blob: 7efcf1525921037d434060f73c0ff77b9981e02c [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/mm/swap_state.c
3 *
4 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
5 * Swap reorganised 29.12.95, Stephen Tweedie
6 *
7 * Rewritten to use page cache, (C) 1998 Stephen Tweedie
8 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/mm.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090010#include <linux/gfp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/kernel_stat.h>
12#include <linux/swap.h>
Hugh Dickins46017e92008-02-04 22:28:41 -080013#include <linux/swapops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/init.h>
15#include <linux/pagemap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/backing-dev.h>
Christian Ehrhardt3fb5c292012-07-31 16:41:44 -070017#include <linux/blkdev.h>
Hugh Dickinsc484d412006-01-06 00:10:55 -080018#include <linux/pagevec.h>
Christoph Lameterb20a3502006-03-22 00:09:12 -080019#include <linux/migrate.h>
KAMEZAWA Hiroyuki8c7c6e342009-01-07 18:08:00 -080020#include <linux/page_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
22#include <asm/pgtable.h>
23
24/*
25 * swapper_space is a fiction, retained to simplify the path through
Jens Axboe7eaceac2011-03-10 08:52:07 +010026 * vmscan's shrink_page_list.
Linus Torvalds1da177e2005-04-16 15:20:36 -070027 */
Christoph Hellwigf5e54d62006-06-28 04:26:44 -070028static const struct address_space_operations swap_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070029 .writepage = swap_writepage,
Mel Gorman62c230b2012-07-31 16:44:55 -070030 .set_page_dirty = swap_set_page_dirty,
Christoph Lametere965f962006-02-01 03:05:41 -080031 .migratepage = migrate_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -070032};
33
34static struct backing_dev_info swap_backing_dev_info = {
Jens Axboed9938312009-06-12 14:45:52 +020035 .name = "swap",
Rik van Riel4f98a2f2008-10-18 20:26:32 -070036 .capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
Linus Torvalds1da177e2005-04-16 15:20:36 -070037};
38
Shaohua Li33806f02013-02-22 16:34:37 -080039struct address_space swapper_spaces[MAX_SWAPFILES] = {
40 [0 ... MAX_SWAPFILES - 1] = {
41 .page_tree = RADIX_TREE_INIT(GFP_ATOMIC|__GFP_NOWARN),
42 .a_ops = &swap_aops,
43 .backing_dev_info = &swap_backing_dev_info,
44 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070045};
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
47#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
48
49static struct {
50 unsigned long add_total;
51 unsigned long del_total;
52 unsigned long find_success;
53 unsigned long find_total;
Linus Torvalds1da177e2005-04-16 15:20:36 -070054} swap_cache_info;
55
Shaohua Li33806f02013-02-22 16:34:37 -080056unsigned long total_swapcache_pages(void)
57{
58 int i;
59 unsigned long ret = 0;
60
61 for (i = 0; i < MAX_SWAPFILES; i++)
62 ret += swapper_spaces[i].nrpages;
63 return ret;
64}
65
Linus Torvalds1da177e2005-04-16 15:20:36 -070066void show_swap_cache_info(void)
67{
Shaohua Li33806f02013-02-22 16:34:37 -080068 printk("%lu pages in swap cache\n", total_swapcache_pages());
Johannes Weiner2c97b7f2008-07-25 19:46:01 -070069 printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 swap_cache_info.add_total, swap_cache_info.del_total,
Hugh Dickinsbb63be02008-02-04 22:28:49 -080071 swap_cache_info.find_success, swap_cache_info.find_total);
Shaohua Liec8acf22013-02-22 16:34:38 -080072 printk("Free swap = %ldkB\n",
73 get_nr_swap_pages() << (PAGE_SHIFT - 10));
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
75}
76
77/*
Daisuke Nishimura31a56392009-09-21 17:02:50 -070078 * __add_to_swap_cache resembles add_to_page_cache_locked on swapper_space,
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 * but sets SwapCache flag and private instead of mapping and index.
