| 1 | /* |
| 2 | * Copyright (C) 2017 - Julien Desfossez <jdesfossez@efficios.com> |
| 3 | * |
| 4 | * This program is free software; you can redistribute it and/or modify it |
| 5 | * under the terms of the GNU General Public License, version 2 only, as |
| 6 | * published by the Free Software Foundation. |
| 7 | * |
| 8 | * This program is distributed in the hope that it will be useful, but WITHOUT |
| 9 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 10 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for |
| 11 | * more details. |
| 12 | * |
| 13 | * You should have received a copy of the GNU General Public License along with |
| 14 | * this program; if not, write to the Free Software Foundation, Inc., 51 |
| 15 | * Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. |
| 16 | */ |
| 17 | |
| 18 | #define _LGPL_SOURCE |
| 19 | #include <lttng/trigger/trigger.h> |
| 20 | #include <common/error.h> |
| 21 | #include <common/config/session-config.h> |
| 22 | #include <common/defaults.h> |
| 23 | #include <common/utils.h> |
| 24 | #include <common/futex.h> |
| 25 | #include <common/align.h> |
| 26 | #include <common/time.h> |
| 27 | #include <common/hashtable/utils.h> |
| 28 | #include <sys/eventfd.h> |
| 29 | #include <sys/stat.h> |
| 30 | #include <time.h> |
| 31 | #include <signal.h> |
| 32 | #include <inttypes.h> |
| 33 | |
| 34 | #include <common/kernel-ctl/kernel-ctl.h> |
| 35 | #include <lttng/notification/channel-internal.h> |
| 36 | |
| 37 | #include "rotation-thread.h" |
| 38 | #include "lttng-sessiond.h" |
| 39 | #include "health-sessiond.h" |
| 40 | #include "rotate.h" |
| 41 | #include "cmd.h" |
| 42 | #include "session.h" |
| 43 | #include "sessiond-timer.h" |
| 44 | |
| 45 | #include <urcu.h> |
| 46 | #include <urcu/list.h> |
| 47 | #include <urcu/rculfhash.h> |
| 48 | |
| 49 | /* |
| 50 | * Store a struct rotation_channel_info for each channel that is currently |
| 51 | * being rotated by the consumer. |
| 52 | */ |
| 53 | struct cds_lfht *channel_pending_rotate_ht; |
| 54 | |
| 55 | struct rotation_thread_state { |
| 56 | struct lttng_poll_event events; |
| 57 | }; |
| 58 | |
| 59 | static |
| 60 | void channel_rotation_info_destroy(struct rotation_channel_info *channel_info) |
| 61 | { |
| 62 | assert(channel_info); |
| 63 | free(channel_info); |
| 64 | } |
| 65 | |
| 66 | static |
| 67 | int match_channel_info(struct cds_lfht_node *node, const void *key) |
| 68 | { |
| 69 | struct rotation_channel_key *channel_key = (struct rotation_channel_key *) key; |
| 70 | struct rotation_channel_info *channel_info; |
| 71 | |
| 72 | channel_info = caa_container_of(node, struct rotation_channel_info, |
| 73 | rotate_channels_ht_node); |
| 74 | |
| 75 | return !!((channel_key->key == channel_info->channel_key.key) && |
| 76 | (channel_key->domain == channel_info->channel_key.domain)); |
| 77 | } |
| 78 | |
| 79 | static |
| 80 | struct rotation_channel_info *lookup_channel_pending(uint64_t key, |
| 81 | enum lttng_domain_type domain) |
| 82 | { |
| 83 | struct cds_lfht_iter iter; |
| 84 | struct cds_lfht_node *node; |
