Eric Lee / linux-smarc-t335x-v3.2

Download as
Browse Code ยป

Commit c8dad2bb6307f5b00f804a686917105206a4d5c9

Authored by Jan Blunck
Committed by Linus Torvalds
1 parent f817ed4853
Exists in master and in 4 other branches v3.2_AM335xPSP_04.06.00.11, v3.2_NEWT335xPSP_04.06.00.11, v3.2_SBC_SMARTMEN, v3.2_SMARCT335xPSP_04.06.00.11

memcg: reduce size of mem_cgroup by using nr_cpu_ids 

As Jan Blunck <jblunck@suse.de> pointed out, allocating per-cpu stat for
memcg to the size of NR_CPUS is not good.

This patch changes mem_cgroup's cpustat allocation not based on NR_CPUS
but based on nr_cpu_ids.

Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Showing 1 changed file with 18 additions and 17 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 #include <linux/mm_inline.h> 35 #include <linux/mm_inline.h>
36 #include <linux/page_cgroup.h> 36 #include <linux/page_cgroup.h>
37 37
38 #include <asm/uaccess.h> 38 #include <asm/uaccess.h>
39 39
40 struct cgroup_subsys mem_cgroup_subsys __read_mostly; 40 struct cgroup_subsys mem_cgroup_subsys __read_mostly;
41 #define MEM_CGROUP_RECLAIM_RETRIES 5 41 #define MEM_CGROUP_RECLAIM_RETRIES 5
42 42
43 /* 43 /*
44 * Statistics for memory cgroup. 44 * Statistics for memory cgroup.
45 */ 45 */
46 enum mem_cgroup_stat_index { 46 enum mem_cgroup_stat_index {
47 /* 47 /*
48 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss. 48 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
49 */ 49 */
50 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */ 50 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
51 MEM_CGROUP_STAT_RSS, /* # of pages charged as rss */ 51 MEM_CGROUP_STAT_RSS, /* # of pages charged as rss */
52 MEM_CGROUP_STAT_PGPGIN_COUNT, /* # of pages paged in */ 52 MEM_CGROUP_STAT_PGPGIN_COUNT, /* # of pages paged in */
53 MEM_CGROUP_STAT_PGPGOUT_COUNT, /* # of pages paged out */ 53 MEM_CGROUP_STAT_PGPGOUT_COUNT, /* # of pages paged out */
54 54
55 MEM_CGROUP_STAT_NSTATS, 55 MEM_CGROUP_STAT_NSTATS,
56 }; 56 };
57 57
58 struct mem_cgroup_stat_cpu { 58 struct mem_cgroup_stat_cpu {
59 s64 count[MEM_CGROUP_STAT_NSTATS]; 59 s64 count[MEM_CGROUP_STAT_NSTATS];
60 } ____cacheline_aligned_in_smp; 60 } ____cacheline_aligned_in_smp;
61 61
62 struct mem_cgroup_stat { 62 struct mem_cgroup_stat {
63 struct mem_cgroup_stat_cpu cpustat[NR_CPUS]; 63 struct mem_cgroup_stat_cpu cpustat[0];
64 }; 64 };
65 65
66 /* 66 /*
67 * For accounting under irq disable, no need for increment preempt count. 67 * For accounting under irq disable, no need for increment preempt count.
68 */ 68 */
69 static inline void __mem_cgroup_stat_add_safe(struct mem_cgroup_stat_cpu *stat, 69 static inline void __mem_cgroup_stat_add_safe(struct mem_cgroup_stat_cpu *stat,
70 enum mem_cgroup_stat_index idx, int val) 70 enum mem_cgroup_stat_index idx, int val)
71 { 71 {
72 stat->count[idx] += val; 72 stat->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 struct mem_cgroup_per_zone { 88 struct mem_cgroup_per_zone {
89 /* 89 /*
90 * spin_lock to protect the per cgroup LRU 90 * spin_lock to protect the per cgroup LRU
91 */ 91 */
92 spinlock_t lru_lock; 92 spinlock_t lru_lock;
93 struct list_head lists[NR_LRU_LISTS]; 93 struct list_head lists[NR_LRU_LISTS];
94 unsigned long count[NR_LRU_LISTS]; 94 unsigned long count[NR_LRU_LISTS];
95 }; 95 };
96 /* Macro for accessing counter */ 96 /* Macro for accessing counter */
97 #define MEM_CGROUP_ZSTAT(mz, idx) ((mz)->count[(idx)]) 97 #define MEM_CGROUP_ZSTAT(mz, idx) ((mz)->count[(idx)])
98 98
99 struct mem_cgroup_per_node { 99 struct mem_cgroup_per_node {
100 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES]; 100 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
101 }; 101 };
102 102
103 struct mem_cgroup_lru_info { 103 struct mem_cgroup_lru_info {
104 struct mem_cgroup_per_node *nodeinfo[MAX_NUMNODES]; 104 struct mem_cgroup_per_node *nodeinfo[MAX_NUMNODES];
105 }; 105 };
106 106
107 /* 107 /*
108 * The memory controller data structure. The memory controller controls both 108 * The memory controller data structure. The memory controller controls both
109 * page cache and RSS per cgroup. We would eventually like to provide 109 * page cache and RSS per cgroup. We would eventually like to provide
110 * statistics based on the statistics developed by Rik Van Riel for clock-pro, 110 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
111 * to help the administrator determine what knobs to tune. 111 * to help the administrator determine what knobs to tune.
112 * 112 *
113 * TODO: Add a water mark for the memory controller. Reclaim will begin when 113 * TODO: Add a water mark for the memory controller. Reclaim will begin when
114 * we hit the water mark. May be even add a low water mark, such that 114 * we hit the water mark. May be even add a low water mark, such that
115 * no reclaim occurs from a cgroup at it's low water mark, this is 115 * no reclaim occurs from a cgroup at it's low water mark, this is
116 * a feature that will be implemented much later in the future. 116 * a feature that will be implemented much later in the future.
117 */ 117 */
118 struct mem_cgroup { 118 struct mem_cgroup {
119 struct cgroup_subsys_state css; 119 struct cgroup_subsys_state css;
120 /* 120 /*
121 * the counter to account for memory usage 121 * the counter to account for memory usage
122 */ 122 */
123 struct res_counter res; 123 struct res_counter res;
124 /* 124 /*
125 * Per cgroup active and inactive list, similar to the 125 * Per cgroup active and inactive list, similar to the
126 * per zone LRU lists. 126 * per zone LRU lists.
127 */ 127 */
128 struct mem_cgroup_lru_info info; 128 struct mem_cgroup_lru_info info;
129 129
130 int prev_priority; /* for recording reclaim priority */ 130 int prev_priority; /* for recording reclaim priority */
131 /* 131 /*
132 * statistics. 132 * statistics. This must be placed at the end of memcg.
133 */ 133 */
134 struct mem_cgroup_stat stat; 134 struct mem_cgroup_stat stat;
135 }; 135 };
136 static struct mem_cgroup init_mem_cgroup;
137 136
138 enum charge_type { 137 enum charge_type {
139 MEM_CGROUP_CHARGE_TYPE_CACHE = 0, 138 MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
140 MEM_CGROUP_CHARGE_TYPE_MAPPED, 139 MEM_CGROUP_CHARGE_TYPE_MAPPED,
141 MEM_CGROUP_CHARGE_TYPE_SHMEM, /* used by page migration of shmem */ 140 MEM_CGROUP_CHARGE_TYPE_SHMEM, /* used by page migration of shmem */
142 MEM_CGROUP_CHARGE_TYPE_FORCE, /* used by force_empty */ 141 MEM_CGROUP_CHARGE_TYPE_FORCE, /* used by force_empty */
143 NR_CHARGE_TYPE, 142 NR_CHARGE_TYPE,
144 }; 143 };
145 144
146 /* only for here (for easy reading.) */ 145 /* only for here (for easy reading.) */
147 #define PCGF_CACHE (1UL << PCG_CACHE) 146 #define PCGF_CACHE (1UL << PCG_CACHE)
148 #define PCGF_USED (1UL << PCG_USED) 147 #define PCGF_USED (1UL << PCG_USED)
149 #define PCGF_ACTIVE (1UL << PCG_ACTIVE) 148 #define PCGF_ACTIVE (1UL << PCG_ACTIVE)
150 #define PCGF_LOCK (1UL << PCG_LOCK) 149 #define PCGF_LOCK (1UL << PCG_LOCK)
151 #define PCGF_FILE (1UL << PCG_FILE) 150 #define PCGF_FILE (1UL << PCG_FILE)
152 static const unsigned long 151 static const unsigned long
153 pcg_default_flags[NR_CHARGE_TYPE] = { 152 pcg_default_flags[NR_CHARGE_TYPE] = {
154 PCGF_CACHE | PCGF_FILE | PCGF_USED | PCGF_LOCK, /* File Cache */ 153 PCGF_CACHE | PCGF_FILE | PCGF_USED | PCGF_LOCK, /* File Cache */
155 PCGF_ACTIVE | PCGF_USED | PCGF_LOCK, /* Anon */ 154 PCGF_ACTIVE | PCGF_USED | PCGF_LOCK, /* Anon */
156 PCGF_ACTIVE | PCGF_CACHE | PCGF_USED | PCGF_LOCK, /* Shmem */ 155 PCGF_ACTIVE | PCGF_CACHE | PCGF_USED | PCGF_LOCK, /* Shmem */
157 0, /* FORCE */ 156 0, /* FORCE */
158 }; 157 };
159 158
160 /* 159 /*
161 * Always modified under lru lock. Then, not necessary to preempt_disable() 160 * Always modified under lru lock. Then, not necessary to preempt_disable()
162 */ 161 */
163 static void mem_cgroup_charge_statistics(struct mem_cgroup *mem, 162 static void mem_cgroup_charge_statistics(struct mem_cgroup *mem,
164 struct page_cgroup *pc, 163 struct page_cgroup *pc,
165 bool charge) 164 bool charge)
166 { 165 {
167 int val = (charge)? 1 : -1; 166 int val = (charge)? 1 : -1;
168 struct mem_cgroup_stat *stat = &mem->stat; 167 struct mem_cgroup_stat *stat = &mem->stat;
169 struct mem_cgroup_stat_cpu *cpustat; 168 struct mem_cgroup_stat_cpu *cpustat;
170 169
171 VM_BUG_ON(!irqs_disabled()); 170 VM_BUG_ON(!irqs_disabled());
172 171
173 cpustat = &stat->cpustat[smp_processor_id()]; 172 cpustat = &stat->cpustat[smp_processor_id()];
174 if (PageCgroupCache(pc)) 173 if (PageCgroupCache(pc))
175 __mem_cgroup_stat_add_safe(cpustat, MEM_CGROUP_STAT_CACHE, val); 174 __mem_cgroup_stat_add_safe(cpustat, MEM_CGROUP_STAT_CACHE, val);
176 else 175 else
177 __mem_cgroup_stat_add_safe(cpustat, MEM_CGROUP_STAT_RSS, val); 176 __mem_cgroup_stat_add_safe(cpustat, MEM_CGROUP_STAT_RSS, val);
178 177
179 if (charge) 178 if (charge)
180 __mem_cgroup_stat_add_safe(cpustat, 179 __mem_cgroup_stat_add_safe(cpustat,
181 MEM_CGROUP_STAT_PGPGIN_COUNT, 1); 180 MEM_CGROUP_STAT_PGPGIN_COUNT, 1);
182 else 181 else
183 __mem_cgroup_stat_add_safe(cpustat, 182 __mem_cgroup_stat_add_safe(cpustat,
184 MEM_CGROUP_STAT_PGPGOUT_COUNT, 1); 183 MEM_CGROUP_STAT_PGPGOUT_COUNT, 1);
185 } 184 }
186 185
187 static struct mem_cgroup_per_zone * 186 static struct mem_cgroup_per_zone *
188 mem_cgroup_zoneinfo(struct mem_cgroup *mem, int nid, int zid) 187 mem_cgroup_zoneinfo(struct mem_cgroup *mem, int nid, int zid)
189 { 188 {
190 return &mem->info.nodeinfo[nid]->zoneinfo[zid]; 189 return &mem->info.nodeinfo[nid]->zoneinfo[zid];
191 } 190 }
192 191
193 static struct mem_cgroup_per_zone * 192 static struct mem_cgroup_per_zone *
194 page_cgroup_zoneinfo(struct page_cgroup *pc) 193 page_cgroup_zoneinfo(struct page_cgroup *pc)
195 { 194 {
196 struct mem_cgroup *mem = pc->mem_cgroup; 195 struct mem_cgroup *mem = pc->mem_cgroup;
197 int nid = page_cgroup_nid(pc); 196 int nid = page_cgroup_nid(pc);
198 int zid = page_cgroup_zid(pc); 197 int zid = page_cgroup_zid(pc);
199 198
200 return mem_cgroup_zoneinfo(mem, nid, zid); 199 return mem_cgroup_zoneinfo(mem, nid, zid);
201 } 200 }
202 201
203 static unsigned long mem_cgroup_get_all_zonestat(struct mem_cgroup *mem, 202 static unsigned long mem_cgroup_get_all_zonestat(struct mem_cgroup *mem,
204 enum lru_list idx) 203 enum lru_list idx)
205 { 204 {
206 int nid, zid; 205 int nid, zid;
207 struct mem_cgroup_per_zone *mz; 206 struct mem_cgroup_per_zone *mz;
208 u64 total = 0; 207 u64 total = 0;
209 208
210 for_each_online_node(nid) 209 for_each_online_node(nid)
211 for (zid = 0; zid < MAX_NR_ZONES; zid++) { 210 for (zid = 0; zid < MAX_NR_ZONES; zid++) {
212 mz = mem_cgroup_zoneinfo(mem, nid, zid); 211 mz = mem_cgroup_zoneinfo(mem, nid, zid);
213 total += MEM_CGROUP_ZSTAT(mz, idx); 212 total += MEM_CGROUP_ZSTAT(mz, idx);
214 } 213 }
215 return total; 214 return total;
216 } 215 }
217 216
218 static struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont) 217 static struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
219 { 218 {
220 return container_of(cgroup_subsys_state(cont, 219 return container_of(cgroup_subsys_state(cont,
221 mem_cgroup_subsys_id), struct mem_cgroup, 220 mem_cgroup_subsys_id), struct mem_cgroup,
222 css); 221 css);
223 } 222 }
224 223
225 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p) 224 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
226 { 225 {
227 /* 226 /*
228 * mm_update_next_owner() may clear mm->owner to NULL 227 * mm_update_next_owner() may clear mm->owner to NULL
229 * if it races with swapoff, page migration, etc. 228 * if it races with swapoff, page migration, etc.
