Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit 9623e078c1f4692a91531af2f639ec8aff8f0472

Authored by Hugh Dickins
Committed by Linus Torvalds
1 parent 3ee1062b4e
Exists in master and in 7 other branches imx_3.0.35_4.1.0, imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga, smarc-l5.0.0_1.0.0-ga

memcg: fix oops in mem_cgroup_shrink_usage 

Got an oops in mem_cgroup_shrink_usage() when testing loop over tmpfs:
yes, of course, loop0 has no mm: other entry points check but this didn't.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Showing 1 changed file with 2 additions and 0 deletions Inline Diff

1 /* memcontrol.c - Memory Controller 1 /* memcontrol.c - Memory Controller
2 * 2 *
3 * Copyright IBM Corporation, 2007 3 * Copyright IBM Corporation, 2007
4 * Author Balbir Singh <balbir@linux.vnet.ibm.com> 4 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
5 * 5 *
6 * Copyright 2007 OpenVZ SWsoft Inc 6 * Copyright 2007 OpenVZ SWsoft Inc
7 * Author: Pavel Emelianov <xemul@openvz.org> 7 * Author: Pavel Emelianov <xemul@openvz.org>
8 * 8 *
9 * This program is free software; you can redistribute it and/or modify 9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by 10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or 11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version. 12 * (at your option) any later version.
13 * 13 *
14 * This program is distributed in the hope that it will be useful, 14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of 15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details. 17 * GNU General Public License for more details.
18 */ 18 */
19 19
20 #include <linux/res_counter.h> 20 #include <linux/res_counter.h>
21 #include <linux/memcontrol.h> 21 #include <linux/memcontrol.h>
22 #include <linux/cgroup.h> 22 #include <linux/cgroup.h>
23 #include <linux/mm.h> 23 #include <linux/mm.h>
24 #include <linux/smp.h> 24 #include <linux/smp.h>
25 #include <linux/page-flags.h> 25 #include <linux/page-flags.h>
26 #include <linux/backing-dev.h> 26 #include <linux/backing-dev.h>
27 #include <linux/bit_spinlock.h> 27 #include <linux/bit_spinlock.h>
28 #include <linux/rcupdate.h> 28 #include <linux/rcupdate.h>
29 #include <linux/slab.h> 29 #include <linux/slab.h>
30 #include <linux/swap.h> 30 #include <linux/swap.h>
31 #include <linux/spinlock.h> 31 #include <linux/spinlock.h>
32 #include <linux/fs.h> 32 #include <linux/fs.h>
33 #include <linux/seq_file.h> 33 #include <linux/seq_file.h>
34 #include <linux/vmalloc.h> 34 #include <linux/vmalloc.h>
35 35
36 #include <asm/uaccess.h> 36 #include <asm/uaccess.h>
37 37
38 struct cgroup_subsys mem_cgroup_subsys __read_mostly; 38 struct cgroup_subsys mem_cgroup_subsys __read_mostly;
39 static struct kmem_cache *page_cgroup_cache __read_mostly; 39 static struct kmem_cache *page_cgroup_cache __read_mostly;
40 #define MEM_CGROUP_RECLAIM_RETRIES 5 40 #define MEM_CGROUP_RECLAIM_RETRIES 5
41 41
42 /* 42 /*
43 * Statistics for memory cgroup. 43 * Statistics for memory cgroup.
44 */ 44 */
45 enum mem_cgroup_stat_index { 45 enum mem_cgroup_stat_index {
46 /* 46 /*
47 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss. 47 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
48 */ 48 */
49 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */ 49 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
50 MEM_CGROUP_STAT_RSS, /* # of pages charged as rss */ 50 MEM_CGROUP_STAT_RSS, /* # of pages charged as rss */
51 MEM_CGROUP_STAT_PGPGIN_COUNT, /* # of pages paged in */ 51 MEM_CGROUP_STAT_PGPGIN_COUNT, /* # of pages paged in */
52 MEM_CGROUP_STAT_PGPGOUT_COUNT, /* # of pages paged out */ 52 MEM_CGROUP_STAT_PGPGOUT_COUNT, /* # of pages paged out */
53 53
54 MEM_CGROUP_STAT_NSTATS, 54 MEM_CGROUP_STAT_NSTATS,
55 }; 55 };
56 56
57 struct mem_cgroup_stat_cpu { 57 struct mem_cgroup_stat_cpu {
58 s64 count[MEM_CGROUP_STAT_NSTATS]; 58 s64 count[MEM_CGROUP_STAT_NSTATS];
59 } ____cacheline_aligned_in_smp; 59 } ____cacheline_aligned_in_smp;
60 60
61 struct mem_cgroup_stat { 61 struct mem_cgroup_stat {
62 struct mem_cgroup_stat_cpu cpustat[NR_CPUS]; 62 struct mem_cgroup_stat_cpu cpustat[NR_CPUS];
63 }; 63 };
64 64
65 /* 65 /*
66 * For accounting under irq disable, no need for increment preempt count. 66 * For accounting under irq disable, no need for increment preempt count.
67 */ 67 */
68 static void __mem_cgroup_stat_add_safe(struct mem_cgroup_stat *stat, 68 static void __mem_cgroup_stat_add_safe(struct mem_cgroup_stat *stat,
69 enum mem_cgroup_stat_index idx, int val) 69 enum mem_cgroup_stat_index idx, int val)
70 { 70 {
71 int cpu = smp_processor_id(); 71 int cpu = smp_processor_id();
72 stat->cpustat[cpu].count[idx] += val; 72 stat->cpustat[cpu].count[idx] += val;
73 } 73 }
74 74
75 static s64 mem_cgroup_read_stat(struct mem_cgroup_stat *stat, 75 static s64 mem_cgroup_read_stat(struct mem_cgroup_stat *stat,
76 enum mem_cgroup_stat_index idx) 76 enum mem_cgroup_stat_index idx)
77 { 77 {
78 int cpu; 78 int cpu;
79 s64 ret = 0; 79 s64 ret = 0;
80 for_each_possible_cpu(cpu) 80 for_each_possible_cpu(cpu)
81 ret += stat->cpustat[cpu].count[idx]; 81 ret += stat->cpustat[cpu].count[idx];
82 return ret; 82 return ret;
83 } 83 }
84 84
85 /* 85 /*
86 * per-zone information in memory controller. 86 * per-zone information in memory controller.
87 */ 87 */
88 88
89 enum mem_cgroup_zstat_index { 89 enum mem_cgroup_zstat_index {
90 MEM_CGROUP_ZSTAT_ACTIVE, 90 MEM_CGROUP_ZSTAT_ACTIVE,
91 MEM_CGROUP_ZSTAT_INACTIVE, 91 MEM_CGROUP_ZSTAT_INACTIVE,
92 92
93 NR_MEM_CGROUP_ZSTAT, 93 NR_MEM_CGROUP_ZSTAT,
94 }; 94 };
95 95
96 struct mem_cgroup_per_zone { 96 struct mem_cgroup_per_zone {
97 /* 97 /*
98 * spin_lock to protect the per cgroup LRU 98 * spin_lock to protect the per cgroup LRU
99 */ 99 */
100 spinlock_t lru_lock; 100 spinlock_t lru_lock;
101 struct list_head active_list; 101 struct list_head active_list;
102 struct list_head inactive_list; 102 struct list_head inactive_list;
103 unsigned long count[NR_MEM_CGROUP_ZSTAT]; 103 unsigned long count[NR_MEM_CGROUP_ZSTAT];
104 }; 104 };
105 /* Macro for accessing counter */ 105 /* Macro for accessing counter */
106 #define MEM_CGROUP_ZSTAT(mz, idx) ((mz)->count[(idx)]) 106 #define MEM_CGROUP_ZSTAT(mz, idx) ((mz)->count[(idx)])
107 107
108 struct mem_cgroup_per_node { 108 struct mem_cgroup_per_node {
109 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES]; 109 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
110 }; 110 };
111 111
112 struct mem_cgroup_lru_info { 112 struct mem_cgroup_lru_info {
113 struct mem_cgroup_per_node *nodeinfo[MAX_NUMNODES]; 113 struct mem_cgroup_per_node *nodeinfo[MAX_NUMNODES];
114 }; 114 };
115 115
116 /* 116 /*
117 * The memory controller data structure. The memory controller controls both 117 * The memory controller data structure. The memory controller controls both
118 * page cache and RSS per cgroup. We would eventually like to provide 118 * page cache and RSS per cgroup. We would eventually like to provide
119 * statistics based on the statistics developed by Rik Van Riel for clock-pro, 119 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
120 * to help the administrator determine what knobs to tune. 120 * to help the administrator determine what knobs to tune.
121 * 121 *
122 * TODO: Add a water mark for the memory controller. Reclaim will begin when 122 * TODO: Add a water mark for the memory controller. Reclaim will begin when
123 * we hit the water mark. May be even add a low water mark, such that 123 * we hit the water mark. May be even add a low water mark, such that
124 * no reclaim occurs from a cgroup at it's low water mark, this is 124 * no reclaim occurs from a cgroup at it's low water mark, this is
125 * a feature that will be implemented much later in the future. 125 * a feature that will be implemented much later in the future.
126 */ 126 */
127 struct mem_cgroup { 127 struct mem_cgroup {
128 struct cgroup_subsys_state css; 128 struct cgroup_subsys_state css;
129 /* 129 /*
130 * the counter to account for memory usage 130 * the counter to account for memory usage
131 */ 131 */
132 struct res_counter res; 132 struct res_counter res;
133 /* 133 /*
134 * Per cgroup active and inactive list, similar to the 134 * Per cgroup active and inactive list, similar to the
135 * per zone LRU lists. 135 * per zone LRU lists.
136 */ 136 */
137 struct mem_cgroup_lru_info info; 137 struct mem_cgroup_lru_info info;
138 138
139 int prev_priority; /* for recording reclaim priority */ 139 int prev_priority; /* for recording reclaim priority */
140 /* 140 /*
141 * statistics. 141 * statistics.
