| 1 |
/* SPDX-License-Identifier: LGPL-2.1-or-later */ |
| 2 |
#pragma once |
| 3 |
|
| 4 |
#include "basic-forward.h" |
| 5 |
|
| 6 |
#define SYSTEMD_CGROUP_CONTROLLER "_systemd" |
| 7 |
|
| 8 |
/* An enum of well known cgroup controllers */ |
| 9 |
typedef enum CGroupController { |
| 10 |
/* Original cgroup controllers */ |
| 11 |
CGROUP_CONTROLLER_CPU, |
| 12 |
CGROUP_CONTROLLER_CPUACCT, /* v1 only */ |
| 13 |
CGROUP_CONTROLLER_CPUSET, /* v2 only */ |
| 14 |
CGROUP_CONTROLLER_IO, /* v2 only */ |
| 15 |
CGROUP_CONTROLLER_BLKIO, /* v1 only */ |
| 16 |
CGROUP_CONTROLLER_MEMORY, |
| 17 |
CGROUP_CONTROLLER_DEVICES, /* v1 only */ |
| 18 |
CGROUP_CONTROLLER_PIDS, |
| 19 |
|
| 20 |
/* BPF-based pseudo-controllers, v2 only */ |
| 21 |
CGROUP_CONTROLLER_BPF_FIREWALL, |
| 22 |
CGROUP_CONTROLLER_BPF_DEVICES, |
| 23 |
CGROUP_CONTROLLER_BPF_FOREIGN, |
| 24 |
CGROUP_CONTROLLER_BPF_SOCKET_BIND, |
| 25 |
CGROUP_CONTROLLER_BPF_RESTRICT_NETWORK_INTERFACES, |
| 26 |
CGROUP_CONTROLLER_BPF_BIND_NETWORK_INTERFACE, |
| 27 |
/* The BPF hook implementing RestrictFileSystems= is not defined here. |
| 28 |
* It's applied as late as possible in exec_invoke() so we don't block |
| 29 |
* our own unit setup code. */ |
| 30 |
|
| 31 |
_CGROUP_CONTROLLER_MAX, |
| 32 |
_CGROUP_CONTROLLER_INVALID = -EINVAL, |
| 33 |
} CGroupController; |
| 34 |
|
| 35 |
#define CGROUP_CONTROLLER_TO_MASK(c) (1U << (c)) |
| 36 |
|
| 37 |
/* A bit mask of well known cgroup controllers */ |
| 38 |
typedef enum CGroupMask { |
| 39 |
CGROUP_MASK_CPU = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPU), |
| 40 |
CGROUP_MASK_CPUACCT = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPUACCT), |
| 41 |
CGROUP_MASK_CPUSET = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_CPUSET), |
| 42 |
CGROUP_MASK_IO = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_IO), |
| 43 |
CGROUP_MASK_BLKIO = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BLKIO), |
| 44 |
CGROUP_MASK_MEMORY = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_MEMORY), |
| 45 |
CGROUP_MASK_DEVICES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_DEVICES), |
| 46 |
CGROUP_MASK_PIDS = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_PIDS), |
| 47 |
CGROUP_MASK_BPF_FIREWALL = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_FIREWALL), |
| 48 |
CGROUP_MASK_BPF_DEVICES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_DEVICES), |
| 49 |
CGROUP_MASK_BPF_FOREIGN = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_FOREIGN), |
| 50 |
CGROUP_MASK_BPF_SOCKET_BIND = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_SOCKET_BIND), |
| 51 |
CGROUP_MASK_BPF_RESTRICT_NETWORK_INTERFACES = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_RESTRICT_NETWORK_INTERFACES), |
| 52 |
CGROUP_MASK_BPF_BIND_NETWORK_INTERFACE = CGROUP_CONTROLLER_TO_MASK(CGROUP_CONTROLLER_BPF_BIND_NETWORK_INTERFACE), |
| 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|CGROUP_MASK_BPF_BIND_NETWORK_INTERFACE, |
| 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 |
DECLARE_STRING_TABLE_LOOKUP(cgroup_io_limit_type, CGroupIOLimitType); |
| 99 |
void cgroup_io_limits_list(void); |
| 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 |
/* |
| 116 |
* General rules: |
| 117 |
* |
| 118 |
* We require absolute cgroup paths. When returning, we will always |
| 119 |
* generate paths with multiple adjacent / removed. |
| 120 |
*/ |
| 121 |
|
| 122 |
int cg_is_available(void); |
| 123 |
|
| 124 |
int cg_path_open(const char *path); |
