1
2
3
4
5
6
7
8
9#include <linux/mm.h>
10#include <linux/gfp.h>
11#include <linux/kernel_stat.h>
12#include <linux/swap.h>
13#include <linux/swapops.h>
14#include <linux/init.h>
15#include <linux/pagemap.h>
16#include <linux/backing-dev.h>
17#include <linux/pagevec.h>
18#include <linux/migrate.h>
19#include <linux/page_cgroup.h>
20
21#include <asm/pgtable.h>
22
23
24
25
26
27static const struct address_space_operations swap_aops = {
28 .writepage = swap_writepage,
29 .set_page_dirty = __set_page_dirty_no_writeback,
30 .migratepage = migrate_page,
31};
32
33static struct backing_dev_info swap_backing_dev_info = {
34 .name = "swap",
35 .capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
36};
37
38struct address_space swapper_space = {
39 .page_tree = RADIX_TREE_INIT(GFP_ATOMIC|__GFP_NOWARN),
40 .tree_lock = __SPIN_LOCK_UNLOCKED(swapper_space.tree_lock),
41 .a_ops = &swap_aops,
42 .i_mmap_nonlinear = LIST_HEAD_INIT(swapper_space.i_mmap_nonlinear),
43 .backing_dev_info = &swap_backing_dev_info,
44};
45
46#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
47
48static struct {
49 unsigned long add_total;
50 unsigned long del_total;
51 unsigned long find_success;
52 unsigned long find_total;
53} swap_cache_info;
54
55void show_swap_cache_info(void)
56{
57 printk("%lu pages in swap cache\n", total_swapcache_pages);
58 printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
59 swap_cache_info.add_total, swap_cache_info.del_total,
60 swap_cache_info.find_success, swap_cache_info.find_total);
61 printk("Free swap = %ldkB\n", nr_swap_pages << (PAGE_SHIFT - 10));
62 printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
63}
64
65
66
67
68
69static int __add_to_swap_cache(struct page *page, swp_entry_t entry)
70{
71 int error;
72
73 VM_BUG_ON(!PageLocked(page));
74 VM_BUG_ON(PageSwapCache(page));
75 VM_BUG_ON(!PageSwapBacked(page));
76
77 page_cache_get(page);
78 SetPageSwapCache(page);
79 set_page_private(page, entry.val);
80
81 spin_lock_irq(&swapper_space.tree_lock);
82 error = radix_tree_insert(&swapper_space.page_tree, entry.val, page);
83 if (likely(!error)) {
84 total_swapcache_pages++;
85 __inc_zone_page_state(page, NR_FILE_PAGES);
86 INC_CACHE_INFO(add_total);
87 }
88 spin_unlock_irq(&swapper_space.tree_lock);
89
90 if (unlikely(error)) {
91
92
93
94
95
96 VM_BUG_ON(error == -EEXIST);
97 set_page_private(page, 0UL);
98 ClearPageSwapCache(page);
99 page_cache_release(page);
100 }
101
102 return error;
103}
104
105
106int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
107{
108 int error;
109
110 error = radix_tree_preload(gfp_mask);
111 if (!error) {
112 error = __add_to_swap_cache(page, entry);
113 radix_tree_preload_end();
114 }
115 return error;
116}
117
118
119
120
121
122void __delete_from_swap_cache(struct page *page)
123{
124 VM_BUG_ON(!PageLocked(page));
125 VM_BUG_ON(!PageSwapCache(page));
126 VM_BUG_ON(PageWriteback(page));
127
128 radix_tree_delete(&swapper_space.page_tree, page_private(page));
129 set_page_private(page, 0);
130 ClearPageSwapCache(page);
131 total_swapcache_pages--;
132 __dec_zone_page_state(page, NR_FILE_PAGES);
133 INC_CACHE_INFO(del_total);
134}
135
136
137
138
139
140
141
142
143int add_to_swap(struct page *page)
144{
145 swp_entry_t entry;
146 int err;
147
148 VM_BUG_ON(!PageLocked(page));
149 VM_BUG_ON(!PageUptodate(page));
150
151 entry = get_swap_page();
152 if (!entry.val)
153 return 0;
154
155 if (unlikely(PageTransHuge(page)))
156 if (unlikely(split_huge_page(page))) {
157 swapcache_free(entry, NULL);
158 return 0;
159 }
160
161
162
163
164
165
166
167
168
169
