]>
Commit | Line | Data |
---|---|---|
2bc64a20 AK |
1 | /* |
2 | * | |
3 | * Copyright IBM Corporation, 2012 | |
4 | * Author Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> | |
5 | * | |
faced7e0 GS |
6 | * Cgroup v2 |
7 | * Copyright (C) 2019 Red Hat, Inc. | |
8 | * Author: Giuseppe Scrivano <gscrivan@redhat.com> | |
9 | * | |
2bc64a20 AK |
10 | * This program is free software; you can redistribute it and/or modify it |
11 | * under the terms of version 2.1 of the GNU Lesser General Public License | |
12 | * as published by the Free Software Foundation. | |
13 | * | |
14 | * This program is distributed in the hope that it would be useful, but | |
15 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. | |
17 | * | |
18 | */ | |
19 | ||
20 | #include <linux/cgroup.h> | |
71f87bee | 21 | #include <linux/page_counter.h> |
2bc64a20 AK |
22 | #include <linux/slab.h> |
23 | #include <linux/hugetlb.h> | |
24 | #include <linux/hugetlb_cgroup.h> | |
25 | ||
faced7e0 GS |
26 | enum hugetlb_memory_event { |
27 | HUGETLB_MAX, | |
28 | HUGETLB_NR_MEMORY_EVENTS, | |
29 | }; | |
30 | ||
2bc64a20 AK |
31 | struct hugetlb_cgroup { |
32 | struct cgroup_subsys_state css; | |
faced7e0 | 33 | |
2bc64a20 AK |
34 | /* |
35 | * the counter to account for hugepages from hugetlb. | |
36 | */ | |
71f87bee | 37 | struct page_counter hugepage[HUGE_MAX_HSTATE]; |
faced7e0 | 38 | |
cdc2fcfe MA |
39 | /* |
40 | * the counter to account for hugepage reservations from hugetlb. | |
41 | */ | |
42 | struct page_counter rsvd_hugepage[HUGE_MAX_HSTATE]; | |
43 | ||
faced7e0 GS |
44 | atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS]; |
45 | atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS]; | |
46 | ||
47 | /* Handle for "hugetlb.events" */ | |
48 | struct cgroup_file events_file[HUGE_MAX_HSTATE]; | |
49 | ||
50 | /* Handle for "hugetlb.events.local" */ | |
51 | struct cgroup_file events_local_file[HUGE_MAX_HSTATE]; | |
2bc64a20 AK |
52 | }; |
53 | ||
abb8206c AK |
54 | #define MEMFILE_PRIVATE(x, val) (((x) << 16) | (val)) |
55 | #define MEMFILE_IDX(val) (((val) >> 16) & 0xffff) | |
56 | #define MEMFILE_ATTR(val) ((val) & 0xffff) | |
57 | ||
faced7e0 GS |
58 | #define hugetlb_cgroup_from_counter(counter, idx) \ |
59 | container_of(counter, struct hugetlb_cgroup, hugepage[idx]) | |
60 | ||
2bc64a20 AK |
61 | static struct hugetlb_cgroup *root_h_cgroup __read_mostly; |
62 | ||
cdc2fcfe MA |
63 | static inline struct page_counter * |
64 | hugetlb_cgroup_counter_from_cgroup(struct hugetlb_cgroup *h_cg, int idx, | |
65 | bool rsvd) | |
66 | { | |
67 | if (rsvd) | |
68 | return &h_cg->rsvd_hugepage[idx]; | |
69 | return &h_cg->hugepage[idx]; | |
70 | } | |
71 | ||
2bc64a20 AK |
72 | static inline |
73 | struct hugetlb_cgroup *hugetlb_cgroup_from_css(struct cgroup_subsys_state *s) | |
74 | { | |
a7c6d554 | 75 | return s ? container_of(s, struct hugetlb_cgroup, css) : NULL; |
2bc64a20 AK |
76 | } |
77 | ||
2bc64a20 AK |
78 | static inline |
79 | struct hugetlb_cgroup *hugetlb_cgroup_from_task(struct task_struct *task) | |
80 | { | |
073219e9 | 81 | return hugetlb_cgroup_from_css(task_css(task, hugetlb_cgrp_id)); |
2bc64a20 AK |
82 | } |
83 | ||
84 | static inline bool hugetlb_cgroup_is_root(struct hugetlb_cgroup *h_cg) | |
85 | { | |
86 | return (h_cg == root_h_cgroup); | |
87 | } | |
88 | ||
3f798518 TH |
89 | static inline struct hugetlb_cgroup * |
90 | parent_hugetlb_cgroup(struct hugetlb_cgroup *h_cg) | |