142 */ 142 */
143 struct mem_cgroup_stat stat; 143 struct mem_cgroup_stat stat;
144 }; 144 };
145 static struct mem_cgroup init_mem_cgroup; 145 static struct mem_cgroup init_mem_cgroup;
146 146
147 /* 147 /*
148 * We use the lower bit of the page->page_cgroup pointer as a bit spin 148 * We use the lower bit of the page->page_cgroup pointer as a bit spin
149 * lock. We need to ensure that page->page_cgroup is at least two 149 * lock. We need to ensure that page->page_cgroup is at least two
150 * byte aligned (based on comments from Nick Piggin). But since 150 * byte aligned (based on comments from Nick Piggin). But since
151 * bit_spin_lock doesn't actually set that lock bit in a non-debug 151 * bit_spin_lock doesn't actually set that lock bit in a non-debug
152 * uniprocessor kernel, we should avoid setting it here too. 152 * uniprocessor kernel, we should avoid setting it here too.
153 */ 153 */
154 #define PAGE_CGROUP_LOCK_BIT 0x0 154 #define PAGE_CGROUP_LOCK_BIT 0x0
155 #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) 155 #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
156 #define PAGE_CGROUP_LOCK (1 << PAGE_CGROUP_LOCK_BIT) 156 #define PAGE_CGROUP_LOCK (1 << PAGE_CGROUP_LOCK_BIT)
157 #else 157 #else
158 #define PAGE_CGROUP_LOCK 0x0 158 #define PAGE_CGROUP_LOCK 0x0
159 #endif 159 #endif
160 160
161 /* 161 /*
162 * A page_cgroup page is associated with every page descriptor. The 162 * A page_cgroup page is associated with every page descriptor. The
163 * page_cgroup helps us identify information about the cgroup 163 * page_cgroup helps us identify information about the cgroup
164 */ 164 */
165 struct page_cgroup { 165 struct page_cgroup {
166 struct list_head lru; /* per cgroup LRU list */ 166 struct list_head lru; /* per cgroup LRU list */
167 struct page *page; 167 struct page *page;
168 struct mem_cgroup *mem_cgroup; 168 struct mem_cgroup *mem_cgroup;
169 int flags; 169 int flags;
170 }; 170 };
171 #define PAGE_CGROUP_FLAG_CACHE (0x1) /* charged as cache */ 171 #define PAGE_CGROUP_FLAG_CACHE (0x1) /* charged as cache */
172 #define PAGE_CGROUP_FLAG_ACTIVE (0x2) /* page is active in this cgroup */ 172 #define PAGE_CGROUP_FLAG_ACTIVE (0x2) /* page is active in this cgroup */
173 173
174 static int page_cgroup_nid(struct page_cgroup *pc) 174 static int page_cgroup_nid(struct page_cgroup *pc)
175 { 175 {
176 return page_to_nid(pc->page); 176 return page_to_nid(pc->page);
177 } 177 }
178 178
179 static enum zone_type page_cgroup_zid(struct page_cgroup *pc) 179 static enum zone_type page_cgroup_zid(struct page_cgroup *pc)
180 { 180 {
181 return page_zonenum(pc->page); 181 return page_zonenum(pc->page);
182 } 182 }
183 183
184 enum charge_type { 184 enum charge_type {
185 MEM_CGROUP_CHARGE_TYPE_CACHE = 0, 185 MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
186 MEM_CGROUP_CHARGE_TYPE_MAPPED, 186 MEM_CGROUP_CHARGE_TYPE_MAPPED,
187 MEM_CGROUP_CHARGE_TYPE_FORCE, /* used by force_empty */ 187 MEM_CGROUP_CHARGE_TYPE_FORCE, /* used by force_empty */
188 }; 188 };
189 189
190 /* 190 /*
191 * Always modified under lru lock. Then, not necessary to preempt_disable() 191 * Always modified under lru lock. Then, not necessary to preempt_disable()
192 */ 192 */
193 static void mem_cgroup_charge_statistics(struct mem_cgroup *mem, int flags, 193 static void mem_cgroup_charge_statistics(struct mem_cgroup *mem, int flags,
194 bool charge) 194 bool charge)
195 { 195 {
196 int val = (charge)? 1 : -1; 196 int val = (charge)? 1 : -1;
197 struct mem_cgroup_stat *stat = &mem->stat; 197 struct mem_cgroup_stat *stat = &mem->stat;
198 198
199 VM_BUG_ON(!irqs_disabled()); 199 VM_BUG_ON(!irqs_disabled());
200 if (flags & PAGE_CGROUP_FLAG_CACHE) 200 if (flags & PAGE_CGROUP_FLAG_CACHE)
201 __mem_cgroup_stat_add_safe(stat, MEM_CGROUP_STAT_CACHE, val); 201 __mem_cgroup_stat_add_safe(stat, MEM_CGROUP_STAT_CACHE, val);
202 else 202 else
203 __mem_cgroup_stat_add_safe(stat, MEM_CGROUP_STAT_RSS, val); 203 __mem_cgroup_stat_add_safe(stat, MEM_CGROUP_STAT_RSS, val);
204 204
205 if (charge) 205 if (charge)
206 __mem_cgroup_stat_add_safe(stat, 206 __mem_cgroup_stat_add_safe(stat,
207 MEM_CGROUP_STAT_PGPGIN_COUNT, 1); 207 MEM_CGROUP_STAT_PGPGIN_COUNT, 1);
208 else 208 else
209 __mem_cgroup_stat_add_safe(stat, 209 __mem_cgroup_stat_add_safe(stat,
210 MEM_CGROUP_STAT_PGPGOUT_COUNT, 1); 210 MEM_CGROUP_STAT_PGPGOUT_COUNT, 1);
211 } 211 }
212 212
213 static struct mem_cgroup_per_zone * 213 static struct mem_cgroup_per_zone *
214 mem_cgroup_zoneinfo(struct mem_cgroup *mem, int nid, int zid) 214 mem_cgroup_zoneinfo(struct mem_cgroup *mem, int nid, int zid)
215 { 215 {
216 return &mem->info.nodeinfo[nid]->zoneinfo[zid]; 216 return &mem->info.nodeinfo[nid]->zoneinfo[zid];
217 } 217 }
218 218
219 static struct mem_cgroup_per_zone * 219 static struct mem_cgroup_per_zone *
220 page_cgroup_zoneinfo(struct page_cgroup *pc) 220 page_cgroup_zoneinfo(struct page_cgroup *pc)
221 { 221 {
222 struct mem_cgroup *mem = pc->mem_cgroup; 222 struct mem_cgroup *mem = pc->mem_cgroup;
223 int nid = page_cgroup_nid(pc); 223 int nid = page_cgroup_nid(pc);
224 int zid = page_cgroup_zid(pc); 224 int zid = page_cgroup_zid(pc);
225 225
226 return mem_cgroup_zoneinfo(mem, nid, zid); 226 return mem_cgroup_zoneinfo(mem, nid, zid);
227 } 227 }
228 228
229 static unsigned long mem_cgroup_get_all_zonestat(struct mem_cgroup *mem, 229 static unsigned long mem_cgroup_get_all_zonestat(struct mem_cgroup *mem,
230 enum mem_cgroup_zstat_index idx) 230 enum mem_cgroup_zstat_index idx)
231 { 231 {
232 int nid, zid; 232 int nid, zid;
233 struct mem_cgroup_per_zone *mz; 233 struct mem_cgroup_per_zone *mz;
234 u64 total = 0; 234 u64 total = 0;
235 235
236 for_each_online_node(nid) 236 for_each_online_node(nid)
237 for (zid = 0; zid < MAX_NR_ZONES; zid++) { 237 for (zid = 0; zid < MAX_NR_ZONES; zid++) {
238 mz = mem_cgroup_zoneinfo(mem, nid, zid); 238 mz = mem_cgroup_zoneinfo(mem, nid, zid);
239 total += MEM_CGROUP_ZSTAT(mz, idx); 239 total += MEM_CGROUP_ZSTAT(mz, idx);
240 } 240 }
241 return total; 241 return total;
242 } 242 }
243 243
244 static struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont) 244 static struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
245 { 245 {
246 return container_of(cgroup_subsys_state(cont, 246 return container_of(cgroup_subsys_state(cont,
247 mem_cgroup_subsys_id), struct mem_cgroup, 247 mem_cgroup_subsys_id), struct mem_cgroup,
248 css); 248 css);
249 } 249 }
250 250
251 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p) 251 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
252 { 252 {
253 return container_of(task_subsys_state(p, mem_cgroup_subsys_id), 253 return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
254 struct mem_cgroup, css); 254 struct mem_cgroup, css);
255 } 255 }
256 256
257 static inline int page_cgroup_locked(struct page *page) 257 static inline int page_cgroup_locked(struct page *page)
258 { 258 {
259 return bit_spin_is_locked(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup); 259 return bit_spin_is_locked(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
260 } 260 }
261 261
262 static void page_assign_page_cgroup(struct page *page, struct page_cgroup *pc) 262 static void page_assign_page_cgroup(struct page *page, struct page_cgroup *pc)
263 { 263 {
264 VM_BUG_ON(!page_cgroup_locked(page)); 264 VM_BUG_ON(!page_cgroup_locked(page));
265 page->page_cgroup = ((unsigned long)pc | PAGE_CGROUP_LOCK); 265 page->page_cgroup = ((unsigned long)pc | PAGE_CGROUP_LOCK);
266 } 266 }
267 267
268 struct page_cgroup *page_get_page_cgroup(struct page *page) 268 struct page_cgroup *page_get_page_cgroup(struct page *page)
269 { 269 {
270 return (struct page_cgroup *) (page->page_cgroup & ~PAGE_CGROUP_LOCK); 270 return (struct page_cgroup *) (page->page_cgroup & ~PAGE_CGROUP_LOCK);
271 } 271 }
272 272
273 static void lock_page_cgroup(struct page *page) 273 static void lock_page_cgroup(struct page *page)
274 { 274 {
275 bit_spin_lock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup); 275 bit_spin_lock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
276 } 276 }
277 277
278 static int try_lock_page_cgroup(struct page *page) 278 static int try_lock_page_cgroup(struct page *page)
279 { 279 {
280 return bit_spin_trylock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup); 280 return bit_spin_trylock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
281 } 281 }
282 282
283 static void unlock_page_cgroup(struct page *page) 283 static void unlock_page_cgroup(struct page *page)
284 { 284 {
285 bit_spin_unlock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup); 285 bit_spin_unlock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
286 } 286 }
287 287
288 static void __mem_cgroup_remove_list(struct mem_cgroup_per_zone *mz, 288 static void __mem_cgroup_remove_list(struct mem_cgroup_per_zone *mz,
289 struct page_cgroup *pc) 289 struct page_cgroup *pc)
290 { 290 {
