b.liu | e958203 | 2025-04-17 19:18:16 +0800 | [diff] [blame^] | 1 | // SPDX-License-Identifier: GPL-2.0 |
| 2 | #define _GNU_SOURCE |
| 3 | #include <sched.h> |
| 4 | #include <sys/mount.h> |
| 5 | #include <sys/stat.h> |
| 6 | #include <sys/types.h> |
| 7 | #include <linux/limits.h> |
| 8 | #include <stdio.h> |
| 9 | #include <stdlib.h> |
| 10 | #include <linux/sched.h> |
| 11 | #include <fcntl.h> |
| 12 | #include <unistd.h> |
| 13 | #include <ftw.h> |
| 14 | |
| 15 | |
| 16 | #include "cgroup_helpers.h" |
| 17 | |
| 18 | /* |
| 19 | * To avoid relying on the system setup, when setup_cgroup_env is called |
| 20 | * we create a new mount namespace, and cgroup namespace. The cgroup2 |
| 21 | * root is mounted at CGROUP_MOUNT_PATH |
| 22 | * |
| 23 | * Unfortunately, most people don't have cgroupv2 enabled at this point in time. |
| 24 | * It's easier to create our own mount namespace and manage it ourselves. |
| 25 | * |
| 26 | * We assume /mnt exists. |
| 27 | */ |
| 28 | |
| 29 | #define WALK_FD_LIMIT 16 |
| 30 | #define CGROUP_MOUNT_PATH "/mnt" |
| 31 | #define CGROUP_WORK_DIR "/cgroup-test-work-dir" |
| 32 | #define format_cgroup_path(buf, path) \ |
| 33 | snprintf(buf, sizeof(buf), "%s%s%s", CGROUP_MOUNT_PATH, \ |
| 34 | CGROUP_WORK_DIR, path) |
| 35 | |
| 36 | /** |
| 37 | * enable_all_controllers() - Enable all available cgroup v2 controllers |
| 38 | * |
| 39 | * Enable all available cgroup v2 controllers in order to increase |
| 40 | * the code coverage. |
| 41 | * |
| 42 | * If successful, 0 is returned. |
| 43 | */ |
| 44 | int enable_all_controllers(char *cgroup_path) |
| 45 | { |
| 46 | char path[PATH_MAX + 1]; |
| 47 | char buf[PATH_MAX]; |
| 48 | char *c, *c2; |
| 49 | int fd, cfd; |
| 50 | ssize_t len; |
| 51 | |
| 52 | snprintf(path, sizeof(path), "%s/cgroup.controllers", cgroup_path); |
| 53 | fd = open(path, O_RDONLY); |
| 54 | if (fd < 0) { |
| 55 | log_err("Opening cgroup.controllers: %s", path); |
| 56 | return 1; |
| 57 | } |
| 58 | |
| 59 | len = read(fd, buf, sizeof(buf) - 1); |
| 60 | if (len < 0) { |
| 61 | close(fd); |
| 62 | log_err("Reading cgroup.controllers: %s", path); |
| 63 | return 1; |
| 64 | } |
| 65 | buf[len] = 0; |
| 66 | close(fd); |
| 67 | |
| 68 | /* No controllers available? We're probably on cgroup v1. */ |
| 69 | if (len == 0) |
| 70 | return 0; |
| 71 | |
| 72 | snprintf(path, sizeof(path), "%s/cgroup.subtree_control", cgroup_path); |
| 73 | cfd = open(path, O_RDWR); |
| 74 | if (cfd < 0) { |
| 75 | log_err("Opening cgroup.subtree_control: %s", path); |
| 76 | return 1; |
| 77 | } |
| 78 | |
| 79 | for (c = strtok_r(buf, " ", &c2); c; c = strtok_r(NULL, " ", &c2)) { |
| 80 | if (dprintf(cfd, "+%s\n", c) <= 0) { |
| 81 | log_err("Enabling controller %s: %s", c, path); |
| 82 | close(cfd); |
| 83 | return 1; |
| 84 | } |
| 85 | } |
| 86 | close(cfd); |
| 87 | return 0; |
| 88 | } |
| 89 | |
| 90 | /** |
| 91 | * setup_cgroup_environment() - Setup the cgroup environment |
| 92 | * |
| 93 | * After calling this function, cleanup_cgroup_environment should be called |
| 94 | * once testing is complete. |
| 95 | * |
| 96 | * This function will print an error to stderr and return 1 if it is unable |
| 97 | * to setup the cgroup environment. If setup is successful, 0 is returned. |
