aboutsummaryrefslogtreecommitdiffstats
path: root/mm/hugetlb.c
diff options
context:
space:
mode:
authorMel Gorman <mel@csn.ul.ie>2009-05-28 17:34:40 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-05-29 11:40:03 -0400
commitf83a275dbc5ca1721143698e844243fcadfabf6a (patch)
treed76da760e6e234d1d445211c59f0f1087d57facf /mm/hugetlb.c
parent32b154c0b0bae2879bf4e549d861caf1759a3546 (diff)
mm: account for MAP_SHARED mappings using VM_MAYSHARE and not VM_SHARED in hugetlbfs
Addresses http://bugzilla.kernel.org/show_bug.cgi?id=13302 hugetlbfs reserves huge pages but does not fault them at mmap() time to ensure that future faults succeed. The reservation behaviour differs depending on whether the mapping was mapped MAP_SHARED or MAP_PRIVATE. For MAP_SHARED mappings, hugepages are reserved when mmap() is first called and are tracked based on information associated with the inode. Other processes mapping MAP_SHARED use the same reservation. MAP_PRIVATE track the reservations based on the VMA created as part of the mmap() operation. Each process mapping MAP_PRIVATE must make its own reservation. hugetlbfs currently checks if a VMA is MAP_SHARED with the VM_SHARED flag and not VM_MAYSHARE. For file-backed mappings, such as hugetlbfs, VM_SHARED is set only if the mapping is MAP_SHARED and the file was opened read-write. If a shared memory mapping was mapped shared-read-write for populating of data and mapped shared-read-only by other processes, then hugetlbfs would account for the mapping as if it was MAP_PRIVATE. This causes processes to fail to map the file MAP_SHARED even though it should succeed as the reservation is there. This patch alters mm/hugetlb.c and replaces VM_SHARED with VM_MAYSHARE when the intent of the code was to check whether the VMA was mapped MAP_SHARED or MAP_PRIVATE. Signed-off-by: Mel Gorman <mel@csn.ul.ie> Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk> Cc: Ingo Molnar <mingo@elte.hu> Cc: <stable@kernel.org> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: <starlight@binnacle.cx> Cc: Eric B Munson <ebmunson@us.ibm.com> Cc: Adam Litke <agl@us.ibm.com> Cc: Andy Whitcroft <apw@canonical.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'mm/hugetlb.c')
-rw-r--r--mm/hugetlb.c26
1 files changed, 13 insertions, 13 deletions
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 28c655ba935..e83ad2c9228 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -316,7 +316,7 @@ static void resv_map_release(struct kref *ref)
316static struct resv_map *vma_resv_map(struct vm_area_struct *vma) 316static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
317{ 317{
318 VM_BUG_ON(!is_vm_hugetlb_page(vma)); 318 VM_BUG_ON(!is_vm_hugetlb_page(vma));
319 if (!(vma->vm_flags & VM_SHARED)) 319 if (!(vma->vm_flags & VM_MAYSHARE))
320 return (struct resv_map *)(get_vma_private_data(vma) & 320 return (struct resv_map *)(get_vma_private_data(vma) &
321 ~HPAGE_RESV_MASK); 321 ~HPAGE_RESV_MASK);
322 return NULL; 322 return NULL;
@@ -325,7 +325,7 @@ static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
325static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map) 325static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
326{ 326{
327 VM_BUG_ON(!is_vm_hugetlb_page(vma)); 327 VM_BUG_ON(!is_vm_hugetlb_page(vma));
328 VM_BUG_ON(vma->vm_flags & VM_SHARED); 328 VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
329 329
330 set_vma_private_data(vma, (get_vma_private_data(vma) & 330 set_vma_private_data(vma, (get_vma_private_data(vma) &
331 HPAGE_RESV_MASK) | (unsigned long)map); 331 HPAGE_RESV_MASK) | (unsigned long)map);
@@ -334,7 +334,7 @@ static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
334static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags) 334static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
335{ 335{
336 VM_BUG_ON(!is_vm_hugetlb_page(vma)); 336 VM_BUG_ON(!is_vm_hugetlb_page(vma));
337 VM_BUG_ON(vma->vm_flags & VM_SHARED); 337 VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
338 338
339 set_vma_private_data(vma, get_vma_private_data(vma) | flags); 339 set_vma_private_data(vma, get_vma_private_data(vma) | flags);
340} 340}
@@ -353,7 +353,7 @@ static void decrement_hugepage_resv_vma(struct hstate *h,
353 if (vma->vm_flags & VM_NORESERVE) 353 if (vma->vm_flags & VM_NORESERVE)
354 return; 354 return;
355 355
356 if (vma->vm_flags & VM_SHARED) { 356 if (vma->vm_flags & VM_MAYSHARE) {
357 /* Shared mappings always use reserves */ 357 /* Shared mappings always use reserves */
358 h->resv_huge_pages--; 358 h->resv_huge_pages--;
359 } else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) { 359 } else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
@@ -369,14 +369,14 @@ static void decrement_hugepage_resv_vma(struct hstate *h,