| 85 | struct rotation_channel_info *channel_info = NULL; |
| 86 | struct rotation_channel_key channel_key = { .key = key, |
| 87 | .domain = domain }; |
| 88 | |
| 89 | cds_lfht_lookup(channel_pending_rotate_ht, |
| 90 | hash_channel_key(&channel_key), |
| 91 | match_channel_info, |
| 92 | &channel_key, &iter); |
| 93 | node = cds_lfht_iter_get_node(&iter); |
| 94 | if (!node) { |
| 95 | goto end; |
| 96 | } |
| 97 | |
| 98 | channel_info = caa_container_of(node, struct rotation_channel_info, |
| 99 | rotate_channels_ht_node); |
| 100 | cds_lfht_del(channel_pending_rotate_ht, node); |
| 101 | end: |
| 102 | return channel_info; |
| 103 | } |
| 104 | |
| 105 | /* |
| 106 | * Destroy the thread data previously created by the init function. |
| 107 | */ |
| 108 | void rotation_thread_handle_destroy( |
| 109 | struct rotation_thread_handle *handle) |
| 110 | { |
| 111 | int ret; |
| 112 | |
| 113 | if (!handle) { |
| 114 | goto end; |
| 115 | } |
| 116 | |
| 117 | if (handle->ust32_consumer >= 0) { |
| 118 | ret = close(handle->ust32_consumer); |
| 119 | if (ret) { |
| 120 | PERROR("close 32-bit consumer channel rotation pipe"); |
| 121 | } |
| 122 | } |
| 123 | if (handle->ust64_consumer >= 0) { |
| 124 | ret = close(handle->ust64_consumer); |
| 125 | if (ret) { |
| 126 | PERROR("close 64-bit consumer channel rotation pipe"); |
| 127 | } |
| 128 | } |
| 129 | if (handle->kernel_consumer >= 0) { |
| 130 | ret = close(handle->kernel_consumer); |
| 131 | if (ret) { |
| 132 | PERROR("close kernel consumer channel rotation pipe"); |
| 133 | } |
| 134 | } |
| 135 | |
| 136 | end: |
| 137 | free(handle); |
| 138 | } |
| 139 | |
| 140 | struct rotation_thread_handle *rotation_thread_handle_create( |
| 141 | struct lttng_pipe *ust32_channel_rotate_pipe, |
| 142 | struct lttng_pipe *ust64_channel_rotate_pipe, |
| 143 | struct lttng_pipe *kernel_channel_rotate_pipe, |
| 144 | int thread_quit_pipe, |
| 145 | struct rotation_thread_timer_queue *rotation_timer_queue) |
| 146 | { |
| 147 | struct rotation_thread_handle *handle; |
| 148 | |
| 149 | handle = zmalloc(sizeof(*handle)); |
| 150 | if (!handle) { |
| 151 | goto end; |
| 152 | } |
| 153 | |
| 154 | if (ust32_channel_rotate_pipe) { |
| 155 | handle->ust32_consumer = |
| 156 | lttng_pipe_release_readfd( |
| 157 | ust32_channel_rotate_pipe); |
| 158 | if (handle->ust32_consumer < 0) { |
| 159 | goto error; |
| 160 | } |
| 161 | } else { |
| 162 | handle->ust32_consumer = -1; |
| 163 | } |
| 164 | if (ust64_channel_rotate_pipe) { |
| 165 | handle->ust64_consumer = |
| 166 | lttng_pipe_release_readfd( |
| 167 | ust64_channel_rotate_pipe); |
| 168 | if (handle->ust64_consumer < 0) { |
| 169 | goto error; |
| 170 | } |
| 171 | } else { |
| 172 | handle->ust64_consumer = -1; |
| 173 | } |
| 174 | if (kernel_channel_rotate_pipe) { |
| 175 | handle->kernel_consumer = |
| 176 | lttng_pipe_release_readfd( |
| 177 | kernel_channel_rotate_pipe); |
| 178 | if (handle->kernel_consumer < 0) { |
| 179 | goto error; |
| 180 | } |
| 181 | } else { |
| 182 | handle->kernel_consumer = -1; |