230 * So this can be called with p == NULL. 229 * So this can be called with p == NULL.
231 */ 230 */
232 if (unlikely(!p)) 231 if (unlikely(!p))
233 return NULL; 232 return NULL;
234 233
235 return container_of(task_subsys_state(p, mem_cgroup_subsys_id), 234 return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
236 struct mem_cgroup, css); 235 struct mem_cgroup, css);
237 } 236 }
238 237
239 static void __mem_cgroup_remove_list(struct mem_cgroup_per_zone *mz, 238 static void __mem_cgroup_remove_list(struct mem_cgroup_per_zone *mz,
240 struct page_cgroup *pc) 239 struct page_cgroup *pc)
241 { 240 {
242 int lru = LRU_BASE; 241 int lru = LRU_BASE;
243 242
244 if (PageCgroupUnevictable(pc)) 243 if (PageCgroupUnevictable(pc))
245 lru = LRU_UNEVICTABLE; 244 lru = LRU_UNEVICTABLE;
246 else { 245 else {
247 if (PageCgroupActive(pc)) 246 if (PageCgroupActive(pc))
248 lru += LRU_ACTIVE; 247 lru += LRU_ACTIVE;
249 if (PageCgroupFile(pc)) 248 if (PageCgroupFile(pc))
250 lru += LRU_FILE; 249 lru += LRU_FILE;
251 } 250 }
252 251
253 MEM_CGROUP_ZSTAT(mz, lru) -= 1; 252 MEM_CGROUP_ZSTAT(mz, lru) -= 1;
254 253
255 mem_cgroup_charge_statistics(pc->mem_cgroup, pc, false); 254 mem_cgroup_charge_statistics(pc->mem_cgroup, pc, false);
256 list_del(&pc->lru); 255 list_del(&pc->lru);
257 } 256 }
258 257
259 static void __mem_cgroup_add_list(struct mem_cgroup_per_zone *mz, 258 static void __mem_cgroup_add_list(struct mem_cgroup_per_zone *mz,
260 struct page_cgroup *pc, bool hot) 259 struct page_cgroup *pc, bool hot)
261 { 260 {
262 int lru = LRU_BASE; 261 int lru = LRU_BASE;
263 262
264 if (PageCgroupUnevictable(pc)) 263 if (PageCgroupUnevictable(pc))
265 lru = LRU_UNEVICTABLE; 264 lru = LRU_UNEVICTABLE;
266 else { 265 else {
267 if (PageCgroupActive(pc)) 266 if (PageCgroupActive(pc))
268 lru += LRU_ACTIVE; 267 lru += LRU_ACTIVE;
269 if (PageCgroupFile(pc)) 268 if (PageCgroupFile(pc))
270 lru += LRU_FILE; 269 lru += LRU_FILE;
271 } 270 }
272 271
273 MEM_CGROUP_ZSTAT(mz, lru) += 1; 272 MEM_CGROUP_ZSTAT(mz, lru) += 1;
274 if (hot) 273 if (hot)
275 list_add(&pc->lru, &mz->lists[lru]); 274 list_add(&pc->lru, &mz->lists[lru]);
276 else 275 else
277 list_add_tail(&pc->lru, &mz->lists[lru]); 276 list_add_tail(&pc->lru, &mz->lists[lru]);
278 277
279 mem_cgroup_charge_statistics(pc->mem_cgroup, pc, true); 278 mem_cgroup_charge_statistics(pc->mem_cgroup, pc, true);
280 } 279 }
281 280
282 static void __mem_cgroup_move_lists(struct page_cgroup *pc, enum lru_list lru) 281 static void __mem_cgroup_move_lists(struct page_cgroup *pc, enum lru_list lru)
283 { 282 {
284 struct mem_cgroup_per_zone *mz = page_cgroup_zoneinfo(pc); 283 struct mem_cgroup_per_zone *mz = page_cgroup_zoneinfo(pc);
285 int active = PageCgroupActive(pc); 284 int active = PageCgroupActive(pc);
286 int file = PageCgroupFile(pc); 285 int file = PageCgroupFile(pc);
287 int unevictable = PageCgroupUnevictable(pc); 286 int unevictable = PageCgroupUnevictable(pc);
288 enum lru_list from = unevictable ? LRU_UNEVICTABLE : 287 enum lru_list from = unevictable ? LRU_UNEVICTABLE :
289 (LRU_FILE * !!file + !!active); 288 (LRU_FILE * !!file + !!active);
290 289
291 if (lru == from) 290 if (lru == from)
292 return; 291 return;
293 292
294 MEM_CGROUP_ZSTAT(mz, from) -= 1; 293 MEM_CGROUP_ZSTAT(mz, from) -= 1;
295 /* 294 /*
296 * However this is done under mz->lru_lock, another flags, which 295 * However this is done under mz->lru_lock, another flags, which
297 * are not related to LRU, will be modified from out-of-lock. 296 * are not related to LRU, will be modified from out-of-lock.
298 * We have to use atomic set/clear flags. 297 * We have to use atomic set/clear flags.
299 */ 298 */
300 if (is_unevictable_lru(lru)) { 299 if (is_unevictable_lru(lru)) {
301 ClearPageCgroupActive(pc); 300 ClearPageCgroupActive(pc);
302 SetPageCgroupUnevictable(pc); 301 SetPageCgroupUnevictable(pc);
303 } else { 302 } else {
304 if (is_active_lru(lru)) 303 if (is_active_lru(lru))
305 SetPageCgroupActive(pc); 304 SetPageCgroupActive(pc);
306 else 305 else
307 ClearPageCgroupActive(pc); 306 ClearPageCgroupActive(pc);
308 ClearPageCgroupUnevictable(pc); 307 ClearPageCgroupUnevictable(pc);
309 } 308 }
310 309
311 MEM_CGROUP_ZSTAT(mz, lru) += 1; 310 MEM_CGROUP_ZSTAT(mz, lru) += 1;
312 list_move(&pc->lru, &mz->lists[lru]); 311 list_move(&pc->lru, &mz->lists[lru]);
313 } 312 }
314 313
315 int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *mem) 314 int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *mem)
316 { 315 {
317 int ret; 316 int ret;
318 317
319 task_lock(task); 318 task_lock(task);
320 ret = task->mm && mm_match_cgroup(task->mm, mem); 319 ret = task->mm && mm_match_cgroup(task->mm, mem);
321 task_unlock(task); 320 task_unlock(task);
322 return ret; 321 return ret;
323 } 322 }
324 323
325 /* 324 /*
326 * This routine assumes that the appropriate zone's lru lock is already held 325 * This routine assumes that the appropriate zone's lru lock is already held
327 */ 326 */
328 void mem_cgroup_move_lists(struct page *page, enum lru_list lru) 327 void mem_cgroup_move_lists(struct page *page, enum lru_list lru)
329 { 328 {
330 struct page_cgroup *pc; 329 struct page_cgroup *pc;
331 struct mem_cgroup_per_zone *mz; 330 struct mem_cgroup_per_zone *mz;
332 unsigned long flags; 331 unsigned long flags;
333 332
334 if (mem_cgroup_subsys.disabled) 333 if (mem_cgroup_subsys.disabled)
335 return; 334 return;
336 335
337 /* 336 /*
338 * We cannot lock_page_cgroup while holding zone's lru_lock, 337 * We cannot lock_page_cgroup while holding zone's lru_lock,
339 * because other holders of lock_page_cgroup can be interrupted 338 * because other holders of lock_page_cgroup can be interrupted
340 * with an attempt to rotate_reclaimable_page. But we cannot 339 * with an attempt to rotate_reclaimable_page. But we cannot
341 * safely get to page_cgroup without it, so just try_lock it: 340 * safely get to page_cgroup without it, so just try_lock it:
342 * mem_cgroup_isolate_pages allows for page left on wrong list. 341 * mem_cgroup_isolate_pages allows for page left on wrong list.
343 */ 342 */
344 pc = lookup_page_cgroup(page); 343 pc = lookup_page_cgroup(page);
345 if (!trylock_page_cgroup(pc)) 344 if (!trylock_page_cgroup(pc))
346 return; 345 return;
347 if (pc && PageCgroupUsed(pc)) { 346 if (pc && PageCgroupUsed(pc)) {
348 mz = page_cgroup_zoneinfo(pc); 347 mz = page_cgroup_zoneinfo(pc);
349 spin_lock_irqsave(&mz->lru_lock, flags); 348 spin_lock_irqsave(&mz->lru_lock, flags);
350 __mem_cgroup_move_lists(pc, lru); 349 __mem_cgroup_move_lists(pc, lru);
351 spin_unlock_irqrestore(&mz->lru_lock, flags); 350 spin_unlock_irqrestore(&mz->lru_lock, flags);
352 } 351 }
353 unlock_page_cgroup(pc); 352 unlock_page_cgroup(pc);
354 } 353 }
355 354
356 /* 355 /*
357 * Calculate mapped_ratio under memory controller. This will be used in 356 * Calculate mapped_ratio under memory controller. This will be used in
358 * vmscan.c for deteremining we have to reclaim mapped pages. 357 * vmscan.c for deteremining we have to reclaim mapped pages.
