src/basic/cgroup-util.h

   1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
   2 #pragma once
   3
   4 #include <dirent.h>
   5 #include <stdbool.h>
   6 #include <stdint.h>
   7 #include <stdio.h>
   8 #include <sys/statfs.h>
   9 #include <sys/types.h>
  10
  11 #include "def.h"
  12 #include "set.h"
  13
  14 #define SYSTEMD_CGROUP_CONTROLLER_LEGACY "name=systemd"
  15 #define SYSTEMD_CGROUP_CONTROLLER_HYBRID "name=unified"
  16 #define SYSTEMD_CGROUP_CONTROLLER "_systemd"
  17
  18 /* An enum of well known cgroup controllers */
  19 typedef enum CGroupController {
  20         /* Original cgroup controllers */
  21         CGROUP_CONTROLLER_CPU,
  22         CGROUP_CONTROLLER_CPUACCT,    /* v1 only */
  23         CGROUP_CONTROLLER_CPUSET,     /* v2 only */
  24         CGROUP_CONTROLLER_IO,         /* v2 only */
  25         CGROUP_CONTROLLER_BLKIO,      /* v1 only */
  26         CGROUP_CONTROLLER_MEMORY,
  27         CGROUP_CONTROLLER_DEVICES,    /* v1 only */
  28         CGROUP_CONTROLLER_PIDS,
  29
  30         /* BPF-based pseudo-controllers, v2 only */
  31         CGROUP_CONTROLLER_BPF_FIREWALL,
  32         CGROUP_CONTROLLER_BPF_DEVICES,
  33
  34         _CGROUP_CONTROLLER_MAX,
  35         _CGROUP_CONTROLLER_INVALID = -EINVAL,
  36 } CGroupController;
  37
  38 #define CGROUP_CONTROLLER_TO_MASK(c) (1U << (c))
  39
  40 /* A bit mask of well known cgroup controllers */
  41 typedef enum CGroupMask {
  42         CGROUP_MASK_CPU = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPU),
  43         CGROUP_MASK_CPUACCT = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPUACCT),
  44         CGROUP_MASK_CPUSET = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPUSET),
  45         CGROUP_MASK_IO = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_IO),
  46         CGROUP_MASK_BLKIO = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BLKIO),
  47         CGROUP_MASK_MEMORY = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_MEMORY),
  48         CGROUP_MASK_DEVICES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_DEVICES),
  49         CGROUP_MASK_PIDS = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_PIDS),
  50         CGROUP_MASK_BPF_FIREWALL = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_FIREWALL),
  51         CGROUP_MASK_BPF_DEVICES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_DEVICES),
  52
  53         /* All real cgroup v1 controllers */
  54         CGROUP_MASK_V1 = CGROUP_MASK_CPU|CGROUP_MASK_CPUACCT|CGROUP_MASK_BLKIO|CGROUP_MASK_MEMORY|CGROUP_MASK_DEVICES|CGROUP_MASK_PIDS,
  55
  56         /* All real cgroup v2 controllers */
  57         CGROUP_MASK_V2 = CGROUP_MASK_CPU|CGROUP_MASK_CPUSET|CGROUP_MASK_IO|CGROUP_MASK_MEMORY|CGROUP_MASK_PIDS,
  58
  59         /* All cgroup v2 BPF pseudo-controllers */
  60         CGROUP_MASK_BPF = CGROUP_MASK_BPF_FIREWALL|CGROUP_MASK_BPF_DEVICES,
  61
  62         _CGROUP_MASK_ALL = CGROUP_CONTROLLER_TO_MASK(_CGROUP_CONTROLLER_MAX) - 1
  63 } CGroupMask;
  64
  65 static inline CGroupMask CGROUP_MASK_EXTEND_JOINED(CGroupMask mask) {