80 */
Daisuke Nishimura31a56392009-09-21 17:02:50 -070081static int __add_to_swap_cache(struct page *page, swp_entry_t entry)
Linus Torvalds1da177e2005-04-16 15:20:36 -070082{
83 int error;
Shaohua Li33806f02013-02-22 16:34:37 -080084 struct address_space *address_space;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
Hugh Dickins51726b12009-01-06 14:39:25 -080086 VM_BUG_ON(!PageLocked(page));
87 VM_BUG_ON(PageSwapCache(page));
88 VM_BUG_ON(!PageSwapBacked(page));
89
Daisuke Nishimura31a56392009-09-21 17:02:50 -070090 page_cache_get(page);
91 SetPageSwapCache(page);
92 set_page_private(page, entry.val);
93
Shaohua Li33806f02013-02-22 16:34:37 -080094 address_space = swap_address_space(entry);
95 spin_lock_irq(&address_space->tree_lock);
96 error = radix_tree_insert(&address_space->page_tree,
97 entry.val, page);
Daisuke Nishimura31a56392009-09-21 17:02:50 -070098 if (likely(!error)) {
Shaohua Li33806f02013-02-22 16:34:37 -080099 address_space->nrpages++;
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700100 __inc_zone_page_state(page, NR_FILE_PAGES);
101 INC_CACHE_INFO(add_total);
102 }
Shaohua Li33806f02013-02-22 16:34:37 -0800103 spin_unlock_irq(&address_space->tree_lock);
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700104
105 if (unlikely(error)) {
Daisuke Nishimura2ca45322009-09-21 17:02:52 -0700106 /*
107 * Only the context which have set SWAP_HAS_CACHE flag
108 * would call add_to_swap_cache().
109 * So add_to_swap_cache() doesn't returns -EEXIST.
110 */
111 VM_BUG_ON(error == -EEXIST);
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700112 set_page_private(page, 0UL);
113 ClearPageSwapCache(page);
114 page_cache_release(page);
115 }
116
117 return error;
118}
119
120
121int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
122{
123 int error;
124
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 error = radix_tree_preload(gfp_mask);
126 if (!error) {
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700127 error = __add_to_swap_cache(page, entry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 radix_tree_preload_end();
Hugh Dickinsfa1de902008-02-07 00:14:13 -0800129 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 return error;
131}
132
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133/*
134 * This must be called only on pages that have
135 * been verified to be in the swap cache.
136 */
137void __delete_from_swap_cache(struct page *page)
138{
Shaohua Li33806f02013-02-22 16:34:37 -0800139 swp_entry_t entry;
140 struct address_space *address_space;
141
Hugh Dickins51726b12009-01-06 14:39:25 -0800142 VM_BUG_ON(!PageLocked(page));
143 VM_BUG_ON(!PageSwapCache(page));
144 VM_BUG_ON(PageWriteback(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145
Shaohua Li33806f02013-02-22 16:34:37 -0800146 entry.val = page_private(page);
147 address_space = swap_address_space(entry);
148 radix_tree_delete(&address_space->page_tree, page_private(page));
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700149 set_page_private(page, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 ClearPageSwapCache(page);
Shaohua Li33806f02013-02-22 16:34:37 -0800151 address_space->nrpages--;
Christoph Lameter347ce432006-06-30 01:55:35 -0700152 __dec_zone_page_state(page, NR_FILE_PAGES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 INC_CACHE_INFO(del_total);
154}
155
156/**
157 * add_to_swap - allocate swap space for a page
158 * @page: page we want to move to swap
159 *
160 * Allocate swap space for the page and add the page to the
161 * swap cache. Caller needs to hold the page lock.
