blob: 74ccff61d1bea8d6e258170280ea5b20523dad23 [file] [log] [blame]
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -07001#include <linux/mm.h>
2#include <linux/mmzone.h>
3#include <linux/bootmem.h>
4#include <linux/bit_spinlock.h>
5#include <linux/page_cgroup.h>
6#include <linux/hash.h>
KAMEZAWA Hiroyuki94b6da52008-10-22 14:15:05 -07007#include <linux/slab.h>
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -07008#include <linux/memory.h>
Paul Mundt4c8210422008-10-22 14:14:58 -07009#include <linux/vmalloc.h>
KAMEZAWA Hiroyuki94b6da52008-10-22 14:15:05 -070010#include <linux/cgroup.h>
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -080011#include <linux/swapops.h>
Catalin Marinas7952f982010-07-19 11:54:14 +010012#include <linux/kmemleak.h>
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070013
Johannes Weiner6b3ae582011-03-23 16:42:30 -070014static void __meminit init_page_cgroup(struct page_cgroup *pc, unsigned long id)
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070015{
16 pc->flags = 0;
Johannes Weiner6b3ae582011-03-23 16:42:30 -070017 set_page_cgroup_array_id(pc, id);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070018 pc->mem_cgroup = NULL;
KAMEZAWA Hiroyuki08e552c2009-01-07 18:08:01 -080019 INIT_LIST_HEAD(&pc->lru);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070020}
21static unsigned long total_usage;
22
23#if !defined(CONFIG_SPARSEMEM)
24
25
Al Viro31168482008-11-22 17:33:24 +000026void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070027{
28 pgdat->node_page_cgroup = NULL;
29}
30
31struct page_cgroup *lookup_page_cgroup(struct page *page)
32{
33 unsigned long pfn = page_to_pfn(page);
34 unsigned long offset;
35 struct page_cgroup *base;
36
37 base = NODE_DATA(page_to_nid(page))->node_page_cgroup;
38 if (unlikely(!base))
39 return NULL;
40
41 offset = pfn - NODE_DATA(page_to_nid(page))->node_start_pfn;
42 return base + offset;
43}
44
Johannes Weiner6b3ae582011-03-23 16:42:30 -070045struct page *lookup_cgroup_page(struct page_cgroup *pc)
46{
47 unsigned long pfn;
48 struct page *page;
49 pg_data_t *pgdat;
50
51 pgdat = NODE_DATA(page_cgroup_array_id(pc));
52 pfn = pc - pgdat->node_page_cgroup + pgdat->node_start_pfn;
53 page = pfn_to_page(pfn);
54 VM_BUG_ON(pc != lookup_page_cgroup(page));
55 return page;
56}
57
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070058static int __init alloc_node_page_cgroup(int nid)
59{
60 struct page_cgroup *base, *pc;
61 unsigned long table_size;
62 unsigned long start_pfn, nr_pages, index;
63
64 start_pfn = NODE_DATA(nid)->node_start_pfn;
65 nr_pages = NODE_DATA(nid)->node_spanned_pages;
66
KAMEZAWA Hiroyuki653d22c2008-12-09 13:14:20 -080067 if (!nr_pages)
68 return 0;
69
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070070 table_size = sizeof(struct page_cgroup) * nr_pages;
KAMEZAWA Hiroyukica371c02009-06-12 10:33:53 +030071
72 base = __alloc_bootmem_node_nopanic(NODE_DATA(nid),
73 table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
74 if (!base)
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070075 return -ENOMEM;
76 for (index = 0; index < nr_pages; index++) {
77 pc = base + index;
Johannes Weiner6b3ae582011-03-23 16:42:30 -070078 init_page_cgroup(pc, nid);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070079 }
80 NODE_DATA(nid)->node_page_cgroup = base;
81 total_usage += table_size;
82 return 0;
83}
84
KAMEZAWA Hiroyukica371c02009-06-12 10:33:53 +030085void __init page_cgroup_init_flatmem(void)
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070086{
87
88 int nid, fail;
89
Hirokazu Takahashif8d665422009-01-07 18:08:02 -080090 if (mem_cgroup_disabled())
KAMEZAWA Hiroyuki94b6da52008-10-22 14:15:05 -070091 return;