  66         /* We always mount "cpu" and "cpuacct" in the same hierarchy. Hence, when one bit is set also set the other */
  67
  68         if (mask & (CGROUP_MASK_CPU|CGROUP_MASK_CPUACCT))
  69                 mask |= (CGROUP_MASK_CPU|CGROUP_MASK_CPUACCT);
  70
  71         return mask;
  72 }
  73
  74 CGroupMask get_cpu_accounting_mask(void);
  75 bool cpu_accounting_is_cheap(void);
  76
  77 /* Special values for all weight knobs on unified hierarchy */
  78 #define CGROUP_WEIGHT_INVALID (UINT64_MAX)
  79 #define CGROUP_WEIGHT_MIN UINT64_C(1)
  80 #define CGROUP_WEIGHT_MAX UINT64_C(10000)
  81 #define CGROUP_WEIGHT_DEFAULT UINT64_C(100)
  82
  83 #define CGROUP_LIMIT_MIN UINT64_C(0)
  84 #define CGROUP_LIMIT_MAX (UINT64_MAX)
  85
  86 static inline bool CGROUP_WEIGHT_IS_OK(uint64_t x) {
  87         return
  88             x == CGROUP_WEIGHT_INVALID ||
  89             (x >= CGROUP_WEIGHT_MIN && x <= CGROUP_WEIGHT_MAX);
  90 }
  91
  92 /* IO limits on unified hierarchy */
  93 typedef enum CGroupIOLimitType {
  94         CGROUP_IO_RBPS_MAX,
  95         CGROUP_IO_WBPS_MAX,
  96         CGROUP_IO_RIOPS_MAX,
  97         CGROUP_IO_WIOPS_MAX,
  98
  99         _CGROUP_IO_LIMIT_TYPE_MAX,
 100         _CGROUP_IO_LIMIT_TYPE_INVALID = -EINVAL,
 101 } CGroupIOLimitType;
 102
 103 extern const uint64_t cgroup_io_limit_defaults[_CGROUP_IO_LIMIT_TYPE_MAX];
 104
 105 const char* cgroup_io_limit_type_to_string(CGroupIOLimitType t) _const_;
 106 CGroupIOLimitType cgroup_io_limit_type_from_string(const char *s) _pure_;
 107
 108 /* Special values for the cpu.shares attribute */
 109 #define CGROUP_CPU_SHARES_INVALID (UINT64_MAX)
 110 #define CGROUP_CPU_SHARES_MIN UINT64_C(2)
 111 #define CGROUP_CPU_SHARES_MAX UINT64_C(262144)
 112 #define CGROUP_CPU_SHARES_DEFAULT UINT64_C(1024)
 113
 114 static inline bool CGROUP_CPU_SHARES_IS_OK(uint64_t x) {
 115         return
 116             x == CGROUP_CPU_SHARES_INVALID ||
 117             (x >= CGROUP_CPU_SHARES_MIN && x <= CGROUP_CPU_SHARES_MAX);
 118 }
 119
 120 /* Special values for the blkio.weight attribute */
 121 #define CGROUP_BLKIO_WEIGHT_INVALID (UINT64_MAX)
 122 #define CGROUP_BLKIO_WEIGHT_MIN UINT64_C(10)
 123 #define CGROUP_BLKIO_WEIGHT_MAX UINT64_C(1000)
 124 #define CGROUP_BLKIO_WEIGHT_DEFAULT UINT64_C(500)
 125
 126 static inline bool CGROUP_BLKIO_WEIGHT_IS_OK(uint64_t x) {
 127         return
 128             x == CGROUP_BLKIO_WEIGHT_INVALID ||
 129             (x >= CGROUP_BLKIO_WEIGHT_MIN && x <= CGROUP_BLKIO_WEIGHT_MAX);
 130 }
 131
 132 typedef enum CGroupUnified {
 133         CGROUP_UNIFIED_UNKNOWN = -1,
 134         CGROUP_UNIFIED_NONE = 0,        /* Both systemd and controllers on legacy */
 135         CGROUP_UNIFIED_SYSTEMD = 1,     /* Only systemd on unified */
 136         CGROUP_UNIFIED_ALL = 2,         /* Both systemd and controllers on unified */
 137 } CGroupUnified;
 138
 139 /*
 140  * General rules:
 141  *
 142  * We accept named hierarchies in the syntax "foo" and "name=foo".
 143  *
 144  * We expect that named hierarchies do not conflict in name with a
 145  * kernel hierarchy, modulo the "name=" prefix.