291 int from = pc->flags & PAGE_CGROUP_FLAG_ACTIVE; 291 int from = pc->flags & PAGE_CGROUP_FLAG_ACTIVE;
292 292
293 if (from) 293 if (from)
294 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) -= 1; 294 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) -= 1;
295 else 295 else
296 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) -= 1; 296 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) -= 1;
297 297
298 mem_cgroup_charge_statistics(pc->mem_cgroup, pc->flags, false); 298 mem_cgroup_charge_statistics(pc->mem_cgroup, pc->flags, false);
299 list_del(&pc->lru); 299 list_del(&pc->lru);
300 } 300 }
301 301
302 static void __mem_cgroup_add_list(struct mem_cgroup_per_zone *mz, 302 static void __mem_cgroup_add_list(struct mem_cgroup_per_zone *mz,
303 struct page_cgroup *pc) 303 struct page_cgroup *pc)
304 { 304 {
305 int to = pc->flags & PAGE_CGROUP_FLAG_ACTIVE; 305 int to = pc->flags & PAGE_CGROUP_FLAG_ACTIVE;
306 306
307 if (!to) { 307 if (!to) {
308 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) += 1; 308 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) += 1;
309 list_add(&pc->lru, &mz->inactive_list); 309 list_add(&pc->lru, &mz->inactive_list);
310 } else { 310 } else {
311 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) += 1; 311 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) += 1;
312 list_add(&pc->lru, &mz->active_list); 312 list_add(&pc->lru, &mz->active_list);
313 } 313 }
314 mem_cgroup_charge_statistics(pc->mem_cgroup, pc->flags, true); 314 mem_cgroup_charge_statistics(pc->mem_cgroup, pc->flags, true);
315 } 315 }
316 316
317 static void __mem_cgroup_move_lists(struct page_cgroup *pc, bool active) 317 static void __mem_cgroup_move_lists(struct page_cgroup *pc, bool active)
318 { 318 {
319 int from = pc->flags & PAGE_CGROUP_FLAG_ACTIVE; 319 int from = pc->flags & PAGE_CGROUP_FLAG_ACTIVE;
320 struct mem_cgroup_per_zone *mz = page_cgroup_zoneinfo(pc); 320 struct mem_cgroup_per_zone *mz = page_cgroup_zoneinfo(pc);
321 321
322 if (from) 322 if (from)
323 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) -= 1; 323 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) -= 1;
324 else 324 else
325 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) -= 1; 325 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) -= 1;
326 326
327 if (active) { 327 if (active) {
328 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) += 1; 328 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) += 1;
329 pc->flags |= PAGE_CGROUP_FLAG_ACTIVE; 329 pc->flags |= PAGE_CGROUP_FLAG_ACTIVE;
330 list_move(&pc->lru, &mz->active_list); 330 list_move(&pc->lru, &mz->active_list);
331 } else { 331 } else {
332 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) += 1; 332 MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) += 1;
333 pc->flags &= ~PAGE_CGROUP_FLAG_ACTIVE; 333 pc->flags &= ~PAGE_CGROUP_FLAG_ACTIVE;
334 list_move(&pc->lru, &mz->inactive_list); 334 list_move(&pc->lru, &mz->inactive_list);
335 } 335 }
336 } 336 }
337 337
338 int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *mem) 338 int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *mem)
339 { 339 {
340 int ret; 340 int ret;
341 341
342 task_lock(task); 342 task_lock(task);
343 ret = task->mm && mm_match_cgroup(task->mm, mem); 343 ret = task->mm && mm_match_cgroup(task->mm, mem);
344 task_unlock(task); 344 task_unlock(task);
345 return ret; 345 return ret;
346 } 346 }
347 347
348 /* 348 /*
349 * This routine assumes that the appropriate zone's lru lock is already held 349 * This routine assumes that the appropriate zone's lru lock is already held
350 */ 350 */
351 void mem_cgroup_move_lists(struct page *page, bool active) 351 void mem_cgroup_move_lists(struct page *page, bool active)
352 { 352 {
353 struct page_cgroup *pc; 353 struct page_cgroup *pc;
354 struct mem_cgroup_per_zone *mz; 354 struct mem_cgroup_per_zone *mz;
355 unsigned long flags; 355 unsigned long flags;
356 356
357 if (mem_cgroup_subsys.disabled) 357 if (mem_cgroup_subsys.disabled)
358 return; 358 return;
359 359
360 /* 360 /*
361 * We cannot lock_page_cgroup while holding zone's lru_lock, 361 * We cannot lock_page_cgroup while holding zone's lru_lock,
362 * because other holders of lock_page_cgroup can be interrupted 362 * because other holders of lock_page_cgroup can be interrupted
363 * with an attempt to rotate_reclaimable_page. But we cannot 363 * with an attempt to rotate_reclaimable_page. But we cannot
364 * safely get to page_cgroup without it, so just try_lock it: 364 * safely get to page_cgroup without it, so just try_lock it:
365 * mem_cgroup_isolate_pages allows for page left on wrong list. 365 * mem_cgroup_isolate_pages allows for page left on wrong list.
366 */ 366 */
367 if (!try_lock_page_cgroup(page)) 367 if (!try_lock_page_cgroup(page))
368 return; 368 return;
369 369
370 pc = page_get_page_cgroup(page); 370 pc = page_get_page_cgroup(page);
371 if (pc) { 371 if (pc) {
372 mz = page_cgroup_zoneinfo(pc); 372 mz = page_cgroup_zoneinfo(pc);
373 spin_lock_irqsave(&mz->lru_lock, flags); 373 spin_lock_irqsave(&mz->lru_lock, flags);
374 __mem_cgroup_move_lists(pc, active); 374 __mem_cgroup_move_lists(pc, active);
375 spin_unlock_irqrestore(&mz->lru_lock, flags); 375 spin_unlock_irqrestore(&mz->lru_lock, flags);
376 } 376 }
377 unlock_page_cgroup(page); 377 unlock_page_cgroup(page);
378 } 378 }
379 379
380 /* 380 /*
381 * Calculate mapped_ratio under memory controller. This will be used in 381 * Calculate mapped_ratio under memory controller. This will be used in
382 * vmscan.c for deteremining we have to reclaim mapped pages. 382 * vmscan.c for deteremining we have to reclaim mapped pages.
383 */ 383 */
384 int mem_cgroup_calc_mapped_ratio(struct mem_cgroup *mem) 384 int mem_cgroup_calc_mapped_ratio(struct mem_cgroup *mem)
385 { 385 {
386 long total, rss; 386 long total, rss;
387 387
388 /* 388 /*
389 * usage is recorded in bytes. But, here, we assume the number of 389 * usage is recorded in bytes. But, here, we assume the number of
390 * physical pages can be represented by "long" on any arch. 390 * physical pages can be represented by "long" on any arch.
391 */ 391 */
392 total = (long) (mem->res.usage >> PAGE_SHIFT) + 1L; 392 total = (long) (mem->res.usage >> PAGE_SHIFT) + 1L;
393 rss = (long)mem_cgroup_read_stat(&mem->stat, MEM_CGROUP_STAT_RSS); 393 rss = (long)mem_cgroup_read_stat(&mem->stat, MEM_CGROUP_STAT_RSS);
394 return (int)((rss * 100L) / total); 394 return (int)((rss * 100L) / total);
395 } 395 }
396 396
397 /* 397 /*
398 * This function is called from vmscan.c. In page reclaiming loop. balance 398 * This function is called from vmscan.c. In page reclaiming loop. balance
399 * between active and inactive list is calculated. For memory controller 399 * between active and inactive list is calculated. For memory controller
400 * page reclaiming, we should use using mem_cgroup's imbalance rather than 400 * page reclaiming, we should use using mem_cgroup's imbalance rather than
401 * zone's global lru imbalance. 401 * zone's global lru imbalance.
402 */ 402 */
403 long mem_cgroup_reclaim_imbalance(struct mem_cgroup *mem) 403 long mem_cgroup_reclaim_imbalance(struct mem_cgroup *mem)
404 { 404 {
405 unsigned long active, inactive; 405 unsigned long active, inactive;
406 /* active and inactive are the number of pages. 'long' is ok.*/ 406 /* active and inactive are the number of pages. 'long' is ok.*/
407 active = mem_cgroup_get_all_zonestat(mem, MEM_CGROUP_ZSTAT_ACTIVE); 407 active = mem_cgroup_get_all_zonestat(mem, MEM_CGROUP_ZSTAT_ACTIVE);
408 inactive = mem_cgroup_get_all_zonestat(mem, MEM_CGROUP_ZSTAT_INACTIVE); 408 inactive = mem_cgroup_get_all_zonestat(mem, MEM_CGROUP_ZSTAT_INACTIVE);
409 return (long) (active / (inactive + 1)); 409 return (long) (active / (inactive + 1));
410 } 410 }
411 411
412 /* 412 /*
413 * prev_priority control...this will be used in memory reclaim path. 413 * prev_priority control...this will be used in memory reclaim path.
414 */ 414 */
415 int mem_cgroup_get_reclaim_priority(struct mem_cgroup *mem) 415 int mem_cgroup_get_reclaim_priority(struct mem_cgroup *mem)
416 { 416 {
417 return mem->prev_priority; 417 return mem->prev_priority;
418 } 418 }
419 419
420 void mem_cgroup_note_reclaim_priority(struct mem_cgroup *mem, int priority) 420 void mem_cgroup_note_reclaim_priority(struct mem_cgroup *mem, int priority)
421 { 421 {
422 if (priority < mem->prev_priority) 422 if (priority < mem->prev_priority)
423 mem->prev_priority = priority; 423 mem->prev_priority = priority;
424 } 424 }
425 425
426 void mem_cgroup_record_reclaim_priority(struct mem_cgroup *mem, int priority) 426 void mem_cgroup_record_reclaim_priority(struct mem_cgroup *mem, int priority)
427 { 427 {
428 mem->prev_priority = priority; 428 mem->prev_priority = priority;
429 } 429 }
430 430
431 /* 431 /*
432 * Calculate # of pages to be scanned in this priority/zone. 432 * Calculate # of pages to be scanned in this priority/zone.