92
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -070093 for_each_online_node(nid) {
94 fail = alloc_node_page_cgroup(nid);
95 if (fail)
96 goto fail;
97 }
98 printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
Randy Dunlap8ca739e2009-06-17 16:26:32 -070099 printk(KERN_INFO "please try 'cgroup_disable=memory' option if you"
100 " don't want memory cgroups\n");
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700101 return;
102fail:
Randy Dunlap8ca739e2009-06-17 16:26:32 -0700103 printk(KERN_CRIT "allocation of page_cgroup failed.\n");
104 printk(KERN_CRIT "please try 'cgroup_disable=memory' boot option\n");
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700105 panic("Out of memory");
106}
107
108#else /* CONFIG_FLAT_NODE_MEM_MAP */
109
110struct page_cgroup *lookup_page_cgroup(struct page *page)
111{
112 unsigned long pfn = page_to_pfn(page);
113 struct mem_section *section = __pfn_to_section(pfn);
114
Balbir Singhd69b0422009-06-17 16:26:34 -0700115 if (!section->page_cgroup)
116 return NULL;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700117 return section->page_cgroup + pfn;
118}
119
Johannes Weiner6b3ae582011-03-23 16:42:30 -0700120struct page *lookup_cgroup_page(struct page_cgroup *pc)
121{
122 struct mem_section *section;
123 struct page *page;
124 unsigned long nr;
125
126 nr = page_cgroup_array_id(pc);
127 section = __nr_to_section(nr);
128 page = pfn_to_page(pc - section->page_cgroup);
129 VM_BUG_ON(pc != lookup_page_cgroup(page));
130 return page;
131}
132
Namhyung Kim268433b2011-05-26 16:25:29 -0700133static void *__meminit alloc_page_cgroup(size_t size, int nid)
Michal Hockodde79e02011-03-23 16:42:40 -0700134{
135 void *addr = NULL;
136
Andi Kleen21a3c962011-05-11 15:13:35 -0700137 addr = alloc_pages_exact_nid(nid, size, GFP_KERNEL | __GFP_NOWARN);
Michal Hockodde79e02011-03-23 16:42:40 -0700138 if (addr)
139 return addr;
140
141 if (node_state(nid, N_HIGH_MEMORY))
142 addr = vmalloc_node(size, nid);
143 else
144 addr = vmalloc(size);
145
146 return addr;
147}
148
149#ifdef CONFIG_MEMORY_HOTPLUG
150static void free_page_cgroup(void *addr)
151{
152 if (is_vmalloc_addr(addr)) {
153 vfree(addr);
154 } else {
155 struct page *page = virt_to_page(addr);
Michal Hocko6cfddb22011-03-23 16:42:41 -0700156 size_t table_size =
157 sizeof(struct page_cgroup) * PAGES_PER_SECTION;
158
159 BUG_ON(PageReserved(page));
160 free_pages_exact(addr, table_size);
Michal Hockodde79e02011-03-23 16:42:40 -0700161 }
162}
163#endif
164
Namhyung Kim268433b2011-05-26 16:25:29 -0700165static int __meminit init_section_page_cgroup(unsigned long pfn)
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700166{
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700167 struct page_cgroup *base, *pc;
Johannes Weiner6b3ae582011-03-23 16:42:30 -0700168 struct mem_section *section;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700169 unsigned long table_size;
Johannes Weiner6b3ae582011-03-23 16:42:30 -0700170 unsigned long nr;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700171 int nid, index;
172
Johannes Weiner6b3ae582011-03-23 16:42:30 -0700173 nr = pfn_to_section_nr(pfn);
174 section = __nr_to_section(nr);
175
176 if (section->page_cgroup)
177 return 0;
178
179 nid = page_to_nid(pfn_to_page(pfn));
180 table_size = sizeof(struct page_cgroup) * PAGES_PER_SECTION;
Michal Hockodde79e02011-03-23 16:42:40 -0700181 base = alloc_page_cgroup(table_size, nid);
182
Johannes Weiner6b3ae582011-03-23 16:42:30 -0700183 /*
184 * The value stored in section->page_cgroup is (base - pfn)
185 * and it does not point to the memory block allocated above,
186 * causing kmemleak false positives.
