src/basic/cgroup-util.h

   1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
   2 #pragma once
   3
   4 #include "forward.h"
   5
   6 #define SYSTEMD_CGROUP_CONTROLLER_LEGACY "name=systemd"
   7 #define SYSTEMD_CGROUP_CONTROLLER_HYBRID "name=unified"
   8 #define SYSTEMD_CGROUP_CONTROLLER "_systemd"
   9
  10 /* An enum of well known cgroup controllers */
  11 typedef enum CGroupController {
  12         /* Original cgroup controllers */
  13         CGROUP_CONTROLLER_CPU,
  14         CGROUP_CONTROLLER_CPUACCT,    /* v1 only */
  15         CGROUP_CONTROLLER_CPUSET,     /* v2 only */
  16         CGROUP_CONTROLLER_IO,         /* v2 only */
  17         CGROUP_CONTROLLER_BLKIO,      /* v1 only */
  18         CGROUP_CONTROLLER_MEMORY,
  19         CGROUP_CONTROLLER_DEVICES,    /* v1 only */
  20         CGROUP_CONTROLLER_PIDS,
  21
  22         /* BPF-based pseudo-controllers, v2 only */
  23         CGROUP_CONTROLLER_BPF_FIREWALL,
  24         CGROUP_CONTROLLER_BPF_DEVICES,
  25         CGROUP_CONTROLLER_BPF_FOREIGN,
  26         CGROUP_CONTROLLER_BPF_SOCKET_BIND,
  27         CGROUP_CONTROLLER_BPF_RESTRICT_NETWORK_INTERFACES,
  28         /* The BPF hook implementing RestrictFileSystems= is not defined here.
  29          * It's applied as late as possible in exec_invoke() so we don't block
  30          * our own unit setup code. */
  31
  32         _CGROUP_CONTROLLER_MAX,
  33         _CGROUP_CONTROLLER_INVALID = -EINVAL,
  34 } CGroupController;
  35
  36 #define CGROUP_CONTROLLER_TO_MASK(c) (1U << (c))
  37
  38 /* A bit mask of well known cgroup controllers */
  39 typedef enum CGroupMask {
  40         CGROUP_MASK_CPU = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPU),
  41         CGROUP_MASK_CPUACCT = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPUACCT),
  42         CGROUP_MASK_CPUSET = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPUSET),
  43         CGROUP_MASK_IO = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_IO),
  44         CGROUP_MASK_BLKIO = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BLKIO),
  45         CGROUP_MASK_MEMORY = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_MEMORY),
  46         CGROUP_MASK_DEVICES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_DEVICES),
  47         CGROUP_MASK_PIDS = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_PIDS),
  48         CGROUP_MASK_BPF_FIREWALL = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_FIREWALL),
  49         CGROUP_MASK_BPF_DEVICES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_DEVICES),
  50         CGROUP_MASK_BPF_FOREIGN = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_FOREIGN),
  51         CGROUP_MASK_BPF_SOCKET_BIND = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_SOCKET_BIND),
  52         CGROUP_MASK_BPF_RESTRICT_NETWORK_INTERFACES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_RESTRICT_NETWORK_INTERFACES),
  53
  54         /* All real cgroup v1 controllers */
  55         CGROUP_MASK_V1 = CGROUP_MASK_CPU|CGROUP_MASK_CPUACCT|CGROUP_MASK_BLKIO|CGROUP_MASK_MEMORY|CGROUP_MASK_DEVICES|CGROUP_MASK_PIDS,
  56
  57         /* All real cgroup v2 controllers */