| 98 | */ |
| 99 | int setup_cgroup_environment(void) |
| 100 | { |
| 101 | char cgroup_workdir[PATH_MAX - 24]; |
| 102 | |
| 103 | format_cgroup_path(cgroup_workdir, ""); |
| 104 | |
| 105 | if (unshare(CLONE_NEWNS)) { |
| 106 | log_err("unshare"); |
| 107 | return 1; |
| 108 | } |
| 109 | |
| 110 | if (mount("none", "/", NULL, MS_REC | MS_PRIVATE, NULL)) { |
| 111 | log_err("mount fakeroot"); |
| 112 | return 1; |
| 113 | } |
| 114 | |
| 115 | if (mount("none", CGROUP_MOUNT_PATH, "cgroup2", 0, NULL) && errno != EBUSY) { |
| 116 | log_err("mount cgroup2"); |
| 117 | return 1; |
| 118 | } |
| 119 | |
| 120 | /* Cleanup existing failed runs, now that the environment is setup */ |
| 121 | cleanup_cgroup_environment(); |
| 122 | |
| 123 | if (mkdir(cgroup_workdir, 0777) && errno != EEXIST) { |
| 124 | log_err("mkdir cgroup work dir"); |
| 125 | return 1; |
| 126 | } |
| 127 | |
| 128 | if (enable_all_controllers(cgroup_workdir)) |
| 129 | return 1; |
| 130 | |
| 131 | return 0; |
| 132 | } |
| 133 | |
| 134 | static int nftwfunc(const char *filename, const struct stat *statptr, |
| 135 | int fileflags, struct FTW *pfwt) |
| 136 | { |
| 137 | if ((fileflags & FTW_D) && rmdir(filename)) |
| 138 | log_err("Removing cgroup: %s", filename); |
| 139 | return 0; |
| 140 | } |
| 141 | |
| 142 | |
| 143 | static int join_cgroup_from_top(char *cgroup_path) |
| 144 | { |
| 145 | char cgroup_procs_path[PATH_MAX + 1]; |
| 146 | pid_t pid = getpid(); |
| 147 | int fd, rc = 0; |
| 148 | |
| 149 | snprintf(cgroup_procs_path, sizeof(cgroup_procs_path), |
| 150 | "%s/cgroup.procs", cgroup_path); |
| 151 | |
| 152 | fd = open(cgroup_procs_path, O_WRONLY); |
| 153 | if (fd < 0) { |
| 154 | log_err("Opening Cgroup Procs: %s", cgroup_procs_path); |
| 155 | return 1; |
| 156 | } |
| 157 | |
| 158 | if (dprintf(fd, "%d\n", pid) < 0) { |
| 159 | log_err("Joining Cgroup"); |
| 160 | rc = 1; |
| 161 | } |
| 162 | |
| 163 | close(fd); |
| 164 | return rc; |
| 165 | } |
| 166 | |
| 167 | /** |
| 168 | * join_cgroup() - Join a cgroup |
| 169 | * @path: The cgroup path, relative to the workdir, to join |
| 170 | * |
| 171 | * This function expects a cgroup to already be created, relative to the cgroup |
| 172 | * work dir, and it joins it. For example, passing "/my-cgroup" as the path |
| 173 | * would actually put the calling process into the cgroup |
| 174 | * "/cgroup-test-work-dir/my-cgroup" |
| 175 | * |
| 176 | * On success, it returns 0, otherwise on failure it returns 1. |
| 177 | */ |
| 178 | int join_cgroup(const char *path) |
| 179 | { |
| 180 | char cgroup_path[PATH_MAX + 1]; |
| 181 | |
| 182 | format_cgroup_path(cgroup_path, path); |
| 183 | return join_cgroup_from_top(cgroup_path); |
| 184 | } |
| 185 | |
| 186 | /** |
| 187 | * cleanup_cgroup_environment() - Cleanup Cgroup Testing Environment |
| 188 | * |
| 189 | * This is an idempotent function to delete all temporary cgroups that |
| 190 | * have been created during the test, including the cgroup testing work |
| 191 | * directory. |
| 192 | * |
| 193 | * At call time, it moves the calling process to the root cgroup, and then |
| 194 | * runs the deletion process. It is idempotent, and should not fail, unless |
| 195 | * a process is lingering. |