187 */
188 kmemleak_not_leak(base);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700189
190 if (!base) {
191 printk(KERN_ERR "page cgroup allocation failure\n");
192 return -ENOMEM;
193 }
194
195 for (index = 0; index < PAGES_PER_SECTION; index++) {
196 pc = base + index;
Johannes Weiner6b3ae582011-03-23 16:42:30 -0700197 init_page_cgroup(pc, nr);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700198 }
199
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700200 section->page_cgroup = base - pfn;
201 total_usage += table_size;
202 return 0;
203}
204#ifdef CONFIG_MEMORY_HOTPLUG
205void __free_page_cgroup(unsigned long pfn)
206{
207 struct mem_section *ms;
208 struct page_cgroup *base;
209
210 ms = __pfn_to_section(pfn);
211 if (!ms || !ms->page_cgroup)
212 return;
213 base = ms->page_cgroup + pfn;
Michal Hockodde79e02011-03-23 16:42:40 -0700214 free_page_cgroup(base);
215 ms->page_cgroup = NULL;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700216}
217
Al Viro31168482008-11-22 17:33:24 +0000218int __meminit online_page_cgroup(unsigned long start_pfn,
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700219 unsigned long nr_pages,
220 int nid)
221{
222 unsigned long start, end, pfn;
223 int fail = 0;
224
KAMEZAWA Hiroyuki33c5d3d2008-11-12 13:27:01 -0800225 start = start_pfn & ~(PAGES_PER_SECTION - 1);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700226 end = ALIGN(start_pfn + nr_pages, PAGES_PER_SECTION);
227
228 for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION) {
229 if (!pfn_present(pfn))
230 continue;
231 fail = init_section_page_cgroup(pfn);
232 }
233 if (!fail)
234 return 0;
235
236 /* rollback */
237 for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
238 __free_page_cgroup(pfn);
239
240 return -ENOMEM;
241}
242
Al Viro31168482008-11-22 17:33:24 +0000243int __meminit offline_page_cgroup(unsigned long start_pfn,
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700244 unsigned long nr_pages, int nid)
245{
246 unsigned long start, end, pfn;
247
KAMEZAWA Hiroyuki33c5d3d2008-11-12 13:27:01 -0800248 start = start_pfn & ~(PAGES_PER_SECTION - 1);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700249 end = ALIGN(start_pfn + nr_pages, PAGES_PER_SECTION);
250
251 for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
252 __free_page_cgroup(pfn);
253 return 0;
254
255}
256
Al Viro31168482008-11-22 17:33:24 +0000257static int __meminit page_cgroup_callback(struct notifier_block *self,
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700258 unsigned long action, void *arg)
259{
260 struct memory_notify *mn = arg;
261 int ret = 0;
262 switch (action) {
263 case MEM_GOING_ONLINE:
264 ret = online_page_cgroup(mn->start_pfn,
265 mn->nr_pages, mn->status_change_nid);
266 break;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700267 case MEM_OFFLINE:
268 offline_page_cgroup(mn->start_pfn,
269 mn->nr_pages, mn->status_change_nid);
270 break;
KAMEZAWA Hiroyukidc19f9d2008-12-01 13:13:48 -0800271 case MEM_CANCEL_ONLINE:
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700272 case MEM_GOING_OFFLINE:
273 break;
274 case MEM_ONLINE:
275 case MEM_CANCEL_OFFLINE:
276 break;
277 }
KAMEZAWA Hiroyukidc19f9d2008-12-01 13:13:48 -0800278
Prarit Bhargava5fda1bd2011-03-22 16:30:49 -0700279 return notifier_from_errno(ret);
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700280}
281
282#endif
283
284void __init page_cgroup_init(void)
285{
286 unsigned long pfn;
287 int fail = 0;
288
Hirokazu Takahashif8d665422009-01-07 18:08:02 -0800289 if (mem_cgroup_disabled())
KAMEZAWA Hiroyuki94b6da52008-10-22 14:15:05 -0700290 return;
291
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700292 for (pfn = 0; !fail && pfn < max_pfn; pfn += PAGES_PER_SECTION) {
293 if (!pfn_present(pfn))
294 continue;
295 fail = init_section_page_cgroup(pfn);
296 }
297 if (fail) {
Randy Dunlap8ca739e2009-06-17 16:26:32 -0700298 printk(KERN_CRIT "try 'cgroup_disable=memory' boot option\n");
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700299 panic("Out of memory");
300 } else {
301 hotplug_memory_notifier(page_cgroup_callback, 0);
302 }
303 printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
Randy Dunlap8ca739e2009-06-17 16:26:32 -0700304 printk(KERN_INFO "please try 'cgroup_disable=memory' option if you don't"
305 " want memory cgroups\n");
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700306}
307
Al Viro31168482008-11-22 17:33:24 +0000308void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700309{
310 return;
311}
312
313#endif
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800314
315
316#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
317
318static DEFINE_MUTEX(swap_cgroup_mutex);
319struct swap_cgroup_ctrl {
320 struct page **map;
321 unsigned long length;
KAMEZAWA Hiroyukie9e58a42010-03-15 00:34:57 -0400322 spinlock_t lock;
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800323};
324
325struct swap_cgroup_ctrl swap_cgroup_ctrl[MAX_SWAPFILES];
326
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800327struct swap_cgroup {
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700328 unsigned short id;
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800329};
330#define SC_PER_PAGE (PAGE_SIZE/sizeof(struct swap_cgroup))
331#define SC_POS_MASK (SC_PER_PAGE - 1)
332
333/*
334 * SwapCgroup implements "lookup" and "exchange" operations.