433 * See also vmscan.c 433 * See also vmscan.c
434 * 434 *
435 * priority starts from "DEF_PRIORITY" and decremented in each loop. 435 * priority starts from "DEF_PRIORITY" and decremented in each loop.
436 * (see include/linux/mmzone.h) 436 * (see include/linux/mmzone.h)
437 */ 437 */
438 438
439 long mem_cgroup_calc_reclaim_active(struct mem_cgroup *mem, 439 long mem_cgroup_calc_reclaim_active(struct mem_cgroup *mem,
440 struct zone *zone, int priority) 440 struct zone *zone, int priority)
441 { 441 {
442 long nr_active; 442 long nr_active;
443 int nid = zone->zone_pgdat->node_id; 443 int nid = zone->zone_pgdat->node_id;
444 int zid = zone_idx(zone); 444 int zid = zone_idx(zone);
445 struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(mem, nid, zid); 445 struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(mem, nid, zid);
446 446
447 nr_active = MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE); 447 nr_active = MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE);
448 return (nr_active >> priority); 448 return (nr_active >> priority);
449 } 449 }
450 450
451 long mem_cgroup_calc_reclaim_inactive(struct mem_cgroup *mem, 451 long mem_cgroup_calc_reclaim_inactive(struct mem_cgroup *mem,
452 struct zone *zone, int priority) 452 struct zone *zone, int priority)
453 { 453 {
454 long nr_inactive; 454 long nr_inactive;
455 int nid = zone->zone_pgdat->node_id; 455 int nid = zone->zone_pgdat->node_id;
456 int zid = zone_idx(zone); 456 int zid = zone_idx(zone);
457 struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(mem, nid, zid); 457 struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(mem, nid, zid);
458 458
459 nr_inactive = MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE); 459 nr_inactive = MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE);
460 return (nr_inactive >> priority); 460 return (nr_inactive >> priority);
461 } 461 }
462 462
463 unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan, 463 unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
464 struct list_head *dst, 464 struct list_head *dst,
465 unsigned long *scanned, int order, 465 unsigned long *scanned, int order,
466 int mode, struct zone *z, 466 int mode, struct zone *z,
467 struct mem_cgroup *mem_cont, 467 struct mem_cgroup *mem_cont,
468 int active) 468 int active)
469 { 469 {
470 unsigned long nr_taken = 0; 470 unsigned long nr_taken = 0;
471 struct page *page; 471 struct page *page;
472 unsigned long scan; 472 unsigned long scan;
473 LIST_HEAD(pc_list); 473 LIST_HEAD(pc_list);
474 struct list_head *src; 474 struct list_head *src;
475 struct page_cgroup *pc, *tmp; 475 struct page_cgroup *pc, *tmp;
476 int nid = z->zone_pgdat->node_id; 476 int nid = z->zone_pgdat->node_id;
477 int zid = zone_idx(z); 477 int zid = zone_idx(z);
478 struct mem_cgroup_per_zone *mz; 478 struct mem_cgroup_per_zone *mz;
479 479
480 BUG_ON(!mem_cont); 480 BUG_ON(!mem_cont);
481 mz = mem_cgroup_zoneinfo(mem_cont, nid, zid); 481 mz = mem_cgroup_zoneinfo(mem_cont, nid, zid);
482 if (active) 482 if (active)
483 src = &mz->active_list; 483 src = &mz->active_list;
484 else 484 else
485 src = &mz->inactive_list; 485 src = &mz->inactive_list;
486 486
487 487
488 spin_lock(&mz->lru_lock); 488 spin_lock(&mz->lru_lock);
489 scan = 0; 489 scan = 0;
490 list_for_each_entry_safe_reverse(pc, tmp, src, lru) { 490 list_for_each_entry_safe_reverse(pc, tmp, src, lru) {
491 if (scan >= nr_to_scan) 491 if (scan >= nr_to_scan)
492 break; 492 break;
493 page = pc->page; 493 page = pc->page;
494 494
495 if (unlikely(!PageLRU(page))) 495 if (unlikely(!PageLRU(page)))
496 continue; 496 continue;
497 497
498 if (PageActive(page) && !active) { 498 if (PageActive(page) && !active) {
499 __mem_cgroup_move_lists(pc, true); 499 __mem_cgroup_move_lists(pc, true);
500 continue; 500 continue;
501 } 501 }
502 if (!PageActive(page) && active) { 502 if (!PageActive(page) && active) {
503 __mem_cgroup_move_lists(pc, false); 503 __mem_cgroup_move_lists(pc, false);
504 continue; 504 continue;
505 } 505 }
506 506
507 scan++; 507 scan++;
508 list_move(&pc->lru, &pc_list); 508 list_move(&pc->lru, &pc_list);
509 509
510 if (__isolate_lru_page(page, mode) == 0) { 510 if (__isolate_lru_page(page, mode) == 0) {
511 list_move(&page->lru, dst); 511 list_move(&page->lru, dst);
512 nr_taken++; 512 nr_taken++;
513 } 513 }
514 } 514 }
515 515
516 list_splice(&pc_list, src); 516 list_splice(&pc_list, src);
517 spin_unlock(&mz->lru_lock); 517 spin_unlock(&mz->lru_lock);
518 518
519 *scanned = scan; 519 *scanned = scan;
520 return nr_taken; 520 return nr_taken;
521 } 521 }
522 522
523 /* 523 /*
524 * Charge the memory controller for page usage. 524 * Charge the memory controller for page usage.
525 * Return 525 * Return
526 * 0 if the charge was successful 526 * 0 if the charge was successful
527 * < 0 if the cgroup is over its limit 527 * < 0 if the cgroup is over its limit
528 */ 528 */
529 static int mem_cgroup_charge_common(struct page *page, struct mm_struct *mm, 529 static int mem_cgroup_charge_common(struct page *page, struct mm_struct *mm,
530 gfp_t gfp_mask, enum charge_type ctype, 530 gfp_t gfp_mask, enum charge_type ctype,
531 struct mem_cgroup *memcg) 531 struct mem_cgroup *memcg)
532 { 532 {
533 struct mem_cgroup *mem; 533 struct mem_cgroup *mem;
534 struct page_cgroup *pc; 534 struct page_cgroup *pc;
535 unsigned long flags; 535 unsigned long flags;
536 unsigned long nr_retries = MEM_CGROUP_RECLAIM_RETRIES; 536 unsigned long nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
537 struct mem_cgroup_per_zone *mz; 537 struct mem_cgroup_per_zone *mz;
538 538
539 pc = kmem_cache_alloc(page_cgroup_cache, gfp_mask); 539 pc = kmem_cache_alloc(page_cgroup_cache, gfp_mask);
540 if (unlikely(pc == NULL)) 540 if (unlikely(pc == NULL))
541 goto err; 541 goto err;
542 542
543 /* 543 /*
544 * We always charge the cgroup the mm_struct belongs to. 544 * We always charge the cgroup the mm_struct belongs to.
545 * The mm_struct's mem_cgroup changes on task migration if the 545 * The mm_struct's mem_cgroup changes on task migration if the
546 * thread group leader migrates. It's possible that mm is not 546 * thread group leader migrates. It's possible that mm is not
547 * set, if so charge the init_mm (happens for pagecache usage). 547 * set, if so charge the init_mm (happens for pagecache usage).
548 */ 548 */
549 if (likely(!memcg)) { 549 if (likely(!memcg)) {
550 rcu_read_lock(); 550 rcu_read_lock();
551 mem = mem_cgroup_from_task(rcu_dereference(mm->owner)); 551 mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
552 /* 552 /*
553 * For every charge from the cgroup, increment reference count 553 * For every charge from the cgroup, increment reference count
554 */ 554 */
555 css_get(&mem->css); 555 css_get(&mem->css);
556 rcu_read_unlock(); 556 rcu_read_unlock();
557 } else { 557 } else {
558 mem = memcg; 558 mem = memcg;
559 css_get(&memcg->css); 559 css_get(&memcg->css);
560 } 560 }
561 561
562 while (res_counter_charge(&mem->res, PAGE_SIZE)) { 562 while (res_counter_charge(&mem->res, PAGE_SIZE)) {
563 if (!(gfp_mask & __GFP_WAIT)) 563 if (!(gfp_mask & __GFP_WAIT))
564 goto out; 564 goto out;
565 565
566 if (try_to_free_mem_cgroup_pages(mem, gfp_mask)) 566 if (try_to_free_mem_cgroup_pages(mem, gfp_mask))
567 continue; 567 continue;
568 568
569 /* 569 /*
570 * try_to_free_mem_cgroup_pages() might not give us a full 570 * try_to_free_mem_cgroup_pages() might not give us a full
571 * picture of reclaim. Some pages are reclaimed and might be 571 * picture of reclaim. Some pages are reclaimed and might be
572 * moved to swap cache or just unmapped from the cgroup. 572 * moved to swap cache or just unmapped from the cgroup.
573 * Check the limit again to see if the reclaim reduced the 573 * Check the limit again to see if the reclaim reduced the
574 * current usage of the cgroup before giving up 574 * current usage of the cgroup before giving up
575 */ 575 */
576 if (res_counter_check_under_limit(&mem->res)) 576 if (res_counter_check_under_limit(&mem->res))
577 continue; 577 continue;
578 578
579 if (!nr_retries--) { 579 if (!nr_retries--) {
580 mem_cgroup_out_of_memory(mem, gfp_mask); 580 mem_cgroup_out_of_memory(mem, gfp_mask);
581 goto out; 581 goto out;
582 } 582 }
583 } 583 }
584 584
585 pc->mem_cgroup = mem; 585 pc->mem_cgroup = mem;
586 pc->page = page; 586 pc->page = page;
587 /* 587 /*
588 * If a page is accounted as a page cache, insert to inactive list. 588 * If a page is accounted as a page cache, insert to inactive list.