162 */
Hugh Dickinsac47b002009-01-06 14:39:39 -0800163int add_to_swap(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164{
165 swp_entry_t entry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 int err;
167
Hugh Dickins51726b12009-01-06 14:39:25 -0800168 VM_BUG_ON(!PageLocked(page));
169 VM_BUG_ON(!PageUptodate(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
Daisuke Nishimura2ca45322009-09-21 17:02:52 -0700171 entry = get_swap_page();
172 if (!entry.val)
173 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174
Andrea Arcangeli3f04f622011-01-13 15:46:47 -0800175 if (unlikely(PageTransHuge(page)))
176 if (unlikely(split_huge_page(page))) {
177 swapcache_free(entry, NULL);
178 return 0;
179 }
180
Daisuke Nishimura2ca45322009-09-21 17:02:52 -0700181 /*
182 * Radix-tree node allocations from PF_MEMALLOC contexts could
183 * completely exhaust the page allocator. __GFP_NOMEMALLOC
184 * stops emergency reserves from being allocated.
185 *
186 * TODO: this could cause a theoretical memory reclaim
187 * deadlock in the swap out path.
188 */
189 /*
190 * Add it to the swap cache and mark it dirty
191 */
192 err = add_to_swap_cache(page, entry,
193 __GFP_HIGH|__GFP_NOMEMALLOC|__GFP_NOWARN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
Daisuke Nishimura2ca45322009-09-21 17:02:52 -0700195 if (!err) { /* Success */
196 SetPageDirty(page);
197 return 1;
198 } else { /* -ENOMEM radix-tree allocation failure */
199 /*
200 * add_to_swap_cache() doesn't return -EEXIST, so we can safely
201 * clear SWAP_HAS_CACHE flag.
202 */
203 swapcache_free(entry, NULL);
204 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 }
206}
207
208/*
209 * This must be called only on pages that have
210 * been verified to be in the swap cache and locked.
211 * It will never put the page into the free list,
212 * the caller has a reference on the page.
213 */
214void delete_from_swap_cache(struct page *page)
215{
216 swp_entry_t entry;
Shaohua Li33806f02013-02-22 16:34:37 -0800217 struct address_space *address_space;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700219 entry.val = page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
Shaohua Li33806f02013-02-22 16:34:37 -0800221 address_space = swap_address_space(entry);
222 spin_lock_irq(&address_space->tree_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 __delete_from_swap_cache(page);
Shaohua Li33806f02013-02-22 16:34:37 -0800224 spin_unlock_irq(&address_space->tree_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225
KAMEZAWA Hiroyukicb4b86b2009-06-16 15:32:52 -0700226 swapcache_free(entry, page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 page_cache_release(page);
228}
229
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230/*
231 * If we are the only user, then try to free up the swap cache.
232 *
233 * Its ok to check for PageSwapCache without the page lock
Hugh Dickinsa2c43ee2009-01-06 14:39:36 -0800234 * here because we are going to recheck again inside
235 * try_to_free_swap() _with_ the lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 * - Marcelo
237 */
238static inline void free_swap_cache(struct page *page)
239{
Hugh Dickinsa2c43ee2009-01-06 14:39:36 -0800240 if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
241 try_to_free_swap(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 unlock_page(page);
243 }
244}
245
246/*
247 * Perform a free_page(), also freeing any swap cache associated with
Hugh Dickinsb8072f02005-10-29 18:16:41 -0700248 * this page if it is the last user of the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 */
250void free_page_and_swap_cache(struct page *page)
251{
252 free_swap_cache(page);
253 page_cache_release(page);
254}
255
256/*
257 * Passed an array of pages, drop them all from swapcache and then release
258 * them. They are removed from the LRU and freed if this is their last use.
