Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
1da177e4 LT |
2 | #ifndef _LINUX_CPUSET_H |
3 | #define _LINUX_CPUSET_H | |
4 | /* | |
5 | * cpuset interface | |
6 | * | |
7 | * Copyright (C) 2003 BULL SA | |
825a46af | 8 | * Copyright (C) 2004-2006 Silicon Graphics, Inc. |
1da177e4 LT |
9 | * |
10 | */ | |
11 | ||
12 | #include <linux/sched.h> | |
105ab3d8 | 13 | #include <linux/sched/topology.h> |
f719ff9b | 14 | #include <linux/sched/task.h> |
1da177e4 LT |
15 | #include <linux/cpumask.h> |
16 | #include <linux/nodemask.h> | |
a1bc5a4e | 17 | #include <linux/mm.h> |
d4b96fb9 | 18 | #include <linux/mmu_context.h> |
664eedde | 19 | #include <linux/jump_label.h> |
1da177e4 LT |
20 | |
21 | #ifdef CONFIG_CPUSETS | |
22 | ||
89affbf5 DZ |
23 | /* |
24 | * Static branch rewrites can happen in an arbitrary order for a given | |
25 | * key. In code paths where we need to loop with read_mems_allowed_begin() and | |
26 | * read_mems_allowed_retry() to get a consistent view of mems_allowed, we need | |
27 | * to ensure that begin() always gets rewritten before retry() in the | |
28 | * disabled -> enabled transition. If not, then if local irqs are disabled | |
29 | * around the loop, we can deadlock since retry() would always be | |
30 | * comparing the latest value of the mems_allowed seqcount against 0 as | |
31 | * begin() still would see cpusets_enabled() as false. The enabled -> disabled | |
32 | * transition should happen in reverse order for the same reasons (want to stop | |
33 | * looking at real value of mems_allowed.sequence in retry() first). | |
34 | */ | |
35 | extern struct static_key_false cpusets_pre_enable_key; | |
002f2906 | 36 | extern struct static_key_false cpusets_enabled_key; |
8ca1b5a4 FT |
37 | extern struct static_key_false cpusets_insane_config_key; |
38 | ||
664eedde MG |
39 | static inline bool cpusets_enabled(void) |
40 | { | |
002f2906 | 41 | return static_branch_unlikely(&cpusets_enabled_key); |
664eedde MG |
42 | } |
43 | ||
664eedde MG |
44 | static inline void cpuset_inc(void) |
45 | { | |
d74b27d6 JL |
46 | static_branch_inc_cpuslocked(&cpusets_pre_enable_key); |
47 | static_branch_inc_cpuslocked(&cpusets_enabled_key); | |
664eedde MG |
48 | } |
49 | ||
50 | static inline void cpuset_dec(void) | |
51 | { | |
d74b27d6 JL |
52 | static_branch_dec_cpuslocked(&cpusets_enabled_key); |
53 | static_branch_dec_cpuslocked(&cpusets_pre_enable_key); | |
664eedde | 54 | } |
202f72d5 | 55 | |
8ca1b5a4 FT |
56 | /* |
57 | * This will get enabled whenever a cpuset configuration is considered | |
58 | * unsupportable in general. E.g. movable only node which cannot satisfy | |
59 | * any non movable allocations (see update_nodemask). Page allocator | |
60 | * needs to make additional checks for those configurations and this | |
61 | * check is meant to guard those checks without any overhead for sane | |
62 | * configurations. | |
63 | */ | |
64 | static inline bool cpusets_insane_config(void) | |
65 | { | |
66 | return static_branch_unlikely(&cpusets_insane_config_key); | |
67 | } | |
68 | ||
1da177e4 LT |
69 | extern int cpuset_init(void); |
70 | extern void cpuset_init_smp(void); | |
50e76632 | 71 | extern void cpuset_force_rebuild(void); |
30e03acd | 72 | extern void cpuset_update_active_cpus(void); |
50e76632 | 73 | extern void cpuset_wait_for_hotplug(void); |
710da3c8 JL |
74 | extern void cpuset_read_lock(void); |
75 | extern void cpuset_read_unlock(void); | |
6af866af | 76 | extern void cpuset_cpus_allowed(struct task_struct *p, struct cpumask *mask); |
97c0054d | 77 | extern bool cpuset_cpus_allowed_fallback(struct task_struct *p); |
909d75a3 | 78 | extern nodemask_t cpuset_mems_allowed(struct task_struct *p); |
9276b1bc | 79 | #define cpuset_current_mems_allowed (current->mems_allowed) |
1da177e4 | 80 | void cpuset_init_current_mems_allowed(void); |
19770b32 | 81 | int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask); |
202f72d5 | 82 | |