589 * If anon, insert to active list. 589 * If anon, insert to active list.
590 */ 590 */
591 if (ctype == MEM_CGROUP_CHARGE_TYPE_CACHE) 591 if (ctype == MEM_CGROUP_CHARGE_TYPE_CACHE)
592 pc->flags = PAGE_CGROUP_FLAG_CACHE; 592 pc->flags = PAGE_CGROUP_FLAG_CACHE;
593 else 593 else
594 pc->flags = PAGE_CGROUP_FLAG_ACTIVE; 594 pc->flags = PAGE_CGROUP_FLAG_ACTIVE;
595 595
596 lock_page_cgroup(page); 596 lock_page_cgroup(page);
597 if (unlikely(page_get_page_cgroup(page))) { 597 if (unlikely(page_get_page_cgroup(page))) {
598 unlock_page_cgroup(page); 598 unlock_page_cgroup(page);
599 res_counter_uncharge(&mem->res, PAGE_SIZE); 599 res_counter_uncharge(&mem->res, PAGE_SIZE);
600 css_put(&mem->css); 600 css_put(&mem->css);
601 kmem_cache_free(page_cgroup_cache, pc); 601 kmem_cache_free(page_cgroup_cache, pc);
602 goto done; 602 goto done;
603 } 603 }
604 page_assign_page_cgroup(page, pc); 604 page_assign_page_cgroup(page, pc);
605 605
606 mz = page_cgroup_zoneinfo(pc); 606 mz = page_cgroup_zoneinfo(pc);
607 spin_lock_irqsave(&mz->lru_lock, flags); 607 spin_lock_irqsave(&mz->lru_lock, flags);
608 __mem_cgroup_add_list(mz, pc); 608 __mem_cgroup_add_list(mz, pc);
609 spin_unlock_irqrestore(&mz->lru_lock, flags); 609 spin_unlock_irqrestore(&mz->lru_lock, flags);
610 610
611 unlock_page_cgroup(page); 611 unlock_page_cgroup(page);
612 done: 612 done:
613 return 0; 613 return 0;
614 out: 614 out:
615 css_put(&mem->css); 615 css_put(&mem->css);
616 kmem_cache_free(page_cgroup_cache, pc); 616 kmem_cache_free(page_cgroup_cache, pc);
617 err: 617 err:
618 return -ENOMEM; 618 return -ENOMEM;
619 } 619 }
620 620
621 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask) 621 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask)
622 { 622 {
623 if (mem_cgroup_subsys.disabled) 623 if (mem_cgroup_subsys.disabled)
624 return 0; 624 return 0;
625 625
626 /* 626 /*
627 * If already mapped, we don't have to account. 627 * If already mapped, we don't have to account.
628 * If page cache, page->mapping has address_space. 628 * If page cache, page->mapping has address_space.
629 * But page->mapping may have out-of-use anon_vma pointer, 629 * But page->mapping may have out-of-use anon_vma pointer,
630 * detecit it by PageAnon() check. newly-mapped-anon's page->mapping 630 * detecit it by PageAnon() check. newly-mapped-anon's page->mapping
631 * is NULL. 631 * is NULL.
632 */ 632 */
633 if (page_mapped(page) || (page->mapping && !PageAnon(page))) 633 if (page_mapped(page) || (page->mapping && !PageAnon(page)))
634 return 0; 634 return 0;
635 if (unlikely(!mm)) 635 if (unlikely(!mm))
636 mm = &init_mm; 636 mm = &init_mm;
637 return mem_cgroup_charge_common(page, mm, gfp_mask, 637 return mem_cgroup_charge_common(page, mm, gfp_mask,
638 MEM_CGROUP_CHARGE_TYPE_MAPPED, NULL); 638 MEM_CGROUP_CHARGE_TYPE_MAPPED, NULL);
639 } 639 }
640 640
641 int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm, 641 int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
642 gfp_t gfp_mask) 642 gfp_t gfp_mask)
643 { 643 {
644 if (mem_cgroup_subsys.disabled) 644 if (mem_cgroup_subsys.disabled)
645 return 0; 645 return 0;
646 646
647 /* 647 /*
648 * Corner case handling. This is called from add_to_page_cache() 648 * Corner case handling. This is called from add_to_page_cache()
649 * in usual. But some FS (shmem) precharges this page before calling it 649 * in usual. But some FS (shmem) precharges this page before calling it
650 * and call add_to_page_cache() with GFP_NOWAIT. 650 * and call add_to_page_cache() with GFP_NOWAIT.
651 * 651 *
652 * For GFP_NOWAIT case, the page may be pre-charged before calling 652 * For GFP_NOWAIT case, the page may be pre-charged before calling
653 * add_to_page_cache(). (See shmem.c) check it here and avoid to call 653 * add_to_page_cache(). (See shmem.c) check it here and avoid to call
654 * charge twice. (It works but has to pay a bit larger cost.) 654 * charge twice. (It works but has to pay a bit larger cost.)
655 */ 655 */
656 if (!(gfp_mask & __GFP_WAIT)) { 656 if (!(gfp_mask & __GFP_WAIT)) {
657 struct page_cgroup *pc; 657 struct page_cgroup *pc;
658 658
659 lock_page_cgroup(page); 659 lock_page_cgroup(page);
660 pc = page_get_page_cgroup(page); 660 pc = page_get_page_cgroup(page);
661 if (pc) { 661 if (pc) {
662 VM_BUG_ON(pc->page != page); 662 VM_BUG_ON(pc->page != page);
663 VM_BUG_ON(!pc->mem_cgroup); 663 VM_BUG_ON(!pc->mem_cgroup);
664 unlock_page_cgroup(page); 664 unlock_page_cgroup(page);
665 return 0; 665 return 0;
666 } 666 }
667 unlock_page_cgroup(page); 667 unlock_page_cgroup(page);
668 } 668 }
669 669
670 if (unlikely(!mm)) 670 if (unlikely(!mm))
671 mm = &init_mm; 671 mm = &init_mm;
672 672
673 return mem_cgroup_charge_common(page, mm, gfp_mask, 673 return mem_cgroup_charge_common(page, mm, gfp_mask,
674 MEM_CGROUP_CHARGE_TYPE_CACHE, NULL); 674 MEM_CGROUP_CHARGE_TYPE_CACHE, NULL);
675 } 675 }
676 676
677 /* 677 /*
678 * uncharge if !page_mapped(page) 678 * uncharge if !page_mapped(page)
679 */ 679 */
680 static void 680 static void
681 __mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype) 681 __mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
682 { 682 {
683 struct page_cgroup *pc; 683 struct page_cgroup *pc;
684 struct mem_cgroup *mem; 684 struct mem_cgroup *mem;
685 struct mem_cgroup_per_zone *mz; 685 struct mem_cgroup_per_zone *mz;
686 unsigned long flags; 686 unsigned long flags;
687 687
688 if (mem_cgroup_subsys.disabled) 688 if (mem_cgroup_subsys.disabled)
689 return; 689 return;
690 690
691 /* 691 /*
692 * Check if our page_cgroup is valid 692 * Check if our page_cgroup is valid
693 */ 693 */
694 lock_page_cgroup(page); 694 lock_page_cgroup(page);
695 pc = page_get_page_cgroup(page); 695 pc = page_get_page_cgroup(page);
696 if (unlikely(!pc)) 696 if (unlikely(!pc))
697 goto unlock; 697 goto unlock;
698 698
699 VM_BUG_ON(pc->page != page); 699 VM_BUG_ON(pc->page != page);
700 700
701 if ((ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED) 701 if ((ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED)
702 && ((pc->flags & PAGE_CGROUP_FLAG_CACHE) 702 && ((pc->flags & PAGE_CGROUP_FLAG_CACHE)
703 || page_mapped(page))) 703 || page_mapped(page)))
704 goto unlock; 704 goto unlock;
705 705
706 mz = page_cgroup_zoneinfo(pc); 706 mz = page_cgroup_zoneinfo(pc);
707 spin_lock_irqsave(&mz->lru_lock, flags); 707 spin_lock_irqsave(&mz->lru_lock, flags);
708 __mem_cgroup_remove_list(mz, pc); 708 __mem_cgroup_remove_list(mz, pc);
709 spin_unlock_irqrestore(&mz->lru_lock, flags); 709 spin_unlock_irqrestore(&mz->lru_lock, flags);
710 710
711 page_assign_page_cgroup(page, NULL); 711 page_assign_page_cgroup(page, NULL);
712 unlock_page_cgroup(page); 712 unlock_page_cgroup(page);
713 713
714 mem = pc->mem_cgroup; 714 mem = pc->mem_cgroup;
715 res_counter_uncharge(&mem->res, PAGE_SIZE); 715 res_counter_uncharge(&mem->res, PAGE_SIZE);
716 css_put(&mem->css); 716 css_put(&mem->css);
717 717
718 kmem_cache_free(page_cgroup_cache, pc); 718 kmem_cache_free(page_cgroup_cache, pc);
719 return; 719 return;
720 unlock: 720 unlock:
721 unlock_page_cgroup(page); 721 unlock_page_cgroup(page);
722 } 722 }
723 723
724 void mem_cgroup_uncharge_page(struct page *page) 724 void mem_cgroup_uncharge_page(struct page *page)
725 { 725 {
726 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_MAPPED); 726 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_MAPPED);
727 } 727 }
728 728
729 void mem_cgroup_uncharge_cache_page(struct page *page) 729 void mem_cgroup_uncharge_cache_page(struct page *page)
730 { 730 {
731 VM_BUG_ON(page_mapped(page)); 731 VM_BUG_ON(page_mapped(page));
732 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE); 732 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
733 } 733 }
734 734
735 /* 735 /*
736 * Before starting migration, account against new page. 736 * Before starting migration, account against new page.