259 */
260void free_pages_and_swap_cache(struct page **pages, int nr)
261{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 struct page **pagep = pages;
263
264 lru_add_drain();
265 while (nr) {
Hugh Dickinsc484d412006-01-06 00:10:55 -0800266 int todo = min(nr, PAGEVEC_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 int i;
268
269 for (i = 0; i < todo; i++)
270 free_swap_cache(pagep[i]);
271 release_pages(pagep, todo, 0);
272 pagep += todo;
273 nr -= todo;
274 }
275}
276
277/*
278 * Lookup a swap entry in the swap cache. A found page will be returned
279 * unlocked and with its refcount incremented - we rely on the kernel
280 * lock getting page table operations atomic even if we drop the page
281 * lock before returning.
282 */
283struct page * lookup_swap_cache(swp_entry_t entry)
284{
285 struct page *page;
286
Shaohua Li33806f02013-02-22 16:34:37 -0800287 page = find_get_page(swap_address_space(entry), entry.val);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
289 if (page)
290 INC_CACHE_INFO(find_success);
291
292 INC_CACHE_INFO(find_total);
293 return page;
294}
295
296/*
297 * Locate a page of swap in physical memory, reserving swap cache space
298 * and reading the disk if it is not already cached.
299 * A failure return means that either the page allocation failed or that
300 * the swap entry is no longer in use.
301 */
Hugh Dickins02098fe2008-02-04 22:28:42 -0800302struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 struct vm_area_struct *vma, unsigned long addr)
304{
305 struct page *found_page, *new_page = NULL;
306 int err;
307
308 do {
309 /*
310 * First check the swap cache. Since this is normally
311 * called after lookup_swap_cache() failed, re-calling
312 * that would confuse statistics.
313 */
Shaohua Li33806f02013-02-22 16:34:37 -0800314 found_page = find_get_page(swap_address_space(entry),
315 entry.val);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 if (found_page)
317 break;
318
319 /*
320 * Get a new page to read into from swap.
321 */
322 if (!new_page) {
Hugh Dickins02098fe2008-02-04 22:28:42 -0800323 new_page = alloc_page_vma(gfp_mask, vma, addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 if (!new_page)
325 break; /* Out of memory */
326 }
327
328 /*
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700329 * call radix_tree_preload() while we can wait.
330 */
331 err = radix_tree_preload(gfp_mask & GFP_KERNEL);
332 if (err)
333 break;
334
335 /*
Hugh Dickinsf0009442008-02-04 22:28:49 -0800336 * Swap entry may have been freed since our caller observed it.
337 */
KAMEZAWA Hiroyuki355cfa72009-06-16 15:32:53 -0700338 err = swapcache_prepare(entry);
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700339 if (err == -EEXIST) { /* seems racy */
340 radix_tree_preload_end();
KAMEZAWA Hiroyuki355cfa72009-06-16 15:32:53 -0700341 continue;
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700342 }
343 if (err) { /* swp entry is obsolete ? */
344 radix_tree_preload_end();
Hugh Dickinsf0009442008-02-04 22:28:49 -0800345 break;
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700346 }
Hugh Dickinsf0009442008-02-04 22:28:49 -0800347
Daisuke Nishimura2ca45322009-09-21 17:02:52 -0700348 /* May fail (-ENOMEM) if radix-tree node allocation failed. */
Nick Pigginf45840b2008-10-18 20:26:57 -0700349 __set_page_locked(new_page);
Rik van Rielb2e18532008-10-18 20:26:30 -0700350 SetPageSwapBacked(new_page);
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700351 err = __add_to_swap_cache(new_page, entry);
Nick Piggin529ae9a2008-08-02 12:01:03 +0200352 if (likely(!err)) {
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700353 radix_tree_preload_end();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 /*
355 * Initiate read into locked page and return.
356 */
Rik van Rielc5fdae42008-10-18 20:26:36 -0700357 lru_cache_add_anon(new_page);
Minchan Kimaca8bf32009-06-16 15:33:02 -0700358 swap_readpage(new_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 return new_page;
360 }
Daisuke Nishimura31a56392009-09-21 17:02:50 -0700361 radix_tree_preload_end();
Rik van Rielb2e18532008-10-18 20:26:30 -0700362 ClearPageSwapBacked(new_page);
Nick Pigginf45840b2008-10-18 20:26:57 -0700363 __clear_page_locked(new_page);
Daisuke Nishimura2ca45322009-09-21 17:02:52 -0700364 /*
365 * add_to_swap_cache() doesn't return -EEXIST, so we can safely
366 * clear SWAP_HAS_CACHE flag.