359 */ 358 */
360 int mem_cgroup_calc_mapped_ratio(struct mem_cgroup *mem) 359 int mem_cgroup_calc_mapped_ratio(struct mem_cgroup *mem)
361 { 360 {
362 long total, rss; 361 long total, rss;
363 362
364 /* 363 /*
365 * usage is recorded in bytes. But, here, we assume the number of 364 * usage is recorded in bytes. But, here, we assume the number of
366 * physical pages can be represented by "long" on any arch. 365 * physical pages can be represented by "long" on any arch.
367 */ 366 */
368 total = (long) (mem->res.usage >> PAGE_SHIFT) + 1L; 367 total = (long) (mem->res.usage >> PAGE_SHIFT) + 1L;
369 rss = (long)mem_cgroup_read_stat(&mem->stat, MEM_CGROUP_STAT_RSS); 368 rss = (long)mem_cgroup_read_stat(&mem->stat, MEM_CGROUP_STAT_RSS);
370 return (int)((rss * 100L) / total); 369 return (int)((rss * 100L) / total);
371 } 370 }
372 371
373 /* 372 /*
374 * prev_priority control...this will be used in memory reclaim path. 373 * prev_priority control...this will be used in memory reclaim path.
375 */ 374 */
376 int mem_cgroup_get_reclaim_priority(struct mem_cgroup *mem) 375 int mem_cgroup_get_reclaim_priority(struct mem_cgroup *mem)
377 { 376 {
378 return mem->prev_priority; 377 return mem->prev_priority;
379 } 378 }
380 379
381 void mem_cgroup_note_reclaim_priority(struct mem_cgroup *mem, int priority) 380 void mem_cgroup_note_reclaim_priority(struct mem_cgroup *mem, int priority)
382 { 381 {
383 if (priority < mem->prev_priority) 382 if (priority < mem->prev_priority)
384 mem->prev_priority = priority; 383 mem->prev_priority = priority;
385 } 384 }
386 385
387 void mem_cgroup_record_reclaim_priority(struct mem_cgroup *mem, int priority) 386 void mem_cgroup_record_reclaim_priority(struct mem_cgroup *mem, int priority)
388 { 387 {
389 mem->prev_priority = priority; 388 mem->prev_priority = priority;
390 } 389 }
391 390
392 /* 391 /*
393 * Calculate # of pages to be scanned in this priority/zone. 392 * Calculate # of pages to be scanned in this priority/zone.
394 * See also vmscan.c 393 * See also vmscan.c
395 * 394 *
396 * priority starts from "DEF_PRIORITY" and decremented in each loop. 395 * priority starts from "DEF_PRIORITY" and decremented in each loop.
397 * (see include/linux/mmzone.h) 396 * (see include/linux/mmzone.h)
398 */ 397 */
399 398
400 long mem_cgroup_calc_reclaim(struct mem_cgroup *mem, struct zone *zone, 399 long mem_cgroup_calc_reclaim(struct mem_cgroup *mem, struct zone *zone,
401 int priority, enum lru_list lru) 400 int priority, enum lru_list lru)
402 { 401 {
403 long nr_pages; 402 long nr_pages;
404 int nid = zone->zone_pgdat->node_id; 403 int nid = zone->zone_pgdat->node_id;
405 int zid = zone_idx(zone); 404 int zid = zone_idx(zone);
406 struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(mem, nid, zid); 405 struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(mem, nid, zid);
407 406
408 nr_pages = MEM_CGROUP_ZSTAT(mz, lru); 407 nr_pages = MEM_CGROUP_ZSTAT(mz, lru);
409 408
410 return (nr_pages >> priority); 409 return (nr_pages >> priority);
411 } 410 }
412 411
413 unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan, 412 unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
414 struct list_head *dst, 413 struct list_head *dst,
415 unsigned long *scanned, int order, 414 unsigned long *scanned, int order,
416 int mode, struct zone *z, 415 int mode, struct zone *z,
417 struct mem_cgroup *mem_cont, 416 struct mem_cgroup *mem_cont,
418 int active, int file) 417 int active, int file)
419 { 418 {
420 unsigned long nr_taken = 0; 419 unsigned long nr_taken = 0;
421 struct page *page; 420 struct page *page;
422 unsigned long scan; 421 unsigned long scan;
423 LIST_HEAD(pc_list); 422 LIST_HEAD(pc_list);
424 struct list_head *src; 423 struct list_head *src;
425 struct page_cgroup *pc, *tmp; 424 struct page_cgroup *pc, *tmp;
426 int nid = z->zone_pgdat->node_id; 425 int nid = z->zone_pgdat->node_id;
427 int zid = zone_idx(z); 426 int zid = zone_idx(z);
428 struct mem_cgroup_per_zone *mz; 427 struct mem_cgroup_per_zone *mz;
429 int lru = LRU_FILE * !!file + !!active; 428 int lru = LRU_FILE * !!file + !!active;
430 429
431 BUG_ON(!mem_cont); 430 BUG_ON(!mem_cont);
432 mz = mem_cgroup_zoneinfo(mem_cont, nid, zid); 431 mz = mem_cgroup_zoneinfo(mem_cont, nid, zid);
433 src = &mz->lists[lru]; 432 src = &mz->lists[lru];
434 433
435 spin_lock(&mz->lru_lock); 434 spin_lock(&mz->lru_lock);
436 scan = 0; 435 scan = 0;
437 list_for_each_entry_safe_reverse(pc, tmp, src, lru) { 436 list_for_each_entry_safe_reverse(pc, tmp, src, lru) {
438 if (scan >= nr_to_scan) 437 if (scan >= nr_to_scan)
439 break; 438 break;
440 if (unlikely(!PageCgroupUsed(pc))) 439 if (unlikely(!PageCgroupUsed(pc)))
441 continue; 440 continue;
442 page = pc->page; 441 page = pc->page;
443 442
444 if (unlikely(!PageLRU(page))) 443 if (unlikely(!PageLRU(page)))
445 continue; 444 continue;
446 445
447 /* 446 /*
448 * TODO: play better with lumpy reclaim, grabbing anything. 447 * TODO: play better with lumpy reclaim, grabbing anything.
449 */ 448 */
450 if (PageUnevictable(page) || 449 if (PageUnevictable(page) ||
451 (PageActive(page) && !active) || 450 (PageActive(page) && !active) ||
452 (!PageActive(page) && active)) { 451 (!PageActive(page) && active)) {
453 __mem_cgroup_move_lists(pc, page_lru(page)); 452 __mem_cgroup_move_lists(pc, page_lru(page));
454 continue; 453 continue;
455 } 454 }
456 455
457 scan++; 456 scan++;
458 list_move(&pc->lru, &pc_list); 457 list_move(&pc->lru, &pc_list);
459 458
460 if (__isolate_lru_page(page, mode, file) == 0) { 459 if (__isolate_lru_page(page, mode, file) == 0) {
461 list_move(&page->lru, dst); 460 list_move(&page->lru, dst);
462 nr_taken++; 461 nr_taken++;
463 } 462 }
464 } 463 }
465 464
466 list_splice(&pc_list, src); 465 list_splice(&pc_list, src);
467 spin_unlock(&mz->lru_lock); 466 spin_unlock(&mz->lru_lock);
468 467
469 *scanned = scan; 468 *scanned = scan;
470 return nr_taken; 469 return nr_taken;
471 } 470 }
472 471
473 /* 472 /*
474 * Unlike exported interface, "oom" parameter is added. if oom==true, 473 * Unlike exported interface, "oom" parameter is added. if oom==true,
475 * oom-killer can be invoked. 474 * oom-killer can be invoked.
476 */ 475 */
477 static int __mem_cgroup_try_charge(struct mm_struct *mm, 476 static int __mem_cgroup_try_charge(struct mm_struct *mm,
478 gfp_t gfp_mask, struct mem_cgroup **memcg, bool oom) 477 gfp_t gfp_mask, struct mem_cgroup **memcg, bool oom)
479 { 478 {
480 struct mem_cgroup *mem; 479 struct mem_cgroup *mem;
481 int nr_retries = MEM_CGROUP_RECLAIM_RETRIES; 480 int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
482 /* 481 /*
483 * We always charge the cgroup the mm_struct belongs to. 482 * We always charge the cgroup the mm_struct belongs to.
484 * The mm_struct's mem_cgroup changes on task migration if the 483 * The mm_struct's mem_cgroup changes on task migration if the
485 * thread group leader migrates. It's possible that mm is not 484 * thread group leader migrates. It's possible that mm is not
486 * set, if so charge the init_mm (happens for pagecache usage). 485 * set, if so charge the init_mm (happens for pagecache usage).
487 */ 486 */
488 if (likely(!*memcg)) { 487 if (likely(!*memcg)) {
489 rcu_read_lock(); 488 rcu_read_lock();
490 mem = mem_cgroup_from_task(rcu_dereference(mm->owner)); 489 mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
491 if (unlikely(!mem)) { 490 if (unlikely(!mem)) {
492 rcu_read_unlock(); 491 rcu_read_unlock();
493 return 0; 492 return 0;
494 } 493 }
495 /* 494 /*
496 * For every charge from the cgroup, increment reference count 495 * For every charge from the cgroup, increment reference count
497 */ 496 */
498 css_get(&mem->css); 497 css_get(&mem->css);
499 *memcg = mem; 498 *memcg = mem;
500 rcu_read_unlock(); 499 rcu_read_unlock();
501 } else { 500 } else {
502 mem = *memcg; 501 mem = *memcg;
503 css_get(&mem->css); 502 css_get(&mem->css);
504 } 503 }
505 504
506 505
507 while (unlikely(res_counter_charge(&mem->res, PAGE_SIZE))) { 506 while (unlikely(res_counter_charge(&mem->res, PAGE_SIZE))) {
508 if (!(gfp_mask & __GFP_WAIT)) 507 if (!(gfp_mask & __GFP_WAIT))
509 goto nomem; 508 goto nomem;
510 509
511 if (try_to_free_mem_cgroup_pages(mem, gfp_mask)) 510 if (try_to_free_mem_cgroup_pages(mem, gfp_mask))
512 continue; 511 continue;
513 512
514 /* 513 /*
515 * try_to_free_mem_cgroup_pages() might not give us a full 514 * try_to_free_mem_cgroup_pages() might not give us a full
516 * picture of reclaim. Some pages are reclaimed and might be 515 * picture of reclaim. Some pages are reclaimed and might be
517 * moved to swap cache or just unmapped from the cgroup. 516 * moved to swap cache or just unmapped from the cgroup.
518 * Check the limit again to see if the reclaim reduced the 517 * Check the limit again to see if the reclaim reduced the
519 * current usage of the cgroup before giving up 518 * current usage of the cgroup before giving up
520 */ 519 */
521 if (res_counter_check_under_limit(&mem->res)) 520 if (res_counter_check_under_limit(&mem->res))
522 continue; 521 continue;
523 522
524 if (!nr_retries--) { 523 if (!nr_retries--) {
525 if (oom) 524 if (oom)
526 mem_cgroup_out_of_memory(mem, gfp_mask); 525 mem_cgroup_out_of_memory(mem, gfp_mask);
527 goto nomem; 526 goto nomem;
528 } 527 }
529 } 528 }
530 return 0; 529 return 0;
531 nomem: 530 nomem:
532 css_put(&mem->css); 531 css_put(&mem->css);
533 return -ENOMEM; 532 return -ENOMEM;
534 } 533 }
535 534
536 /** 535 /**
537 * mem_cgroup_try_charge - get charge of PAGE_SIZE. 536 * mem_cgroup_try_charge - get charge of PAGE_SIZE.
