_unmapped_area to cope with VA holes for userland */ +/* We provide our own get_unmapped_area to cope with VA holes and + * SHM area cache aliasing for userland. + */ #define HAVE_ARCH_UNMAPPED_AREA +#define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN /* We provide a special get_unmapped_area for framebuffer mmaps to try and use * the largest alignment possible such that larget PTEs can be used. diff --git a/include/asm-sparc64/processor.h b/include/asm-sparc64/processor.h index 685479fb43..c6896b8828 100644 --- a/include/asm-sparc64/processor.h +++ b/include/asm-sparc64/processor.h @@ -217,6 +217,8 @@ static inline void prefetchw(const void *x) #define spin_lock_prefetch(x) prefetchw(x) +#define HAVE_ARCH_PICK_MMAP_LAYOUT + #endif /* !(__ASSEMBLY__) */ #endif /* !(__ASM_SPARC64_PROCESSOR_H) */ -- 2.39.5