4 * This code provides the generic "frontend" layer to call a matching
5 * "backend" driver implementation of frontswap. See
6 * Documentation/vm/frontswap.txt for more information.
8 * Copyright (C) 2009-2012 Oracle Corp. All rights reserved.
9 * Author: Dan Magenheimer
11 * This work is licensed under the terms of the GNU GPL, version 2.
15 #include <linux/mman.h>
16 #include <linux/swap.h>
17 #include <linux/swapops.h>
18 #include <linux/proc_fs.h>
19 #include <linux/security.h>
20 #include <linux/capability.h>
21 #include <linux/module.h>
22 #include <linux/uaccess.h>
23 #include <linux/debugfs.h>
24 #include <linux/frontswap.h>
25 #include <linux/swapfile.h>
28 * frontswap_ops is set by frontswap_register_ops to contain the pointers
29 * to the frontswap "backend" implementation functions.
31 static struct frontswap_ops frontswap_ops __read_mostly;
34 * This global enablement flag reduces overhead on systems where frontswap_ops
35 * has not been registered, so is preferred to the slower alternative: a
36 * function call that checks a non-global.
38 bool frontswap_enabled __read_mostly;
39 EXPORT_SYMBOL(frontswap_enabled);
42 * If enabled, frontswap_store will return failure even on success. As
43 * a result, the swap subsystem will always write the page to swap, in
44 * effect converting frontswap into a writethrough cache. In this mode,
45 * there is no direct reduction in swap writes, but a frontswap backend
46 * can unilaterally "reclaim" any pages in use with no data loss, thus
47 * providing increases control over maximum memory usage due to frontswap.
49 static bool frontswap_writethrough_enabled __read_mostly;
51 #ifdef CONFIG_DEBUG_FS
53 * Counters available via /sys/kernel/debug/frontswap (if debugfs is
54 * properly configured). These are for information only so are not protected
55 * against increment races.
57 static u64 frontswap_loads;
58 static u64 frontswap_succ_stores;
59 static u64 frontswap_failed_stores;
60 static u64 frontswap_invalidates;
62 static inline void inc_frontswap_loads(void) {
65 static inline void inc_frontswap_succ_stores(void) {
66 frontswap_succ_stores++;
68 static inline void inc_frontswap_failed_stores(void) {
69 frontswap_failed_stores++;
71 static inline void inc_frontswap_invalidates(void) {
72 frontswap_invalidates++;
75 static inline void inc_frontswap_loads(void) { }
76 static inline void inc_frontswap_succ_stores(void) { }
77 static inline void inc_frontswap_failed_stores(void) { }
78 static inline void inc_frontswap_invalidates(void) { }
81 * Register operations for frontswap, returning previous thus allowing
82 * detection of multiple backends and possible nesting.
84 struct frontswap_ops frontswap_register_ops(struct frontswap_ops *ops)
86 struct frontswap_ops old = frontswap_ops;
89 frontswap_enabled = true;
92 EXPORT_SYMBOL(frontswap_register_ops);
95 * Enable/disable frontswap writethrough (see above).
97 void frontswap_writethrough(bool enable)
99 frontswap_writethrough_enabled = enable;
101 EXPORT_SYMBOL(frontswap_writethrough);
104 * Called when a swap device is swapon'd.
106 void __frontswap_init(unsigned type)
108 struct swap_info_struct *sis = swap_info[type];
111 if (sis->frontswap_map == NULL)
113 frontswap_ops.init(type);
115 EXPORT_SYMBOL(__frontswap_init);
117 static inline void __frontswap_clear(struct swap_info_struct *sis, pgoff_t offset)
119 frontswap_clear(sis, offset);
120 atomic_dec(&sis->frontswap_pages);
124 * "Store" data from a page to frontswap and associate it with the page's
125 * swaptype and offset. Page must be locked and in the swap cache.
126 * If frontswap already contains a page with matching swaptype and
127 * offset, the frontswap implmentation may either overwrite the data and
128 * return success or invalidate the page from frontswap and return failure.
130 int __frontswap_store(struct page *page)
132 int ret = -1, dup = 0;
133 swp_entry_t entry = { .val = page_private(page), };
134 int type = swp_type(entry);
135 struct swap_info_struct *sis = swap_info[type];
136 pgoff_t offset = swp_offset(entry);
138 BUG_ON(!PageLocked(page));
140 if (frontswap_test(sis, offset))
142 ret = frontswap_ops.store(type, offset, page);
144 frontswap_set(sis, offset);
145 inc_frontswap_succ_stores();
147 atomic_inc(&sis->frontswap_pages);
150 failed dup always results in automatic invalidate of
151 the (older) page from frontswap
153 inc_frontswap_failed_stores();
155 __frontswap_clear(sis, offset);
157 if (frontswap_writethrough_enabled)
158 /* report failure so swap also writes to swap device */
162 EXPORT_SYMBOL(__frontswap_store);
165 * "Get" data from frontswap associated with swaptype and offset that were
166 * specified when the data was put to frontswap and use it to fill the
167 * specified page with data. Page must be locked and in the swap cache.