538 * @mm: an mm_struct which is charged against. (when *memcg is NULL) 537 * @mm: an mm_struct which is charged against. (when *memcg is NULL)
539 * @gfp_mask: gfp_mask for reclaim. 538 * @gfp_mask: gfp_mask for reclaim.
540 * @memcg: a pointer to memory cgroup which is charged against. 539 * @memcg: a pointer to memory cgroup which is charged against.
541 * 540 *
542 * charge against memory cgroup pointed by *memcg. if *memcg == NULL, estimated 541 * charge against memory cgroup pointed by *memcg. if *memcg == NULL, estimated
543 * memory cgroup from @mm is got and stored in *memcg. 542 * memory cgroup from @mm is got and stored in *memcg.
544 * 543 *
545 * Returns 0 if success. -ENOMEM at failure. 544 * Returns 0 if success. -ENOMEM at failure.
546 * This call can invoke OOM-Killer. 545 * This call can invoke OOM-Killer.
547 */ 546 */
548 547
549 int mem_cgroup_try_charge(struct mm_struct *mm, 548 int mem_cgroup_try_charge(struct mm_struct *mm,
550 gfp_t mask, struct mem_cgroup **memcg) 549 gfp_t mask, struct mem_cgroup **memcg)
551 { 550 {
552 return __mem_cgroup_try_charge(mm, mask, memcg, true); 551 return __mem_cgroup_try_charge(mm, mask, memcg, true);
553 } 552 }
554 553
555 /* 554 /*
556 * commit a charge got by mem_cgroup_try_charge() and makes page_cgroup to be 555 * commit a charge got by mem_cgroup_try_charge() and makes page_cgroup to be
557 * USED state. If already USED, uncharge and return. 556 * USED state. If already USED, uncharge and return.
558 */ 557 */
559 558
560 static void __mem_cgroup_commit_charge(struct mem_cgroup *mem, 559 static void __mem_cgroup_commit_charge(struct mem_cgroup *mem,
561 struct page_cgroup *pc, 560 struct page_cgroup *pc,
562 enum charge_type ctype) 561 enum charge_type ctype)
563 { 562 {
564 struct mem_cgroup_per_zone *mz; 563 struct mem_cgroup_per_zone *mz;
565 unsigned long flags; 564 unsigned long flags;
566 565
567 /* try_charge() can return NULL to *memcg, taking care of it. */ 566 /* try_charge() can return NULL to *memcg, taking care of it. */
568 if (!mem) 567 if (!mem)
569 return; 568 return;
570 569
571 lock_page_cgroup(pc); 570 lock_page_cgroup(pc);
572 if (unlikely(PageCgroupUsed(pc))) { 571 if (unlikely(PageCgroupUsed(pc))) {
573 unlock_page_cgroup(pc); 572 unlock_page_cgroup(pc);
574 res_counter_uncharge(&mem->res, PAGE_SIZE); 573 res_counter_uncharge(&mem->res, PAGE_SIZE);
575 css_put(&mem->css); 574 css_put(&mem->css);
576 return; 575 return;
577 } 576 }
578 pc->mem_cgroup = mem; 577 pc->mem_cgroup = mem;
579 /* 578 /*
580 * If a page is accounted as a page cache, insert to inactive list. 579 * If a page is accounted as a page cache, insert to inactive list.
581 * If anon, insert to active list. 580 * If anon, insert to active list.
582 */ 581 */
583 pc->flags = pcg_default_flags[ctype]; 582 pc->flags = pcg_default_flags[ctype];
584 583
585 mz = page_cgroup_zoneinfo(pc); 584 mz = page_cgroup_zoneinfo(pc);
586 585
587 spin_lock_irqsave(&mz->lru_lock, flags); 586 spin_lock_irqsave(&mz->lru_lock, flags);
588 __mem_cgroup_add_list(mz, pc, true); 587 __mem_cgroup_add_list(mz, pc, true);
589 spin_unlock_irqrestore(&mz->lru_lock, flags); 588 spin_unlock_irqrestore(&mz->lru_lock, flags);
590 unlock_page_cgroup(pc); 589 unlock_page_cgroup(pc);
591 } 590 }
592 591
593 /** 592 /**
594 * mem_cgroup_move_account - move account of the page 593 * mem_cgroup_move_account - move account of the page
595 * @pc: page_cgroup of the page. 594 * @pc: page_cgroup of the page.
596 * @from: mem_cgroup which the page is moved from. 595 * @from: mem_cgroup which the page is moved from.
597 * @to: mem_cgroup which the page is moved to. @from != @to. 596 * @to: mem_cgroup which the page is moved to. @from != @to.
598 * 597 *
599 * The caller must confirm following. 598 * The caller must confirm following.
600 * 1. disable irq. 599 * 1. disable irq.
601 * 2. lru_lock of old mem_cgroup(@from) should be held. 600 * 2. lru_lock of old mem_cgroup(@from) should be held.
602 * 601 *
603 * returns 0 at success, 602 * returns 0 at success,
604 * returns -EBUSY when lock is busy or "pc" is unstable. 603 * returns -EBUSY when lock is busy or "pc" is unstable.
605 * 604 *
606 * This function does "uncharge" from old cgroup but doesn't do "charge" to 605 * This function does "uncharge" from old cgroup but doesn't do "charge" to
607 * new cgroup. It should be done by a caller. 606 * new cgroup. It should be done by a caller.
608 */ 607 */
609 608
610 static int mem_cgroup_move_account(struct page_cgroup *pc, 609 static int mem_cgroup_move_account(struct page_cgroup *pc,
611 struct mem_cgroup *from, struct mem_cgroup *to) 610 struct mem_cgroup *from, struct mem_cgroup *to)
612 { 611 {
613 struct mem_cgroup_per_zone *from_mz, *to_mz; 612 struct mem_cgroup_per_zone *from_mz, *to_mz;
614 int nid, zid; 613 int nid, zid;
615 int ret = -EBUSY; 614 int ret = -EBUSY;
616 615
617 VM_BUG_ON(!irqs_disabled()); 616 VM_BUG_ON(!irqs_disabled());
618 VM_BUG_ON(from == to); 617 VM_BUG_ON(from == to);
619 618
620 nid = page_cgroup_nid(pc); 619 nid = page_cgroup_nid(pc);
621 zid = page_cgroup_zid(pc); 620 zid = page_cgroup_zid(pc);
622 from_mz = mem_cgroup_zoneinfo(from, nid, zid); 621 from_mz = mem_cgroup_zoneinfo(from, nid, zid);
623 to_mz = mem_cgroup_zoneinfo(to, nid, zid); 622 to_mz = mem_cgroup_zoneinfo(to, nid, zid);
624 623
625 624
626 if (!trylock_page_cgroup(pc)) 625 if (!trylock_page_cgroup(pc))
627 return ret; 626 return ret;
628 627
629 if (!PageCgroupUsed(pc)) 628 if (!PageCgroupUsed(pc))
630 goto out; 629 goto out;
631 630
632 if (pc->mem_cgroup != from) 631 if (pc->mem_cgroup != from)
633 goto out; 632 goto out;
634 633
635 if (spin_trylock(&to_mz->lru_lock)) { 634 if (spin_trylock(&to_mz->lru_lock)) {
636 __mem_cgroup_remove_list(from_mz, pc); 635 __mem_cgroup_remove_list(from_mz, pc);
637 css_put(&from->css); 636 css_put(&from->css);
638 res_counter_uncharge(&from->res, PAGE_SIZE); 637 res_counter_uncharge(&from->res, PAGE_SIZE);
639 pc->mem_cgroup = to; 638 pc->mem_cgroup = to;
640 css_get(&to->css); 639 css_get(&to->css);
641 __mem_cgroup_add_list(to_mz, pc, false); 640 __mem_cgroup_add_list(to_mz, pc, false);
642 ret = 0; 641 ret = 0;
643 spin_unlock(&to_mz->lru_lock); 642 spin_unlock(&to_mz->lru_lock);
644 } 643 }
645 out: 644 out:
646 unlock_page_cgroup(pc); 645 unlock_page_cgroup(pc);
647 return ret; 646 return ret;
648 } 647 }
649 648
650 /* 649 /*
651 * move charges to its parent. 650 * move charges to its parent.
652 */ 651 */
653 652
654 static int mem_cgroup_move_parent(struct page_cgroup *pc, 653 static int mem_cgroup_move_parent(struct page_cgroup *pc,
655 struct mem_cgroup *child, 654 struct mem_cgroup *child,
656 gfp_t gfp_mask) 655 gfp_t gfp_mask)
657 { 656 {
658 struct cgroup *cg = child->css.cgroup; 657 struct cgroup *cg = child->css.cgroup;
659 struct cgroup *pcg = cg->parent; 658 struct cgroup *pcg = cg->parent;
660 struct mem_cgroup *parent; 659 struct mem_cgroup *parent;
661 struct mem_cgroup_per_zone *mz; 660 struct mem_cgroup_per_zone *mz;
662 unsigned long flags; 661 unsigned long flags;
663 int ret; 662 int ret;
664 663
665 /* Is ROOT ? */ 664 /* Is ROOT ? */
666 if (!pcg) 665 if (!pcg)
667 return -EINVAL; 666 return -EINVAL;
668 667
669 parent = mem_cgroup_from_cont(pcg); 668 parent = mem_cgroup_from_cont(pcg);
670 669
671 ret = __mem_cgroup_try_charge(NULL, gfp_mask, &parent, false); 670 ret = __mem_cgroup_try_charge(NULL, gfp_mask, &parent, false);
672 if (ret) 671 if (ret)
673 return ret; 672 return ret;
674 673
675 mz = mem_cgroup_zoneinfo(child, 674 mz = mem_cgroup_zoneinfo(child,
676 page_cgroup_nid(pc), page_cgroup_zid(pc)); 675 page_cgroup_nid(pc), page_cgroup_zid(pc));
677 676
678 spin_lock_irqsave(&mz->lru_lock, flags); 677 spin_lock_irqsave(&mz->lru_lock, flags);
679 ret = mem_cgroup_move_account(pc, child, parent); 678 ret = mem_cgroup_move_account(pc, child, parent);
680 spin_unlock_irqrestore(&mz->lru_lock, flags); 679 spin_unlock_irqrestore(&mz->lru_lock, flags);
681 680
682 /* drop extra refcnt */ 681 /* drop extra refcnt */
683 css_put(&parent->css); 682 css_put(&parent->css);
684 /* uncharge if move fails */ 683 /* uncharge if move fails */
685 if (ret) 684 if (ret)
686 res_counter_uncharge(&parent->res, PAGE_SIZE); 685 res_counter_uncharge(&parent->res, PAGE_SIZE);
687 686
688 return ret; 687 return ret;
689 } 688 }
690 689
691 /* 690 /*
692 * Charge the memory controller for page usage. 691 * Charge the memory controller for page usage.