170
171
172 err = add_to_swap_cache(page, entry,
173 __GFP_HIGH|__GFP_NOMEMALLOC|__GFP_NOWARN);
174
175 if (!err) {
176 SetPageDirty(page);
177 return 1;
178 } else {
179
180
181
182
183 swapcache_free(entry, NULL);
184 return 0;
185 }
186}
187
188
189
190
191
192
193
194void delete_from_swap_cache(struct page *page)
195{
196 swp_entry_t entry;
197
198 entry.val = page_private(page);
199
200 spin_lock_irq(&swapper_space.tree_lock);
201 __delete_from_swap_cache(page);
202 spin_unlock_irq(&swapper_space.tree_lock);
203
204 swapcache_free(entry, page);
205 page_cache_release(page);
206}
207
208
209
210
211
212
213
214
215
216static inline void free_swap_cache(struct page *page)
217{
218 if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
219 try_to_free_swap(page);
220 unlock_page(page);
221 }
222}
223
224
225
226
227
228void free_page_and_swap_cache(struct page *page)
229{
230 free_swap_cache(page);
231 page_cache_release(page);
232}
233
234
235
236
237
238void free_pages_and_swap_cache(struct page **pages, int nr)
239{
240 struct page **pagep = pages;
241
242 lru_add_drain();
243 while (nr) {
244 int todo = min(nr, PAGEVEC_SIZE);
245 int i;
246
247 for (i = 0; i < todo; i++)
248 free_swap_cache(pagep[i]);
249 release_pages(pagep, todo, 0);
250 pagep += todo;
251 nr -= todo;
252 }
253}
254
255
256
257
258
259
260
261struct page * lookup_swap_cache(swp_entry_t entry)
262{
263 struct page *page;
264
265 page = find_get_page(&swapper_space, entry.val);
266
267 if (page)
268 INC_CACHE_INFO(find_success);
269
270 INC_CACHE_INFO(find_total);
271 return page;
272}
273
274
275
276
277
278
279
280struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
281 struct vm_area_struct *vma, unsigned long addr)
282{
283 struct page *found_page, *new_page = NULL;
284 int err;
285
286 do {
287
288
289
290
291
292 found_page = find_get_page(&swapper_space, entry.val);
293 if (found_page)
294 break;
295
296
297
298
299 if (!new_page) {
300 new_page = alloc_page_vma(gfp_mask, vma, addr);
301 if (!new_page)
302 break;
303 }
304
305
306
307
308 err = radix_tree_preload(gfp_mask & GFP_KERNEL);
309 if (err)
310 break;
311
312
313
314
315 err = swapcache_prepare(entry);
316 if (err == -EEXIST) {
317 radix_tree_preload_end();
318 continue;
319 }
320 if (err) {
321 radix_tree_preload_end();
322 break;
323 }
324
325
326 __set_page_locked(new_page);
327 SetPageSwapBacked(new_page);
328 err = __add_to_swap_cache(new_page, entry);
329 if (likely(!err)) {
330 radix_tree_preload_end();
331
332
333
334 lru_cache_add_anon(new_page);
335 swap_readpage(new_page);
336 return new_page;
337 }
338 radix_tree_preload_end();
339 ClearPageSwapBacked(new_page);
340 __clear_page_locked(new_page);
341
342
343
344
345 swapcache_free(entry, NULL);
346 } while (err != -ENOMEM);
347
348 if (new_page)
349 page_cache_release(new_page);
350 return found_page;
351}
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
373 struct vm_area_struct *vma, unsigned long addr)
374{
375 struct page *page;
376 unsigned long offset = swp_offset(entry);
377 unsigned long start_offset, end_offset;
378 unsigned long mask = (1UL << page_cluster) - 1;
379
380
381 start_offset = offset & ~mask;
382 end_offset = offset | mask;
383 if (!start_offset)
384 start_offset++;
385
386 for (offset = start_offset; offset <= end_offset ; offset++) {
387
388 page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
389 gfp_mask, vma, addr);
390 if (!page)
391 continue;
392 page_cache_release(page);
393 }
394 lru_add_drain();
395 return read_swap_cache_async(entry, gfp_mask, vma, addr);
396}
397