| 183 | } |
| 184 | handle->thread_quit_pipe = thread_quit_pipe; |
| 185 | handle->rotation_timer_queue = rotation_timer_queue; |
| 186 | |
| 187 | end: |
| 188 | return handle; |
| 189 | error: |
| 190 | rotation_thread_handle_destroy(handle); |
| 191 | return NULL; |
| 192 | } |
| 193 | |
| 194 | static |
| 195 | int init_poll_set(struct lttng_poll_event *poll_set, |
| 196 | struct rotation_thread_handle *handle) |
| 197 | { |
| 198 | int ret; |
| 199 | |
| 200 | /* |
| 201 | * Create pollset with size 5: |
| 202 | * - sessiond quit pipe |
| 203 | * - sessiond timer pipe, |
| 204 | * - consumerd (32-bit user space) channel rotate pipe, |
| 205 | * - consumerd (64-bit user space) channel rotate pipe, |
| 206 | * - consumerd (kernel) channel rotate pipe, |
| 207 | */ |
| 208 | ret = lttng_poll_create(poll_set, 5, LTTNG_CLOEXEC); |
| 209 | if (ret < 0) { |
| 210 | goto end; |
| 211 | } |
| 212 | |
| 213 | ret = lttng_poll_add(poll_set, handle->thread_quit_pipe, |
| 214 | LPOLLIN | LPOLLERR); |
| 215 | if (ret < 0) { |
| 216 | ERR("[rotation-thread] Failed to add thread_quit_pipe fd to pollset"); |
| 217 | goto error; |
| 218 | } |
| 219 | ret = lttng_poll_add(poll_set, |
| 220 | lttng_pipe_get_readfd(handle->rotation_timer_queue->event_pipe), |
| 221 | LPOLLIN | LPOLLERR); |
| 222 | if (ret < 0) { |
| 223 | ERR("[rotation-thread] Failed to add rotate_pending fd to pollset"); |
| 224 | goto error; |
| 225 | } |
| 226 | ret = lttng_poll_add(poll_set, handle->ust32_consumer, |
| 227 | LPOLLIN | LPOLLERR); |
| 228 | if (ret < 0) { |
| 229 | ERR("[rotation-thread] Failed to add ust-32 channel rotation pipe fd to pollset"); |
| 230 | goto error; |
| 231 | } |
| 232 | ret = lttng_poll_add(poll_set, handle->ust64_consumer, |
| 233 | LPOLLIN | LPOLLERR); |
| 234 | if (ret < 0) { |
| 235 | ERR("[rotation-thread] Failed to add ust-64 channel rotation pipe fd to pollset"); |
| 236 | goto error; |
| 237 | } |
| 238 | if (handle->kernel_consumer >= 0) { |
| 239 | ret = lttng_poll_add(poll_set, handle->kernel_consumer, |
| 240 | LPOLLIN | LPOLLERR); |
| 241 | if (ret < 0) { |
| 242 | ERR("[rotation-thread] Failed to add kernel channel rotation pipe fd to pollset"); |
| 243 | goto error; |
| 244 | } |
| 245 | } |
| 246 | |
| 247 | end: |
| 248 | return ret; |
| 249 | error: |
| 250 | lttng_poll_clean(poll_set); |
| 251 | return ret; |
| 252 | } |
| 253 | |
| 254 | static |
| 255 | void fini_thread_state(struct rotation_thread_state *state) |
| 256 | { |
| 257 | lttng_poll_clean(&state->events); |
| 258 | cds_lfht_destroy(channel_pending_rotate_ht, NULL); |
| 259 | } |
| 260 | |
| 261 | static |
| 262 | int init_thread_state(struct rotation_thread_handle *handle, |
| 263 | struct rotation_thread_state *state) |
| 264 | { |
| 265 | int ret; |
| 266 | |
| 267 | memset(state, 0, sizeof(*state)); |
| 268 | lttng_poll_init(&state->events); |
| 269 | |
| 270 | ret = init_poll_set(&state->events, handle); |
| 271 | if (ret) { |
| 272 | ERR("[rotation-thread] Failed to initialize rotation thread poll set"); |
| 273 | goto end; |