367 */
KAMEZAWA Hiroyukicb4b86b2009-06-16 15:32:52 -0700368 swapcache_free(entry, NULL);
Hugh Dickinsf0009442008-02-04 22:28:49 -0800369 } while (err != -ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
371 if (new_page)
372 page_cache_release(new_page);
373 return found_page;
374}
Hugh Dickins46017e92008-02-04 22:28:41 -0800375
376/**
377 * swapin_readahead - swap in pages in hope we need them soon
378 * @entry: swap entry of this memory
Randy Dunlap76824862008-03-19 17:00:40 -0700379 * @gfp_mask: memory allocation flags
Hugh Dickins46017e92008-02-04 22:28:41 -0800380 * @vma: user vma this address belongs to
381 * @addr: target address for mempolicy
382 *
383 * Returns the struct page for entry and addr, after queueing swapin.
384 *
385 * Primitive swap readahead code. We simply read an aligned block of
386 * (1 << page_cluster) entries in the swap area. This method is chosen
387 * because it doesn't cost us any seek time. We also make sure to queue
388 * the 'original' request together with the readahead ones...
389 *
390 * This has been extended to use the NUMA policies from the mm triggering
391 * the readahead.
392 *
393 * Caller must hold down_read on the vma->vm_mm if vma is not NULL.
394 */
Hugh Dickins02098fe2008-02-04 22:28:42 -0800395struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
Hugh Dickins46017e92008-02-04 22:28:41 -0800396 struct vm_area_struct *vma, unsigned long addr)
397{
Hugh Dickins46017e92008-02-04 22:28:41 -0800398 struct page *page;
Rik van Riel67f96aa2012-03-21 16:33:50 -0700399 unsigned long offset = swp_offset(entry);
400 unsigned long start_offset, end_offset;
401 unsigned long mask = (1UL << page_cluster) - 1;
Christian Ehrhardt3fb5c292012-07-31 16:41:44 -0700402 struct blk_plug plug;
Hugh Dickins46017e92008-02-04 22:28:41 -0800403
Rik van Riel67f96aa2012-03-21 16:33:50 -0700404 /* Read a page_cluster sized and aligned cluster around offset. */
405 start_offset = offset & ~mask;
406 end_offset = offset | mask;
407 if (!start_offset) /* First page is swap header. */
408 start_offset++;
409
Christian Ehrhardt3fb5c292012-07-31 16:41:44 -0700410 blk_start_plug(&plug);
Rik van Riel67f96aa2012-03-21 16:33:50 -0700411 for (offset = start_offset; offset <= end_offset ; offset++) {
Hugh Dickins46017e92008-02-04 22:28:41 -0800412 /* Ok, do the async read-ahead now */
413 page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
Hugh Dickins02098fe2008-02-04 22:28:42 -0800414 gfp_mask, vma, addr);
Hugh Dickins46017e92008-02-04 22:28:41 -0800415 if (!page)
Rik van Riel67f96aa2012-03-21 16:33:50 -0700416 continue;
Hugh Dickins46017e92008-02-04 22:28:41 -0800417 page_cache_release(page);
418 }
Christian Ehrhardt3fb5c292012-07-31 16:41:44 -0700419 blk_finish_plug(&plug);
420
Hugh Dickins46017e92008-02-04 22:28:41 -0800421 lru_add_drain(); /* Push any new pages onto the LRU now */
Hugh Dickins02098fe2008-02-04 22:28:42 -0800422 return read_swap_cache_async(entry, gfp_mask, vma, addr);
Hugh Dickins46017e92008-02-04 22:28:41 -0800423}