2bc64a20 | 91 | { |
5c9d535b | 92 | return hugetlb_cgroup_from_css(h_cg->css.parent); |
2bc64a20 AK |
93 | } |
94 | ||
3f798518 | 95 | static inline bool hugetlb_cgroup_have_usage(struct hugetlb_cgroup *h_cg) |
2bc64a20 AK |
96 | { |
97 | int idx; | |
2bc64a20 AK |
98 | |
99 | for (idx = 0; idx < hugetlb_max_hstate; idx++) { | |
71f87bee | 100 | if (page_counter_read(&h_cg->hugepage[idx])) |
2bc64a20 AK |
101 | return true; |
102 | } | |
103 | return false; | |
104 | } | |
105 | ||
297880f4 DR |
106 | static void hugetlb_cgroup_init(struct hugetlb_cgroup *h_cgroup, |
107 | struct hugetlb_cgroup *parent_h_cgroup) | |
108 | { | |
109 | int idx; | |
110 | ||
111 | for (idx = 0; idx < HUGE_MAX_HSTATE; idx++) { | |
112 | struct page_counter *counter = &h_cgroup->hugepage[idx]; | |
113 | struct page_counter *parent = NULL; | |
114 | unsigned long limit; | |
115 | int ret; | |
116 | ||
117 | if (parent_h_cgroup) | |
118 | parent = &parent_h_cgroup->hugepage[idx]; | |
119 | page_counter_init(counter, parent); | |
120 | ||
121 | limit = round_down(PAGE_COUNTER_MAX, | |
122 | 1 << huge_page_order(&hstates[idx])); | |
bbec2e15 | 123 | ret = page_counter_set_max(counter, limit); |
297880f4 DR |
124 | VM_BUG_ON(ret); |
125 | } | |
126 | } | |
127 | ||
eb95419b TH |
128 | static struct cgroup_subsys_state * |
129 | hugetlb_cgroup_css_alloc(struct cgroup_subsys_state *parent_css) | |
2bc64a20 | 130 | { |
eb95419b TH |
131 | struct hugetlb_cgroup *parent_h_cgroup = hugetlb_cgroup_from_css(parent_css); |
132 | struct hugetlb_cgroup *h_cgroup; | |
2bc64a20 AK |
133 | |
134 | h_cgroup = kzalloc(sizeof(*h_cgroup), GFP_KERNEL); | |
135 | if (!h_cgroup) | |
136 | return ERR_PTR(-ENOMEM); | |
137 | ||
297880f4 | 138 | if (!parent_h_cgroup) |
2bc64a20 | 139 | root_h_cgroup = h_cgroup; |
297880f4 DR |
140 | |
141 | hugetlb_cgroup_init(h_cgroup, parent_h_cgroup); | |
2bc64a20 AK |
142 | return &h_cgroup->css; |
143 | } | |
144 | ||
eb95419b | 145 | static void hugetlb_cgroup_css_free(struct cgroup_subsys_state *css) |
2bc64a20 AK |
146 | { |
147 | struct hugetlb_cgroup *h_cgroup; | |
148 | ||
eb95419b | 149 | h_cgroup = hugetlb_cgroup_from_css(css); |
2bc64a20 AK |
150 | kfree(h_cgroup); |
151 | } | |
152 | ||
da1def55 AK |
153 | |
154 | /* | |
155 | * Should be called with hugetlb_lock held. | |
156 | * Since we are holding hugetlb_lock, pages cannot get moved from | |
157 | * active list or uncharged from the cgroup, So no need to get | |
158 | * page reference and test for page active here. This function | |
159 | * cannot fail. | |
160 | */ | |
3f798518 | 161 | static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg, |
da1def55 AK |
162 | struct page *page) |
163 | { | |
71f87bee JW |
164 | unsigned int nr_pages; |
165 | struct page_counter *counter; | |
da1def55 | 166 | struct hugetlb_cgroup *page_hcg; |
3f798518 | 167 | struct hugetlb_cgroup *parent = parent_hugetlb_cgroup(h_cg); |
da1def55 AK |
168 | |
169 | page_hcg = hugetlb_cgroup_from_page(page); | |
170 | /* | |
171 | * We can have pages in active list without any cgroup | |
172 | * ie, hugepage with less than 3 pages. We can safely | |
173 | * ignore those pages. | |
174 | */ | |
175 | if (!page_hcg || page_hcg != h_cg) | |
176 | goto out; | |
177 | ||
d8c6546b | 178 | nr_pages = compound_nr(page); |
da1def55 AK |
179 | if (!parent) { |
180 | parent = root_h_cgroup; | |
181 | /* root has no limit */ | |
