2 * mm/truncate.c - code for taking down pages from address_spaces
4 * Copyright (C) 2002, Linus Torvalds
6 * 10Sep2002 akpm@zip.com.au
10 #include <linux/kernel.h>
12 #include <linux/swap.h>
13 #include <linux/module.h>
14 #include <linux/pagemap.h>
15 #include <linux/pagevec.h>
16 #include <linux/buffer_head.h> /* grr. try_to_release_page,
21 * do_invalidatepage - invalidate part of all of a page
22 * @page: the page which is affected
23 * @offset: the index of the truncation point
25 * do_invalidatepage() is called when all or part of the page has become
26 * invalidated by a truncate operation.
28 * do_invalidatepage() does not have to release all buffers, but it must
29 * ensure that no dirty buffer is left outside @offset and that no I/O
30 * is underway against any of the blocks which are outside the truncation
31 * point. Because the caller is about to free (and possibly reuse) those
34 void do_invalidatepage(struct page
*page
, unsigned long offset
)
36 void (*invalidatepage
)(struct page
*, unsigned long);
37 #if 0 // mask by Victor Yu. 02-12-2007
38 invalidatepage
= page
->mapping
->a_ops
->invalidatepage
;
40 invalidatepage
= page
->u
.xx
.mapping
->a_ops
->invalidatepage
;
44 invalidatepage
= block_invalidatepage
;
47 (*invalidatepage
)(page
, offset
);
50 static inline void truncate_partial_page(struct page
*page
, unsigned partial
)
52 memclear_highpage_flush(page
, partial
, PAGE_CACHE_SIZE
-partial
);
53 if (PagePrivate(page
))
54 do_invalidatepage(page
, partial
);
58 * If truncate cannot remove the fs-private metadata from the page, the page
59 * becomes anonymous. It will be left on the LRU and may even be mapped into
60 * user pagetables if we're racing with filemap_nopage().
62 * We need to bale out if page->mapping is no longer equal to the original
63 * mapping. This happens a) when the VM reclaimed the page while we waited on
64 * its lock, b) when a concurrent invalidate_inode_pages got there first and
65 * c) when tmpfs swizzles a page between a tmpfs inode and swapper_space.
68 truncate_complete_page(struct address_space
*mapping
, struct page
*page
)
70 #if 0 // mask by Victor Yu. 02-12-2007
71 if (page
->mapping
!= mapping
)
73 if (page
->u
.xx
.mapping
!= mapping
)
77 if (PagePrivate(page
))
78 do_invalidatepage(page
, 0);
80 clear_page_dirty(page
);
81 ClearPageUptodate(page
);
82 ClearPageMappedToDisk(page
);
83 remove_from_page_cache(page
);
84 page_cache_release(page
); /* pagecache ref */
88 * This is for invalidate_inode_pages(). That function can be called at
89 * any time, and is not supposed to throw away dirty pages. But pages can
90 * be marked dirty at any time too, so use remove_mapping which safely
91 * discards clean, unused pages.
93 * Returns non-zero if the page was successfully invalidated.
96 invalidate_complete_page(struct address_space
*mapping
, struct page
*page
)
100 #if 0 // mask by Victor Yu. 02-12-2007
101 if (page
->mapping
!= mapping
)
103 if (page
->u
.xx
.mapping
!= mapping
)
107 if (PagePrivate(page
) && !try_to_release_page(page
, 0))
110 ret
= remove_mapping(mapping
, page
);
116 * truncate_inode_pages - truncate range of pages specified by start and
118 * @mapping: mapping to truncate
119 * @lstart: offset from which to truncate
120 * @lend: offset to which to truncate
122 * Truncate the page cache, removing the pages that are between
123 * specified offsets (and zeroing out partial page
124 * (if lstart is not page aligned)).
126 * Truncate takes two passes - the first pass is nonblocking. It will not
127 * block on page locks and it will not block on writeback. The second pass
128 * will wait. This is to prevent as much IO as possible in the affected region.
