Root/
1 | #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> |
7 | #include <linux/slab.h> |
8 | #include <linux/memory.h> |
9 | #include <linux/vmalloc.h> |
10 | #include <linux/cgroup.h> |
11 | #include <linux/swapops.h> |
12 | |
13 | static void __meminit |
14 | __init_page_cgroup(struct page_cgroup *pc, unsigned long pfn) |
15 | { |
16 | pc->flags = 0; |
17 | pc->mem_cgroup = NULL; |
18 | pc->page = pfn_to_page(pfn); |
19 | INIT_LIST_HEAD(&pc->lru); |
20 | } |
21 | static unsigned long total_usage; |
22 | |
23 | #if !defined(CONFIG_SPARSEMEM) |
24 | |
25 | |
26 | void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat) |
27 | { |
28 | pgdat->node_page_cgroup = NULL; |
29 | } |
30 | |
31 | struct 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 | |
45 | static int __init alloc_node_page_cgroup(int nid) |
46 | { |
47 | struct page_cgroup *base, *pc; |
48 | unsigned long table_size; |
49 | unsigned long start_pfn, nr_pages, index; |
50 | |
51 | start_pfn = NODE_DATA(nid)->node_start_pfn; |
52 | nr_pages = NODE_DATA(nid)->node_spanned_pages; |
53 | |
54 | if (!nr_pages) |
55 | return 0; |
56 | |
57 | table_size = sizeof(struct page_cgroup) * nr_pages; |
58 | |
59 | base = __alloc_bootmem_node_nopanic(NODE_DATA(nid), |
60 | table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS)); |
61 | if (!base) |
62 | return -ENOMEM; |
63 | for (index = 0; index < nr_pages; index++) { |
64 | pc = base + index; |
65 | __init_page_cgroup(pc, start_pfn + index); |
66 | } |
67 | NODE_DATA(nid)->node_page_cgroup = base; |
68 | total_usage += table_size; |
69 | return 0; |
70 | } |
71 | |
72 | void __init page_cgroup_init_flatmem(void) |
73 | { |
74 | |
75 | int nid, fail; |
76 | |
77 | if (mem_cgroup_disabled()) |
78 | return; |
79 | |
80 | for_each_online_node(nid) { |
81 | fail = alloc_node_page_cgroup(nid); |
82 | if (fail) |
83 | goto fail; |
84 | } |
85 | printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage); |
86 | printk(KERN_INFO "please try 'cgroup_disable=memory' option if you" |
87 | " don't want memory cgroups\n"); |
88 | return; |
89 | fail: |
90 | printk(KERN_CRIT "allocation of page_cgroup failed.\n"); |
91 | printk(KERN_CRIT "please try 'cgroup_disable=memory' boot option\n"); |
92 | panic("Out of memory"); |
93 | } |
94 | |
95 | #else /* CONFIG_FLAT_NODE_MEM_MAP */ |
96 | |
97 | struct page_cgroup *lookup_page_cgroup(struct page *page) |
98 | { |
99 | unsigned long pfn = page_to_pfn(page); |
100 | struct mem_section *section = __pfn_to_section(pfn); |
101 | |
102 | if (!section->page_cgroup) |
103 | return NULL; |
104 | return section->page_cgroup + pfn; |
105 | } |
106 | |
107 | /* __alloc_bootmem...() is protected by !slab_available() */ |
108 | static int __init_refok init_section_page_cgroup(unsigned long pfn) |
109 | { |
110 | struct mem_section *section = __pfn_to_section(pfn); |
111 | struct page_cgroup *base, *pc; |
112 | unsigned long table_size; |
113 | int nid, index; |
114 | |
115 | if (!section->page_cgroup) { |
116 | nid = page_to_nid(pfn_to_page(pfn)); |
117 | table_size = sizeof(struct page_cgroup) * PAGES_PER_SECTION; |
118 | VM_BUG_ON(!slab_is_available()); |
119 | if (node_state(nid, N_HIGH_MEMORY)) { |
120 | base = kmalloc_node(table_size, |
121 | GFP_KERNEL | __GFP_NOWARN, nid); |
122 | if (!base) |
123 | base = vmalloc_node(table_size, nid); |
124 | } else { |
125 | base = kmalloc(table_size, GFP_KERNEL | __GFP_NOWARN); |
126 | if (!base) |
127 | base = vmalloc(table_size); |
128 | } |
129 | } else { |
130 | /* |