| 274 | } |
| 275 | |
| 276 | channel_pending_rotate_ht = cds_lfht_new(DEFAULT_HT_SIZE, |
| 277 | 1, 0, CDS_LFHT_AUTO_RESIZE | CDS_LFHT_ACCOUNTING, NULL); |
| 278 | if (!channel_pending_rotate_ht) { |
| 279 | ERR("[rotation-thread] Failed to create channel pending rotation hash table"); |
| 280 | ret = -1; |
| 281 | goto end; |
| 282 | } |
| 283 | |
| 284 | end: |
| 285 | return ret; |
| 286 | } |
| 287 | |
| 288 | static |
| 289 | int handle_channel_rotation_pipe(int fd, uint32_t revents, |
| 290 | struct rotation_thread_handle *handle, |
| 291 | struct rotation_thread_state *state) |
| 292 | { |
| 293 | int ret = 0; |
| 294 | enum lttng_domain_type domain; |
| 295 | struct rotation_channel_info *channel_info; |
| 296 | struct ltt_session *session = NULL; |
| 297 | uint64_t key; |
| 298 | |
| 299 | if (fd == handle->ust32_consumer || |
| 300 | fd == handle->ust64_consumer) { |
| 301 | domain = LTTNG_DOMAIN_UST; |
| 302 | } else if (fd == handle->kernel_consumer) { |
| 303 | domain = LTTNG_DOMAIN_KERNEL; |
| 304 | } else { |
| 305 | ERR("[rotation-thread] Unknown channel rotation pipe fd %d", |
| 306 | fd); |
| 307 | abort(); |
| 308 | } |
| 309 | |
| 310 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
| 311 | ret = lttng_poll_del(&state->events, fd); |
| 312 | if (ret) { |
| 313 | ERR("[rotation-thread] Failed to remove consumer " |
| 314 | "rotation pipe from poll set"); |
| 315 | } |
| 316 | goto end; |
| 317 | } |
| 318 | |
| 319 | do { |
| 320 | ret = read(fd, &key, sizeof(key)); |
| 321 | } while (ret == -1 && errno == EINTR); |
| 322 | if (ret != sizeof(key)) { |
| 323 | ERR("[rotation-thread] Failed to read from pipe (fd = %i)", |
| 324 | fd); |
| 325 | ret = -1; |
| 326 | goto end; |
| 327 | } |
| 328 | |
| 329 | DBG("[rotation-thread] Received notification for chan %" PRIu64 |
| 330 | ", domain %d\n", key, domain); |
| 331 | |
| 332 | channel_info = lookup_channel_pending(key, domain); |
| 333 | if (!channel_info) { |
| 334 | ERR("[rotation-thread] Failed to find channel_info (key = %" |
| 335 | PRIu64 ")", key); |
| 336 | ret = -1; |
| 337 | goto end; |
| 338 | } |
| 339 | rcu_read_lock(); |
| 340 | session_lock_list(); |
| 341 | session = session_find_by_id(channel_info->session_id); |
| 342 | if (!session) { |
| 343 | /* |
| 344 | * The session may have been destroyed before we had a chance to |
| 345 | * perform this action, return gracefully. |
| 346 | */ |
| 347 | DBG("[rotation-thread] Session %" PRIu64 " not found", |
| 348 | channel_info->session_id); |
| 349 | ret = 0; |
| 350 | goto end_unlock_session_list; |
| 351 | } |
| 352 | |
| 353 | session_lock(session); |
| 354 | if (--session->nr_chan_rotate_pending == 0) { |
| 355 | time_t now = time(NULL); |
| 356 | |
| 357 | if (now == (time_t) -1) { |
| 358 | session->rotation_status = LTTNG_ROTATION_STATUS_ERROR; |
| 359 | ret = LTTNG_ERR_UNK; |
| 360 | goto end_unlock_session; |
| 361 | } |
| 362 | |
| 363 | ret = rename_complete_chunk(session, now); |
| 364 | if (ret < 0) { |
| 365 | ERR("Failed to rename completed rotation chunk"); |