| 125 |
int cg_cgroupid_open(int cgroupfs_fd, uint64_t id); |
| 126 |
|
| 127 |
int cg_path_from_cgroupid(int cgroupfs_fd, uint64_t id, char **ret); |
| 128 |
|
| 129 |
typedef enum CGroupFlags { |
| 130 |
CGROUP_SIGCONT = 1 << 0, |
| 131 |
CGROUP_IGNORE_SELF = 1 << 1, |
| 132 |
CGROUP_DONT_SKIP_UNMAPPED = 1 << 2, |
| 133 |
} CGroupFlags; |
| 134 |
|
| 135 |
int cg_enumerate_processes(const char *path, FILE **ret); |
| 136 |
int cg_read_pid(FILE *f, pid_t *ret, CGroupFlags flags); |
| 137 |
int cg_read_pidref(FILE *f, PidRef *ret, CGroupFlags flags); |
| 138 |
|
| 139 |
int cg_enumerate_subgroups(const char *path, DIR **ret); |
| 140 |
int cg_read_subgroup(DIR *d, char **ret); |
| 141 |
|
| 142 |
typedef int (*cg_kill_log_func_t)(const PidRef *pid, int sig, void *userdata); |
| 143 |
|
| 144 |
int cg_kill(const char *path, int sig, CGroupFlags flags, Set *killed_pids, cg_kill_log_func_t log_kill, void *userdata); |
| 145 |
int cg_kill_kernel_sigkill(const char *path, uint64_t *ret_n_pids_killed); |
| 146 |
int cg_kill_recursive(const char *path, int sig, CGroupFlags flags, Set *killed_pids, cg_kill_log_func_t log_kill, void *userdata); |
| 147 |
|
| 148 |
int cg_split_spec(const char *spec, char **ret_controller, char **ret_path); |
| 149 |
|
| 150 |
int cg_get_path(const char *path, const char *suffix, char **ret); |
| 151 |
|
| 152 |
int cg_pid_get_path(pid_t pid, char **ret); |
| 153 |
int cg_pidref_get_path(const PidRef *pidref, char **ret); |
| 154 |
|
| 155 |
int cg_is_threaded(const char *path); |
| 156 |
|
| 157 |
int cg_is_delegated(const char *path); |
| 158 |
int cg_is_delegated_fd(int fd); |
| 159 |
|
| 160 |
int cg_has_coredump_receive(const char *path); |
| 161 |
|
| 162 |
int cg_set_attribute(const char *path, const char *attribute, const char *value); |
| 163 |
int cg_get_attribute(const char *path, const char *attribute, char **ret); |
| 164 |
int cg_get_attribute_as_uint64(const char *path, const char *attribute, uint64_t *ret); |
| 165 |
int cg_get_attribute_as_bool(const char *path, const char *attribute); |
| 166 |
|
| 167 |
int cg_get_keyed_attribute(const char *path, const char *attribute, char * const *keys, char **values); |
| 168 |
int cg_get_keyed_attribute_uint64(const char *path, const char *attribute, const char *key, uint64_t *ret); |
| 169 |
|
| 170 |
int cg_get_owner(const char *path, uid_t *ret_uid); |
| 171 |
|
| 172 |
int cg_set_xattr(const char *path, const char *name, const void *value, size_t size, int flags); |
| 173 |
int cg_get_xattr(const char *path, const char *name, char **ret, size_t *ret_size); |
| 174 |
/* Returns negative on error, and 0 or 1 on success for the bool value */ |
| 175 |
int cg_get_xattr_bool(const char *path, const char *name); |
| 176 |
int cg_remove_xattr(const char *path, const char *name); |
| 177 |
|
| 178 |
int cg_is_empty(const char *path); |
| 179 |
|
| 180 |
int cg_get_root_path(char **path); |
| 181 |
|
| 182 |
int cg_path_get_session(const char *path, char **ret_session); |
| 183 |
int cg_path_get_owner_uid(const char *path, uid_t *ret_uid); |
| 184 |
int cg_path_get_unit_full(const char *path, char **ret_unit, char **ret_subgroup); |
| 185 |
static inline int cg_path_get_unit(const char *path, char **ret_unit) { |
| 186 |
return cg_path_get_unit_full(path, ret_unit, NULL); |
| 187 |
} |
| 188 |
int cg_path_get_unit_path(const char *path, char **ret_unit); |
| 189 |
int cg_path_get_user_unit_full(const char *path, char **ret_unit, char **ret_subgroup); |
| 190 |