335 * In typical usage, this swap_cgroup is accessed via memcg's charge/uncharge
336 * against SwapCache. At swap_free(), this is accessed directly from swap.
337 *
338 * This means,
339 * - we have no race in "exchange" when we're accessed via SwapCache because
340 * SwapCache(and its swp_entry) is under lock.
341 * - When called via swap_free(), there is no user of this entry and no race.
342 * Then, we don't need lock around "exchange".
343 *
344 * TODO: we can push these buffers out to HIGHMEM.
345 */
346
347/*
348 * allocate buffer for swap_cgroup.
349 */
350static int swap_cgroup_prepare(int type)
351{
352 struct page *page;
353 struct swap_cgroup_ctrl *ctrl;
354 unsigned long idx, max;
355
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800356 ctrl = &swap_cgroup_ctrl[type];
357
358 for (idx = 0; idx < ctrl->length; idx++) {
359 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
360 if (!page)
361 goto not_enough_page;
362 ctrl->map[idx] = page;
363 }
364 return 0;
365not_enough_page:
366 max = idx;
367 for (idx = 0; idx < max; idx++)
368 __free_page(ctrl->map[idx]);
369
370 return -ENOMEM;
371}
372
373/**
Daisuke Nishimura02491442010-03-10 15:22:17 -0800374 * swap_cgroup_cmpxchg - cmpxchg mem_cgroup's id for this swp_entry.
375 * @end: swap entry to be cmpxchged
376 * @old: old id
377 * @new: new id
378 *
379 * Returns old id at success, 0 at failure.
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300380 * (There is no mem_cgroup using 0 as its id)
Daisuke Nishimura02491442010-03-10 15:22:17 -0800381 */
382unsigned short swap_cgroup_cmpxchg(swp_entry_t ent,
383 unsigned short old, unsigned short new)
384{
385 int type = swp_type(ent);
386 unsigned long offset = swp_offset(ent);
387 unsigned long idx = offset / SC_PER_PAGE;
388 unsigned long pos = offset & SC_POS_MASK;
389 struct swap_cgroup_ctrl *ctrl;
390 struct page *mappage;
391 struct swap_cgroup *sc;
KAMEZAWA Hiroyukie9e58a42010-03-15 00:34:57 -0400392 unsigned long flags;
393 unsigned short retval;
Daisuke Nishimura02491442010-03-10 15:22:17 -0800394
395 ctrl = &swap_cgroup_ctrl[type];
396
397 mappage = ctrl->map[idx];
398 sc = page_address(mappage);
399 sc += pos;
KAMEZAWA Hiroyukie9e58a42010-03-15 00:34:57 -0400400 spin_lock_irqsave(&ctrl->lock, flags);
401 retval = sc->id;
402 if (retval == old)
403 sc->id = new;
Daisuke Nishimura02491442010-03-10 15:22:17 -0800404 else
KAMEZAWA Hiroyukie9e58a42010-03-15 00:34:57 -0400405 retval = 0;
406 spin_unlock_irqrestore(&ctrl->lock, flags);
407 return retval;
Daisuke Nishimura02491442010-03-10 15:22:17 -0800408}
409
410/**
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800411 * swap_cgroup_record - record mem_cgroup for this swp_entry.
412 * @ent: swap entry to be recorded into
413 * @mem: mem_cgroup to be recorded
414 *
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700415 * Returns old value at success, 0 at failure.
416 * (Of course, old value can be 0.)