71f87bee | 182 | page_counter_charge(&parent->hugepage[idx], nr_pages); |
da1def55 AK |
183 | } |
184 | counter = &h_cg->hugepage[idx]; | |
71f87bee JW |
185 | /* Take the pages off the local counter */ |
186 | page_counter_cancel(counter, nr_pages); | |
da1def55 AK |
187 | |
188 | set_hugetlb_cgroup(page, parent); | |
189 | out: | |
190 | return; | |
191 | } | |
192 | ||
193 | /* | |
194 | * Force the hugetlb cgroup to empty the hugetlb resources by moving them to | |
195 | * the parent cgroup. | |
196 | */ | |
eb95419b | 197 | static void hugetlb_cgroup_css_offline(struct cgroup_subsys_state *css) |
2bc64a20 | 198 | { |
eb95419b | 199 | struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(css); |
da1def55 AK |
200 | struct hstate *h; |
201 | struct page *page; | |
9d093cb1 | 202 | int idx = 0; |
da1def55 AK |
203 | |
204 | do { | |
da1def55 AK |
205 | for_each_hstate(h) { |
206 | spin_lock(&hugetlb_lock); | |
207 | list_for_each_entry(page, &h->hugepage_activelist, lru) | |
3f798518 | 208 | hugetlb_cgroup_move_parent(idx, h_cg, page); |
da1def55 AK |
209 | |
210 | spin_unlock(&hugetlb_lock); | |
211 | idx++; | |
212 | } | |
213 | cond_resched(); | |
3f798518 | 214 | } while (hugetlb_cgroup_have_usage(h_cg)); |
2bc64a20 AK |
215 | } |
216 | ||
faced7e0 GS |
217 | static inline void hugetlb_event(struct hugetlb_cgroup *hugetlb, int idx, |
218 | enum hugetlb_memory_event event) | |
219 | { | |
220 | atomic_long_inc(&hugetlb->events_local[idx][event]); | |
221 | cgroup_file_notify(&hugetlb->events_local_file[idx]); | |
222 | ||
223 | do { | |
224 | atomic_long_inc(&hugetlb->events[idx][event]); | |
225 | cgroup_file_notify(&hugetlb->events_file[idx]); | |
226 | } while ((hugetlb = parent_hugetlb_cgroup(hugetlb)) && | |
227 | !hugetlb_cgroup_is_root(hugetlb)); | |
228 | } | |
229 | ||
6d76dcf4 AK |
230 | int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages, |
231 | struct hugetlb_cgroup **ptr) | |
232 | { | |
233 | int ret = 0; | |
71f87bee | 234 | struct page_counter *counter; |
6d76dcf4 | 235 | struct hugetlb_cgroup *h_cg = NULL; |
6d76dcf4 AK |
236 | |
237 | if (hugetlb_cgroup_disabled()) | |
238 | goto done; | |
239 | /* | |
240 | * We don't charge any cgroup if the compound page have less | |
241 | * than 3 pages. | |
242 | */ | |
243 | if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER) | |
244 | goto done; | |
245 | again: | |
246 | rcu_read_lock(); | |
247 | h_cg = hugetlb_cgroup_from_task(current); | |
0362f326 | 248 | if (!css_tryget(&h_cg->css)) { |
6d76dcf4 AK |
249 | rcu_read_unlock(); |
250 | goto again; | |
251 | } | |
252 | rcu_read_unlock(); | |
253 | ||
faced7e0 GS |
254 | if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages, |
255 | &counter)) { | |
6071ca52 | 256 | ret = -ENOMEM; |
726b7bbe | 257 | hugetlb_event(h_cg, idx, HUGETLB_MAX); |
faced7e0 | 258 | } |
6d76dcf4 AK |
259 | css_put(&h_cg->css); |
260 | done: | |
261 | *ptr = h_cg; | |
262 | return ret; | |
263 | } | |
264 | ||
94ae8ba7 | 265 | /* Should be called with hugetlb_lock held */ |
6d76dcf4 AK |
266 | void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages, |
267 | struct hugetlb_cgroup *h_cg, | |
268 | struct page *page) | |
269 | { | |
270 | if (hugetlb_cgroup_disabled() || !h_cg) | |
271 | return; | |
272 | ||
6d76dcf4 | 273 | set_hugetlb_cgroup(page, h_cg); |
6d76dcf4 AK |
274 | return; |
275 | } | |
276 | ||
277 | /* | |
278 | * Should be called with hugetlb_lock held | |
