hugetlb: enforce quotas during reservation for shared mappings
authorAdam Litke <agl@us.ibm.com>
Thu, 15 Nov 2007 00:59:42 +0000 (16:59 -0800)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Thu, 15 Nov 2007 02:45:40 +0000 (18:45 -0800)
When a MAP_SHARED mmap of a hugetlbfs file succeeds, huge pages are reserved
to guarantee no problems will occur later when instantiating pages.  If quotas
are in force, page instantiation could fail due to a race with another process
or an oversized (but approved) shared mapping.

To prevent these scenarios, debit the quota for the full reservation amount up
front and credit the unused quota when the reservation is released.

Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/hugetlb.c

index 1e31746..b52b6dd 100644 (file)
@@ -367,7 +367,7 @@ static struct page *alloc_huge_page_shared(struct vm_area_struct *vma,
        spin_lock(&hugetlb_lock);
        page = dequeue_huge_page(vma, addr);
        spin_unlock(&hugetlb_lock);
-       return page;
+       return page ? page : ERR_PTR(-VM_FAULT_OOM);
 }
 
 static struct page *alloc_huge_page_private(struct vm_area_struct *vma,
@@ -375,13 +375,16 @@ static struct page *alloc_huge_page_private(struct vm_area_struct *vma,
 {
        struct page *page = NULL;
 
+       if (hugetlb_get_quota(vma->vm_file->f_mapping, 1))
+               return ERR_PTR(-VM_FAULT_SIGBUS);
+
        spin_lock(&hugetlb_lock);
        if (free_huge_pages > resv_huge_pages)
                page = dequeue_huge_page(vma, addr);
        spin_unlock(&hugetlb_lock);
        if (!page)
                page = alloc_buddy_huge_page(vma, addr);
-       return page;
+       return page ? page : ERR_PTR(-VM_FAULT_OOM);
 }
 
 static struct page *alloc_huge_page(struct vm_area_struct *vma,
@@ -390,19 +393,16 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
        struct page *page;
        struct address_space *mapping = vma->vm_file->f_mapping;
 
-       if (hugetlb_get_quota(mapping, 1))
-               return ERR_PTR(-VM_FAULT_SIGBUS);
-
        if (vma->vm_flags & VM_MAYSHARE)
                page = alloc_huge_page_shared(vma, addr);
        else
                page = alloc_huge_page_private(vma, addr);
-       if (page) {
+
+       if (!IS_ERR(page)) {
                set_page_refcounted(page);
                set_page_private(page, (unsigned long) mapping);
-               return page;
-       } else
-               return ERR_PTR(-VM_FAULT_OOM);
+       }
+       return page;
 }
 
 static int __init hugetlb_init(void)
@@ -1148,6 +1148,8 @@ int hugetlb_reserve_pages(struct inode *inode, long from, long to)
        if (chg < 0)
                return chg;
 
+       if (hugetlb_get_quota(inode->i_mapping, chg))
+               return -ENOSPC;
        ret = hugetlb_acct_memory(chg);
        if (ret < 0)
                return ret;
@@ -1158,5 +1160,6 @@ int hugetlb_reserve_pages(struct inode *inode, long from, long to)
 void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
 {
        long chg = region_truncate(&inode->i_mapping->private_list, offset);
-       hugetlb_acct_memory(freed - chg);
+       hugetlb_put_quota(inode->i_mapping, (chg - freed));
+       hugetlb_acct_memory(-(chg - freed));
 }