| 196 | * |
| 197 | * On failure, it will print an error to stderr, and try to continue. |
| 198 | */ |
| 199 | void cleanup_cgroup_environment(void) |
| 200 | { |
| 201 | char cgroup_workdir[PATH_MAX + 1]; |
| 202 | |
| 203 | format_cgroup_path(cgroup_workdir, ""); |
| 204 | join_cgroup_from_top(CGROUP_MOUNT_PATH); |
| 205 | nftw(cgroup_workdir, nftwfunc, WALK_FD_LIMIT, FTW_DEPTH | FTW_MOUNT); |
| 206 | } |
| 207 | |
| 208 | /** |
| 209 | * create_and_get_cgroup() - Create a cgroup, relative to workdir, and get the FD |
| 210 | * @path: The cgroup path, relative to the workdir, to join |
| 211 | * |
| 212 | * This function creates a cgroup under the top level workdir and returns the |
| 213 | * file descriptor. It is idempotent. |
| 214 | * |
| 215 | * On success, it returns the file descriptor. On failure it returns -1. |
| 216 | * If there is a failure, it prints the error to stderr. |
| 217 | */ |
| 218 | int create_and_get_cgroup(const char *path) |
| 219 | { |
| 220 | char cgroup_path[PATH_MAX + 1]; |
| 221 | int fd; |
| 222 | |
| 223 | format_cgroup_path(cgroup_path, path); |
| 224 | if (mkdir(cgroup_path, 0777) && errno != EEXIST) { |
| 225 | log_err("mkdiring cgroup %s .. %s", path, cgroup_path); |
| 226 | return -1; |
| 227 | } |
| 228 | |
| 229 | fd = open(cgroup_path, O_RDONLY); |
| 230 | if (fd < 0) { |
| 231 | log_err("Opening Cgroup"); |
| 232 | return -1; |
| 233 | } |
| 234 | |
| 235 | return fd; |
| 236 | } |
| 237 | |
| 238 | /** |
| 239 | * get_cgroup_id() - Get cgroup id for a particular cgroup path |
| 240 | * @path: The cgroup path, relative to the workdir, to join |
| 241 | * |
| 242 | * On success, it returns the cgroup id. On failure it returns 0, |
| 243 | * which is an invalid cgroup id. |
| 244 | * If there is a failure, it prints the error to stderr. |
| 245 | */ |
| 246 | unsigned long long get_cgroup_id(const char *path) |
| 247 | { |
| 248 | int dirfd, err, flags, mount_id, fhsize; |
| 249 | union { |
| 250 | unsigned long long cgid; |
| 251 | unsigned char raw_bytes[8]; |
| 252 | } id; |
| 253 | char cgroup_workdir[PATH_MAX + 1]; |
| 254 | struct file_handle *fhp, *fhp2; |
| 255 | unsigned long long ret = 0; |
| 256 | |
| 257 | format_cgroup_path(cgroup_workdir, path); |
| 258 | |
| 259 | dirfd = AT_FDCWD; |
| 260 | flags = 0; |
| 261 | fhsize = sizeof(*fhp); |
| 262 | fhp = calloc(1, fhsize); |
| 263 | if (!fhp) { |
| 264 | log_err("calloc"); |
| 265 | return 0; |
| 266 | } |
| 267 | err = name_to_handle_at(dirfd, cgroup_workdir, fhp, &mount_id, flags); |
| 268 | if (err >= 0 || fhp->handle_bytes != 8) { |
| 269 | log_err("name_to_handle_at"); |
| 270 | goto free_mem; |
| 271 | } |
| 272 | |
| 273 | fhsize = sizeof(struct file_handle) + fhp->handle_bytes; |
| 274 | fhp2 = realloc(fhp, fhsize); |
| 275 | if (!fhp2) { |
| 276 | log_err("realloc"); |
| 277 | goto free_mem; |
| 278 | } |
| 279 | err = name_to_handle_at(dirfd, cgroup_workdir, fhp2, &mount_id, flags); |
| 280 | fhp = fhp2; |
| 281 | if (err < 0) { |
| 282 | log_err("name_to_handle_at"); |
| 283 | goto free_mem; |
| 284 | } |
| 285 | |
| 286 | memcpy(id.raw_bytes, fhp->f_handle, 8); |
| 287 | ret = id.cgid; |
| 288 | |
| 289 | free_mem: |
| 290 | free(fhp); |
| 291 | return ret; |
| 292 | } |