279 | */ | |
280 | void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages, | |
281 | struct page *page) | |
282 | { | |
283 | struct hugetlb_cgroup *h_cg; | |
6d76dcf4 AK |
284 | |
285 | if (hugetlb_cgroup_disabled()) | |
286 | return; | |
7ea8574e | 287 | lockdep_assert_held(&hugetlb_lock); |
6d76dcf4 AK |
288 | h_cg = hugetlb_cgroup_from_page(page); |
289 | if (unlikely(!h_cg)) | |
290 | return; | |
291 | set_hugetlb_cgroup(page, NULL); | |
71f87bee | 292 | page_counter_uncharge(&h_cg->hugepage[idx], nr_pages); |
6d76dcf4 AK |
293 | return; |
294 | } | |
295 | ||
296 | void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages, | |
297 | struct hugetlb_cgroup *h_cg) | |
298 | { | |
6d76dcf4 AK |
299 | if (hugetlb_cgroup_disabled() || !h_cg) |
300 | return; | |
301 | ||
302 | if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER) | |
303 | return; | |
304 | ||
71f87bee | 305 | page_counter_uncharge(&h_cg->hugepage[idx], nr_pages); |
6d76dcf4 AK |
306 | return; |
307 | } | |
308 | ||
71f87bee JW |
309 | enum { |
310 | RES_USAGE, | |
cdc2fcfe | 311 | RES_RSVD_USAGE, |
71f87bee | 312 | RES_LIMIT, |
cdc2fcfe | 313 | RES_RSVD_LIMIT, |
71f87bee | 314 | RES_MAX_USAGE, |
cdc2fcfe | 315 | RES_RSVD_MAX_USAGE, |
71f87bee | 316 | RES_FAILCNT, |
cdc2fcfe | 317 | RES_RSVD_FAILCNT, |
71f87bee JW |
318 | }; |
319 | ||
716f479d TH |
320 | static u64 hugetlb_cgroup_read_u64(struct cgroup_subsys_state *css, |
321 | struct cftype *cft) | |
abb8206c | 322 | { |
71f87bee | 323 | struct page_counter *counter; |
cdc2fcfe | 324 | struct page_counter *rsvd_counter; |
182446d0 | 325 | struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(css); |
abb8206c | 326 | |
71f87bee | 327 | counter = &h_cg->hugepage[MEMFILE_IDX(cft->private)]; |
cdc2fcfe | 328 | rsvd_counter = &h_cg->rsvd_hugepage[MEMFILE_IDX(cft->private)]; |
abb8206c | 329 | |
71f87bee JW |
330 | switch (MEMFILE_ATTR(cft->private)) { |
331 | case RES_USAGE: | |
332 | return (u64)page_counter_read(counter) * PAGE_SIZE; | |
cdc2fcfe MA |
333 | case RES_RSVD_USAGE: |
334 | return (u64)page_counter_read(rsvd_counter) * PAGE_SIZE; | |
71f87bee | 335 | case RES_LIMIT: |
bbec2e15 | 336 | return (u64)counter->max * PAGE_SIZE; |
cdc2fcfe MA |
337 | case RES_RSVD_LIMIT: |
338 | return (u64)rsvd_counter->max * PAGE_SIZE; | |
71f87bee JW |
339 | case RES_MAX_USAGE: |
340 | return (u64)counter->watermark * PAGE_SIZE; | |
cdc2fcfe MA |
341 | case RES_RSVD_MAX_USAGE: |
342 | return (u64)rsvd_counter->watermark * PAGE_SIZE; | |
71f87bee JW |
343 | case RES_FAILCNT: |
344 | return counter->failcnt; | |
cdc2fcfe MA |
345 | case RES_RSVD_FAILCNT: |
346 | return rsvd_counter->failcnt; | |
71f87bee JW |
347 | default: |
348 | BUG(); | |
349 | } | |
abb8206c AK |
350 | } |
351 | ||
faced7e0 GS |
352 | static int hugetlb_cgroup_read_u64_max(struct seq_file *seq, void *v) |
353 | { | |
354 | int idx; | |
355 | u64 val; | |
356 | struct cftype *cft = seq_cft(seq); | |
357 | unsigned long limit; | |
358 | struct page_counter *counter; | |
359 | struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(seq_css(seq)); | |
360 | ||
361 | idx = MEMFILE_IDX(cft->private); | |
362 | counter = &h_cg->hugepage[idx]; | |
363 | ||
364 | limit = round_down(PAGE_COUNTER_MAX, | |
365 | 1 << huge_page_order(&hstates[idx])); | |
366 | ||
367 | switch (MEMFILE_ATTR(cft->private)) { | |
cdc2fcfe MA |
368 | case RES_RSVD_USAGE: |
369 | counter = &h_cg->rsvd_hugepage[idx]; | |