693 * Return 692 * Return
694 * 0 if the charge was successful 693 * 0 if the charge was successful
695 * < 0 if the cgroup is over its limit 694 * < 0 if the cgroup is over its limit
696 */ 695 */
697 static int mem_cgroup_charge_common(struct page *page, struct mm_struct *mm, 696 static int mem_cgroup_charge_common(struct page *page, struct mm_struct *mm,
698 gfp_t gfp_mask, enum charge_type ctype, 697 gfp_t gfp_mask, enum charge_type ctype,
699 struct mem_cgroup *memcg) 698 struct mem_cgroup *memcg)
700 { 699 {
701 struct mem_cgroup *mem; 700 struct mem_cgroup *mem;
702 struct page_cgroup *pc; 701 struct page_cgroup *pc;
703 int ret; 702 int ret;
704 703
705 pc = lookup_page_cgroup(page); 704 pc = lookup_page_cgroup(page);
706 /* can happen at boot */ 705 /* can happen at boot */
707 if (unlikely(!pc)) 706 if (unlikely(!pc))
708 return 0; 707 return 0;
709 prefetchw(pc); 708 prefetchw(pc);
710 709
711 mem = memcg; 710 mem = memcg;
712 ret = __mem_cgroup_try_charge(mm, gfp_mask, &mem, true); 711 ret = __mem_cgroup_try_charge(mm, gfp_mask, &mem, true);
713 if (ret) 712 if (ret)
714 return ret; 713 return ret;
715 714
716 __mem_cgroup_commit_charge(mem, pc, ctype); 715 __mem_cgroup_commit_charge(mem, pc, ctype);
717 return 0; 716 return 0;
718 } 717 }
719 718
720 int mem_cgroup_newpage_charge(struct page *page, 719 int mem_cgroup_newpage_charge(struct page *page,
721 struct mm_struct *mm, gfp_t gfp_mask) 720 struct mm_struct *mm, gfp_t gfp_mask)
722 { 721 {
723 if (mem_cgroup_subsys.disabled) 722 if (mem_cgroup_subsys.disabled)
724 return 0; 723 return 0;
725 if (PageCompound(page)) 724 if (PageCompound(page))
726 return 0; 725 return 0;
727 /* 726 /*
728 * If already mapped, we don't have to account. 727 * If already mapped, we don't have to account.
729 * If page cache, page->mapping has address_space. 728 * If page cache, page->mapping has address_space.
730 * But page->mapping may have out-of-use anon_vma pointer, 729 * But page->mapping may have out-of-use anon_vma pointer,
731 * detecit it by PageAnon() check. newly-mapped-anon's page->mapping 730 * detecit it by PageAnon() check. newly-mapped-anon's page->mapping
732 * is NULL. 731 * is NULL.
733 */ 732 */
734 if (page_mapped(page) || (page->mapping && !PageAnon(page))) 733 if (page_mapped(page) || (page->mapping && !PageAnon(page)))
735 return 0; 734 return 0;
736 if (unlikely(!mm)) 735 if (unlikely(!mm))
737 mm = &init_mm; 736 mm = &init_mm;
738 return mem_cgroup_charge_common(page, mm, gfp_mask, 737 return mem_cgroup_charge_common(page, mm, gfp_mask,
739 MEM_CGROUP_CHARGE_TYPE_MAPPED, NULL); 738 MEM_CGROUP_CHARGE_TYPE_MAPPED, NULL);
740 } 739 }
741 740
742 int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm, 741 int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
743 gfp_t gfp_mask) 742 gfp_t gfp_mask)
744 { 743 {
745 if (mem_cgroup_subsys.disabled) 744 if (mem_cgroup_subsys.disabled)
746 return 0; 745 return 0;
747 if (PageCompound(page)) 746 if (PageCompound(page))
748 return 0; 747 return 0;
749 /* 748 /*
750 * Corner case handling. This is called from add_to_page_cache() 749 * Corner case handling. This is called from add_to_page_cache()
751 * in usual. But some FS (shmem) precharges this page before calling it 750 * in usual. But some FS (shmem) precharges this page before calling it
752 * and call add_to_page_cache() with GFP_NOWAIT. 751 * and call add_to_page_cache() with GFP_NOWAIT.
753 * 752 *
754 * For GFP_NOWAIT case, the page may be pre-charged before calling 753 * For GFP_NOWAIT case, the page may be pre-charged before calling
755 * add_to_page_cache(). (See shmem.c) check it here and avoid to call 754 * add_to_page_cache(). (See shmem.c) check it here and avoid to call
756 * charge twice. (It works but has to pay a bit larger cost.) 755 * charge twice. (It works but has to pay a bit larger cost.)
757 */ 756 */
758 if (!(gfp_mask & __GFP_WAIT)) { 757 if (!(gfp_mask & __GFP_WAIT)) {
759 struct page_cgroup *pc; 758 struct page_cgroup *pc;
760 759
761 760
762 pc = lookup_page_cgroup(page); 761 pc = lookup_page_cgroup(page);
763 if (!pc) 762 if (!pc)
764 return 0; 763 return 0;
765 lock_page_cgroup(pc); 764 lock_page_cgroup(pc);
766 if (PageCgroupUsed(pc)) { 765 if (PageCgroupUsed(pc)) {
767 unlock_page_cgroup(pc); 766 unlock_page_cgroup(pc);
768 return 0; 767 return 0;
769 } 768 }
770 unlock_page_cgroup(pc); 769 unlock_page_cgroup(pc);
771 } 770 }
772 771
773 if (unlikely(!mm)) 772 if (unlikely(!mm))
774 mm = &init_mm; 773 mm = &init_mm;
775 774
776 if (page_is_file_cache(page)) 775 if (page_is_file_cache(page))
777 return mem_cgroup_charge_common(page, mm, gfp_mask, 776 return mem_cgroup_charge_common(page, mm, gfp_mask,
778 MEM_CGROUP_CHARGE_TYPE_CACHE, NULL); 777 MEM_CGROUP_CHARGE_TYPE_CACHE, NULL);
779 else 778 else
780 return mem_cgroup_charge_common(page, mm, gfp_mask, 779 return mem_cgroup_charge_common(page, mm, gfp_mask,
781 MEM_CGROUP_CHARGE_TYPE_SHMEM, NULL); 780 MEM_CGROUP_CHARGE_TYPE_SHMEM, NULL);
782 } 781 }
783 782
784 void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr) 783 void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
785 { 784 {
786 struct page_cgroup *pc; 785 struct page_cgroup *pc;
787 786
788 if (mem_cgroup_subsys.disabled) 787 if (mem_cgroup_subsys.disabled)
789 return; 788 return;
790 if (!ptr) 789 if (!ptr)
791 return; 790 return;
792 pc = lookup_page_cgroup(page); 791 pc = lookup_page_cgroup(page);
793 __mem_cgroup_commit_charge(ptr, pc, MEM_CGROUP_CHARGE_TYPE_MAPPED); 792 __mem_cgroup_commit_charge(ptr, pc, MEM_CGROUP_CHARGE_TYPE_MAPPED);
794 } 793 }
795 794
796 void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *mem) 795 void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *mem)
797 { 796 {
798 if (mem_cgroup_subsys.disabled) 797 if (mem_cgroup_subsys.disabled)
799 return; 798 return;
800 if (!mem) 799 if (!mem)
801 return; 800 return;
802 res_counter_uncharge(&mem->res, PAGE_SIZE); 801 res_counter_uncharge(&mem->res, PAGE_SIZE);
803 css_put(&mem->css); 802 css_put(&mem->css);
804 } 803 }
805 804
806 805
807 /* 806 /*
808 * uncharge if !page_mapped(page) 807 * uncharge if !page_mapped(page)
809 */ 808 */
810 static void 809 static void
811 __mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype) 810 __mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
812 { 811 {
813 struct page_cgroup *pc; 812 struct page_cgroup *pc;
814 struct mem_cgroup *mem; 813 struct mem_cgroup *mem;
815 struct mem_cgroup_per_zone *mz; 814 struct mem_cgroup_per_zone *mz;
816 unsigned long flags; 815 unsigned long flags;
817 816
818 if (mem_cgroup_subsys.disabled) 817 if (mem_cgroup_subsys.disabled)
819 return; 818 return;
820 819
821 /* 820 /*
822 * Check if our page_cgroup is valid 821 * Check if our page_cgroup is valid
823 */ 822 */
824 pc = lookup_page_cgroup(page); 823 pc = lookup_page_cgroup(page);
825 if (unlikely(!pc || !PageCgroupUsed(pc))) 824 if (unlikely(!pc || !PageCgroupUsed(pc)))
826 return; 825 return;
827 826
828 lock_page_cgroup(pc); 827 lock_page_cgroup(pc);
829 if ((ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED && page_mapped(page)) 828 if ((ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED && page_mapped(page))
830 || !PageCgroupUsed(pc)) { 829 || !PageCgroupUsed(pc)) {
831 /* This happens at race in zap_pte_range() and do_swap_page()*/ 830 /* This happens at race in zap_pte_range() and do_swap_page()*/
832 unlock_page_cgroup(pc); 831 unlock_page_cgroup(pc);
833 return; 832 return;
834 } 833 }
835 ClearPageCgroupUsed(pc); 834 ClearPageCgroupUsed(pc);
836 mem = pc->mem_cgroup; 835 mem = pc->mem_cgroup;
837 836
838 mz = page_cgroup_zoneinfo(pc); 837 mz = page_cgroup_zoneinfo(pc);
839 spin_lock_irqsave(&mz->lru_lock, flags); 838 spin_lock_irqsave(&mz->lru_lock, flags);
840 __mem_cgroup_remove_list(mz, pc); 839 __mem_cgroup_remove_list(mz, pc);
841 spin_unlock_irqrestore(&mz->lru_lock, flags); 840 spin_unlock_irqrestore(&mz->lru_lock, flags);
842 unlock_page_cgroup(pc); 841 unlock_page_cgroup(pc);
843 842
844 res_counter_uncharge(&mem->res, PAGE_SIZE); 843 res_counter_uncharge(&mem->res, PAGE_SIZE);
845 css_put(&mem->css); 844 css_put(&mem->css);
846 845
847 return; 846 return;
848 } 847 }
849 848
850 void mem_cgroup_uncharge_page(struct page *page) 849 void mem_cgroup_uncharge_page(struct page *page)
851 { 850 {
852 /* early check. */ 851 /* early check. */
853 if (page_mapped(page)) 852 if (page_mapped(page))
854 return; 853 return;
855 if (page->mapping && !PageAnon(page)) 854 if (page->mapping && !PageAnon(page))
856 return; 855 return;
857 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_MAPPED); 856 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_MAPPED);
858 } 857 }
859 858
860 void mem_cgroup_uncharge_cache_page(struct page *page) 859 void mem_cgroup_uncharge_cache_page(struct page *page)
861 { 860 {
862 VM_BUG_ON(page_mapped(page)); 861 VM_BUG_ON(page_mapped(page));
863 VM_BUG_ON(page->mapping); 862 VM_BUG_ON(page->mapping);
864 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE); 863 __mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
865 } 864 }
866 865
867 /* 866 /*
868 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old 867 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
869 * page belongs to. 868 * page belongs to.
