1 /*- 2 * Copyright (c) 2009-2010 The FreeBSD Foundation 3 * All rights reserved. 4 * 5 * This software was developed by Pawel Jakub Dawidek under sponsorship from 6 * the FreeBSD Foundation. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 */ 29 30 #include <sys/cdefs.h> 31 __FBSDID("$FreeBSD$"); 32 33 #include <sys/types.h> 34 #include <sys/wait.h> 35 36 #include <errno.h> 37 #include <pthread.h> 38 #include <signal.h> 39 #include <stdio.h> 40 #include <string.h> 41 #include <unistd.h> 42 43 #include "hast.h" 44 #include "hastd.h" 45 #include "hast_checksum.h" 46 #include "hast_compression.h" 47 #include "hast_proto.h" 48 #include "hooks.h" 49 #include "nv.h" 50 #include "pjdlog.h" 51 #include "proto.h" 52 #include "subr.h" 53 54 #include "control.h" 55 56 void 57 child_cleanup(struct hast_resource *res) 58 { 59 60 proto_close(res->hr_ctrl); 61 res->hr_ctrl = NULL; 62 if (res->hr_event != NULL) { 63 proto_close(res->hr_event); 64 res->hr_event = NULL; 65 } 66 if (res->hr_conn != NULL) { 67 proto_close(res->hr_conn); 68 res->hr_conn = NULL; 69 } 70 res->hr_workerpid = 0; 71 } 72 73 static void 74 control_set_role_common(struct hastd_config *cfg, struct nv *nvout, 75 uint8_t role, struct hast_resource *res, const char *name, unsigned int no) 76 { 77 int oldrole; 78 79 /* Name is always needed. */ 80 if (name != NULL) 81 nv_add_string(nvout, name, "resource%u", no); 82 83 if (res == NULL) { 84 PJDLOG_ASSERT(cfg != NULL); 85 PJDLOG_ASSERT(name != NULL); 86 87 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) { 88 if (strcmp(res->hr_name, name) == 0) 89 break; 90 } 91 if (res == NULL) { 92 nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no); 93 return; 94 } 95 } 96 PJDLOG_ASSERT(res != NULL); 97 98 /* Send previous role back. */ 99 nv_add_string(nvout, role2str(res->hr_role), "role%u", no); 100 101 /* Nothing changed, return here. */ 102 if (role == res->hr_role) 103 return; 104 105 pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role)); 106 pjdlog_info("Role changed to %s.", role2str(role)); 107 108 /* Change role to the new one. */ 109 oldrole = res->hr_role; 110 res->hr_role = role; 111 pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role)); 112 113 /* 114 * If previous role was primary or secondary we have to kill process 115 * doing that work. 116 */ 117 if (res->hr_workerpid != 0) { 118 if (kill(res->hr_workerpid, SIGTERM) < 0) { 119 pjdlog_errno(LOG_WARNING, 120 "Unable to kill worker process %u", 121 (unsigned int)res->hr_workerpid); 122 } else if (waitpid(res->hr_workerpid, NULL, 0) != 123 res->hr_workerpid) { 124 pjdlog_errno(LOG_WARNING, 125 "Error while waiting for worker process %u", 126 (unsigned int)res->hr_workerpid); 127 } else { 128 pjdlog_debug(1, "Worker process %u stopped.", 129 (unsigned int)res->hr_workerpid); 130 } 131 child_cleanup(res); 132 } 133 134 /* Start worker process if we are changing to primary. */ 135 if (role == HAST_ROLE_PRIMARY) 136 hastd_primary(res); 137 pjdlog_prefix_set("%s", ""); 138 hook_exec(res->hr_exec, "role", res->hr_name, role2str(oldrole), 139 role2str(res->hr_role), NULL); 140 } 141 142 void 143 control_set_role(struct hast_resource *res, uint8_t role) 144 { 145 146 control_set_role_common(NULL, NULL, role, res, NULL, 0); 147 } 148 149 static void 150 control_status_worker(struct hast_resource *res, struct nv *nvout, 151 unsigned int no) 152 { 153 struct nv *cnvin, *cnvout; 154 const char *str; 155 int error; 156 157 cnvin = NULL; 158 159 /* 160 * Prepare and send command to worker process. 