370 | /* Fall through. */ | |
faced7e0 GS |
371 | case RES_USAGE: |
372 | val = (u64)page_counter_read(counter); | |
373 | seq_printf(seq, "%llu\n", val * PAGE_SIZE); | |
374 | break; | |
cdc2fcfe MA |
375 | case RES_RSVD_LIMIT: |
376 | counter = &h_cg->rsvd_hugepage[idx]; | |
377 | /* Fall through. */ | |
faced7e0 GS |
378 | case RES_LIMIT: |
379 | val = (u64)counter->max; | |
380 | if (val == limit) | |
381 | seq_puts(seq, "max\n"); | |
382 | else | |
383 | seq_printf(seq, "%llu\n", val * PAGE_SIZE); | |
384 | break; | |
385 | default: | |
386 | BUG(); | |
387 | } | |
388 | ||
389 | return 0; | |
390 | } | |
391 | ||
71f87bee JW |
392 | static DEFINE_MUTEX(hugetlb_limit_mutex); |
393 | ||
451af504 | 394 | static ssize_t hugetlb_cgroup_write(struct kernfs_open_file *of, |
faced7e0 GS |
395 | char *buf, size_t nbytes, loff_t off, |
396 | const char *max) | |
abb8206c | 397 | { |
71f87bee JW |
398 | int ret, idx; |
399 | unsigned long nr_pages; | |
451af504 | 400 | struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of)); |
cdc2fcfe | 401 | bool rsvd = false; |
abb8206c | 402 | |
71f87bee JW |
403 | if (hugetlb_cgroup_is_root(h_cg)) /* Can't set limit on root */ |
404 | return -EINVAL; | |
405 | ||
451af504 | 406 | buf = strstrip(buf); |
faced7e0 | 407 | ret = page_counter_memparse(buf, max, &nr_pages); |
71f87bee JW |
408 | if (ret) |
409 | return ret; | |
410 | ||
451af504 | 411 | idx = MEMFILE_IDX(of_cft(of)->private); |
297880f4 | 412 | nr_pages = round_down(nr_pages, 1 << huge_page_order(&hstates[idx])); |
abb8206c | 413 | |
71f87bee | 414 | switch (MEMFILE_ATTR(of_cft(of)->private)) { |
cdc2fcfe MA |
415 | case RES_RSVD_LIMIT: |
416 | rsvd = true; | |
417 | /* Fall through. */ | |
abb8206c | 418 | case RES_LIMIT: |
71f87bee | 419 | mutex_lock(&hugetlb_limit_mutex); |
cdc2fcfe MA |
420 | ret = page_counter_set_max( |
421 | hugetlb_cgroup_counter_from_cgroup(h_cg, idx, rsvd), | |
422 | nr_pages); | |
71f87bee | 423 | mutex_unlock(&hugetlb_limit_mutex); |
abb8206c AK |
424 | break; |
425 | default: | |
426 | ret = -EINVAL; | |
427 | break; | |
428 | } | |
451af504 | 429 | return ret ?: nbytes; |
abb8206c AK |
430 | } |
431 | ||
faced7e0 GS |
432 | static ssize_t hugetlb_cgroup_write_legacy(struct kernfs_open_file *of, |
433 | char *buf, size_t nbytes, loff_t off) | |
434 | { | |
435 | return hugetlb_cgroup_write(of, buf, nbytes, off, "-1"); | |
436 | } | |
437 | ||
438 | static ssize_t hugetlb_cgroup_write_dfl(struct kernfs_open_file *of, | |
439 | char *buf, size_t nbytes, loff_t off) | |
440 | { | |
441 | return hugetlb_cgroup_write(of, buf, nbytes, off, "max"); | |
442 | } | |
443 | ||
6770c64e TH |
444 | static ssize_t hugetlb_cgroup_reset(struct kernfs_open_file *of, |
445 | char *buf, size_t nbytes, loff_t off) | |
abb8206c | 446 | { |
71f87bee | 447 | int ret = 0; |
cdc2fcfe | 448 | struct page_counter *counter, *rsvd_counter; |
6770c64e | 449 | struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of)); |
abb8206c | 450 | |
71f87bee | 451 | counter = &h_cg->hugepage[MEMFILE_IDX(of_cft(of)->private)]; |
cdc2fcfe | 452 | rsvd_counter = &h_cg->rsvd_hugepage[MEMFILE_IDX(of_cft(of)->private)]; |
abb8206c | 453 | |
71f87bee | 454 | switch (MEMFILE_ATTR(of_cft(of)->private)) { |
abb8206c | 455 | case RES_MAX_USAGE: |
71f87bee | 456 | page_counter_reset_watermark(counter); |
abb8206c | 457 | break; |
cdc2fcfe MA |
458 | case RES_RSVD_MAX_USAGE: |