169 int __frontswap_load(struct page *page)
172 swp_entry_t entry = { .val = page_private(page), };
173 int type = swp_type(entry);
174 struct swap_info_struct *sis = swap_info[type];
175 pgoff_t offset = swp_offset(entry);
177 BUG_ON(!PageLocked(page));
179 if (frontswap_test(sis, offset))
180 ret = frontswap_ops.load(type, offset, page);
182 inc_frontswap_loads();
185 EXPORT_SYMBOL(__frontswap_load);
188 * Invalidate any data from frontswap associated with the specified swaptype
189 * and offset so that a subsequent "get" will fail.
191 void __frontswap_invalidate_page(unsigned type, pgoff_t offset)
193 struct swap_info_struct *sis = swap_info[type];
196 if (frontswap_test(sis, offset)) {
197 frontswap_ops.invalidate_page(type, offset);
198 __frontswap_clear(sis, offset);
199 inc_frontswap_invalidates();
202 EXPORT_SYMBOL(__frontswap_invalidate_page);
205 * Invalidate all data from frontswap associated with all offsets for the
206 * specified swaptype.
208 void __frontswap_invalidate_area(unsigned type)
210 struct swap_info_struct *sis = swap_info[type];
213 if (sis->frontswap_map == NULL)
215 frontswap_ops.invalidate_area(type);
216 atomic_set(&sis->frontswap_pages, 0);
217 memset(sis->frontswap_map, 0, sis->max / sizeof(long));
219 EXPORT_SYMBOL(__frontswap_invalidate_area);
221 static unsigned long __frontswap_curr_pages(void)
224 unsigned long totalpages = 0;
225 struct swap_info_struct *si = NULL;
227 assert_spin_locked(&swap_lock);
228 for (type = swap_list.head; type >= 0; type = si->next) {
229 si = swap_info[type];
230 totalpages += atomic_read(&si->frontswap_pages);
235 static int __frontswap_unuse_pages(unsigned long total, unsigned long *unused,
239 struct swap_info_struct *si = NULL;
240 int si_frontswap_pages;
241 unsigned long total_pages_to_unuse = total;
242 unsigned long pages = 0, pages_to_unuse = 0;
245 assert_spin_locked(&swap_lock);
246 for (type = swap_list.head; type >= 0; type = si->next) {
247 si = swap_info[type];
248 si_frontswap_pages = atomic_read(&si->frontswap_pages);
249 if (total_pages_to_unuse < si_frontswap_pages) {
250 pages = pages_to_unuse = total_pages_to_unuse;
252 pages = si_frontswap_pages;
253 pages_to_unuse = 0; /* unuse all */
255 /* ensure there is enough RAM to fetch pages from frontswap */
256 if (security_vm_enough_memory_mm(current->mm, pages)) {
260 vm_unacct_memory(pages);
261 *unused = pages_to_unuse;
270 static int __frontswap_shrink(unsigned long target_pages,
271 unsigned long *pages_to_unuse,
274 unsigned long total_pages = 0, total_pages_to_unuse;
276 assert_spin_locked(&swap_lock);
278 total_pages = __frontswap_curr_pages();
279 if (total_pages <= target_pages) {
284 total_pages_to_unuse = total_pages - target_pages;
285 return __frontswap_unuse_pages(total_pages_to_unuse, pages_to_unuse, type);
289 * Frontswap, like a true swap device, may unnecessarily retain pages
290 * under certain circumstances; "shrink" frontswap is essentially a
291 * "partial swapoff" and works by calling try_to_unuse to attempt to
292 * unuse enough frontswap pages to attempt to -- subject to memory
293 * constraints -- reduce the number of pages in frontswap to the
294 * number given in the parameter target_pages.
296 void frontswap_shrink(unsigned long target_pages)
298 unsigned long pages_to_unuse = 0;
302 * we don't want to hold swap_lock while doing a very
303 * lengthy try_to_unuse, but swap_list may change
304 * so restart scan from swap_list.head each time
306 spin_lock(&swap_lock);
307 ret = __frontswap_shrink(target_pages, &pages_to_unuse, &type);
308 spin_unlock(&swap_lock);
309 if (ret == 0 && pages_to_unuse)
310 try_to_unuse(type, true, pages_to_unuse);
313 EXPORT_SYMBOL(frontswap_shrink);
316 * Count and return the number of frontswap pages across all
317 * swap devices. This is exported so that backend drivers can
318 * determine current usage without reading debugfs.
320 unsigned long frontswap_curr_pages(void)
322 unsigned long totalpages = 0;
324 spin_lock(&swap_lock);
325 totalpages = __frontswap_curr_pages();
326 spin_unlock(&swap_lock);
330 EXPORT_SYMBOL(frontswap_curr_pages);
332 static int __init init_frontswap(void)
334 #ifdef CONFIG_DEBUG_FS
335 struct dentry *root = debugfs_create_dir("frontswap", NULL);
338 debugfs_create_u64("loads", S_IRUGO, root, &frontswap_loads);
339 debugfs_create_u64("succ_stores", S_IRUGO, root, &frontswap_succ_stores);
340 debugfs_create_u64("failed_stores", S_IRUGO, root,
341 &frontswap_failed_stores);
342 debugfs_create_u64("invalidates", S_IRUGO,
343 root, &frontswap_invalidates);
348 module_init(init_frontswap);