static inline int cg_path_get_user_unit(const char *path, char **ret_unit) { |
| 191 |
return cg_path_get_user_unit_full(path, ret_unit, NULL); |
| 192 |
} |
| 193 |
int cg_path_get_machine_name(const char *path, char **ret_machine); |
| 194 |
int cg_path_get_slice(const char *path, char **ret_slice); |
| 195 |
int cg_path_get_user_slice(const char *path, char **ret_slice); |
| 196 |
|
| 197 |
int cg_shift_path(const char *cgroup, const char *cached_root, const char **ret_shifted); |
| 198 |
int cg_pid_get_path_shifted(pid_t pid, const char *cached_root, char **ret_cgroup); |
| 199 |
|
| 200 |
int cg_pid_get_session(pid_t pid, char **ret_session); |
| 201 |
int cg_pidref_get_session(const PidRef *pidref, char **ret); |
| 202 |
int cg_pid_get_owner_uid(pid_t pid, uid_t *ret_uid); |
| 203 |
int cg_pidref_get_owner_uid(const PidRef *pidref, uid_t *ret); |
| 204 |
int cg_pid_get_unit_full(pid_t pid, char **ret_unit, char **ret_subgroup); |
| 205 |
static inline int cg_pid_get_unit(pid_t pid, char **ret_unit) { |
| 206 |
return cg_pid_get_unit_full(pid, ret_unit, NULL); |
| 207 |
} |
| 208 |
int cg_pidref_get_unit_full(const PidRef *pidref, char **ret_unit, char **ret_subgroup); |
| 209 |
static inline int cg_pidref_get_unit(const PidRef *pidref, char **ret_unit) { |
| 210 |
return cg_pidref_get_unit_full(pidref, ret_unit, NULL); |
| 211 |
} |
| 212 |
int cg_pid_get_user_unit_full(pid_t pid, char **ret_unit, char **ret_subgroup); |
| 213 |
static inline int cg_pid_get_user_unit(pid_t pid, char **ret_unit) { |
| 214 |
return cg_pid_get_user_unit_full(pid, ret_unit, NULL); |
| 215 |
} |
| 216 |
int cg_pidref_get_user_unit_full(const PidRef *pidref, char **ret_unit, char **ret_subgroup); |
| 217 |
static inline int cg_pidref_get_user_unit(const PidRef *pidref, char **ret_unit) { |
| 218 |
return cg_pidref_get_user_unit_full(pidref, ret_unit, NULL); |
| 219 |
} |
| 220 |
int cg_pid_get_machine_name(pid_t pid, char **ret_machine); |
| 221 |
int cg_pid_get_slice(pid_t pid, char **ret_slice); |
| 222 |
int cg_pid_get_user_slice(pid_t pid, char **ret_slice); |
| 223 |
|
| 224 |
int cg_path_decode_unit(const char *cgroup, char **ret_unit); |
| 225 |
|
| 226 |
bool cg_needs_escape(const char *p) _pure_; |
| 227 |
int cg_escape(const char *p, char **ret); |
| 228 |
char* cg_unescape(const char *p) _pure_; |
| 229 |
|
| 230 |
int cg_slice_to_path(const char *unit, char **ret); |
| 231 |
|
| 232 |
int cg_mask_supported(CGroupMask *ret); |
| 233 |
int cg_mask_supported_subtree(const char *root, CGroupMask *ret); |
| 234 |
int cg_mask_from_string(const char *s, CGroupMask *ret); |
| 235 |
int cg_mask_to_string(CGroupMask mask, char **ret); |
| 236 |
|
| 237 |
bool cg_kill_supported(void); |
| 238 |
|
| 239 |
DECLARE_STRING_TABLE_LOOKUP(cgroup_controller, CGroupController); |
| 240 |
|
| 241 |
typedef enum ManagedOOMMode { |
| 242 |
MANAGED_OOM_AUTO, |
| 243 |
MANAGED_OOM_KILL, |
| 244 |
_MANAGED_OOM_MODE_MAX, |
| 245 |
_MANAGED_OOM_MODE_INVALID = -EINVAL, |
| 246 |
} ManagedOOMMode; |
| 247 |
|
| 248 |
DECLARE_STRING_TABLE_LOOKUP(managed_oom_mode, ManagedOOMMode); |
| 249 |
|
| 250 |
typedef enum ManagedOOMPreference { |
| 251 |
MANAGED_OOM_PREFERENCE_NONE = 0, |
| 252 |
MANAGED_OOM_PREFERENCE_AVOID = 1, |
| 253 |
MANAGED_OOM_PREFERENCE_OMIT = 2, |
| 254 |
_MANAGED_OOM_PREFERENCE_MAX, |
| 255 |
_MANAGED_OOM_PREFERENCE_INVALID = -EINVAL, |
| 256 |
} ManagedOOMPreference; |
| 257 |
|
| 258 |
DECLARE_STRING_TABLE_LOOKUP(managed_oom_preference, ManagedOOMPreference); |