459 | page_counter_reset_watermark(rsvd_counter); | |
460 | break; | |
abb8206c | 461 | case RES_FAILCNT: |
71f87bee | 462 | counter->failcnt = 0; |
abb8206c | 463 | break; |
cdc2fcfe MA |
464 | case RES_RSVD_FAILCNT: |
465 | rsvd_counter->failcnt = 0; | |
466 | break; | |
abb8206c AK |
467 | default: |
468 | ret = -EINVAL; | |
469 | break; | |
470 | } | |
6770c64e | 471 | return ret ?: nbytes; |
abb8206c AK |
472 | } |
473 | ||
474 | static char *mem_fmt(char *buf, int size, unsigned long hsize) | |
475 | { | |
476 | if (hsize >= (1UL << 30)) | |
477 | snprintf(buf, size, "%luGB", hsize >> 30); | |
478 | else if (hsize >= (1UL << 20)) | |
479 | snprintf(buf, size, "%luMB", hsize >> 20); | |
480 | else | |
481 | snprintf(buf, size, "%luKB", hsize >> 10); | |
482 | return buf; | |
483 | } | |
484 | ||
faced7e0 GS |
485 | static int __hugetlb_events_show(struct seq_file *seq, bool local) |
486 | { | |
487 | int idx; | |
488 | long max; | |
489 | struct cftype *cft = seq_cft(seq); | |
490 | struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(seq_css(seq)); | |
491 | ||
492 | idx = MEMFILE_IDX(cft->private); | |
493 | ||
494 | if (local) | |
495 | max = atomic_long_read(&h_cg->events_local[idx][HUGETLB_MAX]); | |
496 | else | |
497 | max = atomic_long_read(&h_cg->events[idx][HUGETLB_MAX]); | |
498 | ||
499 | seq_printf(seq, "max %lu\n", max); | |
500 | ||
501 | return 0; | |
502 | } | |
503 | ||
504 | static int hugetlb_events_show(struct seq_file *seq, void *v) | |
505 | { | |
506 | return __hugetlb_events_show(seq, false); | |
507 | } | |
508 | ||
509 | static int hugetlb_events_local_show(struct seq_file *seq, void *v) | |
510 | { | |
511 | return __hugetlb_events_show(seq, true); | |
512 | } | |
513 | ||
514 | static void __init __hugetlb_cgroup_file_dfl_init(int idx) | |
abb8206c AK |
515 | { |
516 | char buf[32]; | |
517 | struct cftype *cft; | |
518 | struct hstate *h = &hstates[idx]; | |
519 | ||
520 | /* format the size */ | |
cdc2fcfe | 521 | mem_fmt(buf, sizeof(buf), huge_page_size(h)); |
abb8206c AK |
522 | |
523 | /* Add the limit file */ | |
faced7e0 GS |
524 | cft = &h->cgroup_files_dfl[0]; |
525 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.max", buf); | |
526 | cft->private = MEMFILE_PRIVATE(idx, RES_LIMIT); | |
527 | cft->seq_show = hugetlb_cgroup_read_u64_max; | |
528 | cft->write = hugetlb_cgroup_write_dfl; | |
529 | cft->flags = CFTYPE_NOT_ON_ROOT; | |
530 | ||
cdc2fcfe | 531 | /* Add the reservation limit file */ |
faced7e0 | 532 | cft = &h->cgroup_files_dfl[1]; |
cdc2fcfe MA |
533 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.rsvd.max", buf); |
534 | cft->private = MEMFILE_PRIVATE(idx, RES_RSVD_LIMIT); | |
535 | cft->seq_show = hugetlb_cgroup_read_u64_max; | |
536 | cft->write = hugetlb_cgroup_write_dfl; | |
537 | cft->flags = CFTYPE_NOT_ON_ROOT; | |
538 | ||
539 | /* Add the current usage file */ | |
540 | cft = &h->cgroup_files_dfl[2]; | |
faced7e0 GS |
541 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.current", buf); |
542 | cft->private = MEMFILE_PRIVATE(idx, RES_USAGE); | |
543 | cft->seq_show = hugetlb_cgroup_read_u64_max; | |
544 | cft->flags = CFTYPE_NOT_ON_ROOT; | |
545 | ||
cdc2fcfe MA |
546 | /* Add the current reservation usage file */ |
547 | cft = &h->cgroup_files_dfl[3]; | |
548 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.rsvd.current", buf); | |
549 | cft->private = MEMFILE_PRIVATE(idx, RES_RSVD_USAGE); | |