| 366 | goto end_unlock_session; |
| 367 | } |
| 368 | session->rotate_pending = false; |
| 369 | session->rotation_status = LTTNG_ROTATION_STATUS_COMPLETED; |
| 370 | session->last_chunk_start_ts = session->current_chunk_start_ts; |
| 371 | DBG("Rotation completed for session %s", session->name); |
| 372 | } |
| 373 | |
| 374 | ret = 0; |
| 375 | |
| 376 | end_unlock_session: |
| 377 | channel_rotation_info_destroy(channel_info); |
| 378 | session_unlock(session); |
| 379 | end_unlock_session_list: |
| 380 | session_unlock_list(); |
| 381 | rcu_read_unlock(); |
| 382 | end: |
| 383 | return ret; |
| 384 | } |
| 385 | |
| 386 | void *thread_rotation(void *data) |
| 387 | { |
| 388 | int ret; |
| 389 | struct rotation_thread_handle *handle = data; |
| 390 | struct rotation_thread_state state; |
| 391 | |
| 392 | DBG("[rotation-thread] Started rotation thread"); |
| 393 | |
| 394 | if (!handle) { |
| 395 | ERR("[rotation-thread] Invalid thread context provided"); |
| 396 | goto end; |
| 397 | } |
| 398 | |
| 399 | rcu_register_thread(); |
| 400 | rcu_thread_online(); |
| 401 | |
| 402 | health_register(health_sessiond, HEALTH_SESSIOND_TYPE_ROTATION); |
| 403 | health_code_update(); |
| 404 | |
| 405 | ret = init_thread_state(handle, &state); |
| 406 | if (ret) { |
| 407 | goto end; |
| 408 | } |
| 409 | |
| 410 | /* Ready to handle client connections. */ |
| 411 | sessiond_notify_ready(); |
| 412 | |
| 413 | while (true) { |
| 414 | int fd_count, i; |
| 415 | |
| 416 | health_poll_entry(); |
| 417 | DBG("[rotation-thread] Entering poll wait"); |
| 418 | ret = lttng_poll_wait(&state.events, -1); |
| 419 | DBG("[rotation-thread] Poll wait returned (%i)", ret); |
| 420 | health_poll_exit(); |
| 421 | if (ret < 0) { |
| 422 | /* |
| 423 | * Restart interrupted system call. |
| 424 | */ |
| 425 | if (errno == EINTR) { |
| 426 | continue; |
| 427 | } |
| 428 | ERR("[rotation-thread] Error encountered during lttng_poll_wait (%i)", ret); |
| 429 | goto error; |
| 430 | } |
| 431 | |
| 432 | fd_count = ret; |
| 433 | for (i = 0; i < fd_count; i++) { |
| 434 | int fd = LTTNG_POLL_GETFD(&state.events, i); |
| 435 | uint32_t revents = LTTNG_POLL_GETEV(&state.events, i); |
| 436 | |
| 437 | DBG("[rotation-thread] Handling fd (%i) activity (%u)", |
| 438 | fd, revents); |
| 439 | |
| 440 | if (fd == handle->thread_quit_pipe) { |
| 441 | DBG("[rotation-thread] Quit pipe activity"); |
| 442 | goto exit; |
| 443 | } else if (fd == handle->ust32_consumer || |
| 444 | fd == handle->ust64_consumer || |
| 445 | fd == handle->kernel_consumer) { |
| 446 | ret = handle_channel_rotation_pipe(fd, |
| 447 | revents, handle, &state); |
| 448 | if (ret) { |
| 449 | ERR("[rotation-thread] Handle channel rotation pipe"); |
| 450 | goto error; |
| 451 | } |
| 452 | } |
| 453 | } |
| 454 | } |
| 455 | exit: |
| 456 | error: |
| 457 | DBG("[rotation-thread] Exit"); |
| 458 | fini_thread_state(&state); |
| 459 | health_unregister(health_sessiond); |
| 460 | rcu_thread_offline(); |
| 461 | rcu_unregister_thread(); |
| 462 | end: |
| 463 | return NULL; |
| 464 | } |