8e464522 | 83 | extern bool cpuset_node_allowed(int node, gfp_t gfp_mask); |
02a0e53d | 84 | |
002f2906 | 85 | static inline bool __cpuset_zone_allowed(struct zone *z, gfp_t gfp_mask) |
202f72d5 | 86 | { |
8e464522 | 87 | return cpuset_node_allowed(zone_to_nid(z), gfp_mask); |
002f2906 VB |
88 | } |
89 | ||
90 | static inline bool cpuset_zone_allowed(struct zone *z, gfp_t gfp_mask) | |
91 | { | |
92 | if (cpusets_enabled()) | |
93 | return __cpuset_zone_allowed(z, gfp_mask); | |
94 | return true; | |
202f72d5 PJ |
95 | } |
96 | ||
bbe373f2 DR |
97 | extern int cpuset_mems_allowed_intersects(const struct task_struct *tsk1, |
98 | const struct task_struct *tsk2); | |
3e0d98b9 PJ |
99 | |
100 | #define cpuset_memory_pressure_bump() \ | |
101 | do { \ | |
102 | if (cpuset_memory_pressure_enabled) \ | |
103 | __cpuset_memory_pressure_bump(); \ | |
104 | } while (0) | |
105 | extern int cpuset_memory_pressure_enabled; | |
106 | extern void __cpuset_memory_pressure_bump(void); | |
107 | ||
df5f8314 EB |
108 | extern void cpuset_task_status_allowed(struct seq_file *m, |
109 | struct task_struct *task); | |
52de4779 ZL |
110 | extern int proc_cpuset_show(struct seq_file *m, struct pid_namespace *ns, |
111 | struct pid *pid, struct task_struct *tsk); | |
1da177e4 | 112 | |
825a46af | 113 | extern int cpuset_mem_spread_node(void); |
6adef3eb | 114 | extern int cpuset_slab_spread_node(void); |
825a46af PJ |
115 | |
116 | static inline int cpuset_do_page_mem_spread(void) | |
117 | { | |
2ad654bc | 118 | return task_spread_page(current); |
825a46af PJ |
119 | } |
120 | ||
121 | static inline int cpuset_do_slab_mem_spread(void) | |
122 | { | |
2ad654bc | 123 | return task_spread_slab(current); |
825a46af PJ |
124 | } |
125 | ||
77ef80c6 | 126 | extern bool current_cpuset_is_being_rebound(void); |
8793d854 | 127 | |
e761b772 MK |
128 | extern void rebuild_sched_domains(void); |
129 | ||
da39da3a | 130 | extern void cpuset_print_current_mems_allowed(void); |
75aa1994 | 131 | |
c0ff7453 | 132 | /* |
d26914d1 MG |
133 | * read_mems_allowed_begin is required when making decisions involving |
134 | * mems_allowed such as during page allocation. mems_allowed can be updated in | |
135 | * parallel and depending on the new value an operation can fail potentially | |
136 | * causing process failure. A retry loop with read_mems_allowed_begin and | |
137 | * read_mems_allowed_retry prevents these artificial failures. | |
c0ff7453 | 138 | */ |
d26914d1 | 139 | static inline unsigned int read_mems_allowed_begin(void) |
c0ff7453 | 140 | { |
89affbf5 | 141 | if (!static_branch_unlikely(&cpusets_pre_enable_key)) |
46e700ab MG |
142 | return 0; |
143 | ||
cc9a6c87 | 144 | return read_seqcount_begin(¤t->mems_allowed_seq); |
c0ff7453 MX |
145 | } |
146 | ||
cc9a6c87 | 147 | /* |
d26914d1 MG |
148 | * If this returns true, the operation that took place after |
149 | * read_mems_allowed_begin may have failed artificially due to a concurrent | |
150 | * update of mems_allowed. It is up to the caller to retry the operation if | |
cc9a6c87 MG |
151 | * appropriate. |
152 | */ | |
d26914d1 | 153 | static inline bool read_mems_allowed_retry(unsigned int seq) |
c0ff7453 | 154 | { |
89affbf5 | 155 | if (!static_branch_unlikely(&cpusets_enabled_key)) |
46e700ab MG |
156 | return false; |
157 | ||
d26914d1 | 158 | return read_seqcount_retry(¤t->mems_allowed_seq, seq); |
c0ff7453 MX |
159 | } |
160 | ||
58568d2a MX |
161 | static inline void set_mems_allowed(nodemask_t nodemask) |
162 | { | |
db751fe3 JS |
163 | unsigned long flags; |
164 | ||
c0ff7453 | 165 | task_lock(current); |
db751fe3 | 166 | local_irq_save(flags); |
cc9a6c87 | 167 | write_seqcount_begin(¤t->mems_allowed_seq); |
58568d2a | 168 | current->mems_allowed = nodemask; |
cc9a6c87 | 169 | write_seqcount_end(¤t->mems_allowed_seq); |
db751fe3 | 170 | local_irq_restore(flags); |
c0ff7453 | 171 | task_unlock(current); |
58568d2a MX |
172 | } |
173 | ||
1da177e4 LT |
174 | #else /* !CONFIG_CPUSETS */ |
175 | ||