 146  *
 147  * We always generate "normalized" controller names, i.e. without the
 148  * "name=" prefix.
 149  *
 150  * We require absolute cgroup paths. When returning, we will always
 151  * generate paths with multiple adjacent / removed.
 152  */
 153
 154 int cg_enumerate_processes(const char *controller, const char *path, FILE **_f);
 155 int cg_read_pid(FILE *f, pid_t *_pid);
 156 int cg_read_event(const char *controller, const char *path, const char *event,
 157                   char **val);
 158
 159 int cg_enumerate_subgroups(const char *controller, const char *path, DIR **_d);
 160 int cg_read_subgroup(DIR *d, char **fn);
 161
 162 typedef enum CGroupFlags {
 163         CGROUP_SIGCONT     = 1 << 0,
 164         CGROUP_IGNORE_SELF = 1 << 1,
 165         CGROUP_REMOVE      = 1 << 2,
 166 } CGroupFlags;
 167
 168 typedef int (*cg_kill_log_func_t)(pid_t pid, int sig, void *userdata);
 169
 170 int cg_kill(const char *controller, const char *path, int sig, CGroupFlags flags, Set *s, cg_kill_log_func_t kill_log, void *userdata);
 171 int cg_kill_recursive(const char *controller, const char *path, int sig, CGroupFlags flags, Set *s, cg_kill_log_func_t kill_log, void *userdata);
 172
 173 int cg_split_spec(const char *spec, char **ret_controller, char **ret_path);
 174 int cg_mangle_path(const char *path, char **result);
 175
 176 int cg_get_path(const char *controller, const char *path, const char *suffix, char **fs);
 177 int cg_get_path_and_check(const char *controller, const char *path, const char *suffix, char **fs);
 178
 179 int cg_pid_get_path(const char *controller, pid_t pid, char **path);
 180
 181 int cg_rmdir(const char *controller, const char *path);
 182
 183 typedef enum  {
 184         CG_KEY_MODE_GRACEFUL = 1 << 0,
 185 } CGroupKeyMode;
 186
 187 int cg_set_attribute(const char *controller, const char *path, const char *attribute, const char *value);
 188 int cg_get_attribute(const char *controller, const char *path, const char *attribute, char **ret);
 189 int cg_get_keyed_attribute_full(const char *controller, const char *path, const char *attribute, char **keys, char **values, CGroupKeyMode mode);
 190
 191 static inline int cg_get_keyed_attribute(
 192                 const char *controller,
 193                 const char *path,
 194                 const char *attribute,
 195                 char **keys,
 196                 char **ret_values) {
 197         return cg_get_keyed_attribute_full(controller, path, attribute, keys, ret_values, 0);
 198 }
 199
 200 static inline int cg_get_keyed_attribute_graceful(
 201                 const char *controller,
 202                 const char *path,
 203                 const char *attribute,
 204                 char **keys,
 205                 char **ret_values) {
 206         return cg_get_keyed_attribute_full(controller, path, attribute, keys, ret_values, CG_KEY_MODE_GRACEFUL);
 207 }
 208
 209 int cg_get_attribute_as_uint64(const char *controller, const char *path, const char *attribute, uint64_t *ret);
 210
 211 /* Does a parse_boolean() on the attribute contents and sets ret accordingly */
 212 int cg_get_attribute_as_bool(const char *controller, const char *path, const char *attribute, bool *ret);
 213
 214 int cg_set_access(const char *controller, const char *path, uid_t uid, gid_t gid);
 215 int cg_get_owner(const char *controller, const char *path, uid_t *ret_uid);
 216
 217 int cg_set_xattr(const char *controller, const char *path, const char *name, const void *value, size_t size, int flags);
 218 int cg_get_xattr(const char *controller, const char *path, const char *name, void *value, size_t size);
