]> err.no Git - linux-2.6/commitdiff
[PATCH] mm: msync() cleanup
authorPeter Zijlstra <a.p.zijlstra@chello.nl>
Tue, 26 Sep 2006 06:31:01 +0000 (23:31 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Tue, 26 Sep 2006 15:48:45 +0000 (08:48 -0700)
With the tracking of dirty pages properly done now, msync doesn't need to scan
the PTEs anymore to determine the dirty status.

From: Hugh Dickins <hugh@veritas.com>

In looking to do that, I made some other tidyups: can remove several
#includes, and sys_msync loop termination not quite right.

Most of those points are criticisms of the existing sys_msync, not of your
patch.  In particular, the loop termination errors were introduced in 2.6.17:
I did notice this shortly before it came out, but decided I was more likely to
get it wrong myself, and make matters worse if I tried to rush a last-minute
fix in.  And it's not terribly likely to go wrong, nor disastrous if it does
go wrong (may miss reporting an unmapped area; may also fsync file of a
following vma).

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
mm/msync.c

index d083544df21bdf7a7f2229d5846a02ba22e1218f..358d73cf7b78949c2e62567b1ec89d000dd2cdf7 100644 (file)
 /*
  * The msync() system call.
  */
-#include <linux/slab.h>
-#include <linux/pagemap.h>
 #include <linux/fs.h>
 #include <linux/mm.h>
 #include <linux/mman.h>
-#include <linux/hugetlb.h>
-#include <linux/writeback.h>
 #include <linux/file.h>
 #include <linux/syscalls.h>
 
-#include <asm/pgtable.h>
-#include <asm/tlbflush.h>
-
-static unsigned long msync_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
-                               unsigned long addr, unsigned long end)
-{
-       pte_t *pte;
-       spinlock_t *ptl;
-       int progress = 0;
-       unsigned long ret = 0;
-
-again:
-       pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
-       do {
-               struct page *page;
-
-               if (progress >= 64) {
-                       progress = 0;
-                       if (need_resched() || need_lockbreak(ptl))
-                               break;
-               }
-               progress++;
-               if (!pte_present(*pte))
-                       continue;
-               if (!pte_maybe_dirty(*pte))
-                       continue;
-               page = vm_normal_page(vma, addr, *pte);
-               if (!page)
-                       continue;
-               if (ptep_clear_flush_dirty(vma, addr, pte) ||
-                               page_test_and_clear_dirty(page))
-                       ret += set_page_dirty(page);
-               progress += 3;
-       } while (pte++, addr += PAGE_SIZE, addr != end);
-       pte_unmap_unlock(pte - 1, ptl);
-       cond_resched();
-       if (addr != end)
-               goto again;
-       return ret;
-}
-
-static inline unsigned long msync_pmd_range(struct vm_area_struct *vma,
-                       pud_t *pud, unsigned long addr, unsigned long end)
-{
-       pmd_t *pmd;
-       unsigned long next;
-       unsigned long ret = 0;
-
-       pmd = pmd_offset(pud, addr);
-       do {
-               next = pmd_addr_end(addr, end);
-               if (pmd_none_or_clear_bad(pmd))
-                       continue;
-               ret += msync_pte_range(vma, pmd, addr, next);
-       } while (pmd++, addr = next, addr != end);
-       return ret;
-}
-
-static inline unsigned long msync_pud_range(struct vm_area_struct *vma,
-                       pgd_t *pgd, unsigned long addr, unsigned long end)
-{
-       pud_t *pud;
-       unsigned long next;
-       unsigned long ret = 0;
-
-       pud = pud_offset(pgd, addr);
-       do {
-               next = pud_addr_end(addr, end);
-               if (pud_none_or_clear_bad(pud))
-                       continue;
-               ret += msync_pmd_range(vma, pud, addr, next);
-       } while (pud++, addr = next, addr != end);
-       return ret;
-}
-
-static unsigned long msync_page_range(struct vm_area_struct *vma,
-                               unsigned long addr, unsigned long end)
-{
-       pgd_t *pgd;
-       unsigned long next;
-       unsigned long ret = 0;
-
-       /* For hugepages we can't go walking the page table normally,
-        * but that's ok, hugetlbfs is memory based, so we don't need
-        * to do anything more on an msync().
-        */
-       if (vma->vm_flags & VM_HUGETLB)
-               return 0;
-
-       BUG_ON(addr >= end);
-       pgd = pgd_offset(vma->vm_mm, addr);
-       flush_cache_range(vma, addr, end);
-       do {
-               next = pgd_addr_end(addr, end);
-               if (pgd_none_or_clear_bad(pgd))
-                       continue;
-               ret += msync_pud_range(vma, pgd, addr, next);
-       } while (pgd++, addr = next, addr != end);
-       return ret;
-}
-
 /*
  * MS_SYNC syncs the entire file - including mappings.
  *
- * MS_ASYNC does not start I/O (it used to, up to 2.5.67).  Instead, it just
- * marks the relevant pages dirty.  The application may now run fsync() to
+ * MS_ASYNC does not start I/O (it used to, up to 2.5.67).
+ * Nor does it marks the relevant pages dirty (it used to up to 2.6.17).
+ * Now it doesn't do anything, since dirty pages are properly tracked.
+ *
+ * The application may now run fsync() to
  * write out the dirty pages and wait on the writeout and check the result.
  * Or the application may run fadvise(FADV_DONTNEED) against the fd to start
  * async writeout immediately.
  * So by _not_ starting I/O in MS_ASYNC we provide complete flexibility to
  * applications.
  */
-static int msync_interval(struct vm_area_struct *vma, unsigned long addr,
-                       unsigned long end, int flags,
-                       unsigned long *nr_pages_dirtied)
-{
-       struct file *file = vma->vm_file;
-
-       if ((flags & MS_INVALIDATE) && (vma->vm_flags & VM_LOCKED))
-               return -EBUSY;
-
-       if (file && (vma->vm_flags & VM_SHARED))
-               *nr_pages_dirtied = msync_page_range(vma, addr, end);
-       return 0;
-}
-
 asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
 {
        unsigned long end;
+       struct mm_struct *mm = current->mm;
        struct vm_area_struct *vma;
        int unmapped_error = 0;
        int error = -EINVAL;
-       int done = 0;
 