161 */ 162 cnvout = nv_alloc(); 163 nv_add_uint8(cnvout, CONTROL_STATUS, "cmd"); 164 error = nv_error(cnvout); 165 if (error != 0) { 166 pjdlog_common(LOG_ERR, 0, error, 167 "Unable to prepare control header"); 168 goto end; 169 } 170 if (hast_proto_send(res, res->hr_ctrl, cnvout, NULL, 0) < 0) { 171 error = errno; 172 pjdlog_errno(LOG_ERR, "Unable to send control header"); 173 goto end; 174 } 175 176 /* 177 * Receive response. 178 */ 179 if (hast_proto_recv_hdr(res->hr_ctrl, &cnvin) < 0) { 180 error = errno; 181 pjdlog_errno(LOG_ERR, "Unable to receive control header"); 182 goto end; 183 } 184 185 error = nv_get_int16(cnvin, "error"); 186 if (error != 0) 187 goto end; 188 189 if ((str = nv_get_string(cnvin, "status")) == NULL) { 190 error = ENOENT; 191 pjdlog_errno(LOG_ERR, "Field 'status' is missing."); 192 goto end; 193 } 194 nv_add_string(nvout, str, "status%u", no); 195 nv_add_uint64(nvout, nv_get_uint64(cnvin, "dirty"), "dirty%u", no); 196 nv_add_uint32(nvout, nv_get_uint32(cnvin, "extentsize"), 197 "extentsize%u", no); 198 nv_add_uint32(nvout, nv_get_uint32(cnvin, "keepdirty"), 199 "keepdirty%u", no); 200 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_read"), 201 "stat_read%u", no); 202 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_write"), 203 "stat_write%u", no); 204 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_delete"), 205 "stat_delete%u", no); 206 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_flush"), 207 "stat_flush%u", no); 208 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_activemap_update"), 209 "stat_activemap_update%u", no); 210 end: 211 if (cnvin != NULL) 212 nv_free(cnvin); 213 if (cnvout != NULL) 214 nv_free(cnvout); 215 if (error != 0) 216 nv_add_int16(nvout, error, "error"); 217 } 218 219 static void 220 control_status(struct hastd_config *cfg, struct nv *nvout, 221 struct hast_resource *res, const char *name, unsigned int no) 222 { 223 224 PJDLOG_ASSERT(cfg != NULL); 225 PJDLOG_ASSERT(nvout != NULL); 226 PJDLOG_ASSERT(name != NULL); 227 228 /* Name is always needed. */ 229 nv_add_string(nvout, name, "resource%u", no); 230 231 if (res == NULL) { 232 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) { 233 if (strcmp(res->hr_name, name) == 0) 234 break; 235 } 236 if (res == NULL) { 237 nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no); 238 return; 239 } 240 } 241 PJDLOG_ASSERT(res != NULL); 242 nv_add_string(nvout, res->hr_provname, "provname%u", no); 243 nv_add_string(nvout, res->hr_localpath, "localpath%u", no); 244 nv_add_string(nvout, res->hr_remoteaddr, "remoteaddr%u", no); 245 if (res->hr_sourceaddr[0] != '\0') 246 nv_add_string(nvout, res->hr_sourceaddr, "sourceaddr%u", no); 247 switch (res->hr_replication) { 248 case HAST_REPLICATION_FULLSYNC: 249 nv_add_string(nvout, "fullsync", "replication%u", no); 250 break; 251 case