550 | cft->seq_show = hugetlb_cgroup_read_u64_max; | |
551 | cft->flags = CFTYPE_NOT_ON_ROOT; | |
552 | ||
faced7e0 | 553 | /* Add the events file */ |
cdc2fcfe | 554 | cft = &h->cgroup_files_dfl[4]; |
faced7e0 GS |
555 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.events", buf); |
556 | cft->private = MEMFILE_PRIVATE(idx, 0); | |
557 | cft->seq_show = hugetlb_events_show; | |
558 | cft->file_offset = offsetof(struct hugetlb_cgroup, events_file[idx]), | |
559 | cft->flags = CFTYPE_NOT_ON_ROOT; | |
560 | ||
561 | /* Add the events.local file */ | |
cdc2fcfe | 562 | cft = &h->cgroup_files_dfl[5]; |
faced7e0 GS |
563 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.events.local", buf); |
564 | cft->private = MEMFILE_PRIVATE(idx, 0); | |
565 | cft->seq_show = hugetlb_events_local_show; | |
566 | cft->file_offset = offsetof(struct hugetlb_cgroup, | |
567 | events_local_file[idx]), | |
568 | cft->flags = CFTYPE_NOT_ON_ROOT; | |
569 | ||
570 | /* NULL terminate the last cft */ | |
cdc2fcfe | 571 | cft = &h->cgroup_files_dfl[6]; |
faced7e0 GS |
572 | memset(cft, 0, sizeof(*cft)); |
573 | ||
574 | WARN_ON(cgroup_add_dfl_cftypes(&hugetlb_cgrp_subsys, | |
575 | h->cgroup_files_dfl)); | |
576 | } | |
577 | ||
578 | static void __init __hugetlb_cgroup_file_legacy_init(int idx) | |
579 | { | |
580 | char buf[32]; | |
581 | struct cftype *cft; | |
582 | struct hstate *h = &hstates[idx]; | |
583 | ||
584 | /* format the size */ | |
cdc2fcfe | 585 | mem_fmt(buf, sizeof(buf), huge_page_size(h)); |
faced7e0 GS |
586 | |
587 | /* Add the limit file */ | |
588 | cft = &h->cgroup_files_legacy[0]; | |
abb8206c AK |
589 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.limit_in_bytes", buf); |
590 | cft->private = MEMFILE_PRIVATE(idx, RES_LIMIT); | |
716f479d | 591 | cft->read_u64 = hugetlb_cgroup_read_u64; |
faced7e0 | 592 | cft->write = hugetlb_cgroup_write_legacy; |
abb8206c | 593 | |
cdc2fcfe | 594 | /* Add the reservation limit file */ |
faced7e0 | 595 | cft = &h->cgroup_files_legacy[1]; |
cdc2fcfe MA |
596 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.rsvd.limit_in_bytes", buf); |
597 | cft->private = MEMFILE_PRIVATE(idx, RES_RSVD_LIMIT); | |
598 | cft->read_u64 = hugetlb_cgroup_read_u64; | |
599 | cft->write = hugetlb_cgroup_write_legacy; | |
600 | ||
601 | /* Add the usage file */ | |
602 | cft = &h->cgroup_files_legacy[2]; | |
abb8206c AK |
603 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.usage_in_bytes", buf); |
604 | cft->private = MEMFILE_PRIVATE(idx, RES_USAGE); | |
716f479d | 605 | cft->read_u64 = hugetlb_cgroup_read_u64; |
abb8206c | 606 | |
cdc2fcfe MA |
607 | /* Add the reservation usage file */ |
608 | cft = &h->cgroup_files_legacy[3]; | |
609 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.rsvd.usage_in_bytes", buf); | |
610 | cft->private = MEMFILE_PRIVATE(idx, RES_RSVD_USAGE); | |
611 | cft->read_u64 = hugetlb_cgroup_read_u64; | |
612 | ||
abb8206c | 613 | /* Add the MAX usage file */ |
cdc2fcfe | 614 | cft = &h->cgroup_files_legacy[4]; |
abb8206c AK |
615 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.max_usage_in_bytes", buf); |
616 | cft->private = MEMFILE_PRIVATE(idx, RES_MAX_USAGE); | |
6770c64e | 617 | cft->write = hugetlb_cgroup_reset; |
716f479d | 618 | cft->read_u64 = hugetlb_cgroup_read_u64; |
abb8206c | 619 | |
cdc2fcfe MA |
620 | /* Add the MAX reservation usage file */ |
621 | cft = &h->cgroup_files_legacy[5]; | |