  58         CGROUP_MASK_V2 = CGROUP_MASK_CPU|CGROUP_MASK_CPUSET|CGROUP_MASK_IO|CGROUP_MASK_MEMORY|CGROUP_MASK_PIDS,
  59
  60         /* All controllers we want to delegate in case of Delegate=yes. Which are pretty much the v2 controllers only, as delegation on v1 is not safe, and bpf stuff isn't a real controller */
  61         CGROUP_MASK_DELEGATE = CGROUP_MASK_V2,
  62
  63         /* All cgroup v2 BPF pseudo-controllers */
  64         CGROUP_MASK_BPF = CGROUP_MASK_BPF_FIREWALL|CGROUP_MASK_BPF_DEVICES|CGROUP_MASK_BPF_FOREIGN|CGROUP_MASK_BPF_SOCKET_BIND|CGROUP_MASK_BPF_RESTRICT_NETWORK_INTERFACES,
  65
  66         _CGROUP_MASK_ALL = CGROUP_CONTROLLER_TO_MASK(_CGROUP_CONTROLLER_MAX) - 1,
  67 } CGroupMask;
  68
  69 /* Special values for all weight knobs on unified hierarchy */
  70 #define CGROUP_WEIGHT_INVALID UINT64_MAX
  71 #define CGROUP_WEIGHT_IDLE UINT64_C(0)
  72 #define CGROUP_WEIGHT_MIN UINT64_C(1)
  73 #define CGROUP_WEIGHT_MAX UINT64_C(10000)
  74 #define CGROUP_WEIGHT_DEFAULT UINT64_C(100)
  75
  76 #define CGROUP_LIMIT_MIN UINT64_C(0)
  77 #define CGROUP_LIMIT_MAX UINT64_MAX
  78
  79 static inline bool CGROUP_WEIGHT_IS_OK(uint64_t x) {
  80         return
  81             x == CGROUP_WEIGHT_INVALID ||
  82             (x >= CGROUP_WEIGHT_MIN && x <= CGROUP_WEIGHT_MAX);
  83 }
  84
  85 /* IO limits on unified hierarchy */
  86 typedef enum CGroupIOLimitType {
  87         CGROUP_IO_RBPS_MAX,
  88         CGROUP_IO_WBPS_MAX,
  89         CGROUP_IO_RIOPS_MAX,
  90         CGROUP_IO_WIOPS_MAX,
  91
  92         _CGROUP_IO_LIMIT_TYPE_MAX,
  93         _CGROUP_IO_LIMIT_TYPE_INVALID = -EINVAL,
  94 } CGroupIOLimitType;
  95
  96 extern const uint64_t cgroup_io_limit_defaults[_CGROUP_IO_LIMIT_TYPE_MAX];
  97
  98 const char* cgroup_io_limit_type_to_string(CGroupIOLimitType t) _const_;
  99 CGroupIOLimitType cgroup_io_limit_type_from_string(const char *s) _pure_;
 100
 101 /* Special values for the io.bfq.weight attribute */
 102 #define CGROUP_BFQ_WEIGHT_INVALID UINT64_MAX
 103 #define CGROUP_BFQ_WEIGHT_MIN UINT64_C(1)
 104 #define CGROUP_BFQ_WEIGHT_MAX UINT64_C(1000)
 105 #define CGROUP_BFQ_WEIGHT_DEFAULT UINT64_C(100)
 106
 107 /* Convert the normal io.weight value to io.bfq.weight */
 108 static inline uint64_t BFQ_WEIGHT(uint64_t io_weight) {
 109         return
 110             io_weight <= CGROUP_WEIGHT_DEFAULT ?
 111             CGROUP_BFQ_WEIGHT_DEFAULT - (CGROUP_WEIGHT_DEFAULT - io_weight) * (CGROUP_BFQ_WEIGHT_DEFAULT - CGROUP_BFQ_WEIGHT_MIN) / (CGROUP_WEIGHT_DEFAULT - CGROUP_WEIGHT_MIN) :
 112             CGROUP_BFQ_WEIGHT_DEFAULT + (io_weight - CGROUP_WEIGHT_DEFAULT) * (CGROUP_BFQ_WEIGHT_MAX - CGROUP_BFQ_WEIGHT_DEFAULT) / (CGROUP_WEIGHT_MAX - CGROUP_WEIGHT_DEFAULT);
 113 }
 114
 115 typedef enum CGroupUnified {
 116         CGROUP_UNIFIED_UNKNOWN = -1,
 117         CGROUP_UNIFIED_NONE = 0,        /* Both systemd and controllers on legacy */
 118         CGROUP_UNIFIED_SYSTEMD = 1,     /* Only systemd on unified */
 119         CGROUP_UNIFIED_ALL = 2,         /* Both systemd and controllers on unified */
 120 } CGroupUnified;
 121
 122 /*
 123  * General rules:
 124  *
 125  * We accept named hierarchies in the syntax "foo" and "name=foo".