369void reset_vma_resv_huge_pages(struct vm_area_struct *vma) 369void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
370{ 370{
371 VM_BUG_ON(!is_vm_hugetlb_page(vma)); 371 VM_BUG_ON(!is_vm_hugetlb_page(vma));
372 if (!(vma->vm_flags & VM_SHARED)) 372 if (!(vma->vm_flags & VM_MAYSHARE))
373 vma->vm_private_data = (void *)0; 373 vma->vm_private_data = (void *)0;
374} 374}
375 375
376/* Returns true if the VMA has associated reserve pages */ 376/* Returns true if the VMA has associated reserve pages */
377static int vma_has_reserves(struct vm_area_struct *vma) 377static int vma_has_reserves(struct vm_area_struct *vma)
378{ 378{
379 if (vma->vm_flags & VM_SHARED) 379 if (vma->vm_flags & VM_MAYSHARE)
380 return 1; 380 return 1;
381 if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) 381 if (is_vma_resv_set(vma, HPAGE_RESV_OWNER))
382 return 1; 382 return 1;
@@ -924,7 +924,7 @@ static long vma_needs_reservation(struct hstate *h,
924 struct address_space *mapping = vma->vm_file->f_mapping; 924 struct address_space *mapping = vma->vm_file->f_mapping;
925 struct inode *inode = mapping->host; 925 struct inode *inode = mapping->host;
926 926
927 if (vma->vm_flags & VM_SHARED) { 927 if (vma->vm_flags & VM_MAYSHARE) {
928 pgoff_t idx = vma_hugecache_offset(h, vma, addr); 928 pgoff_t idx = vma_hugecache_offset(h, vma, addr);
929 return region_chg(&inode->i_mapping->private_list, 929 return region_chg(&inode->i_mapping->private_list,
930 idx, idx + 1); 930 idx, idx + 1);
@@ -949,7 +949,7 @@ static void vma_commit_reservation(struct hstate *h,
949 struct address_space *mapping = vma->vm_file->f_mapping; 949 struct address_space *mapping = vma->vm_file->f_mapping;
950 struct inode *inode = mapping->host; 950 struct inode *inode = mapping->host;
951 951
952 if (vma->vm_flags & VM_SHARED) { 952 if (vma->vm_flags & VM_MAYSHARE) {
953 pgoff_t idx = vma_hugecache_offset(h, vma, addr); 953 pgoff_t idx = vma_hugecache_offset(h, vma, addr);
954 region_add(&inode->i_mapping->private_list, idx, idx + 1); 954 region_add(&inode->i_mapping->private_list, idx, idx + 1);
955 955
@@ -1893,7 +1893,7 @@ retry_avoidcopy:
1893 * at the time of fork() could consume its reserves on COW instead 1893 * at the time of fork() could consume its reserves on COW instead
1894 * of the full address range. 1894 * of the full address range.
1895 */ 1895 */
1896 if (!(vma->vm_flags & VM_SHARED) && 1896 if (!(vma->vm_flags & VM_MAYSHARE) &&
1897 is_vma_resv_set(vma, HPAGE_RESV_OWNER) && 1897 is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
1898 old_page != pagecache_page) 1898 old_page != pagecache_page)
1899 outside_reserve = 1; 1899 outside_reserve = 1;
@@ -2000,7 +2000,7 @@ retry:
2000 clear_huge_page(page, address, huge_page_size(h)); 2000 clear_huge_page(page, address, huge_page_size(h));
2001 __SetPageUptodate(page); 2001 __SetPageUptodate(page);
2002 2002
2003 if (vma->vm_flags & VM_SHARED) { 2003 if (vma->vm_flags & VM_MAYSHARE) {
2004 int err; 2004 int err;
2005 struct inode *inode = mapping->host; 2005 struct inode *inode = mapping->host;
2006 2006
@@ -2104,7 +2104,7 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2104 goto out_mutex; 2104 goto out_mutex;
2105 } 2105 }
2106 2106
2107 if (!(vma->vm_flags & VM_SHARED)) 2107 if (!(vma->vm_flags & VM_MAYSHARE))
2108 pagecache_page = hugetlbfs_pagecache_page(h, 2108 pagecache_page = hugetlbfs_pagecache_page(h,
2109 vma, address); 2109 vma, address);
2110 } 2110 }
@@ -2289,7 +2289,7 @@ int hugetlb_reserve_pages(struct inode *inode,
2289 * to reserve the full area even if read-only as mprotect() may be 2289 * to reserve the full area even if read-only as mprotect() may be
2290 * called to make the mapping read-write. Assume !vma is a shm mapping 2290 * called to make the mapping read-write. Assume !vma is a shm mapping
2291 */ 2291 */
2292 if (!vma || vma->vm_flags & VM_SHARED) 2292 if (!vma || vma->vm_flags & VM_MAYSHARE)
2293 chg = region_chg(&inode->i_mapping->private_list, from, to); 2293 chg = region_chg(&inode->i_mapping->private_list, from, to);
2294 else { 2294 else {
2295 struct resv_map *resv_map = resv_map_alloc(); 2295 struct resv_map *resv_map = resv_map_alloc();
@@ -2330,7 +2330,7 @@ int hugetlb_reserve_pages(struct inode *inode,
2330 * consumed reservations are stored in the map. Hence, nothing 2330 * consumed reservations are stored in the map. Hence, nothing
2331 * else has to be done for private mappings here 2331 * else has to be done for private mappings here
2332 */ 2332 */
2333 if (!vma || vma->vm_flags & VM_SHARED) 2333 if (!vma || vma->vm_flags & VM_MAYSHARE)
2334 region_add(&inode->i_mapping->private_list, from, to); 2334 region_add(&inode->i_mapping->private_list, from, to);
2335 return 0; 2335 return 0;
2336} 2336}