129 * The first pass will remove most pages, so the search cost of the second pass
132 * When looking at page->index outside the page lock we need to be careful to
133 * copy it into a local to avoid races (it could change at any time).
135 * We pass down the cache-hot hint to the page freeing code. Even if the
136 * mapping is large, it is probably the case that the final pages are the most
137 * recently touched, and freeing happens in ascending file offset order.
139 void truncate_inode_pages_range(struct address_space
*mapping
,
140 loff_t lstart
, loff_t lend
)
142 const pgoff_t start
= (lstart
+ PAGE_CACHE_SIZE
-1) >> PAGE_CACHE_SHIFT
;
144 const unsigned partial
= lstart
& (PAGE_CACHE_SIZE
- 1);
149 if (mapping
->nrpages
== 0)
152 BUG_ON((lend
& (PAGE_CACHE_SIZE
- 1)) != (PAGE_CACHE_SIZE
- 1));
153 end
= (lend
>> PAGE_CACHE_SHIFT
);
155 pagevec_init(&pvec
, 0);
157 while (next
<= end
&&
158 pagevec_lookup(&pvec
, mapping
, next
, PAGEVEC_SIZE
)) {
159 for (i
= 0; i
< pagevec_count(&pvec
); i
++) {
160 struct page
*page
= pvec
.pages
[i
];
161 pgoff_t page_index
= page
->index
;
163 if (page_index
> end
) {
168 if (page_index
> next
)
171 if (TestSetPageLocked(page
))
173 if (PageWriteback(page
)) {
177 truncate_complete_page(mapping
, page
);
180 pagevec_release(&pvec
);
185 struct page
*page
= find_lock_page(mapping
, start
- 1);
187 wait_on_page_writeback(page
);
188 truncate_partial_page(page
, partial
);
190 page_cache_release(page
);
197 if (!pagevec_lookup(&pvec
, mapping
, next
, PAGEVEC_SIZE
)) {
203 if (pvec
.pages
[0]->index
> end
) {
204 pagevec_release(&pvec
);
207 for (i
= 0; i
< pagevec_count(&pvec
); i
++) {
208 struct page
*page
= pvec
.pages
[i
];
210 if (page
->index
> end
)
213 wait_on_page_writeback(page
);
214 if (page
->index
> next
)
217 truncate_complete_page(mapping
, page
);
220 pagevec_release(&pvec
);
223 EXPORT_SYMBOL(truncate_inode_pages_range
);
226 * truncate_inode_pages - truncate *all* the pages from an offset
227 * @mapping: mapping to truncate
228 * @lstart: offset from which to truncate
230 * Called under (and serialised by) inode->i_mutex.
232 void truncate_inode_pages(struct address_space
*mapping
, loff_t lstart
)
234 truncate_inode_pages_range(mapping
, lstart
, (loff_t
)-1);
236 EXPORT_SYMBOL(truncate_inode_pages
);
239 * invalidate_mapping_pages - Invalidate all the unlocked pages of one inode
240 * @mapping: the address_space which holds the pages to invalidate
241 * @start: the offset 'from' which to invalidate
242 * @end: the offset 'to' which to invalidate (inclusive)
244 * This function only removes the unlocked pages, if you want to
245 * remove all the pages of one inode, you must call truncate_inode_pages.
247 * invalidate_mapping_pages() will not block on IO activity. It will not
248 * invalidate pages which are dirty, locked, under writeback or mapped into
251 unsigned long invalidate_mapping_pages(struct address_space
*mapping
,
252 pgoff_t start
, pgoff_t end
)
255 pgoff_t next
= start
;
256 unsigned long ret
= 0;
259 pagevec_init(&pvec
, 0);
260 while (next
<= end
&&
261 pagevec_lookup(&pvec
, mapping
, next
, PAGEVEC_SIZE
)) {
262 for (i
= 0; i
< pagevec_count(&pvec
); i
++) {
263 struct page
*page
= pvec
.pages
[i
];
267 lock_failed
= TestSetPageLocked(page
);
270 * We really shouldn't be looking at the ->index of an
271 * unlocked page. But we're not allowed to lock these
272 * pages. So we rely upon nobody altering the ->index
273 * of this (pinned-by-us) page.