 126  *
 127  * We expect that named hierarchies do not conflict in name with a
 128  * kernel hierarchy, modulo the "name=" prefix.
 129  *
 130  * We always generate "normalized" controller names, i.e. without the
 131  * "name=" prefix.
 132  *
 133  * We require absolute cgroup paths. When returning, we will always
 134  * generate paths with multiple adjacent / removed.
 135  */
 136
 137 int cg_path_open(const char *controller, const char *path);
 138 int cg_cgroupid_open(int cgroupfs_fd, uint64_t id);
 139
 140 int cg_path_from_cgroupid(int cgroupfs_fd, uint64_t id, char **ret);
 141 int cg_get_cgroupid_at(int dfd, const char *path, uint64_t *ret);
 142 static inline int cg_path_get_cgroupid(const char *path, uint64_t *ret) {
 143         return cg_get_cgroupid_at(AT_FDCWD, path, ret);
 144 }
 145 static inline int cg_fd_get_cgroupid(int fd, uint64_t *ret) {
 146         return cg_get_cgroupid_at(fd, NULL, ret);
 147 }
 148
 149 typedef enum CGroupFlags {
 150         CGROUP_SIGCONT            = 1 << 0,
 151         CGROUP_IGNORE_SELF        = 1 << 1,
 152         CGROUP_DONT_SKIP_UNMAPPED = 1 << 2,
 153 } CGroupFlags;
 154
 155 int cg_enumerate_processes(const char *controller, const char *path, FILE **ret);
 156 int cg_read_pid(FILE *f, pid_t *ret, CGroupFlags flags);
 157 int cg_read_pidref(FILE *f, PidRef *ret, CGroupFlags flags);
 158
 159 int cg_enumerate_subgroups(const char *controller, const char *path, DIR **ret);
 160 int cg_read_subgroup(DIR *d, char **ret);
 161
 162 typedef int (*cg_kill_log_func_t)(const PidRef *pid, int sig, void *userdata);
 163
 164 int cg_kill(const char *path, int sig, CGroupFlags flags, Set *killed_pids, cg_kill_log_func_t log_kill, void *userdata);
 165 int cg_kill_kernel_sigkill(const char *path);
 166 int cg_kill_recursive(const char *path, int sig, CGroupFlags flags, Set *killed_pids, cg_kill_log_func_t log_kill, void *userdata);
 167
 168 int cg_split_spec(const char *spec, char **ret_controller, char **ret_path);
 169 int cg_mangle_path(const char *path, char **ret);
 170
 171 int cg_get_path(const char *controller, const char *path, const char *suffix, char **ret);
 172 int cg_get_path_and_check(const char *controller, const char *path, const char *suffix, char **ret);
 173
 174 int cg_pid_get_path(const char *controller, pid_t pid, char **ret);
 175 int cg_pidref_get_path(const char *controller, const PidRef *pidref, char **ret);
 176
 177 int cg_is_threaded(const char *path);
 178
 179 int cg_is_delegated(const char *path);
 180 int cg_is_delegated_fd(int fd);
 181
 182 int cg_has_coredump_receive(const char *path);
 183
 184 int cg_set_attribute(const char *controller, const char *path, const char *attribute, const char *value);
 185 int cg_get_attribute(const char *controller, const char *path, const char *attribute, char **ret);
 186 int cg_get_keyed_attribute(const char *controller, const char *path, const char *attribute, char * const *keys, char **values);
 187
 188 int cg_get_attribute_as_uint64(const char *controller, const char *path, const char *attribute, uint64_t *ret);
 189 int cg_get_attribute_as_bool(const char *controller, const char *path, const char *attribute);
 190
 191 int cg_get_owner(const char *path, uid_t *ret_uid);
 192
 193 int cg_set_xattr(const char *path, const char *name, const void *value, size_t size, int flags);
 194 int cg_get_xattr(const char *path, const char *name, char **ret, size_t *ret_size);