        if (flags & ~(MS_ASYNC | MS_INVALIDATE | MS_SYNC))
                goto out;
@@ -169,64 +53,50 @@ asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
         * If the interval [start,end) covers some unmapped address ranges,
         * just ignore them, but return -ENOMEM at the end.
         */
-       down_read(&current->mm->mmap_sem);
-       vma = find_vma(current->mm, start);
-       if (!vma) {
-               error = -ENOMEM;
-               goto out_unlock;
-       }
-       do {
-               unsigned long nr_pages_dirtied = 0;
+       down_read(&mm->mmap_sem);
+       vma = find_vma(mm, start);
+       for (;;) {
                struct file *file;
 
+               /* Still start < end. */
+               error = -ENOMEM;
+               if (!vma)
+                       goto out_unlock;
                /* Here start < vma->vm_end. */
                if (start < vma->vm_start) {
-                       unmapped_error = -ENOMEM;
                        start = vma->vm_start;
+                       if (start >= end)
+                               goto out_unlock;
+                       unmapped_error = -ENOMEM;
                }
                /* Here vma->vm_start <= start < vma->vm_end. */
-               if (end <= vma->vm_end) {
-                       if (start < end) {
-                               error = msync_interval(vma, start, end, flags,
-                                                       &nr_pages_dirtied);
-                               if (error)
-                                       goto out_unlock;
-                       }
-                       error = unmapped_error;
-                       done = 1;
-               } else {
-                       /* Here vma->vm_start <= start < vma->vm_end < end. */
-                       error = msync_interval(vma, start, vma->vm_end, flags,
-                                               &nr_pages_dirtied);
-                       if (error)
-                               goto out_unlock;
+               if ((flags & MS_INVALIDATE) &&
+                               (vma->vm_flags & VM_LOCKED)) {
+                       error = -EBUSY;
+                       goto out_unlock;
                }
                file = vma->vm_file;
                start = vma->vm_end;
-               if ((flags & MS_ASYNC) && file && nr_pages_dirtied) {
-                       get_file(file);
-                       up_read(&current->mm->mmap_sem);
-                       balance_dirty_pages_ratelimited_nr(file->f_mapping,
-                                                       nr_pages_dirtied);
-                       fput(file);
-                       down_read(&current->mm->mmap_sem);
-                       vma = find_vma(current->mm, start);
-               } else if ((flags & MS_SYNC) && file &&
+               if ((flags & MS_SYNC) && file &&
                                (vma->vm_flags & VM_SHARED)) {
                        get_file(file);
-                       up_read(&current->mm->mmap_sem);
+                       up_read(&mm->mmap_sem);
                        error = do_fsync(file, 0);
                        fput(file);
-                       down_read(&current->mm->mmap_sem);
-                       if (error)
-                               goto out_unlock;
-                       vma = find_vma(current->mm, start);
+                       if (error || start >= end)
+                               goto out;
+                       down_read(&mm->mmap_sem);
+                       vma = find_vma(mm, start);
                } else {
+                       if (start >= end) {
+                               error = 0;
+                               goto out_unlock;
+                       }
                        vma = vma->vm_next;
                }
-       } while (vma && !done);
+       }
 out_unlock:
-       up_read(&current->mm->mmap_sem);
+       up_read(&mm->mmap_sem);
 out:
-       return error;
+       return error ? : unmapped_error;
 }