870 */ 869 */
871 int mem_cgroup_prepare_migration(struct page *page, struct mem_cgroup **ptr) 870 int mem_cgroup_prepare_migration(struct page *page, struct mem_cgroup **ptr)
872 { 871 {
873 struct page_cgroup *pc; 872 struct page_cgroup *pc;
874 struct mem_cgroup *mem = NULL; 873 struct mem_cgroup *mem = NULL;
875 int ret = 0; 874 int ret = 0;
876 875
877 if (mem_cgroup_subsys.disabled) 876 if (mem_cgroup_subsys.disabled)
878 return 0; 877 return 0;
879 878
880 pc = lookup_page_cgroup(page); 879 pc = lookup_page_cgroup(page);
881 lock_page_cgroup(pc); 880 lock_page_cgroup(pc);
882 if (PageCgroupUsed(pc)) { 881 if (PageCgroupUsed(pc)) {
883 mem = pc->mem_cgroup; 882 mem = pc->mem_cgroup;
884 css_get(&mem->css); 883 css_get(&mem->css);
885 } 884 }
886 unlock_page_cgroup(pc); 885 unlock_page_cgroup(pc);
887 886
888 if (mem) { 887 if (mem) {
889 ret = mem_cgroup_try_charge(NULL, GFP_HIGHUSER_MOVABLE, &mem); 888 ret = mem_cgroup_try_charge(NULL, GFP_HIGHUSER_MOVABLE, &mem);
890 css_put(&mem->css); 889 css_put(&mem->css);
891 } 890 }
892 *ptr = mem; 891 *ptr = mem;
893 return ret; 892 return ret;
894 } 893 }
895 894
896 /* remove redundant charge if migration failed*/ 895 /* remove redundant charge if migration failed*/
897 void mem_cgroup_end_migration(struct mem_cgroup *mem, 896 void mem_cgroup_end_migration(struct mem_cgroup *mem,
898 struct page *oldpage, struct page *newpage) 897 struct page *oldpage, struct page *newpage)
899 { 898 {
900 struct page *target, *unused; 899 struct page *target, *unused;
901 struct page_cgroup *pc; 900 struct page_cgroup *pc;
902 enum charge_type ctype; 901 enum charge_type ctype;
903 902
904 if (!mem) 903 if (!mem)
905 return; 904 return;
906 905
907 /* at migration success, oldpage->mapping is NULL. */ 906 /* at migration success, oldpage->mapping is NULL. */
908 if (oldpage->mapping) { 907 if (oldpage->mapping) {
909 target = oldpage; 908 target = oldpage;
910 unused = NULL; 909 unused = NULL;
911 } else { 910 } else {
912 target = newpage; 911 target = newpage;
913 unused = oldpage; 912 unused = oldpage;
914 } 913 }
915 914
916 if (PageAnon(target)) 915 if (PageAnon(target))
917 ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED; 916 ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED;
918 else if (page_is_file_cache(target)) 917 else if (page_is_file_cache(target))
919 ctype = MEM_CGROUP_CHARGE_TYPE_CACHE; 918 ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
920 else 919 else
921 ctype = MEM_CGROUP_CHARGE_TYPE_SHMEM; 920 ctype = MEM_CGROUP_CHARGE_TYPE_SHMEM;
922 921
923 /* unused page is not on radix-tree now. */ 922 /* unused page is not on radix-tree now. */
924 if (unused && ctype != MEM_CGROUP_CHARGE_TYPE_MAPPED) 923 if (unused && ctype != MEM_CGROUP_CHARGE_TYPE_MAPPED)
925 __mem_cgroup_uncharge_common(unused, ctype); 924 __mem_cgroup_uncharge_common(unused, ctype);
926 925
927 pc = lookup_page_cgroup(target); 926 pc = lookup_page_cgroup(target);
928 /* 927 /*
929 * __mem_cgroup_commit_charge() check PCG_USED bit of page_cgroup. 928 * __mem_cgroup_commit_charge() check PCG_USED bit of page_cgroup.
930 * So, double-counting is effectively avoided. 929 * So, double-counting is effectively avoided.
931 */ 930 */
932 __mem_cgroup_commit_charge(mem, pc, ctype); 931 __mem_cgroup_commit_charge(mem, pc, ctype);
933 932
934 /* 933 /*
935 * Both of oldpage and newpage are still under lock_page(). 934 * Both of oldpage and newpage are still under lock_page().
936 * Then, we don't have to care about race in radix-tree. 935 * Then, we don't have to care about race in radix-tree.
937 * But we have to be careful that this page is unmapped or not. 936 * But we have to be careful that this page is unmapped or not.
938 * 937 *
939 * There is a case for !page_mapped(). At the start of 938 * There is a case for !page_mapped(). At the start of
940 * migration, oldpage was mapped. But now, it's zapped. 939 * migration, oldpage was mapped. But now, it's zapped.
941 * But we know *target* page is not freed/reused under us. 940 * But we know *target* page is not freed/reused under us.
942 * mem_cgroup_uncharge_page() does all necessary checks. 941 * mem_cgroup_uncharge_page() does all necessary checks.
943 */ 942 */
944 if (ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED) 943 if (ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED)
945 mem_cgroup_uncharge_page(target); 944 mem_cgroup_uncharge_page(target);
946 } 945 }
947 946
948 /* 947 /*
949 * A call to try to shrink memory usage under specified resource controller. 948 * A call to try to shrink memory usage under specified resource controller.
950 * This is typically used for page reclaiming for shmem for reducing side 949 * This is typically used for page reclaiming for shmem for reducing side
951 * effect of page allocation from shmem, which is used by some mem_cgroup. 950 * effect of page allocation from shmem, which is used by some mem_cgroup.
952 */ 951 */
953 int mem_cgroup_shrink_usage(struct mm_struct *mm, gfp_t gfp_mask) 952 int mem_cgroup_shrink_usage(struct mm_struct *mm, gfp_t gfp_mask)
954 { 953 {
955 struct mem_cgroup *mem; 954 struct mem_cgroup *mem;
956 int progress = 0; 955 int progress = 0;
957 int retry = MEM_CGROUP_RECLAIM_RETRIES; 956 int retry = MEM_CGROUP_RECLAIM_RETRIES;
958 957
959 if (mem_cgroup_subsys.disabled) 958 if (mem_cgroup_subsys.disabled)
960 return 0; 959 return 0;
961 if (!mm) 960 if (!mm)
962 return 0; 961 return 0;
963 962
964 rcu_read_lock(); 963 rcu_read_lock();
965 mem = mem_cgroup_from_task(rcu_dereference(mm->owner)); 964 mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
966 if (unlikely(!mem)) { 965 if (unlikely(!mem)) {
967 rcu_read_unlock(); 966 rcu_read_unlock();
968 return 0; 967 return 0;
969 } 968 }
970 css_get(&mem->css); 969 css_get(&mem->css);
971 rcu_read_unlock(); 970 rcu_read_unlock();
972 971
973 do { 972 do {
974 progress = try_to_free_mem_cgroup_pages(mem, gfp_mask); 973 progress = try_to_free_mem_cgroup_pages(mem, gfp_mask);
975 progress += res_counter_check_under_limit(&mem->res); 974 progress += res_counter_check_under_limit(&mem->res);
976 } while (!progress && --retry); 975 } while (!progress && --retry);
977 976
978 css_put(&mem->css); 977 css_put(&mem->css);
979 if (!retry) 978 if (!retry)
980 return -ENOMEM; 979 return -ENOMEM;
981 return 0; 980 return 0;
982 } 981 }
983 982
984 static int mem_cgroup_resize_limit(struct mem_cgroup *memcg, 983 static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
985 unsigned long long val) 984 unsigned long long val)
986 { 985 {
987 986
988 int retry_count = MEM_CGROUP_RECLAIM_RETRIES; 987 int retry_count = MEM_CGROUP_RECLAIM_RETRIES;
989 int progress; 988 int progress;
990 int ret = 0; 989 int ret = 0;
991 990
992 while (res_counter_set_limit(&memcg->res, val)) { 991 while (res_counter_set_limit(&memcg->res, val)) {
993 if (signal_pending(current)) { 992 if (signal_pending(current)) {
994 ret = -EINTR; 993 ret = -EINTR;
995 break; 994 break;
996 } 995 }
997 if (!retry_count) { 996 if (!retry_count) {
998 ret = -EBUSY; 997 ret = -EBUSY;
999 break; 998 break;
1000 } 999 }
1001 progress = try_to_free_mem_cgroup_pages(memcg, 1000 progress = try_to_free_mem_cgroup_pages(memcg,
1002 GFP_HIGHUSER_MOVABLE); 1001 GFP_HIGHUSER_MOVABLE);
1003 if (!progress) 1002 if (!progress)
1004 retry_count--; 1003 retry_count--;
1005 } 1004 }
1006 return ret; 1005 return ret;
1007 } 1006 }
1008 1007
1009 1008
1010 /* 1009 /*
1011 * This routine traverse page_cgroup in given list and drop them all. 1010 * This routine traverse page_cgroup in given list and drop them all.
1012 * *And* this routine doesn't reclaim page itself, just removes page_cgroup. 1011 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
1013 */ 1012 */
1014 static int mem_cgroup_force_empty_list(struct mem_cgroup *mem, 1013 static int mem_cgroup_force_empty_list(struct mem_cgroup *mem,
1015 struct mem_cgroup_per_zone *mz, 1014 struct mem_cgroup_per_zone *mz,
1016 enum lru_list lru) 1015 enum lru_list lru)
1017 { 1016 {
1018 struct page_cgroup *pc, *busy; 1017 struct page_cgroup *pc, *busy;
1019 unsigned long flags; 1018 unsigned long flags;
1020 unsigned long loop; 1019 unsigned long loop;
1021 struct list_head *list; 1020 struct list_head *list;
1022 int ret = 0; 1021 int ret = 0;
1023 1022
1024 list = &mz->lists[lru]; 1023 list = &mz->lists[lru];
1025 1024
1026 loop = MEM_CGROUP_ZSTAT(mz, lru); 1025 loop = MEM_CGROUP_ZSTAT(mz, lru);
1027 /* give some margin against EBUSY etc...*/ 1026 /* give some margin against EBUSY etc...*/
1028 loop += 256; 1027 loop += 256;
1029 busy = NULL; 1028 busy = NULL;
1030 while (loop--) { 1029 while (loop--) {
1031 ret = 0; 1030 ret = 0;
1032 spin_lock_irqsave(&mz->lru_lock, flags); 1031 spin_lock_irqsave(&mz->lru_lock, flags);
1033 if (list_empty(list)) { 1032 if (list_empty(list)) {
1034 spin_unlock_irqrestore(&mz->lru_lock, flags); 1033 spin_unlock_irqrestore(&mz->lru_lock, flags);
1035 break; 1034 break;
1036 } 1035 }
1037 pc = list_entry(list->prev, struct page_cgroup, lru); 1036 pc = list_entry(list->prev, struct page_cgroup, lru);
1038 if (busy == pc) { 1037 if (busy == pc) {
1039 list_move(&pc->lru, list); 1038 list_move(&pc->lru, list);
1040 busy = 0; 1039 busy = 0;
1041 spin_unlock_irqrestore(&mz->lru_lock, flags); 1040 spin_unlock_irqrestore(&mz->lru_lock, flags);
1042 continue; 1041 continue;
1043 } 1042 }
1044 spin_unlock_irqrestore(&mz->lru_lock, flags); 1043 spin_unlock_irqrestore(&mz->lru_lock, flags);
1045 1044
1046 ret = mem_cgroup_move_parent(pc, mem, GFP_HIGHUSER_MOVABLE); 1045 ret = mem_cgroup_move_parent(pc, mem, GFP_HIGHUSER_MOVABLE);
1047 if (ret == -ENOMEM) 1046 if (ret == -ENOMEM)
1048 break; 1047 break;
1049 1048
1050 if (ret == -EBUSY || ret == -EINVAL) { 1049 if (ret == -EBUSY || ret == -EINVAL) {
1051 /* found lock contention or "pc" is obsolete. */ 1050 /* found lock contention or "pc" is obsolete. */
1052 busy = pc; 1051 busy = pc;
1053 cond_resched(); 1052 cond_resched();
1054 } else 1053 } else
1055 busy = NULL; 1054 busy = NULL;
1056 } 1055 }
1057 if (!ret && !list_empty(list)) 1056 if (!ret && !list_empty(list))
1058 return -EBUSY; 1057 return -EBUSY;
1059 return ret; 1058 return ret;
1060 } 1059 }
1061 1060
1062 /* 1061 /*
1063 * make mem_cgroup's charge to be 0 if there is no task. 1062 * make mem_cgroup's charge to be 0 if there is no task.