737 */ 737 */
738 int mem_cgroup_prepare_migration(struct page *page, struct page *newpage) 738 int mem_cgroup_prepare_migration(struct page *page, struct page *newpage)
739 { 739 {
740 struct page_cgroup *pc; 740 struct page_cgroup *pc;
741 struct mem_cgroup *mem = NULL; 741 struct mem_cgroup *mem = NULL;
742 enum charge_type ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED; 742 enum charge_type ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED;
743 int ret = 0; 743 int ret = 0;
744 744
745 if (mem_cgroup_subsys.disabled) 745 if (mem_cgroup_subsys.disabled)
746 return 0; 746 return 0;
747 747
748 lock_page_cgroup(page); 748 lock_page_cgroup(page);
749 pc = page_get_page_cgroup(page); 749 pc = page_get_page_cgroup(page);
750 if (pc) { 750 if (pc) {
751 mem = pc->mem_cgroup; 751 mem = pc->mem_cgroup;
752 css_get(&mem->css); 752 css_get(&mem->css);
753 if (pc->flags & PAGE_CGROUP_FLAG_CACHE) 753 if (pc->flags & PAGE_CGROUP_FLAG_CACHE)
754 ctype = MEM_CGROUP_CHARGE_TYPE_CACHE; 754 ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
755 } 755 }
756 unlock_page_cgroup(page); 756 unlock_page_cgroup(page);
757 if (mem) { 757 if (mem) {
758 ret = mem_cgroup_charge_common(newpage, NULL, GFP_KERNEL, 758 ret = mem_cgroup_charge_common(newpage, NULL, GFP_KERNEL,
759 ctype, mem); 759 ctype, mem);
760 css_put(&mem->css); 760 css_put(&mem->css);
761 } 761 }
762 return ret; 762 return ret;
763 } 763 }
764 764
765 /* remove redundant charge if migration failed*/ 765 /* remove redundant charge if migration failed*/
766 void mem_cgroup_end_migration(struct page *newpage) 766 void mem_cgroup_end_migration(struct page *newpage)
767 { 767 {
768 /* 768 /*
769 * At success, page->mapping is not NULL. 769 * At success, page->mapping is not NULL.
770 * special rollback care is necessary when 770 * special rollback care is necessary when
771 * 1. at migration failure. (newpage->mapping is cleared in this case) 771 * 1. at migration failure. (newpage->mapping is cleared in this case)
772 * 2. the newpage was moved but not remapped again because the task 772 * 2. the newpage was moved but not remapped again because the task
773 * exits and the newpage is obsolete. In this case, the new page 773 * exits and the newpage is obsolete. In this case, the new page
774 * may be a swapcache. So, we just call mem_cgroup_uncharge_page() 774 * may be a swapcache. So, we just call mem_cgroup_uncharge_page()
775 * always for avoiding mess. The page_cgroup will be removed if 775 * always for avoiding mess. The page_cgroup will be removed if
776 * unnecessary. File cache pages is still on radix-tree. Don't 776 * unnecessary. File cache pages is still on radix-tree. Don't
777 * care it. 777 * care it.
778 */ 778 */
779 if (!newpage->mapping) 779 if (!newpage->mapping)
780 __mem_cgroup_uncharge_common(newpage, 780 __mem_cgroup_uncharge_common(newpage,
781 MEM_CGROUP_CHARGE_TYPE_FORCE); 781 MEM_CGROUP_CHARGE_TYPE_FORCE);
782 else if (PageAnon(newpage)) 782 else if (PageAnon(newpage))
783 mem_cgroup_uncharge_page(newpage); 783 mem_cgroup_uncharge_page(newpage);
784 } 784 }
785 785
786 /* 786 /*
787 * A call to try to shrink memory usage under specified resource controller. 787 * A call to try to shrink memory usage under specified resource controller.
788 * This is typically used for page reclaiming for shmem for reducing side 788 * This is typically used for page reclaiming for shmem for reducing side
789 * effect of page allocation from shmem, which is used by some mem_cgroup. 789 * effect of page allocation from shmem, which is used by some mem_cgroup.
790 */ 790 */
791 int mem_cgroup_shrink_usage(struct mm_struct *mm, gfp_t gfp_mask) 791 int mem_cgroup_shrink_usage(struct mm_struct *mm, gfp_t gfp_mask)
792 { 792 {
793 struct mem_cgroup *mem; 793 struct mem_cgroup *mem;
794 int progress = 0; 794 int progress = 0;
795 int retry = MEM_CGROUP_RECLAIM_RETRIES; 795 int retry = MEM_CGROUP_RECLAIM_RETRIES;
796 796
797 if (mem_cgroup_subsys.disabled) 797 if (mem_cgroup_subsys.disabled)
798 return 0; 798 return 0;
799 if (!mm)
800 return 0;
799 801
800 rcu_read_lock(); 802 rcu_read_lock();
801 mem = mem_cgroup_from_task(rcu_dereference(mm->owner)); 803 mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
802 css_get(&mem->css); 804 css_get(&mem->css);
803 rcu_read_unlock(); 805 rcu_read_unlock();
804 806
805 do { 807 do {
806 progress = try_to_free_mem_cgroup_pages(mem, gfp_mask); 808 progress = try_to_free_mem_cgroup_pages(mem, gfp_mask);
807 } while (!progress && --retry); 809 } while (!progress && --retry);
808 810
809 css_put(&mem->css); 811 css_put(&mem->css);
810 if (!retry) 812 if (!retry)
811 return -ENOMEM; 813 return -ENOMEM;
812 return 0; 814 return 0;
813 } 815 }
814 816
815 int mem_cgroup_resize_limit(struct mem_cgroup *memcg, unsigned long long val) 817 int mem_cgroup_resize_limit(struct mem_cgroup *memcg, unsigned long long val)
816 { 818 {
817 819
818 int retry_count = MEM_CGROUP_RECLAIM_RETRIES; 820 int retry_count = MEM_CGROUP_RECLAIM_RETRIES;
819 int progress; 821 int progress;
820 int ret = 0; 822 int ret = 0;
821 823
822 while (res_counter_set_limit(&memcg->res, val)) { 824 while (res_counter_set_limit(&memcg->res, val)) {
823 if (signal_pending(current)) { 825 if (signal_pending(current)) {
824 ret = -EINTR; 826 ret = -EINTR;
825 break; 827 break;
826 } 828 }
827 if (!retry_count) { 829 if (!retry_count) {
828 ret = -EBUSY; 830 ret = -EBUSY;
829 break; 831 break;
830 } 832 }
831 progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL); 833 progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL);
832 if (!progress) 834 if (!progress)
833 retry_count--; 835 retry_count--;
834 } 836 }
835 return ret; 837 return ret;
836 } 838 }
837 839
838 840
839 /* 841 /*
840 * This routine traverse page_cgroup in given list and drop them all. 842 * This routine traverse page_cgroup in given list and drop them all.
841 * *And* this routine doesn't reclaim page itself, just removes page_cgroup. 843 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
842 */ 844 */
843 #define FORCE_UNCHARGE_BATCH (128) 845 #define FORCE_UNCHARGE_BATCH (128)
844 static void mem_cgroup_force_empty_list(struct mem_cgroup *mem, 846 static void mem_cgroup_force_empty_list(struct mem_cgroup *mem,
845 struct mem_cgroup_per_zone *mz, 847 struct mem_cgroup_per_zone *mz,
846 int active) 848 int active)
847 { 849 {
848 struct page_cgroup *pc; 850 struct page_cgroup *pc;
849 struct page *page; 851 struct page *page;
850 int count = FORCE_UNCHARGE_BATCH; 852 int count = FORCE_UNCHARGE_BATCH;
851 unsigned long flags; 853 unsigned long flags;
852 struct list_head *list; 854 struct list_head *list;
853 855
854 if (active) 856 if (active)
855 list = &mz->active_list; 857 list = &mz->active_list;
856 else 858 else
857 list = &mz->inactive_list; 859 list = &mz->inactive_list;
858 860
859 spin_lock_irqsave(&mz->lru_lock, flags); 861 spin_lock_irqsave(&mz->lru_lock, flags);
860 while (!list_empty(list)) { 862 while (!list_empty(list)) {
861 pc = list_entry(list->prev, struct page_cgroup, lru); 863 pc = list_entry(list->prev, struct page_cgroup, lru);
862 page = pc->page; 864 page = pc->page;
863 get_page(page); 865 get_page(page);
864 spin_unlock_irqrestore(&mz->lru_lock, flags); 866 spin_unlock_irqrestore(&mz->lru_lock, flags);
865 /* 867 /*
866 * Check if this page is on LRU. !LRU page can be found 868 * Check if this page is on LRU. !LRU page can be found
867 * if it's under page migration. 869 * if it's under page migration.
868 */ 870 */
869 if (PageLRU(page)) { 871 if (PageLRU(page)) {
870 __mem_cgroup_uncharge_common(page, 872 __mem_cgroup_uncharge_common(page,
871 MEM_CGROUP_CHARGE_TYPE_FORCE); 873 MEM_CGROUP_CHARGE_TYPE_FORCE);
872 put_page(page); 874 put_page(page);
873 if (--count <= 0) { 875 if (--count <= 0) {
874 count = FORCE_UNCHARGE_BATCH; 876 count = FORCE_UNCHARGE_BATCH;
875 cond_resched(); 877 cond_resched();
876 } 878 }
877 } else 879 } else
878 cond_resched(); 880 cond_resched();
879 spin_lock_irqsave(&mz->lru_lock, flags); 881 spin_lock_irqsave(&mz->lru_lock, flags);
880 } 882 }
881 spin_unlock_irqrestore(&mz->lru_lock, flags); 883 spin_unlock_irqrestore(&mz->lru_lock, flags);
882 } 884 }
883 885
884 /* 886 /*
885 * make mem_cgroup's charge to be 0 if there is no task. 887 * make mem_cgroup's charge to be 0 if there is no task.
886 * This enables deleting this mem_cgroup. 888 * This enables deleting this mem_cgroup.
887 */ 889 */
888 static int mem_cgroup_force_empty(struct mem_cgroup *mem) 890 static int mem_cgroup_force_empty(struct mem_cgroup *mem)
889 { 891 {
890 int ret = -EBUSY; 892 int ret = -EBUSY;
891 int node, zid; 893 int node, zid;
892 894
893 css_get(&mem->css); 895 css_get(&mem->css);
894 /* 896 /*
895 * page reclaim code (kswapd etc..) will move pages between 897 * page reclaim code (kswapd etc..) will move pages between
896 * active_list <-> inactive_list while we don't take a lock. 898 * active_list <-> inactive_list while we don't take a lock.