131 | * We don't have to allocate page_cgroup again, but |
132 | * address of memmap may be changed. So, we have to initialize |
133 | * again. |
134 | */ |
135 | base = section->page_cgroup + pfn; |
136 | table_size = 0; |
137 | /* check address of memmap is changed or not. */ |
138 | if (base->page == pfn_to_page(pfn)) |
139 | return 0; |
140 | } |
141 | |
142 | if (!base) { |
143 | printk(KERN_ERR "page cgroup allocation failure\n"); |
144 | return -ENOMEM; |
145 | } |
146 | |
147 | for (index = 0; index < PAGES_PER_SECTION; index++) { |
148 | pc = base + index; |
149 | __init_page_cgroup(pc, pfn + index); |
150 | } |
151 | |
152 | section->page_cgroup = base - pfn; |
153 | total_usage += table_size; |
154 | return 0; |
155 | } |
156 | #ifdef CONFIG_MEMORY_HOTPLUG |
157 | void __free_page_cgroup(unsigned long pfn) |
158 | { |
159 | struct mem_section *ms; |
160 | struct page_cgroup *base; |
161 | |
162 | ms = __pfn_to_section(pfn); |
163 | if (!ms || !ms->page_cgroup) |
164 | return; |
165 | base = ms->page_cgroup + pfn; |
166 | if (is_vmalloc_addr(base)) { |
167 | vfree(base); |
168 | ms->page_cgroup = NULL; |
169 | } else { |
170 | struct page *page = virt_to_page(base); |
171 | if (!PageReserved(page)) { /* Is bootmem ? */ |
172 | kfree(base); |
173 | ms->page_cgroup = NULL; |
174 | } |
175 | } |
176 | } |
177 | |
178 | int __meminit online_page_cgroup(unsigned long start_pfn, |
179 | unsigned long nr_pages, |
180 | int nid) |
181 | { |
182 | unsigned long start, end, pfn; |
183 | int fail = 0; |
184 | |
185 | start = start_pfn & ~(PAGES_PER_SECTION - 1); |
186 | end = ALIGN(start_pfn + nr_pages, PAGES_PER_SECTION); |
187 | |
188 | for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION) { |
189 | if (!pfn_present(pfn)) |
190 | continue; |
191 | fail = init_section_page_cgroup(pfn); |
192 | } |
193 | if (!fail) |
194 | return 0; |
195 | |
196 | /* rollback */ |
197 | for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION) |
198 | __free_page_cgroup(pfn); |
199 | |
200 | return -ENOMEM; |
201 | } |
202 | |
203 | int __meminit offline_page_cgroup(unsigned long start_pfn, |
204 | unsigned long nr_pages, int nid) |
205 | { |
206 | unsigned long start, end, pfn; |
207 | |
208 | start = start_pfn & ~(PAGES_PER_SECTION - 1); |
209 | end = ALIGN(start_pfn + nr_pages, PAGES_PER_SECTION); |
210 | |
211 | for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION) |
212 | __free_page_cgroup(pfn); |
213 | return 0; |
214 | |
215 | } |
216 | |
217 | static int __meminit page_cgroup_callback(struct notifier_block *self, |
218 | unsigned long action, void *arg) |
219 | { |
220 | struct memory_notify *mn = arg; |
221 | int ret = 0; |
222 | switch (action) { |
223 | case MEM_GOING_ONLINE: |
224 | ret = online_page_cgroup(mn->start_pfn, |
225 | mn->nr_pages, mn->status_change_nid); |
226 | break; |
227 | case MEM_OFFLINE: |
228 | offline_page_cgroup(mn->start_pfn, |
229 | mn->nr_pages, mn->status_change_nid); |
230 | break; |
231 | case MEM_CANCEL_ONLINE: |
232 | case MEM_GOING_OFFLINE: |
233 | break; |
234 | case MEM_ONLINE: |
235 | case MEM_CANCEL_OFFLINE: |
236 | break; |
237 | } |
238 | |
239 | if (ret) |
240 | ret = notifier_from_errno(ret); |
241 | else |
242 | ret = NOTIFY_OK; |
243 | |
244 | return ret; |
245 | } |
246 | |
247 | #endif |
248 | |
249 | void __init page_cgroup_init(void) |
250 | { |
251 | unsigned long pfn; |
252 | int fail = 0; |
253 | |
254 | if (mem_cgroup_disabled()) |
255 | return; |
256 | |
257 | for (pfn = 0; !fail && pfn < max_pfn; pfn += PAGES_PER_SECTION) { |