1064 * This enables deleting this mem_cgroup. 1063 * This enables deleting this mem_cgroup.
1065 */ 1064 */
1066 static int mem_cgroup_force_empty(struct mem_cgroup *mem) 1065 static int mem_cgroup_force_empty(struct mem_cgroup *mem)
1067 { 1066 {
1068 int ret; 1067 int ret;
1069 int node, zid, shrink; 1068 int node, zid, shrink;
1070 int nr_retries = MEM_CGROUP_RECLAIM_RETRIES; 1069 int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
1071 1070
1072 css_get(&mem->css); 1071 css_get(&mem->css);
1073 1072
1074 shrink = 0; 1073 shrink = 0;
1075 move_account: 1074 move_account:
1076 while (mem->res.usage > 0) { 1075 while (mem->res.usage > 0) {
1077 ret = -EBUSY; 1076 ret = -EBUSY;
1078 if (atomic_read(&mem->css.cgroup->count) > 0) 1077 if (atomic_read(&mem->css.cgroup->count) > 0)
1079 goto out; 1078 goto out;
1080 1079
1081 /* This is for making all *used* pages to be on LRU. */ 1080 /* This is for making all *used* pages to be on LRU. */
1082 lru_add_drain_all(); 1081 lru_add_drain_all();
1083 ret = 0; 1082 ret = 0;
1084 for_each_node_state(node, N_POSSIBLE) { 1083 for_each_node_state(node, N_POSSIBLE) {
1085 for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) { 1084 for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
1086 struct mem_cgroup_per_zone *mz; 1085 struct mem_cgroup_per_zone *mz;
1087 enum lru_list l; 1086 enum lru_list l;
1088 mz = mem_cgroup_zoneinfo(mem, node, zid); 1087 mz = mem_cgroup_zoneinfo(mem, node, zid);
1089 for_each_lru(l) { 1088 for_each_lru(l) {
1090 ret = mem_cgroup_force_empty_list(mem, 1089 ret = mem_cgroup_force_empty_list(mem,
1091 mz, l); 1090 mz, l);
1092 if (ret) 1091 if (ret)
1093 break; 1092 break;
1094 } 1093 }
1095 } 1094 }
1096 if (ret) 1095 if (ret)
1097 break; 1096 break;
1098 } 1097 }
1099 /* it seems parent cgroup doesn't have enough mem */ 1098 /* it seems parent cgroup doesn't have enough mem */
1100 if (ret == -ENOMEM) 1099 if (ret == -ENOMEM)
1101 goto try_to_free; 1100 goto try_to_free;
1102 cond_resched(); 1101 cond_resched();
1103 } 1102 }
1104 ret = 0; 1103 ret = 0;
1105 out: 1104 out:
1106 css_put(&mem->css); 1105 css_put(&mem->css);
1107 return ret; 1106 return ret;
1108 1107
1109 try_to_free: 1108 try_to_free:
1110 /* returns EBUSY if we come here twice. */ 1109 /* returns EBUSY if we come here twice. */
1111 if (shrink) { 1110 if (shrink) {
1112 ret = -EBUSY; 1111 ret = -EBUSY;
1113 goto out; 1112 goto out;
1114 } 1113 }
1115 /* try to free all pages in this cgroup */ 1114 /* try to free all pages in this cgroup */
1116 shrink = 1; 1115 shrink = 1;
1117 while (nr_retries && mem->res.usage > 0) { 1116 while (nr_retries && mem->res.usage > 0) {
1118 int progress; 1117 int progress;
1119 progress = try_to_free_mem_cgroup_pages(mem, 1118 progress = try_to_free_mem_cgroup_pages(mem,
1120 GFP_HIGHUSER_MOVABLE); 1119 GFP_HIGHUSER_MOVABLE);
1121 if (!progress) 1120 if (!progress)
1122 nr_retries--; 1121 nr_retries--;
1123 1122
1124 } 1123 }
1125 /* try move_account...there may be some *locked* pages. */ 1124 /* try move_account...there may be some *locked* pages. */
1126 if (mem->res.usage) 1125 if (mem->res.usage)
1127 goto move_account; 1126 goto move_account;
1128 ret = 0; 1127 ret = 0;
1129 goto out; 1128 goto out;
1130 } 1129 }
1131 1130
1132 static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft) 1131 static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
1133 { 1132 {
1134 return res_counter_read_u64(&mem_cgroup_from_cont(cont)->res, 1133 return res_counter_read_u64(&mem_cgroup_from_cont(cont)->res,
1135 cft->private); 1134 cft->private);
1136 } 1135 }
1137 /* 1136 /*
1138 * The user of this function is... 1137 * The user of this function is...
1139 * RES_LIMIT. 1138 * RES_LIMIT.
1140 */ 1139 */
1141 static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft, 1140 static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
1142 const char *buffer) 1141 const char *buffer)
1143 { 1142 {
1144 struct mem_cgroup *memcg = mem_cgroup_from_cont(cont); 1143 struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
1145 unsigned long long val; 1144 unsigned long long val;
1146 int ret; 1145 int ret;
1147 1146
1148 switch (cft->private) { 1147 switch (cft->private) {
1149 case RES_LIMIT: 1148 case RES_LIMIT:
1150 /* This function does all necessary parse...reuse it */ 1149 /* This function does all necessary parse...reuse it */
1151 ret = res_counter_memparse_write_strategy(buffer, &val); 1150 ret = res_counter_memparse_write_strategy(buffer, &val);
1152 if (!ret) 1151 if (!ret)
1153 ret = mem_cgroup_resize_limit(memcg, val); 1152 ret = mem_cgroup_resize_limit(memcg, val);
1154 break; 1153 break;
1155 default: 1154 default:
1156 ret = -EINVAL; /* should be BUG() ? */ 1155 ret = -EINVAL; /* should be BUG() ? */
1157 break; 1156 break;
1158 } 1157 }
1159 return ret; 1158 return ret;
1160 } 1159 }
1161 1160
1162 static int mem_cgroup_reset(struct cgroup *cont, unsigned int event) 1161 static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
1163 { 1162 {
1164 struct mem_cgroup *mem; 1163 struct mem_cgroup *mem;
1165 1164
1166 mem = mem_cgroup_from_cont(cont); 1165 mem = mem_cgroup_from_cont(cont);
1167 switch (event) { 1166 switch (event) {
1168 case RES_MAX_USAGE: 1167 case RES_MAX_USAGE:
1169 res_counter_reset_max(&mem->res); 1168 res_counter_reset_max(&mem->res);
1170 break; 1169 break;
1171 case RES_FAILCNT: 1170 case RES_FAILCNT:
1172 res_counter_reset_failcnt(&mem->res); 1171 res_counter_reset_failcnt(&mem->res);
1173 break; 1172 break;
1174 } 1173 }
1175 return 0; 1174 return 0;
1176 } 1175 }
1177 1176
1178 static const struct mem_cgroup_stat_desc { 1177 static const struct mem_cgroup_stat_desc {
1179 const char *msg; 1178 const char *msg;
1180 u64 unit; 1179 u64 unit;
1181 } mem_cgroup_stat_desc[] = { 1180 } mem_cgroup_stat_desc[] = {
1182 [MEM_CGROUP_STAT_CACHE] = { "cache", PAGE_SIZE, }, 1181 [MEM_CGROUP_STAT_CACHE] = { "cache", PAGE_SIZE, },
1183 [MEM_CGROUP_STAT_RSS] = { "rss", PAGE_SIZE, }, 1182 [MEM_CGROUP_STAT_RSS] = { "rss", PAGE_SIZE, },
1184 [MEM_CGROUP_STAT_PGPGIN_COUNT] = {"pgpgin", 1, }, 1183 [MEM_CGROUP_STAT_PGPGIN_COUNT] = {"pgpgin", 1, },
1185 [MEM_CGROUP_STAT_PGPGOUT_COUNT] = {"pgpgout", 1, }, 1184 [MEM_CGROUP_STAT_PGPGOUT_COUNT] = {"pgpgout", 1, },
1186 }; 1185 };
1187 1186
1188 static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft, 1187 static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
1189 struct cgroup_map_cb *cb) 1188 struct cgroup_map_cb *cb)
1190 { 1189 {
1191 struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont); 1190 struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);
1192 struct mem_cgroup_stat *stat = &mem_cont->stat; 1191 struct mem_cgroup_stat *stat = &mem_cont->stat;
1193 int i; 1192 int i;
1194 1193
1195 for (i = 0; i < ARRAY_SIZE(stat->cpustat[0].count); i++) { 1194 for (i = 0; i < ARRAY_SIZE(stat->cpustat[0].count); i++) {
1196 s64 val; 1195 s64 val;
1197 1196
1198 val = mem_cgroup_read_stat(stat, i); 1197 val = mem_cgroup_read_stat(stat, i);
1199 val *= mem_cgroup_stat_desc[i].unit; 1198 val *= mem_cgroup_stat_desc[i].unit;
1200 cb->fill(cb, mem_cgroup_stat_desc[i].msg, val); 1199 cb->fill(cb, mem_cgroup_stat_desc[i].msg, val);
1201 } 1200 }
1202 /* showing # of active pages */ 1201 /* showing # of active pages */
1203 { 1202 {
1204 unsigned long active_anon, inactive_anon; 1203 unsigned long active_anon, inactive_anon;
1205 unsigned long active_file, inactive_file; 1204 unsigned long active_file, inactive_file;
1206 unsigned long unevictable; 1205 unsigned long unevictable;
1207 1206
1208 inactive_anon = mem_cgroup_get_all_zonestat(mem_cont, 1207 inactive_anon = mem_cgroup_get_all_zonestat(mem_cont,
1209 LRU_INACTIVE_ANON); 1208 LRU_INACTIVE_ANON);
1210 active_anon = mem_cgroup_get_all_zonestat(mem_cont, 1209 active_anon = mem_cgroup_get_all_zonestat(mem_cont,
1211 LRU_ACTIVE_ANON); 1210 LRU_ACTIVE_ANON);
1212 inactive_file = mem_cgroup_get_all_zonestat(mem_cont, 1211 inactive_file = mem_cgroup_get_all_zonestat(mem_cont,
1213 LRU_INACTIVE_FILE); 1212 LRU_INACTIVE_FILE);
1214 active_file = mem_cgroup_get_all_zonestat(mem_cont, 1213 active_file = mem_cgroup_get_all_zonestat(mem_cont,
1215 LRU_ACTIVE_FILE); 1214 LRU_ACTIVE_FILE);
1216 unevictable = mem_cgroup_get_all_zonestat(mem_cont, 1215 unevictable = mem_cgroup_get_all_zonestat(mem_cont,
1217 LRU_UNEVICTABLE); 1216 LRU_UNEVICTABLE);
1218 1217
1219 cb->fill(cb, "active_anon", (active_anon) * PAGE_SIZE); 1218 cb->fill(cb, "active_anon", (active_anon) * PAGE_SIZE);
1220 cb->fill(cb, "inactive_anon", (inactive_anon) * PAGE_SIZE); 1219 cb->fill(cb, "inactive_anon", (inactive_anon) * PAGE_SIZE);
1221 cb->fill(cb, "active_file", (active_file) * PAGE_SIZE); 1220 cb->fill(cb, "active_file", (active_file) * PAGE_SIZE);
1222 cb->fill(cb, "inactive_file", (inactive_file) * PAGE_SIZE); 1221 cb->fill(cb, "inactive_file", (inactive_file) * PAGE_SIZE);
1223 cb->fill(cb, "unevictable", unevictable * PAGE_SIZE); 1222 cb->fill(cb, "unevictable", unevictable * PAGE_SIZE);
1224 1223
1225 } 1224 }
1226 return 0; 1225 return 0;
1227 } 1226 }
1228 1227
1229 static struct cftype mem_cgroup_files[] = { 1228 static struct cftype mem_cgroup_files[] = {
1230 { 1229 {
1231 .name = "usage_in_bytes", 1230 .name = "usage_in_bytes",
1232 .private = RES_USAGE, 1231 .private = RES_USAGE,
1233 .read_u64 = mem_cgroup_read, 1232 .read_u64 = mem_cgroup_read,
1234 }, 1233 },
1235 { 1234 {
1236 .name = "max_usage_in_bytes", 1235 .name = "max_usage_in_bytes",
1237 .private = RES_MAX_USAGE, 1236 .private = RES_MAX_USAGE,
1238 .trigger = mem_cgroup_reset, 1237 .trigger = mem_cgroup_reset,
1239 .read_u64 = mem_cgroup_read, 1238 .read_u64 = mem_cgroup_read,
1240 }, 1239 },
1241 { 1240 {
1242 .name = "limit_in_bytes", 1241 .name = "limit_in_bytes",
1243 .private = RES_LIMIT, 1242 .private = RES_LIMIT,
1244 .write_string = mem_cgroup_write, 1243 .write_string = mem_cgroup_write,
1245 .read_u64 = mem_cgroup_read, 1244 .read_u64 = mem_cgroup_read,
1246 }, 1245 },
1247 { 1246 {
1248 .name = "failcnt", 1247 .name = "failcnt",
1249 .private = RES_FAILCNT, 1248 .private = RES_FAILCNT,
1250 .trigger = mem_cgroup_reset, 1249 .trigger = mem_cgroup_reset,
1251 .read_u64 = mem_cgroup_read, 1250 .read_u64 = mem_cgroup_read,
1252 }, 1251 },
1253 { 1252 {
1254 .name = "stat", 1253 .name = "stat",
1255 .read_map = mem_control_stat_show, 1254 .read_map = mem_control_stat_show,
1256 }, 1255 },
1257 }; 1256 };
1258 1257
1259 static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node) 1258 static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
1260 { 1259 {
1261 struct mem_cgroup_per_node *pn; 1260 struct mem_cgroup_per_node *pn;
1262 struct mem_cgroup_per_zone *mz; 1261 struct mem_cgroup_per_zone *mz;
1263 enum lru_list l; 1262 enum lru_list l;
1264 int zone, tmp = node; 1263 int zone, tmp = node;
1265 /* 1264 /*
1266 * This routine is called against possible nodes. 1265 * This routine is called against possible nodes.