HAST_REPLICATION_MEMSYNC: 252 nv_add_string(nvout, "memsync", "replication%u", no); 253 break; 254 case HAST_REPLICATION_ASYNC: 255 nv_add_string(nvout, "async", "replication%u", no); 256 break; 257 default: 258 nv_add_string(nvout, "unknown", "replication%u", no); 259 break; 260 } 261 nv_add_string(nvout, checksum_name(res->hr_checksum), 262 "checksum%u", no); 263 nv_add_string(nvout, compression_name(res->hr_compression), 264 "compression%u", no); 265 nv_add_string(nvout, role2str(res->hr_role), "role%u", no); 266 267 switch (res->hr_role) { 268 case HAST_ROLE_PRIMARY: 269 PJDLOG_ASSERT(res->hr_workerpid != 0); 270 /* FALLTHROUGH */ 271 case HAST_ROLE_SECONDARY: 272 if (res->hr_workerpid != 0) 273 break; 274 /* FALLTHROUGH */ 275 default: 276 return; 277 } 278 279 /* 280 * If we are here, it means that we have a worker process, which we 281 * want to ask some questions. 282 */ 283 control_status_worker(res, nvout, no); 284 } 285 286 void 287 control_handle(struct hastd_config *cfg) 288 { 289 struct proto_conn *conn; 290 struct nv *nvin, *nvout; 291 unsigned int ii; 292 const char *str; 293 uint8_t cmd, role; 294 int error; 295 296 if (proto_accept(cfg->hc_controlconn, &conn) < 0) { 297 pjdlog_errno(LOG_ERR, "Unable to accept control connection"); 298 return; 299 } 300 301 cfg->hc_controlin = conn; 302 nvin = nvout = NULL; 303 role = HAST_ROLE_UNDEF; 304 305 if (hast_proto_recv_hdr(conn, &nvin) < 0) { 306 pjdlog_errno(LOG_ERR, "Unable to receive control header"); 307 nvin = NULL; 308 goto close; 309 } 310 311 /* Obtain command code. 0 means that nv_get_uint8() failed. */ 312 cmd = nv_get_uint8(nvin, "cmd"); 313 if (cmd == 0) { 314 pjdlog_error("Control header is missing 'cmd' field."); 315 error = EHAST_INVALID; 316 goto close; 317 } 318 319 /* Allocate outgoing nv structure. */ 320 nvout = nv_alloc(); 321 if (nvout == NULL) { 322 pjdlog_error("Unable to allocate header for control response."); 323 error = EHAST_NOMEMORY; 324 goto close; 325 } 326 327 error = 0; 328 329 str = nv_get_string(nvin, "resource0"); 330 if (str == NULL) { 331 pjdlog_error("Control header is missing 'resource0' field."); 332 error = EHAST_INVALID; 333 goto fail; 334 } 335 if (cmd == HASTCTL_CMD_SETROLE) { 336 role = nv_get_uint8(nvin, "role"); 337 switch (role) { 338 case HAST_ROLE_INIT: 339 case HAST_ROLE_PRIMARY: 340 case HAST_ROLE_SECONDARY: 341 break; 342 default: 343 pjdlog_error("Invalid role received (%hhu).", role); 344 error = EHAST_INVALID; 345 goto fail; 346 } 347 } 348 if (strcmp(str, "all") == 0) { 349 struct hast_resource *res; 350 351 /* All configured resources. */ 352 353 ii = 0; 354 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) { 355 switch (cmd) { 356 case HASTCTL_CMD_SETROLE: 357 control_set_role_common(cfg, nvout, role, res, 358 res->hr_name, ii++); 359 break; 360 case HASTCTL_CMD_STATUS: 361 control_status(cfg, nvout, res, res->hr_name, 362 ii++); 363 break; 364 default: 365 pjdlog_error("Invalid command received (%hhu).", 366 cmd); 367 error = EHAST_UNIMPLEMENTED; 368 goto fail; 369 } 370 } 371 } else { 372 /* Only selected resources. */ 373 374 for (ii = 0; ; ii++) { 375 str = nv_get_string(nvin, "resource%u", ii); 376 if (str == NULL) 377 break; 378 switch (cmd) { 379 case HASTCTL_CMD_SETROLE: 380 control_set_role_common(cfg, nvout, role, NULL, 381 str, ii); 382 break; 383 case HASTCTL_CMD_STATUS: 384 control_status(cfg, nvout, NULL, str, ii); 385 break; 386 default: 387 pjdlog_error("Invalid command received (%hhu).", 388 cmd); 389 error = EHAST_UNIMPLEMENTED; 390 goto fail; 391 } 392 } 393 } 394 if (nv_error(nvout) != 0) 395 goto close; 396 fail: 397 if (error != 0) 398 nv_add_int16(nvout, error, "error"); 399 400 if (hast_proto_send(NULL, conn, nvout, NULL, 0) < 0) 401 pjdlog_errno(LOG_ERR, "Unable to send control response"); 402 close: 403 if (nvin != NULL) 404 nv_free(nvin); 405 if (nvout != NULL) 406 nv_free(nvout); 407 proto_close(conn); 408 cfg->hc_controlin = NULL; 409 } 410 411 /* 412 * Thread handles control requests from the parent. 413 */ 414 void * 415 ctrl_thread(void *arg) 416 { 417 struct hast_resource *res = arg; 418 struct nv *nvin, *nvout; 419 uint8_t cmd; 420 421 for (;;) { 422 if (hast_proto_recv_hdr(res->hr_ctrl, &nvin) < 0) { 423 if (sigexit_received) 424 pthread_exit(NULL); 425 pjdlog_errno(LOG_ERR, 426 "Unable to receive control message"); 427 kill(getpid(), SIGTERM); 428 pthread_exit(NULL); 429 } 430 cmd = nv_get_uint8(nvin, "cmd"); 431 if (cmd == 0) { 432 pjdlog_error("Control message is missing 'cmd' field."); 433 nv_free(nvin); 434 continue; 435 } 436 nvout = nv_alloc(); 437 switch (cmd) { 438 case CONTROL_STATUS: 439 if (res->hr_remotein != NULL && 440 res->hr_remoteout != NULL) { 441 nv_add_string(nvout, "complete", "status"); 442 } else { 443 nv_add_string(nvout, "degraded", "status"); 444 } 445 nv_add_uint32(nvout, (uint32_t)res->hr_extentsize, 446 "extentsize"); 447 if (res->hr_role == HAST_ROLE_PRIMARY) { 448 nv_add_uint32(nvout, 449 (uint32_t)res->hr_keepdirty, "keepdirty"); 450 nv_add_uint64(nvout, 451 (uint64_t)(activemap_ndirty(res->hr_amp) * 452 res->hr_extentsize), "dirty"); 453 } else { 454 nv_add_uint32(nvout, (uint32_t)0, "keepdirty"); 455 nv_add_uint64(nvout, (uint64_t)0, "dirty"); 456 } 457 nv_add_uint64(nvout, res->hr_stat_read, "stat_read"); 458 nv_add_uint64(nvout, res->hr_stat_write, "stat_write"); 459 nv_add_uint64(nvout, res->hr_stat_delete, 460 "stat_delete"); 461 nv_add_uint64(nvout, res->hr_stat_flush, "stat_flush"); 462 nv_add_uint64(nvout, res->hr_stat_activemap_update, 463 "stat_activemap_update"); 464 nv_add_int16(nvout, 0, "error"); 465 break; 466 case CONTROL_RELOAD: 467 /* 468 * When parent receives SIGHUP and discovers that 469 * something related to us has changes, it sends reload 470 * message to us. 471 */ 472 PJDLOG_ASSERT(res->hr_role == HAST_ROLE_PRIMARY); 473 primary_config_reload(res, nvin); 474 nv_add_int16(nvout, 0, "error"); 475 break; 476 default: 477 nv_add_int16(nvout, EINVAL, "error"); 478 break; 479 } 480 nv_free(nvin); 481 if (nv_error(nvout) != 0) { 482 pjdlog_error("Unable to create answer on control message."); 483 nv_free(nvout); 484 continue; 485 } 486 if (hast_proto_send(NULL, res->hr_ctrl, nvout, NULL, 0) < 0) { 487 pjdlog_errno(LOG_ERR, 488 "Unable to send reply to control message"); 489 } 490 nv_free(nvout); 491 } 492 /* NOTREACHED */ 493 return (NULL); 494 } 495