897 * So, we have to do loop here until all lists are empty. 899 * So, we have to do loop here until all lists are empty.
898 */ 900 */
899 while (mem->res.usage > 0) { 901 while (mem->res.usage > 0) {
900 if (atomic_read(&mem->css.cgroup->count) > 0) 902 if (atomic_read(&mem->css.cgroup->count) > 0)
901 goto out; 903 goto out;
902 for_each_node_state(node, N_POSSIBLE) 904 for_each_node_state(node, N_POSSIBLE)
903 for (zid = 0; zid < MAX_NR_ZONES; zid++) { 905 for (zid = 0; zid < MAX_NR_ZONES; zid++) {
904 struct mem_cgroup_per_zone *mz; 906 struct mem_cgroup_per_zone *mz;
905 mz = mem_cgroup_zoneinfo(mem, node, zid); 907 mz = mem_cgroup_zoneinfo(mem, node, zid);
906 /* drop all page_cgroup in active_list */ 908 /* drop all page_cgroup in active_list */
907 mem_cgroup_force_empty_list(mem, mz, 1); 909 mem_cgroup_force_empty_list(mem, mz, 1);
908 /* drop all page_cgroup in inactive_list */ 910 /* drop all page_cgroup in inactive_list */
909 mem_cgroup_force_empty_list(mem, mz, 0); 911 mem_cgroup_force_empty_list(mem, mz, 0);
910 } 912 }
911 } 913 }
912 ret = 0; 914 ret = 0;
913 out: 915 out:
914 css_put(&mem->css); 916 css_put(&mem->css);
915 return ret; 917 return ret;
916 } 918 }
917 919
918 static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft) 920 static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
919 { 921 {
920 return res_counter_read_u64(&mem_cgroup_from_cont(cont)->res, 922 return res_counter_read_u64(&mem_cgroup_from_cont(cont)->res,
921 cft->private); 923 cft->private);
922 } 924 }
923 /* 925 /*
924 * The user of this function is... 926 * The user of this function is...
925 * RES_LIMIT. 927 * RES_LIMIT.
926 */ 928 */
927 static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft, 929 static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
928 const char *buffer) 930 const char *buffer)
929 { 931 {
930 struct mem_cgroup *memcg = mem_cgroup_from_cont(cont); 932 struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
931 unsigned long long val; 933 unsigned long long val;
932 int ret; 934 int ret;
933 935
934 switch (cft->private) { 936 switch (cft->private) {
935 case RES_LIMIT: 937 case RES_LIMIT:
936 /* This function does all necessary parse...reuse it */ 938 /* This function does all necessary parse...reuse it */
937 ret = res_counter_memparse_write_strategy(buffer, &val); 939 ret = res_counter_memparse_write_strategy(buffer, &val);
938 if (!ret) 940 if (!ret)
939 ret = mem_cgroup_resize_limit(memcg, val); 941 ret = mem_cgroup_resize_limit(memcg, val);
940 break; 942 break;
941 default: 943 default:
942 ret = -EINVAL; /* should be BUG() ? */ 944 ret = -EINVAL; /* should be BUG() ? */
943 break; 945 break;
944 } 946 }
945 return ret; 947 return ret;
946 } 948 }
947 949
948 static int mem_cgroup_reset(struct cgroup *cont, unsigned int event) 950 static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
949 { 951 {
950 struct mem_cgroup *mem; 952 struct mem_cgroup *mem;
951 953
952 mem = mem_cgroup_from_cont(cont); 954 mem = mem_cgroup_from_cont(cont);
953 switch (event) { 955 switch (event) {
954 case RES_MAX_USAGE: 956 case RES_MAX_USAGE:
955 res_counter_reset_max(&mem->res); 957 res_counter_reset_max(&mem->res);
956 break; 958 break;
957 case RES_FAILCNT: 959 case RES_FAILCNT:
958 res_counter_reset_failcnt(&mem->res); 960 res_counter_reset_failcnt(&mem->res);
959 break; 961 break;
960 } 962 }
961 return 0; 963 return 0;
962 } 964 }
963 965
964 static int mem_force_empty_write(struct cgroup *cont, unsigned int event) 966 static int mem_force_empty_write(struct cgroup *cont, unsigned int event)
965 { 967 {
966 return mem_cgroup_force_empty(mem_cgroup_from_cont(cont)); 968 return mem_cgroup_force_empty(mem_cgroup_from_cont(cont));
967 } 969 }
968 970
969 static const struct mem_cgroup_stat_desc { 971 static const struct mem_cgroup_stat_desc {
970 const char *msg; 972 const char *msg;
971 u64 unit; 973 u64 unit;
972 } mem_cgroup_stat_desc[] = { 974 } mem_cgroup_stat_desc[] = {
973 [MEM_CGROUP_STAT_CACHE] = { "cache", PAGE_SIZE, }, 975 [MEM_CGROUP_STAT_CACHE] = { "cache", PAGE_SIZE, },
974 [MEM_CGROUP_STAT_RSS] = { "rss", PAGE_SIZE, }, 976 [MEM_CGROUP_STAT_RSS] = { "rss", PAGE_SIZE, },
975 [MEM_CGROUP_STAT_PGPGIN_COUNT] = {"pgpgin", 1, }, 977 [MEM_CGROUP_STAT_PGPGIN_COUNT] = {"pgpgin", 1, },
976 [MEM_CGROUP_STAT_PGPGOUT_COUNT] = {"pgpgout", 1, }, 978 [MEM_CGROUP_STAT_PGPGOUT_COUNT] = {"pgpgout", 1, },
977 }; 979 };
978 980
979 static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft, 981 static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
980 struct cgroup_map_cb *cb) 982 struct cgroup_map_cb *cb)
981 { 983 {
982 struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont); 984 struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);
983 struct mem_cgroup_stat *stat = &mem_cont->stat; 985 struct mem_cgroup_stat *stat = &mem_cont->stat;
984 int i; 986 int i;
985 987
986 for (i = 0; i < ARRAY_SIZE(stat->cpustat[0].count); i++) { 988 for (i = 0; i < ARRAY_SIZE(stat->cpustat[0].count); i++) {
987 s64 val; 989 s64 val;
988 990
989 val = mem_cgroup_read_stat(stat, i); 991 val = mem_cgroup_read_stat(stat, i);
990 val *= mem_cgroup_stat_desc[i].unit; 992 val *= mem_cgroup_stat_desc[i].unit;
991 cb->fill(cb, mem_cgroup_stat_desc[i].msg, val); 993 cb->fill(cb, mem_cgroup_stat_desc[i].msg, val);
992 } 994 }
993 /* showing # of active pages */ 995 /* showing # of active pages */
994 { 996 {
995 unsigned long active, inactive; 997 unsigned long active, inactive;
996 998
997 inactive = mem_cgroup_get_all_zonestat(mem_cont, 999 inactive = mem_cgroup_get_all_zonestat(mem_cont,
998 MEM_CGROUP_ZSTAT_INACTIVE); 1000 MEM_CGROUP_ZSTAT_INACTIVE);
999 active = mem_cgroup_get_all_zonestat(mem_cont, 1001 active = mem_cgroup_get_all_zonestat(mem_cont,
1000 MEM_CGROUP_ZSTAT_ACTIVE); 1002 MEM_CGROUP_ZSTAT_ACTIVE);
1001 cb->fill(cb, "active", (active) * PAGE_SIZE); 1003 cb->fill(cb, "active", (active) * PAGE_SIZE);
1002 cb->fill(cb, "inactive", (inactive) * PAGE_SIZE); 1004 cb->fill(cb, "inactive", (inactive) * PAGE_SIZE);
1003 } 1005 }
1004 return 0; 1006 return 0;
1005 } 1007 }
1006 1008
1007 static struct cftype mem_cgroup_files[] = { 1009 static struct cftype mem_cgroup_files[] = {
1008 { 1010 {
1009 .name = "usage_in_bytes", 1011 .name = "usage_in_bytes",
1010 .private = RES_USAGE, 1012 .private = RES_USAGE,
1011 .read_u64 = mem_cgroup_read, 1013 .read_u64 = mem_cgroup_read,
1012 }, 1014 },
1013 { 1015 {
1014 .name = "max_usage_in_bytes", 1016 .name = "max_usage_in_bytes",
1015 .private = RES_MAX_USAGE, 1017 .private = RES_MAX_USAGE,
1016 .trigger = mem_cgroup_reset, 1018 .trigger = mem_cgroup_reset,
1017 .read_u64 = mem_cgroup_read, 1019 .read_u64 = mem_cgroup_read,
1018 }, 1020 },
1019 { 1021 {
1020 .name = "limit_in_bytes", 1022 .name = "limit_in_bytes",
1021 .private = RES_LIMIT, 1023 .private = RES_LIMIT,
1022 .write_string = mem_cgroup_write, 1024 .write_string = mem_cgroup_write,
1023 .read_u64 = mem_cgroup_read, 1025 .read_u64 = mem_cgroup_read,
1024 }, 1026 },
1025 { 1027 {
1026 .name = "failcnt", 1028 .name = "failcnt",
1027 .private = RES_FAILCNT, 1029 .private = RES_FAILCNT,
1028 .trigger = mem_cgroup_reset, 1030 .trigger = mem_cgroup_reset,
1029 .read_u64 = mem_cgroup_read, 1031 .read_u64 = mem_cgroup_read,
1030 }, 1032 },
1031 { 1033 {
1032 .name = "force_empty", 1034 .name = "force_empty",
1033 .trigger = mem_force_empty_write, 1035 .trigger = mem_force_empty_write,
1034 }, 1036 },
1035 { 1037 {
1036 .name = "stat", 1038 .name = "stat",
1037 .read_map = mem_control_stat_show, 1039 .read_map = mem_control_stat_show,
1038 }, 1040 },
1039 }; 1041 };
1040 1042
1041 static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node) 1043 static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
1042 { 1044 {
1043 struct mem_cgroup_per_node *pn; 1045 struct mem_cgroup_per_node *pn;
1044 struct mem_cgroup_per_zone *mz; 1046 struct mem_cgroup_per_zone *mz;
1045 int zone, tmp = node; 1047 int zone, tmp = node;
1046 /* 1048 /*
1047 * This routine is called against possible nodes. 1049 * This routine is called against possible nodes.