1267 * But it's BUG to call kmalloc() against offline node. 1266 * But it's BUG to call kmalloc() against offline node.
1268 * 1267 *
1269 * TODO: this routine can waste much memory for nodes which will 1268 * TODO: this routine can waste much memory for nodes which will
1270 * never be onlined. It's better to use memory hotplug callback 1269 * never be onlined. It's better to use memory hotplug callback
1271 * function. 1270 * function.
1272 */ 1271 */
1273 if (!node_state(node, N_NORMAL_MEMORY)) 1272 if (!node_state(node, N_NORMAL_MEMORY))
1274 tmp = -1; 1273 tmp = -1;
1275 pn = kmalloc_node(sizeof(*pn), GFP_KERNEL, tmp); 1274 pn = kmalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
1276 if (!pn) 1275 if (!pn)
1277 return 1; 1276 return 1;
1278 1277
1279 mem->info.nodeinfo[node] = pn; 1278 mem->info.nodeinfo[node] = pn;
1280 memset(pn, 0, sizeof(*pn)); 1279 memset(pn, 0, sizeof(*pn));
1281 1280
1282 for (zone = 0; zone < MAX_NR_ZONES; zone++) { 1281 for (zone = 0; zone < MAX_NR_ZONES; zone++) {
1283 mz = &pn->zoneinfo[zone]; 1282 mz = &pn->zoneinfo[zone];
1284 spin_lock_init(&mz->lru_lock); 1283 spin_lock_init(&mz->lru_lock);
1285 for_each_lru(l) 1284 for_each_lru(l)
1286 INIT_LIST_HEAD(&mz->lists[l]); 1285 INIT_LIST_HEAD(&mz->lists[l]);
1287 } 1286 }
1288 return 0; 1287 return 0;
1289 } 1288 }
1290 1289
1291 static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node) 1290 static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
1292 { 1291 {
1293 kfree(mem->info.nodeinfo[node]); 1292 kfree(mem->info.nodeinfo[node]);
1294 } 1293 }
1295 1294
1295 static int mem_cgroup_size(void)
1296 {
1297 int cpustat_size = nr_cpu_ids * sizeof(struct mem_cgroup_stat_cpu);
1298 return sizeof(struct mem_cgroup) + cpustat_size;
1299 }
1300
1296 static struct mem_cgroup *mem_cgroup_alloc(void) 1301 static struct mem_cgroup *mem_cgroup_alloc(void)
1297 { 1302 {
1298 struct mem_cgroup *mem; 1303 struct mem_cgroup *mem;
1304 int size = mem_cgroup_size();
1299 1305
1300 if (sizeof(*mem) < PAGE_SIZE) 1306 if (size < PAGE_SIZE)
1301 mem = kmalloc(sizeof(*mem), GFP_KERNEL); 1307 mem = kmalloc(size, GFP_KERNEL);
1302 else 1308 else
1303 mem = vmalloc(sizeof(*mem)); 1309 mem = vmalloc(size);
1304 1310
1305 if (mem) 1311 if (mem)
1306 memset(mem, 0, sizeof(*mem)); 1312 memset(mem, 0, size);
1307 return mem; 1313 return mem;
1308 } 1314 }
1309 1315
1310 static void mem_cgroup_free(struct mem_cgroup *mem) 1316 static void mem_cgroup_free(struct mem_cgroup *mem)
1311 { 1317 {
1312 if (sizeof(*mem) < PAGE_SIZE) 1318 if (mem_cgroup_size() < PAGE_SIZE)
1313 kfree(mem); 1319 kfree(mem);
1314 else 1320 else
1315 vfree(mem); 1321 vfree(mem);
1316 } 1322 }
1317 1323
1318 1324
1319 static struct cgroup_subsys_state * 1325 static struct cgroup_subsys_state *
1320 mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont) 1326 mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
1321 { 1327 {
1322 struct mem_cgroup *mem; 1328 struct mem_cgroup *mem;
1323 int node; 1329 int node;
1324 1330
1325 if (unlikely((cont->parent) == NULL)) { 1331 mem = mem_cgroup_alloc();
1326 mem = &init_mem_cgroup; 1332 if (!mem)
1327 } else { 1333 return ERR_PTR(-ENOMEM);
1328 mem = mem_cgroup_alloc();
1329 if (!mem)
1330 return ERR_PTR(-ENOMEM);
1331 }
1332 1334
1333 res_counter_init(&mem->res); 1335 res_counter_init(&mem->res);
1334 1336
1335 for_each_node_state(node, N_POSSIBLE) 1337 for_each_node_state(node, N_POSSIBLE)
1336 if (alloc_mem_cgroup_per_zone_info(mem, node)) 1338 if (alloc_mem_cgroup_per_zone_info(mem, node))
1337 goto free_out; 1339 goto free_out;
1338 1340
1339 return &mem->css; 1341 return &mem->css;
1340 free_out: 1342 free_out:
1341 for_each_node_state(node, N_POSSIBLE) 1343 for_each_node_state(node, N_POSSIBLE)
1342 free_mem_cgroup_per_zone_info(mem, node); 1344 free_mem_cgroup_per_zone_info(mem, node);
1343 if (cont->parent != NULL) 1345 mem_cgroup_free(mem);
1344 mem_cgroup_free(mem);
1345 return ERR_PTR(-ENOMEM); 1346 return ERR_PTR(-ENOMEM);
1346 } 1347 }
1347 1348
1348 static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss, 1349 static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
1349 struct cgroup *cont) 1350 struct cgroup *cont)
1350 { 1351 {
1351 struct mem_cgroup *mem = mem_cgroup_from_cont(cont); 1352 struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
1352 mem_cgroup_force_empty(mem); 1353 mem_cgroup_force_empty(mem);
1353 } 1354 }
1354 1355
1355 static void mem_cgroup_destroy(struct cgroup_subsys *ss, 1356 static void mem_cgroup_destroy(struct cgroup_subsys *ss,
1356 struct cgroup *cont) 1357 struct cgroup *cont)
1357 { 1358 {
1358 int node; 1359 int node;
1359 struct mem_cgroup *mem = mem_cgroup_from_cont(cont); 1360 struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
1360 1361
1361 for_each_node_state(node, N_POSSIBLE) 1362 for_each_node_state(node, N_POSSIBLE)
1362 free_mem_cgroup_per_zone_info(mem, node); 1363 free_mem_cgroup_per_zone_info(mem, node);
1363 1364
1364 mem_cgroup_free(mem_cgroup_from_cont(cont)); 1365 mem_cgroup_free(mem_cgroup_from_cont(cont));
1365 } 1366 }
1366 1367
1367 static int mem_cgroup_populate(struct cgroup_subsys *ss, 1368 static int mem_cgroup_populate(struct cgroup_subsys *ss,
1368 struct cgroup *cont) 1369 struct cgroup *cont)
1369 { 1370 {
1370 return cgroup_add_files(cont, ss, mem_cgroup_files, 1371 return cgroup_add_files(cont, ss, mem_cgroup_files,
1371 ARRAY_SIZE(mem_cgroup_files)); 1372 ARRAY_SIZE(mem_cgroup_files));
1372 } 1373 }
1373 1374
1374 static void mem_cgroup_move_task(struct cgroup_subsys *ss, 1375 static void mem_cgroup_move_task(struct cgroup_subsys *ss,
1375 struct cgroup *cont, 1376 struct cgroup *cont,
1376 struct cgroup *old_cont, 1377 struct cgroup *old_cont,
1377 struct task_struct *p) 1378 struct task_struct *p)
1378 { 1379 {
1379 struct mm_struct *mm; 1380 struct mm_struct *mm;
1380 struct mem_cgroup *mem, *old_mem; 1381 struct mem_cgroup *mem, *old_mem;
1381 1382
1382 mm = get_task_mm(p); 1383 mm = get_task_mm(p);
1383 if (mm == NULL) 1384 if (mm == NULL)
1384 return; 1385 return;
1385 1386
1386 mem = mem_cgroup_from_cont(cont); 1387 mem = mem_cgroup_from_cont(cont);
1387 old_mem = mem_cgroup_from_cont(old_cont); 1388 old_mem = mem_cgroup_from_cont(old_cont);
1388 1389
1389 /* 1390 /*
1390 * Only thread group leaders are allowed to migrate, the mm_struct is 1391 * Only thread group leaders are allowed to migrate, the mm_struct is
1391 * in effect owned by the leader 1392 * in effect owned by the leader
1392 */ 1393 */
1393 if (!thread_group_leader(p)) 1394 if (!thread_group_leader(p))
1394 goto out; 1395 goto out;
1395 1396
1396 out: 1397 out:
1397 mmput(mm); 1398 mmput(mm);
1398 } 1399 }
1399 1400
1400 struct cgroup_subsys mem_cgroup_subsys = { 1401 struct cgroup_subsys mem_cgroup_subsys = {
1401 .name = "memory", 1402 .name = "memory",
1402 .subsys_id = mem_cgroup_subsys_id, 1403 .subsys_id = mem_cgroup_subsys_id,
1403 .create = mem_cgroup_create, 1404 .create = mem_cgroup_create,
1404 .pre_destroy = mem_cgroup_pre_destroy, 1405 .pre_destroy = mem_cgroup_pre_destroy,