664eedde MG |
176 | static inline bool cpusets_enabled(void) { return false; } |
177 | ||
8ca1b5a4 FT |
178 | static inline bool cpusets_insane_config(void) { return false; } |
179 | ||
1da177e4 LT |
180 | static inline int cpuset_init(void) { return 0; } |
181 | static inline void cpuset_init_smp(void) {} | |
1da177e4 | 182 | |
50e76632 PZ |
183 | static inline void cpuset_force_rebuild(void) { } |
184 | ||
30e03acd | 185 | static inline void cpuset_update_active_cpus(void) |
3a101d05 TH |
186 | { |
187 | partition_sched_domains(1, NULL, NULL); | |
188 | } | |
189 | ||
50e76632 PZ |
190 | static inline void cpuset_wait_for_hotplug(void) { } |
191 | ||
710da3c8 JL |
192 | static inline void cpuset_read_lock(void) { } |
193 | static inline void cpuset_read_unlock(void) { } | |
194 | ||
6af866af LZ |
195 | static inline void cpuset_cpus_allowed(struct task_struct *p, |
196 | struct cpumask *mask) | |
1da177e4 | 197 | { |
431c69fa | 198 | cpumask_copy(mask, task_cpu_possible_mask(p)); |
1da177e4 LT |
199 | } |
200 | ||
97c0054d | 201 | static inline bool cpuset_cpus_allowed_fallback(struct task_struct *p) |
9084bb82 | 202 | { |
97c0054d | 203 | return false; |
9084bb82 ON |
204 | } |
205 | ||
909d75a3 PJ |
206 | static inline nodemask_t cpuset_mems_allowed(struct task_struct *p) |
207 | { | |
208 | return node_possible_map; | |
209 | } | |
210 | ||
38d7bee9 | 211 | #define cpuset_current_mems_allowed (node_states[N_MEMORY]) |
1da177e4 | 212 | static inline void cpuset_init_current_mems_allowed(void) {} |
1da177e4 | 213 | |
19770b32 | 214 | static inline int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask) |
1da177e4 LT |
215 | { |
216 | return 1; | |
217 | } | |
218 | ||
002f2906 | 219 | static inline bool __cpuset_zone_allowed(struct zone *z, gfp_t gfp_mask) |
1da177e4 | 220 | { |
002f2906 VB |
221 | return true; |
222 | } | |
223 | ||
224 | static inline bool cpuset_zone_allowed(struct zone *z, gfp_t gfp_mask) | |
225 | { | |
226 | return true; | |
1da177e4 LT |
227 | } |
228 | ||
bbe373f2 DR |
229 | static inline int cpuset_mems_allowed_intersects(const struct task_struct *tsk1, |
230 | const struct task_struct *tsk2) | |
ef08e3b4 PJ |
231 | { |
232 | return 1; | |
233 | } | |
234 | ||
3e0d98b9 PJ |
235 | static inline void cpuset_memory_pressure_bump(void) {} |
236 | ||
df5f8314 EB |
237 | static inline void cpuset_task_status_allowed(struct seq_file *m, |
238 | struct task_struct *task) | |
1da177e4 | 239 | { |
1da177e4 LT |
240 | } |
241 | ||
825a46af PJ |
242 | static inline int cpuset_mem_spread_node(void) |
243 | { | |
244 | return 0; | |
245 | } | |
246 | ||
6adef3eb JS |
247 | static inline int cpuset_slab_spread_node(void) |
248 | { | |
249 | return 0; | |
250 | } | |
251 | ||
825a46af PJ |
252 | static inline int cpuset_do_page_mem_spread(void) |
253 | { | |
254 | return 0; | |
255 | } | |
256 | ||
257 | static inline int cpuset_do_slab_mem_spread(void) | |
258 | { | |
259 | return 0; | |
260 | } | |
261 | ||
77ef80c6 | 262 | static inline bool current_cpuset_is_being_rebound(void) |
8793d854 | 263 | { |
77ef80c6 | 264 | return false; |
8793d854 PM |
265 | } |
266 | ||
e761b772 MK |
267 | static inline void rebuild_sched_domains(void) |
268 | { | |
dfb512ec | 269 | partition_sched_domains(1, NULL, NULL); |
e761b772 MK |
270 | } |
271 | ||
da39da3a | 272 | static inline void cpuset_print_current_mems_allowed(void) |
75aa1994 DR |
273 | { |
274 | } | |
275 | ||
58568d2a MX |
276 | static inline void set_mems_allowed(nodemask_t nodemask) |
277 | { | |
278 | } | |
279 | ||
d26914d1 | 280 | static inline unsigned int read_mems_allowed_begin(void) |
c0ff7453 | 281 | { |
cc9a6c87 | 282 | return 0; |
c0ff7453 MX |
283 | } |
284 | ||
d26914d1 | 285 | static inline bool read_mems_allowed_retry(unsigned int seq) |
c0ff7453 | 286 | { |
d26914d1 | 287 | return false; |
c0ff7453 MX |
288 | } |
289 | ||
1da177e4 LT |
290 | #endif /* !CONFIG_CPUSETS */ |
291 | ||
292 | #endif /* _LINUX_CPUSET_H */ |