allocate structures for reservation tracking in hugetlbfs outside of spinlocks v2
Andy Whitcroft [Tue, 12 Aug 2008 22:08:49 +0000 (15:08 -0700)]
[Andrew this should replace the previous version which did not check
the returns from the region prepare for errors.  This has been tested by
us and Gerald and it looks good.

Bah, while reviewing the locking based on your previous email I spotted
that we need to check the return from the vma_needs_reservation call for
allocation errors.  Here is an updated patch to correct this.  This passes
testing here.]

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Tested-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/hugetlb.c

index 4c97c17..67a7119 100644 (file)
@@ -1949,7 +1949,10 @@ retry:
         * the spinlock.
         */
        if (write_access && !(vma->vm_flags & VM_SHARED))
-               vma_needs_reservation(h, vma, address);
+               if (vma_needs_reservation(h, vma, address) < 0) {
+                       ret = VM_FAULT_OOM;
+                       goto backout_unlocked;
+               }
 
        spin_lock(&mm->page_table_lock);
        size = i_size_read(mapping->host) >> huge_page_shift(h);
@@ -1976,6 +1979,7 @@ out:
 
 backout:
        spin_unlock(&mm->page_table_lock);
+backout_unlocked:
        unlock_page(page);
        put_page(page);
        goto out;
@@ -2004,8 +2008,7 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
        entry = huge_ptep_get(ptep);
        if (huge_pte_none(entry)) {
                ret = hugetlb_no_page(mm, vma, address, ptep, write_access);
-               mutex_unlock(&hugetlb_instantiation_mutex);
-               return ret;
+               goto out_unlock;
        }
 
        ret = 0;
@@ -2019,7 +2022,10 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
         * consumed.
         */
        if (write_access && !pte_write(entry)) {
-               vma_needs_reservation(h, vma, address);
+               if (vma_needs_reservation(h, vma, address) < 0) {
+                       ret = VM_FAULT_OOM;
+                       goto out_unlock;
+               }
 
                if (!(vma->vm_flags & VM_SHARED))
                        pagecache_page = hugetlbfs_pagecache_page(h,
@@ -2039,6 +2045,7 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
                put_page(pagecache_page);
        }
 
+out_unlock:
        mutex_unlock(&hugetlb_instantiation_mutex);
 
        return ret;