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800417 */
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700418unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id)
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800419{
420 int type = swp_type(ent);
421 unsigned long offset = swp_offset(ent);
422 unsigned long idx = offset / SC_PER_PAGE;
423 unsigned long pos = offset & SC_POS_MASK;
424 struct swap_cgroup_ctrl *ctrl;
425 struct page *mappage;
426 struct swap_cgroup *sc;
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700427 unsigned short old;
KAMEZAWA Hiroyukie9e58a42010-03-15 00:34:57 -0400428 unsigned long flags;
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800429
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800430 ctrl = &swap_cgroup_ctrl[type];
431
432 mappage = ctrl->map[idx];
433 sc = page_address(mappage);
434 sc += pos;
KAMEZAWA Hiroyukie9e58a42010-03-15 00:34:57 -0400435 spin_lock_irqsave(&ctrl->lock, flags);
436 old = sc->id;
437 sc->id = id;
438 spin_unlock_irqrestore(&ctrl->lock, flags);
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800439
440 return old;
441}
442
443/**
444 * lookup_swap_cgroup - lookup mem_cgroup tied to swap entry
445 * @ent: swap entry to be looked up.
446 *
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700447 * Returns CSS ID of mem_cgroup at success. 0 at failure. (0 is invalid ID)
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800448 */
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700449unsigned short lookup_swap_cgroup(swp_entry_t ent)
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800450{
451 int type = swp_type(ent);
452 unsigned long offset = swp_offset(ent);
453 unsigned long idx = offset / SC_PER_PAGE;
454 unsigned long pos = offset & SC_POS_MASK;
455 struct swap_cgroup_ctrl *ctrl;
456 struct page *mappage;
457 struct swap_cgroup *sc;
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700458 unsigned short ret;
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800459
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800460 ctrl = &swap_cgroup_ctrl[type];
461 mappage = ctrl->map[idx];
462 sc = page_address(mappage);
463 sc += pos;
KAMEZAWA Hiroyukia3b2d692009-04-02 16:57:45 -0700464 ret = sc->id;
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800465 return ret;
466}
467
468int swap_cgroup_swapon(int type, unsigned long max_pages)
469{
470 void *array;
471 unsigned long array_size;
472 unsigned long length;
473 struct swap_cgroup_ctrl *ctrl;
474
475 if (!do_swap_account)
476 return 0;
477
Namhyung Kim33278f72011-05-26 16:25:30 -0700478 length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800479 array_size = length * sizeof(void *);
480
481 array = vmalloc(array_size);
482 if (!array)
483 goto nomem;
484
485 memset(array, 0, array_size);
486 ctrl = &swap_cgroup_ctrl[type];
487 mutex_lock(&swap_cgroup_mutex);
488 ctrl->length = length;
489 ctrl->map = array;
KAMEZAWA Hiroyukie9e58a42010-03-15 00:34:57 -0400490 spin_lock_init(&ctrl->lock);
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800491 if (swap_cgroup_prepare(type)) {
492 /* memory shortage */
493 ctrl->map = NULL;
494 ctrl->length = 0;
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800495 mutex_unlock(&swap_cgroup_mutex);
Namhyung Kim6a5b18d2011-05-26 16:25:31 -0700496 vfree(array);
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800497 goto nomem;
498 }
499 mutex_unlock(&swap_cgroup_mutex);
500
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800501 return 0;
502nomem:
503 printk(KERN_INFO "couldn't allocate enough memory for swap_cgroup.\n");
504 printk(KERN_INFO
505 "swap_cgroup can be disabled by noswapaccount boot option\n");
506 return -ENOMEM;
507}
508
509void swap_cgroup_swapoff(int type)
510{
Namhyung Kim6a5b18d2011-05-26 16:25:31 -0700511 struct page **map;
512 unsigned long i, length;
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800513 struct swap_cgroup_ctrl *ctrl;
514
515 if (!do_swap_account)
516 return;
517
518 mutex_lock(&swap_cgroup_mutex);
519 ctrl = &swap_cgroup_ctrl[type];
Namhyung Kim6a5b18d2011-05-26 16:25:31 -0700520 map = ctrl->map;
521 length = ctrl->length;
522 ctrl->map = NULL;
523 ctrl->length = 0;
524 mutex_unlock(&swap_cgroup_mutex);
525
526 if (map) {
527 for (i = 0; i < length; i++) {
528 struct page *page = map[i];
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800529 if (page)
530 __free_page(page);
531 }
Namhyung Kim6a5b18d2011-05-26 16:25:31 -0700532 vfree(map);
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800533 }
KAMEZAWA Hiroyuki27a7faa2009-01-07 18:07:58 -0800534}
535
536#endif