258 | if (!pfn_present(pfn)) |
259 | continue; |
260 | fail = init_section_page_cgroup(pfn); |
261 | } |
262 | if (fail) { |
263 | printk(KERN_CRIT "try 'cgroup_disable=memory' boot option\n"); |
264 | panic("Out of memory"); |
265 | } else { |
266 | hotplug_memory_notifier(page_cgroup_callback, 0); |
267 | } |
268 | printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage); |
269 | printk(KERN_INFO "please try 'cgroup_disable=memory' option if you don't" |
270 | " want memory cgroups\n"); |
271 | } |
272 | |
273 | void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat) |
274 | { |
275 | return; |
276 | } |
277 | |
278 | #endif |
279 | |
280 | |
281 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP |
282 | |
283 | static DEFINE_MUTEX(swap_cgroup_mutex); |
284 | struct swap_cgroup_ctrl { |
285 | struct page **map; |
286 | unsigned long length; |
287 | spinlock_t lock; |
288 | }; |
289 | |
290 | struct swap_cgroup_ctrl swap_cgroup_ctrl[MAX_SWAPFILES]; |
291 | |
292 | struct swap_cgroup { |
293 | unsigned short id; |
294 | }; |
295 | #define SC_PER_PAGE (PAGE_SIZE/sizeof(struct swap_cgroup)) |
296 | #define SC_POS_MASK (SC_PER_PAGE - 1) |
297 | |
298 | /* |
299 | * SwapCgroup implements "lookup" and "exchange" operations. |
300 | * In typical usage, this swap_cgroup is accessed via memcg's charge/uncharge |
301 | * against SwapCache. At swap_free(), this is accessed directly from swap. |
302 | * |
303 | * This means, |
304 | * - we have no race in "exchange" when we're accessed via SwapCache because |
305 | * SwapCache(and its swp_entry) is under lock. |
306 | * - When called via swap_free(), there is no user of this entry and no race. |
307 | * Then, we don't need lock around "exchange". |
308 | * |
309 | * TODO: we can push these buffers out to HIGHMEM. |
310 | */ |
311 | |
312 | /* |
313 | * allocate buffer for swap_cgroup. |
314 | */ |
315 | static int swap_cgroup_prepare(int type) |
316 | { |
317 | struct page *page; |
318 | struct swap_cgroup_ctrl *ctrl; |
319 | unsigned long idx, max; |
320 | |
321 | ctrl = &swap_cgroup_ctrl[type]; |
322 | |
323 | for (idx = 0; idx < ctrl->length; idx++) { |
324 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
325 | if (!page) |
326 | goto not_enough_page; |
327 | ctrl->map[idx] = page; |
328 | } |
329 | return 0; |
330 | not_enough_page: |
331 | max = idx; |
332 | for (idx = 0; idx < max; idx++) |
333 | __free_page(ctrl->map[idx]); |
334 | |
335 | return -ENOMEM; |
336 | } |
337 | |
338 | /** |
339 | * swap_cgroup_cmpxchg - cmpxchg mem_cgroup's id for this swp_entry. |
340 | * @end: swap entry to be cmpxchged |
341 | * @old: old id |
342 | * @new: new id |
343 | * |
344 | * Returns old id at success, 0 at failure. |
345 | * (There is no mem_cgroup useing 0 as its id) |
346 | */ |
347 | unsigned short swap_cgroup_cmpxchg(swp_entry_t ent, |
348 | unsigned short old, unsigned short new) |
349 | { |
350 | int type = swp_type(ent); |
351 | unsigned long offset = swp_offset(ent); |
352 | unsigned long idx = offset / SC_PER_PAGE; |
353 | unsigned long pos = offset & SC_POS_MASK; |
354 | struct swap_cgroup_ctrl *ctrl; |
355 | struct page *mappage; |
356 | struct swap_cgroup *sc; |
357 | unsigned long flags; |
358 | unsigned short retval; |
359 | |
360 | ctrl = &swap_cgroup_ctrl[type]; |
361 | |
362 | mappage = ctrl->map[idx]; |
363 | sc = page_address(mappage); |
364 | sc += pos; |
365 | spin_lock_irqsave(&ctrl->lock, flags); |
366 | retval = sc->id; |
367 | if (retval == old) |
368 | sc->id = new; |
369 | else |