282 if (PageDirty(page
) || PageWriteback(page
))
284 if (page_mapped(page
))
286 ret
+= invalidate_complete_page(mapping
, page
);
292 pagevec_release(&pvec
);
297 unsigned long invalidate_inode_pages(struct address_space
*mapping
)
299 return invalidate_mapping_pages(mapping
, 0, ~0UL);
301 EXPORT_SYMBOL(invalidate_inode_pages
);
304 * This is like invalidate_complete_page(), except it ignores the page's
305 * refcount. We do this because invalidate_inode_pages2() needs stronger
306 * invalidation guarantees, and cannot afford to leave pages behind because
307 * shrink_list() has a temp ref on them, or because they're transiently sitting
308 * in the lru_cache_add() pagevecs.
311 invalidate_complete_page2(struct address_space
*mapping
, struct page
*page
)
313 #if 0 // mask by Victor Yu. 02-12-2007
314 if (page
->mapping
!= mapping
)
316 if (page
->u
.xx
.mapping
!= mapping
)
320 if (PagePrivate(page
) && !try_to_release_page(page
, GFP_KERNEL
))
323 write_lock_irq(&mapping
->tree_lock
);
327 BUG_ON(PagePrivate(page
));
328 __remove_from_page_cache(page
);
329 write_unlock_irq(&mapping
->tree_lock
);
330 ClearPageUptodate(page
);
331 page_cache_release(page
); /* pagecache ref */
334 write_unlock_irq(&mapping
->tree_lock
);
339 * invalidate_inode_pages2_range - remove range of pages from an address_space
340 * @mapping: the address_space
341 * @start: the page offset 'from' which to invalidate
342 * @end: the page offset 'to' which to invalidate (inclusive)
344 * Any pages which are found to be mapped into pagetables are unmapped prior to
347 * Returns -EIO if any pages could not be invalidated.
349 int invalidate_inode_pages2_range(struct address_space
*mapping
,
350 pgoff_t start
, pgoff_t end
)
356 int did_range_unmap
= 0;
359 pagevec_init(&pvec
, 0);
361 while (next
<= end
&& !ret
&& !wrapped
&&
362 pagevec_lookup(&pvec
, mapping
, next
,
363 min(end
- next
, (pgoff_t
)PAGEVEC_SIZE
- 1) + 1)) {
364 for (i
= 0; !ret
&& i
< pagevec_count(&pvec
); i
++) {
365 struct page
*page
= pvec
.pages
[i
];
370 #if 0 // mask by Victor Yu. 02-12-2007
371 if (page
->mapping
!= mapping
) {
373 if (page
->u
.xx
.mapping
!= mapping
) {
378 page_index
= page
->index
;
379 next
= page_index
+ 1;
382 if (page_index
> end
) {
386 wait_on_page_writeback(page
);
387 while (page_mapped(page
)) {
388 if (!did_range_unmap
) {
390 * Zap the rest of the file in one hit.
392 unmap_mapping_range(mapping
,
393 (loff_t
)page_index
<<PAGE_CACHE_SHIFT
,
394 (loff_t
)(end
- page_index
+ 1)
402 unmap_mapping_range(mapping
,
403 (loff_t
)page_index
<<PAGE_CACHE_SHIFT
,
407 was_dirty
= test_clear_page_dirty(page
);
408 if (!invalidate_complete_page2(mapping
, page
)) {
410 set_page_dirty(page
);
415 pagevec_release(&pvec
);
421 EXPORT_SYMBOL_GPL(invalidate_inode_pages2_range
);
424 * invalidate_inode_pages2 - remove all pages from an address_space
425 * @mapping: the address_space
427 * Any pages which are found to be mapped into pagetables are unmapped prior to
430 * Returns -EIO if any pages could not be invalidated.
432 int invalidate_inode_pages2(struct address_space
*mapping
)
434 return invalidate_inode_pages2_range(mapping
, 0, -1);
436 EXPORT_SYMBOL_GPL(invalidate_inode_pages2
);