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/module.h>
13 #include <linux/pagemap.h>
14 #include <linux/pagevec.h>
15 #include <linux/buffer_head.h> /* grr. try_to_release_page,
16 block_invalidatepage */
19 static int do_invalidatepage(struct page *page, unsigned long offset)
21 int (*invalidatepage)(struct page *, unsigned long);
22 invalidatepage = page->mapping->a_ops->invalidatepage;
23 if (invalidatepage == NULL)
24 invalidatepage = block_invalidatepage;
25 return (*invalidatepage)(page, offset);
28 static inline void truncate_partial_page(struct page *page, unsigned partial)
30 memclear_highpage_flush(page, partial, PAGE_CACHE_SIZE-partial);
31 if (PagePrivate(page))
32 do_invalidatepage(page, partial);
36 * If truncate cannot remove the fs-private metadata from the page, the page
37 * becomes anonymous. It will be left on the LRU and may even be mapped into
38 * user pagetables if we're racing with filemap_nopage().
40 * We need to bale out if page->mapping is no longer equal to the original
41 * mapping. This happens a) when the VM reclaimed the page while we waited on
42 * its lock, b) when a concurrent invalidate_inode_pages got there first and
43 * c) when tmpfs swizzles a page between a tmpfs inode and swapper_space.
46 truncate_complete_page(struct address_space *mapping, struct page *page)
48 if (page->mapping != mapping)
51 if (PagePrivate(page))
52 do_invalidatepage(page, 0);
54 clear_page_dirty(page);
55 ClearPageUptodate(page);
56 ClearPageMappedToDisk(page);
57 remove_from_page_cache(page);
58 page_cache_release(page); /* pagecache ref */
62 * This is for invalidate_inode_pages(). That function can be called at
63 * any time, and is not supposed to throw away dirty pages. But pages can
64 * be marked dirty at any time too. So we re-check the dirtiness inside
65 * ->tree_lock. That provides exclusion against the __set_page_dirty
69 invalidate_complete_page(struct address_space *mapping, struct page *page)
71 if (page->mapping != mapping)
74 if (PagePrivate(page) && !try_to_release_page(page, 0))
77 spin_lock_irq(&mapping->tree_lock);
78 if (PageDirty(page)) {
79 spin_unlock_irq(&mapping->tree_lock);
83 BUG_ON(PagePrivate(page));
84 if (page_count(page) != 2) {
85 spin_unlock_irq(&mapping->tree_lock);
88 __remove_from_page_cache(page);
89 spin_unlock_irq(&mapping->tree_lock);
90 ClearPageUptodate(page);
91 page_cache_release(page); /* pagecache ref */
96 * truncate_inode_pages - truncate *all* the pages from an offset
97 * @mapping: mapping to truncate
98 * @lstart: offset from which to truncate
100 * Truncate the page cache at a set offset, removing the pages that are beyond
101 * that offset (and zeroing out partial pages).
103 * Truncate takes two passes - the first pass is nonblocking. It will not
104 * block on page locks and it will not block on writeback. The second pass
105 * will wait. This is to prevent as much IO as possible in the affected region.
106 * The first pass will remove most pages, so the search cost of the second pass
109 * When looking at page->index outside the page lock we need to be careful to
110 * copy it into a local to avoid races (it could change at any time).
112 * We pass down the cache-hot hint to the page freeing code. Even if the
113 * mapping is large, it is probably the case that the final pages are the most
114 * recently touched, and freeing happens in ascending file offset order.
116 * Called under (and serialised by) inode->i_sem.
118 void truncate_inode_pages(struct address_space *mapping, loff_t lstart)
120 const pgoff_t start = (lstart + PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT;
121 const unsigned partial = lstart & (PAGE_CACHE_SIZE - 1);
126 if (mapping->nrpages == 0)
129 pagevec_init(&pvec, 0);
131 while (pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
132 for (i = 0; i < pagevec_count(&pvec); i++) {
133 struct page *page = pvec.pages[i];
134 pgoff_t page_index = page->index;
136 if (page_index > next)
139 if (TestSetPageLocked(page))
141 if (PageWriteback(page)) {
145 truncate_complete_page(mapping, page);
148 pagevec_release(&pvec);
153 struct page *page = find_lock_page(mapping, start - 1);
155 wait_on_page_writeback(page);
156 truncate_partial_page(page, partial);
158 page_cache_release(page);
165 if (!pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
171 for (i = 0; i < pagevec_count(&pvec); i++) {
172 struct page *page = pvec.pages[i];
175 wait_on_page_writeback(page);
176 if (page->index > next)
179 truncate_complete_page(mapping, page);
182 pagevec_release(&pvec);
186 EXPORT_SYMBOL(truncate_inode_pages);
189 * invalidate_mapping_pages - Invalidate all the unlocked pages of one inode
190 * @mapping: the address_space which holds the pages to invalidate
191 * @start: the offset 'from' which to invalidate
192 * @end: the offset 'to' which to invalidate (inclusive)
194 * This function only removes the unlocked pages, if you want to
195 * remove all the pages of one inode, you must call truncate_inode_pages.
197 * invalidate_mapping_pages() will not block on IO activity. It will not
198 * invalidate pages which are dirty, locked, under writeback or mapped into
201 unsigned long invalidate_mapping_pages(struct address_space *mapping,
202 pgoff_t start, pgoff_t end)
205 pgoff_t next = start;
206 unsigned long ret = 0;
209 pagevec_init(&pvec, 0);
210 while (next <= end &&
211 pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
212 for (i = 0; i < pagevec_count(&pvec); i++) {
213 struct page *page = pvec.pages[i];
215 if (TestSetPageLocked(page)) {
219 if (page->index > next)
222 if (PageDirty(page) || PageWriteback(page))
224 if (page_mapped(page))
226 ret += invalidate_complete_page(mapping, page);
232 pagevec_release(&pvec);
238 EXPORT_SYMBOL_GPL(invalidate_mapping_pages);
240 unsigned long invalidate_inode_pages(struct address_space *mapping)
242 return invalidate_mapping_pages(mapping, 0, ~0UL);
245 EXPORT_SYMBOL(invalidate_inode_pages);
248 * invalidate_inode_pages2 - remove all unmapped pages from an address_space
249 * @mapping - the address_space
251 * invalidate_inode_pages2() is like truncate_inode_pages(), except for the case
252 * where the page is seen to be mapped into process pagetables. In that case,
253 * the page is marked clean but is left attached to its address_space.
255 * The page is also marked not uptodate so that a subsequent pagefault will
256 * perform I/O to bringthe page's contents back into sync with its backing
259 * FIXME: invalidate_inode_pages2() is probably trivially livelockable.
261 void invalidate_inode_pages2(struct address_space *mapping)
267 pagevec_init(&pvec, 0);
268 while (pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
269 for (i = 0; i < pagevec_count(&pvec); i++) {
270 struct page *page = pvec.pages[i];
273 if (page->mapping == mapping) { /* truncate race? */
274 wait_on_page_writeback(page);
275 next = page->index + 1;
276 if (page_mapped(page)) {
277 clear_page_dirty(page);
278 ClearPageUptodate(page);
280 if (!invalidate_complete_page(mapping,
282 clear_page_dirty(page);
283 ClearPageUptodate(page);
289 pagevec_release(&pvec);
294 EXPORT_SYMBOL_GPL(invalidate_inode_pages2);