370 | retval = 0; |
371 | spin_unlock_irqrestore(&ctrl->lock, flags); |
372 | return retval; |
373 | } |
374 | |
375 | /** |
376 | * swap_cgroup_record - record mem_cgroup for this swp_entry. |
377 | * @ent: swap entry to be recorded into |
378 | * @mem: mem_cgroup to be recorded |
379 | * |
380 | * Returns old value at success, 0 at failure. |
381 | * (Of course, old value can be 0.) |
382 | */ |
383 | unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id) |
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; |
392 | unsigned short old; |
393 | unsigned long flags; |
394 | |
395 | ctrl = &swap_cgroup_ctrl[type]; |
396 | |
397 | mappage = ctrl->map[idx]; |
398 | sc = page_address(mappage); |
399 | sc += pos; |
400 | spin_lock_irqsave(&ctrl->lock, flags); |
401 | old = sc->id; |
402 | sc->id = id; |
403 | spin_unlock_irqrestore(&ctrl->lock, flags); |
404 | |
405 | return old; |
406 | } |
407 | |
408 | /** |
409 | * lookup_swap_cgroup - lookup mem_cgroup tied to swap entry |
410 | * @ent: swap entry to be looked up. |
411 | * |
412 | * Returns CSS ID of mem_cgroup at success. 0 at failure. (0 is invalid ID) |
413 | */ |
414 | unsigned short lookup_swap_cgroup(swp_entry_t ent) |
415 | { |
416 | int type = swp_type(ent); |
417 | unsigned long offset = swp_offset(ent); |
418 | unsigned long idx = offset / SC_PER_PAGE; |
419 | unsigned long pos = offset & SC_POS_MASK; |
420 | struct swap_cgroup_ctrl *ctrl; |
421 | struct page *mappage; |
422 | struct swap_cgroup *sc; |
423 | unsigned short ret; |
424 | |
425 | ctrl = &swap_cgroup_ctrl[type]; |
426 | mappage = ctrl->map[idx]; |
427 | sc = page_address(mappage); |
428 | sc += pos; |
429 | ret = sc->id; |
430 | return ret; |
431 | } |
432 | |
433 | int swap_cgroup_swapon(int type, unsigned long max_pages) |
434 | { |
435 | void *array; |
436 | unsigned long array_size; |
437 | unsigned long length; |
438 | struct swap_cgroup_ctrl *ctrl; |
439 | |
440 | if (!do_swap_account) |
441 | return 0; |
442 | |
443 | length = ((max_pages/SC_PER_PAGE) + 1); |
444 | array_size = length * sizeof(void *); |
445 | |
446 | array = vmalloc(array_size); |
447 | if (!array) |
448 | goto nomem; |
449 | |
450 | memset(array, 0, array_size); |
451 | ctrl = &swap_cgroup_ctrl[type]; |
452 | mutex_lock(&swap_cgroup_mutex); |
453 | ctrl->length = length; |
454 | ctrl->map = array; |
455 | spin_lock_init(&ctrl->lock); |
456 | if (swap_cgroup_prepare(type)) { |
457 | /* memory shortage */ |
458 | ctrl->map = NULL; |
459 | ctrl->length = 0; |
460 | vfree(array); |
461 | mutex_unlock(&swap_cgroup_mutex); |
462 | goto nomem; |
463 | } |
464 | mutex_unlock(&swap_cgroup_mutex); |
465 | |
466 | return 0; |
467 | nomem: |
468 | printk(KERN_INFO "couldn't allocate enough memory for swap_cgroup.\n"); |
469 | printk(KERN_INFO |
470 | "swap_cgroup can be disabled by noswapaccount boot option\n"); |
471 | return -ENOMEM; |
472 | } |
473 | |
474 | void swap_cgroup_swapoff(int type) |
475 | { |
476 | int i; |
477 | struct swap_cgroup_ctrl *ctrl; |
478 | |
479 | if (!do_swap_account) |
480 | return; |
481 | |
482 | mutex_lock(&swap_cgroup_mutex); |
483 | ctrl = &swap_cgroup_ctrl[type]; |
484 | if (ctrl->map) { |
485 | for (i = 0; i < ctrl->length; i++) { |
486 | struct page *page = ctrl->map[i]; |
487 | if (page) |
488 | __free_page(page); |
489 | } |
490 | vfree(ctrl->map); |
491 | ctrl->map = NULL; |
492 | ctrl->length = 0; |
493 | } |
494 | mutex_unlock(&swap_cgroup_mutex); |
495 | } |
496 | |
497 | #endif |
498 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9