 219 int cg_get_xattr_malloc(const char *controller, const char *path, const char *name, char **ret);
 220 /* Returns negative on error, and 0 or 1 on success for the bool value */
 221 int cg_get_xattr_bool(const char *controller, const char *path, const char *name);
 222 int cg_remove_xattr(const char *controller, const char *path, const char *name);
 223
 224 int cg_install_release_agent(const char *controller, const char *agent);
 225 int cg_uninstall_release_agent(const char *controller);
 226
 227 int cg_is_empty(const char *controller, const char *path);
 228 int cg_is_empty_recursive(const char *controller, const char *path);
 229
 230 int cg_get_root_path(char **path);
 231
 232 int cg_path_get_session(const char *path, char **session);
 233 int cg_path_get_owner_uid(const char *path, uid_t *uid);
 234 int cg_path_get_unit(const char *path, char **unit);
 235 int cg_path_get_user_unit(const char *path, char **unit);
 236 int cg_path_get_machine_name(const char *path, char **machine);
 237 int cg_path_get_slice(const char *path, char **slice);
 238 int cg_path_get_user_slice(const char *path, char **slice);
 239
 240 int cg_shift_path(const char *cgroup, const char *cached_root, const char **shifted);
 241 int cg_pid_get_path_shifted(pid_t pid, const char *cached_root, char **cgroup);
 242
 243 int cg_pid_get_session(pid_t pid, char **session);
 244 int cg_pid_get_owner_uid(pid_t pid, uid_t *uid);
 245 int cg_pid_get_unit(pid_t pid, char **unit);
 246 int cg_pid_get_user_unit(pid_t pid, char **unit);
 247 int cg_pid_get_machine_name(pid_t pid, char **machine);
 248 int cg_pid_get_slice(pid_t pid, char **slice);
 249 int cg_pid_get_user_slice(pid_t pid, char **slice);
 250
 251 int cg_path_decode_unit(const char *cgroup, char **unit);
 252
 253 char *cg_escape(const char *p);
 254 char *cg_unescape(const char *p) _pure_;
 255
 256 bool cg_controller_is_valid(const char *p);
 257
 258 int cg_slice_to_path(const char *unit, char **ret);
 259
 260 typedef const char* (*cg_migrate_callback_t)(CGroupMask mask, void *userdata);
 261
 262 int cg_mask_supported(CGroupMask *ret);
 263 int cg_mask_from_string(const char *s, CGroupMask *ret);
 264 int cg_mask_to_string(CGroupMask mask, char **ret);
 265
 266 int cg_kernel_controllers(Set **controllers);
 267
 268 bool cg_ns_supported(void);
 269 bool cg_freezer_supported(void);
 270
 271 int cg_all_unified(void);
 272 int cg_hybrid_unified(void);
 273 int cg_unified_controller(const char *controller);
 274 int cg_unified_cached(bool flush);
 275 static inline int cg_unified(void) {
 276         return cg_unified_cached(true);
 277 }
 278
 279 const char* cgroup_controller_to_string(CGroupController c) _const_;
 280 CGroupController cgroup_controller_from_string(const char *s) _pure_;
 281
 282 bool is_cgroup_fs(const struct statfs *s);
 283 bool fd_is_cgroup_fs(int fd);
 284
 285 typedef enum ManagedOOMMode {
 286         MANAGED_OOM_AUTO,
 287         MANAGED_OOM_KILL,
 288         _MANAGED_OOM_MODE_MAX,
 289         _MANAGED_OOM_MODE_INVALID = -EINVAL,
 290 } ManagedOOMMode;
 291
 292 const char* managed_oom_mode_to_string(ManagedOOMMode m) _const_;
 293 ManagedOOMMode managed_oom_mode_from_string(const char *s) _pure_;
 294
 295 typedef enum ManagedOOMPreference {
 296         MANAGED_OOM_PREFERENCE_NONE = 0,
 297         MANAGED_OOM_PREFERENCE_AVOID = 1,
 298         MANAGED_OOM_PREFERENCE_OMIT = 2,
 299         _MANAGED_OOM_PREFERENCE_MAX,
 300         _MANAGED_OOM_PREFERENCE_INVALID = -EINVAL,
 301 } ManagedOOMPreference;
 302
 303 const char* managed_oom_preference_to_string(ManagedOOMPreference a) _const_;
 304 ManagedOOMPreference managed_oom_preference_from_string(const char *s) _pure_;