1048 * But it's BUG to call kmalloc() against offline node. 1050 * But it's BUG to call kmalloc() against offline node.
1049 * 1051 *
1050 * TODO: this routine can waste much memory for nodes which will 1052 * TODO: this routine can waste much memory for nodes which will
1051 * never be onlined. It's better to use memory hotplug callback 1053 * never be onlined. It's better to use memory hotplug callback
1052 * function. 1054 * function.
1053 */ 1055 */
1054 if (!node_state(node, N_NORMAL_MEMORY)) 1056 if (!node_state(node, N_NORMAL_MEMORY))
1055 tmp = -1; 1057 tmp = -1;
1056 pn = kmalloc_node(sizeof(*pn), GFP_KERNEL, tmp); 1058 pn = kmalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
1057 if (!pn) 1059 if (!pn)
1058 return 1; 1060 return 1;
1059 1061
1060 mem->info.nodeinfo[node] = pn; 1062 mem->info.nodeinfo[node] = pn;
1061 memset(pn, 0, sizeof(*pn)); 1063 memset(pn, 0, sizeof(*pn));
1062 1064
1063 for (zone = 0; zone < MAX_NR_ZONES; zone++) { 1065 for (zone = 0; zone < MAX_NR_ZONES; zone++) {
1064 mz = &pn->zoneinfo[zone]; 1066 mz = &pn->zoneinfo[zone];
1065 INIT_LIST_HEAD(&mz->active_list); 1067 INIT_LIST_HEAD(&mz->active_list);
1066 INIT_LIST_HEAD(&mz->inactive_list); 1068 INIT_LIST_HEAD(&mz->inactive_list);
1067 spin_lock_init(&mz->lru_lock); 1069 spin_lock_init(&mz->lru_lock);
1068 } 1070 }
1069 return 0; 1071 return 0;
1070 } 1072 }
1071 1073
1072 static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node) 1074 static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
1073 { 1075 {
1074 kfree(mem->info.nodeinfo[node]); 1076 kfree(mem->info.nodeinfo[node]);
1075 } 1077 }
1076 1078
1077 static struct mem_cgroup *mem_cgroup_alloc(void) 1079 static struct mem_cgroup *mem_cgroup_alloc(void)
1078 { 1080 {
1079 struct mem_cgroup *mem; 1081 struct mem_cgroup *mem;
1080 1082
1081 if (sizeof(*mem) < PAGE_SIZE) 1083 if (sizeof(*mem) < PAGE_SIZE)
1082 mem = kmalloc(sizeof(*mem), GFP_KERNEL); 1084 mem = kmalloc(sizeof(*mem), GFP_KERNEL);
1083 else 1085 else
1084 mem = vmalloc(sizeof(*mem)); 1086 mem = vmalloc(sizeof(*mem));
1085 1087
1086 if (mem) 1088 if (mem)
1087 memset(mem, 0, sizeof(*mem)); 1089 memset(mem, 0, sizeof(*mem));
1088 return mem; 1090 return mem;
1089 } 1091 }
1090 1092
1091 static void mem_cgroup_free(struct mem_cgroup *mem) 1093 static void mem_cgroup_free(struct mem_cgroup *mem)
1092 { 1094 {
1093 if (sizeof(*mem) < PAGE_SIZE) 1095 if (sizeof(*mem) < PAGE_SIZE)
1094 kfree(mem); 1096 kfree(mem);
1095 else 1097 else
1096 vfree(mem); 1098 vfree(mem);
1097 } 1099 }
1098 1100
1099 1101
1100 static struct cgroup_subsys_state * 1102 static struct cgroup_subsys_state *
1101 mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont) 1103 mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
1102 { 1104 {
1103 struct mem_cgroup *mem; 1105 struct mem_cgroup *mem;
1104 int node; 1106 int node;
1105 1107
1106 if (unlikely((cont->parent) == NULL)) { 1108 if (unlikely((cont->parent) == NULL)) {
1107 mem = &init_mem_cgroup; 1109 mem = &init_mem_cgroup;
1108 page_cgroup_cache = KMEM_CACHE(page_cgroup, SLAB_PANIC); 1110 page_cgroup_cache = KMEM_CACHE(page_cgroup, SLAB_PANIC);
1109 } else { 1111 } else {
1110 mem = mem_cgroup_alloc(); 1112 mem = mem_cgroup_alloc();
1111 if (!mem) 1113 if (!mem)
1112 return ERR_PTR(-ENOMEM); 1114 return ERR_PTR(-ENOMEM);
1113 } 1115 }
1114 1116
1115 res_counter_init(&mem->res); 1117 res_counter_init(&mem->res);
1116 1118
1117 for_each_node_state(node, N_POSSIBLE) 1119 for_each_node_state(node, N_POSSIBLE)
1118 if (alloc_mem_cgroup_per_zone_info(mem, node)) 1120 if (alloc_mem_cgroup_per_zone_info(mem, node))
1119 goto free_out; 1121 goto free_out;
1120 1122
1121 return &mem->css; 1123 return &mem->css;
1122 free_out: 1124 free_out:
1123 for_each_node_state(node, N_POSSIBLE) 1125 for_each_node_state(node, N_POSSIBLE)
1124 free_mem_cgroup_per_zone_info(mem, node); 1126 free_mem_cgroup_per_zone_info(mem, node);
1125 if (cont->parent != NULL) 1127 if (cont->parent != NULL)
1126 mem_cgroup_free(mem); 1128 mem_cgroup_free(mem);
1127 return ERR_PTR(-ENOMEM); 1129 return ERR_PTR(-ENOMEM);
1128 } 1130 }
1129 1131
1130 static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss, 1132 static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
1131 struct cgroup *cont) 1133 struct cgroup *cont)
1132 { 1134 {
1133 struct mem_cgroup *mem = mem_cgroup_from_cont(cont); 1135 struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
1134 mem_cgroup_force_empty(mem); 1136 mem_cgroup_force_empty(mem);
1135 } 1137 }
1136 1138
1137 static void mem_cgroup_destroy(struct cgroup_subsys *ss, 1139 static void mem_cgroup_destroy(struct cgroup_subsys *ss,
1138 struct cgroup *cont) 1140 struct cgroup *cont)
1139 { 1141 {
1140 int node; 1142 int node;
1141 struct mem_cgroup *mem = mem_cgroup_from_cont(cont); 1143 struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
1142 1144
1143 for_each_node_state(node, N_POSSIBLE) 1145 for_each_node_state(node, N_POSSIBLE)
1144 free_mem_cgroup_per_zone_info(mem, node); 1146 free_mem_cgroup_per_zone_info(mem, node);
1145 1147
1146 mem_cgroup_free(mem_cgroup_from_cont(cont)); 1148 mem_cgroup_free(mem_cgroup_from_cont(cont));
1147 } 1149 }
1148 1150
1149 static int mem_cgroup_populate(struct cgroup_subsys *ss, 1151 static int mem_cgroup_populate(struct cgroup_subsys *ss,
1150 struct cgroup *cont) 1152 struct cgroup *cont)
1151 { 1153 {
1152 return cgroup_add_files(cont, ss, mem_cgroup_files, 1154 return cgroup_add_files(cont, ss, mem_cgroup_files,
1153 ARRAY_SIZE(mem_cgroup_files)); 1155 ARRAY_SIZE(mem_cgroup_files));
1154 } 1156 }
1155 1157
1156 static void mem_cgroup_move_task(struct cgroup_subsys *ss, 1158 static void mem_cgroup_move_task(struct cgroup_subsys *ss,
1157 struct cgroup *cont, 1159 struct cgroup *cont,
1158 struct cgroup *old_cont, 1160 struct cgroup *old_cont,
1159 struct task_struct *p) 1161 struct task_struct *p)
1160 { 1162 {
1161 struct mm_struct *mm; 1163 struct mm_struct *mm;
1162 struct mem_cgroup *mem, *old_mem; 1164 struct mem_cgroup *mem, *old_mem;
1163 1165
1164 mm = get_task_mm(p); 1166 mm = get_task_mm(p);
1165 if (mm == NULL) 1167 if (mm == NULL)
1166 return; 1168 return;
1167 1169
1168 mem = mem_cgroup_from_cont(cont); 1170 mem = mem_cgroup_from_cont(cont);
1169 old_mem = mem_cgroup_from_cont(old_cont); 1171 old_mem = mem_cgroup_from_cont(old_cont);
1170 1172
1171 /* 1173 /*
1172 * Only thread group leaders are allowed to migrate, the mm_struct is 1174 * Only thread group leaders are allowed to migrate, the mm_struct is
1173 * in effect owned by the leader 1175 * in effect owned by the leader
1174 */ 1176 */
1175 if (!thread_group_leader(p)) 1177 if (!thread_group_leader(p))
1176 goto out; 1178 goto out;
1177 1179
1178 out: 1180 out:
1179 mmput(mm); 1181 mmput(mm);
1180 } 1182 }
1181 1183
1182 struct cgroup_subsys mem_cgroup_subsys = { 1184 struct cgroup_subsys mem_cgroup_subsys = {
1183 .name = "memory", 1185 .name = "memory",
1184 .subsys_id = mem_cgroup_subsys_id, 1186 .subsys_id = mem_cgroup_subsys_id,
1185 .create = mem_cgroup_create, 1187 .create = mem_cgroup_create,
1186 .pre_destroy = mem_cgroup_pre_destroy, 1188 .pre_destroy = mem_cgroup_pre_destroy,
1187 .destroy = mem_cgroup_destroy, 1189 .destroy = mem_cgroup_destroy,
1188 .populate = mem_cgroup_populate, 1190 .populate = mem_cgroup_populate,
1189 .attach = mem_cgroup_move_task, 1191 .attach = mem_cgroup_move_task,
1190 .early_init = 0, 1192 .early_init = 0,
1191 }; 1193 };
1192 1194