memcg: mem+swap controller Kconfig
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / mm / page_cgroup.c
CommitLineData
52d4b9ac
KH
1#include <linux/mm.h>
2#include <linux/mmzone.h>
3#include <linux/bootmem.h>
4#include <linux/bit_spinlock.h>
5#include <linux/page_cgroup.h>
6#include <linux/hash.h>
94b6da5a 7#include <linux/slab.h>
52d4b9ac 8#include <linux/memory.h>
4c821042 9#include <linux/vmalloc.h>
94b6da5a 10#include <linux/cgroup.h>
52d4b9ac
KH
11
12static void __meminit
13__init_page_cgroup(struct page_cgroup *pc, unsigned long pfn)
14{
15 pc->flags = 0;
16 pc->mem_cgroup = NULL;
17 pc->page = pfn_to_page(pfn);
18}
19static unsigned long total_usage;
20
21#if !defined(CONFIG_SPARSEMEM)
22
23
31168481 24void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
52d4b9ac
KH
25{
26 pgdat->node_page_cgroup = NULL;
27}
28
29struct page_cgroup *lookup_page_cgroup(struct page *page)
30{
31 unsigned long pfn = page_to_pfn(page);
32 unsigned long offset;
33 struct page_cgroup *base;
34
35 base = NODE_DATA(page_to_nid(page))->node_page_cgroup;
36 if (unlikely(!base))
37 return NULL;
38
39 offset = pfn - NODE_DATA(page_to_nid(page))->node_start_pfn;
40 return base + offset;
41}
42
43static int __init alloc_node_page_cgroup(int nid)
44{
45 struct page_cgroup *base, *pc;
46 unsigned long table_size;
47 unsigned long start_pfn, nr_pages, index;
48
49 start_pfn = NODE_DATA(nid)->node_start_pfn;
50 nr_pages = NODE_DATA(nid)->node_spanned_pages;
51
653d22c0
KH
52 if (!nr_pages)
53 return 0;
54
52d4b9ac
KH
55 table_size = sizeof(struct page_cgroup) * nr_pages;
56
57 base = __alloc_bootmem_node_nopanic(NODE_DATA(nid),
58 table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
59 if (!base)
60 return -ENOMEM;
61 for (index = 0; index < nr_pages; index++) {
62 pc = base + index;
63 __init_page_cgroup(pc, start_pfn + index);
64 }
65 NODE_DATA(nid)->node_page_cgroup = base;
66 total_usage += table_size;
67 return 0;
68}
69
70void __init page_cgroup_init(void)
71{
72
73 int nid, fail;
74
94b6da5a
KH
75 if (mem_cgroup_subsys.disabled)
76 return;
77
52d4b9ac
KH
78 for_each_online_node(nid) {
79 fail = alloc_node_page_cgroup(nid);
80 if (fail)
81 goto fail;
82 }
83 printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
84 printk(KERN_INFO "please try cgroup_disable=memory option if you"
85 " don't want\n");
86 return;
87fail:
88 printk(KERN_CRIT "allocation of page_cgroup was failed.\n");
89 printk(KERN_CRIT "please try cgroup_disable=memory boot option\n");
90 panic("Out of memory");
91}
92
93#else /* CONFIG_FLAT_NODE_MEM_MAP */
94
95struct page_cgroup *lookup_page_cgroup(struct page *page)
96{
97 unsigned long pfn = page_to_pfn(page);
98 struct mem_section *section = __pfn_to_section(pfn);
99
100 return section->page_cgroup + pfn;
101}
102
31168481 103/* __alloc_bootmem...() is protected by !slab_available() */
feb16694 104static int __init_refok init_section_page_cgroup(unsigned long pfn)
52d4b9ac 105{
0753b0ef 106 struct mem_section *section = __pfn_to_section(pfn);
52d4b9ac
KH
107 struct page_cgroup *base, *pc;
108 unsigned long table_size;
109 int nid, index;
110
dc19f9db
KH
111 if (!section->page_cgroup) {
112 nid = page_to_nid(pfn_to_page(pfn));
113 table_size = sizeof(struct page_cgroup) * PAGES_PER_SECTION;
114 if (slab_is_available()) {
115 base = kmalloc_node(table_size, GFP_KERNEL, nid);
116 if (!base)
117 base = vmalloc_node(table_size, nid);
118 } else {
119 base = __alloc_bootmem_node_nopanic(NODE_DATA(nid),
120 table_size,
94b6da5a 121 PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
dc19f9db
KH
122 }
123 } else {
124 /*
125 * We don't have to allocate page_cgroup again, but
126 * address of memmap may be changed. So, we have to initialize
127 * again.