 195 /* Returns negative on error, and 0 or 1 on success for the bool value */
 196 int cg_get_xattr_bool(const char *path, const char *name);
 197 int cg_remove_xattr(const char *path, const char *name);
 198
 199 int cg_is_empty(const char *controller, const char *path);
 200
 201 int cg_get_root_path(char **path);
 202
 203 int cg_path_get_session(const char *path, char **ret_session);
 204 int cg_path_get_owner_uid(const char *path, uid_t *ret_uid);
 205 int cg_path_get_unit(const char *path, char **ret_unit);
 206 int cg_path_get_unit_path(const char *path, char **ret_unit);
 207 int cg_path_get_user_unit(const char *path, char **ret_unit);
 208 int cg_path_get_machine_name(const char *path, char **ret_machine);
 209 int cg_path_get_slice(const char *path, char **ret_slice);
 210 int cg_path_get_user_slice(const char *path, char **ret_slice);
 211
 212 int cg_shift_path(const char *cgroup, const char *cached_root, const char **ret_shifted);
 213 int cg_pid_get_path_shifted(pid_t pid, const char *cached_root, char **ret_cgroup);
 214
 215 int cg_pid_get_session(pid_t pid, char **ret_session);
 216 int cg_pidref_get_session(const PidRef *pidref, char **ret);
 217 int cg_pid_get_owner_uid(pid_t pid, uid_t *ret_uid);
 218 int cg_pidref_get_owner_uid(const PidRef *pidref, uid_t *ret);
 219 int cg_pid_get_unit(pid_t pid, char **ret_unit);
 220 int cg_pidref_get_unit(const PidRef *pidref, char **ret);
 221 int cg_pid_get_user_unit(pid_t pid, char **ret_unit);
 222 int cg_pid_get_machine_name(pid_t pid, char **ret_machine);
 223 int cg_pid_get_slice(pid_t pid, char **ret_slice);
 224 int cg_pid_get_user_slice(pid_t pid, char **ret_slice);
 225
 226 int cg_path_decode_unit(const char *cgroup, char **ret_unit);
 227
 228 bool cg_needs_escape(const char *p);
 229 int cg_escape(const char *p, char **ret);
 230 char* cg_unescape(const char *p) _pure_;
 231
 232 bool cg_controller_is_valid(const char *p);
 233
 234 int cg_slice_to_path(const char *unit, char **ret);
 235
 236 int cg_mask_supported(CGroupMask *ret);
 237 int cg_mask_supported_subtree(const char *root, CGroupMask *ret);
 238 int cg_mask_from_string(const char *s, CGroupMask *ret);
 239 int cg_mask_to_string(CGroupMask mask, char **ret);
 240
 241 bool cg_kill_supported(void);
 242
 243 int cg_all_unified(void);
 244 int cg_hybrid_unified(void);
 245 int cg_unified_controller(const char *controller);
 246 int cg_unified_cached(bool flush);
 247 static inline int cg_unified(void) {
 248         return cg_unified_cached(true);
 249 }
 250
 251 const char* cgroup_controller_to_string(CGroupController c) _const_;
 252 CGroupController cgroup_controller_from_string(const char *s) _pure_;
 253
 254 typedef enum ManagedOOMMode {
 255         MANAGED_OOM_AUTO,
 256         MANAGED_OOM_KILL,
 257         _MANAGED_OOM_MODE_MAX,
 258         _MANAGED_OOM_MODE_INVALID = -EINVAL,
 259 } ManagedOOMMode;
 260
 261 const char* managed_oom_mode_to_string(ManagedOOMMode m) _const_;
 262 ManagedOOMMode managed_oom_mode_from_string(const char *s) _pure_;
 263
 264 typedef enum ManagedOOMPreference {
 265         MANAGED_OOM_PREFERENCE_NONE = 0,
 266         MANAGED_OOM_PREFERENCE_AVOID = 1,
 267         MANAGED_OOM_PREFERENCE_OMIT = 2,
 268         _MANAGED_OOM_PREFERENCE_MAX,
 269         _MANAGED_OOM_PREFERENCE_INVALID = -EINVAL,
 270 } ManagedOOMPreference;
 271
 272 const char* managed_oom_preference_to_string(ManagedOOMPreference a) _const_;
 273 ManagedOOMPreference managed_oom_preference_from_string(const char *s) _pure_;