622 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.rsvd.max_usage_in_bytes", buf); | |
623 | cft->private = MEMFILE_PRIVATE(idx, RES_RSVD_MAX_USAGE); | |
624 | cft->write = hugetlb_cgroup_reset; | |
625 | cft->read_u64 = hugetlb_cgroup_read_u64; | |
626 | ||
abb8206c | 627 | /* Add the failcntfile */ |
cdc2fcfe | 628 | cft = &h->cgroup_files_legacy[6]; |
abb8206c | 629 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.failcnt", buf); |
cdc2fcfe MA |
630 | cft->private = MEMFILE_PRIVATE(idx, RES_FAILCNT); |
631 | cft->write = hugetlb_cgroup_reset; | |
632 | cft->read_u64 = hugetlb_cgroup_read_u64; | |
633 | ||
634 | /* Add the reservation failcntfile */ | |
635 | cft = &h->cgroup_files_legacy[7]; | |
636 | snprintf(cft->name, MAX_CFTYPE_NAME, "%s.rsvd.failcnt", buf); | |
637 | cft->private = MEMFILE_PRIVATE(idx, RES_RSVD_FAILCNT); | |
6770c64e | 638 | cft->write = hugetlb_cgroup_reset; |
716f479d | 639 | cft->read_u64 = hugetlb_cgroup_read_u64; |
abb8206c AK |
640 | |
641 | /* NULL terminate the last cft */ | |
cdc2fcfe | 642 | cft = &h->cgroup_files_legacy[8]; |
abb8206c AK |
643 | memset(cft, 0, sizeof(*cft)); |
644 | ||
2cf669a5 | 645 | WARN_ON(cgroup_add_legacy_cftypes(&hugetlb_cgrp_subsys, |
faced7e0 GS |
646 | h->cgroup_files_legacy)); |
647 | } | |
648 | ||
649 | static void __init __hugetlb_cgroup_file_init(int idx) | |
650 | { | |
651 | __hugetlb_cgroup_file_dfl_init(idx); | |
652 | __hugetlb_cgroup_file_legacy_init(idx); | |
7179e7bf JW |
653 | } |
654 | ||
655 | void __init hugetlb_cgroup_file_init(void) | |
656 | { | |
657 | struct hstate *h; | |
658 | ||
659 | for_each_hstate(h) { | |
660 | /* | |
661 | * Add cgroup control files only if the huge page consists | |
662 | * of more than two normal pages. This is because we use | |
1d798ca3 | 663 | * page[2].private for storing cgroup details. |
7179e7bf JW |
664 | */ |
665 | if (huge_page_order(h) >= HUGETLB_CGROUP_MIN_ORDER) | |
666 | __hugetlb_cgroup_file_init(hstate_index(h)); | |
667 | } | |
abb8206c AK |
668 | } |
669 | ||
75754681 AK |
670 | /* |
671 | * hugetlb_lock will make sure a parallel cgroup rmdir won't happen | |
672 | * when we migrate hugepages | |
673 | */ | |
8e6ac7fa AK |
674 | void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage) |
675 | { | |
676 | struct hugetlb_cgroup *h_cg; | |
94ae8ba7 | 677 | struct hstate *h = page_hstate(oldhpage); |
8e6ac7fa AK |
678 | |
679 | if (hugetlb_cgroup_disabled()) | |
680 | return; | |
681 | ||
309381fe | 682 | VM_BUG_ON_PAGE(!PageHuge(oldhpage), oldhpage); |
8e6ac7fa AK |
683 | spin_lock(&hugetlb_lock); |
684 | h_cg = hugetlb_cgroup_from_page(oldhpage); | |
685 | set_hugetlb_cgroup(oldhpage, NULL); | |
8e6ac7fa AK |
686 | |
687 | /* move the h_cg details to new cgroup */ | |
688 | set_hugetlb_cgroup(newhpage, h_cg); | |
94ae8ba7 | 689 | list_move(&newhpage->lru, &h->hugepage_activelist); |
8e6ac7fa | 690 | spin_unlock(&hugetlb_lock); |
8e6ac7fa AK |
691 | return; |
692 | } | |
693 | ||
faced7e0 GS |
694 | static struct cftype hugetlb_files[] = { |
695 | {} /* terminate */ | |
696 | }; | |
697 | ||
073219e9 | 698 | struct cgroup_subsys hugetlb_cgrp_subsys = { |
92fb9748 TH |
699 | .css_alloc = hugetlb_cgroup_css_alloc, |
700 | .css_offline = hugetlb_cgroup_css_offline, | |
701 | .css_free = hugetlb_cgroup_css_free, | |
faced7e0 GS |
702 | .dfl_cftypes = hugetlb_files, |
703 | .legacy_cftypes = hugetlb_files, | |
2bc64a20 | 704 | }; |