128 */
129 base = section->page_cgroup + pfn;
130 table_size = 0;
131 /* check address of memmap is changed or not. */
132 if (base->page == pfn_to_page(pfn))
133 return 0;
94b6da5a 134 }
52d4b9ac
KH
135
136 if (!base) {
137 printk(KERN_ERR "page cgroup allocation failure\n");
138 return -ENOMEM;
139 }
140
141 for (index = 0; index < PAGES_PER_SECTION; index++) {
142 pc = base + index;
143 __init_page_cgroup(pc, pfn + index);
144 }
145
52d4b9ac
KH
146 section->page_cgroup = base - pfn;
147 total_usage += table_size;
148 return 0;
149}
150#ifdef CONFIG_MEMORY_HOTPLUG
151void __free_page_cgroup(unsigned long pfn)
152{
153 struct mem_section *ms;
154 struct page_cgroup *base;
155
156 ms = __pfn_to_section(pfn);
157 if (!ms || !ms->page_cgroup)
158 return;
159 base = ms->page_cgroup + pfn;
94b6da5a 160 if (is_vmalloc_addr(base)) {
52d4b9ac 161 vfree(base);
94b6da5a
KH
162 ms->page_cgroup = NULL;
163 } else {
164 struct page *page = virt_to_page(base);
165 if (!PageReserved(page)) { /* Is bootmem ? */
166 kfree(base);
167 ms->page_cgroup = NULL;
168 }
169 }
52d4b9ac
KH
170}
171
31168481 172int __meminit online_page_cgroup(unsigned long start_pfn,
52d4b9ac
KH
173 unsigned long nr_pages,
174 int nid)
175{
176 unsigned long start, end, pfn;
177 int fail = 0;
178
33c5d3d6 179 start = start_pfn & ~(PAGES_PER_SECTION - 1);
52d4b9ac
KH
180 end = ALIGN(start_pfn + nr_pages, PAGES_PER_SECTION);
181
182 for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION) {
183 if (!pfn_present(pfn))
184 continue;
185 fail = init_section_page_cgroup(pfn);
186 }
187 if (!fail)
188 return 0;
189
190 /* rollback */
191 for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
192 __free_page_cgroup(pfn);
193
194 return -ENOMEM;
195}
196
31168481 197int __meminit offline_page_cgroup(unsigned long start_pfn,
52d4b9ac
KH
198 unsigned long nr_pages, int nid)
199{
200 unsigned long start, end, pfn;
201
33c5d3d6 202 start = start_pfn & ~(PAGES_PER_SECTION - 1);
52d4b9ac
KH
203 end = ALIGN(start_pfn + nr_pages, PAGES_PER_SECTION);
204
205 for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
206 __free_page_cgroup(pfn);
207 return 0;
208
209}
210
31168481 211static int __meminit page_cgroup_callback(struct notifier_block *self,
52d4b9ac
KH
212 unsigned long action, void *arg)
213{
214 struct memory_notify *mn = arg;
215 int ret = 0;
216 switch (action) {
217 case MEM_GOING_ONLINE:
218 ret = online_page_cgroup(mn->start_pfn,
219 mn->nr_pages, mn->status_change_nid);
220 break;
52d4b9ac
KH
221 case MEM_OFFLINE:
222 offline_page_cgroup(mn->start_pfn,
223 mn->nr_pages, mn->status_change_nid);
224 break;
dc19f9db 225 case MEM_CANCEL_ONLINE:
52d4b9ac
KH
226 case MEM_GOING_OFFLINE:
227 break;
228 case MEM_ONLINE:
229 case MEM_CANCEL_OFFLINE:
230 break;
231 }
dc19f9db
KH
232
233 if (ret)
234 ret = notifier_from_errno(ret);
235 else
236 ret = NOTIFY_OK;
237
52d4b9ac
KH
238 return ret;
239}
240
241#endif
242
243void __init page_cgroup_init(void)
244{
245 unsigned long pfn;
246 int fail = 0;
247
94b6da5a
KH
248 if (mem_cgroup_subsys.disabled)
249 return;
250
52d4b9ac
KH
251 for (pfn = 0; !fail && pfn < max_pfn; pfn += PAGES_PER_SECTION) {
252 if (!pfn_present(pfn))
253 continue;
254 fail = init_section_page_cgroup(pfn);
255 }
256 if (fail) {
257 printk(KERN_CRIT "try cgroup_disable=memory boot option\n");
258 panic("Out of memory");
259 } else {
260 hotplug_memory_notifier(page_cgroup_callback, 0);
261 }
262 printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
263 printk(KERN_INFO "please try cgroup_disable=memory option if you don't"
264 " want\n");
265}
266
31168481 267void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
52d4b9ac
KH
268{
269 return;
270}
271
272#endif