control.c revision 222228
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: head/sbin/hastd/control.c 222228 2011-05-23 21:15:19Z pjd $"); 32 33#include <sys/types.h> 34#include <sys/wait.h> 35 36#include <assert.h> 37#include <errno.h> 38#include <pthread.h> 39#include <signal.h> 40#include <stdio.h> 41#include <string.h> 42#include <unistd.h> 43 44#include "hast.h" 45#include "hastd.h" 46#include "hast_checksum.h" 47#include "hast_compression.h" 48#include "hast_proto.h" 49#include "hooks.h" 50#include "nv.h" 51#include "pjdlog.h" 52#include "proto.h" 53#include "subr.h" 54 55#include "control.h" 56 57void 58child_cleanup(struct hast_resource *res) 59{ 60 61 proto_close(res->hr_ctrl); 62 res->hr_ctrl = NULL; 63 if (res->hr_event != NULL) { 64 proto_close(res->hr_event); 65 res->hr_event = NULL; 66 } 67 if (res->hr_conn != NULL) { 68 proto_close(res->hr_conn); 69 res->hr_conn = NULL; 70 } 71 res->hr_workerpid = 0; 72} 73 74static void 75control_set_role_common(struct hastd_config *cfg, struct nv *nvout, 76 uint8_t role, struct hast_resource *res, const char *name, unsigned int no) 77{ 78 int oldrole; 79 80 /* Name is always needed. */ 81 if (name != NULL) 82 nv_add_string(nvout, name, "resource%u", no); 83 84 if (res == NULL) { 85 assert(cfg != NULL); 86 assert(name != NULL); 87 88 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) { 89 if (strcmp(res->hr_name, name) == 0) 90 break; 91 } 92 if (res == NULL) { 93 nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no); 94 return; 95 } 96 } 97 assert(res != NULL); 98 99 /* Send previous role back. */ 100 nv_add_string(nvout, role2str(res->hr_role), "role%u", no); 101 102 /* Nothing changed, return here. */ 103 if (role == res->hr_role) 104 return; 105 106 pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role)); 107 pjdlog_info("Role changed to %s.", role2str(role)); 108 109 /* Change role to the new one. */ 110 oldrole = res->hr_role; 111 res->hr_role = role; 112 pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role)); 113 114 /* 115 * If previous role was primary or secondary we have to kill process 116 * doing that work. 117 */ 118 if (res->hr_workerpid != 0) { 119 if (kill(res->hr_workerpid, SIGTERM) < 0) { 120 pjdlog_errno(LOG_WARNING, 121 "Unable to kill worker process %u", 122 (unsigned int)res->hr_workerpid); 123 } else if (waitpid(res->hr_workerpid, NULL, 0) != 124 res->hr_workerpid) { 125 pjdlog_errno(LOG_WARNING, 126 "Error while waiting for worker process %u", 127 (unsigned int)res->hr_workerpid); 128 } else { 129 pjdlog_debug(1, "Worker process %u stopped.", 130 (unsigned int)res->hr_workerpid); 131 } 132 child_cleanup(res); 133 } 134 135 /* Start worker process if we are changing to primary. */ 136 if (role == HAST_ROLE_PRIMARY) 137 hastd_primary(res); 138 pjdlog_prefix_set("%s", ""); 139 hook_exec(res->hr_exec, "role", res->hr_name, role2str(oldrole), 140 role2str(res->hr_role), NULL); 141} 142 143void 144control_set_role(struct hast_resource *res, uint8_t role) 145{ 146 147 control_set_role_common(NULL, NULL, role, res, NULL, 0); 148} 149 150static void 151control_status_worker(struct hast_resource *res, struct nv *nvout, 152 unsigned int no) 153{ 154 struct nv *cnvin, *cnvout; 155 const char *str; 156 int error; 157 158 cnvin = cnvout = NULL; 159 error = 0; 160 161 /* 162 * Prepare and send command to worker process. 163 */ 164 cnvout = nv_alloc(); 165 nv_add_uint8(cnvout, CONTROL_STATUS, "cmd"); 166 error = nv_error(cnvout); 167 if (error != 0) { 168 pjdlog_common(LOG_ERR, 0, error, 169 "Unable to prepare control header"); 170 goto end; 171 } 172 if (hast_proto_send(res, res->hr_ctrl, cnvout, NULL, 0) < 0) { 173 error = errno; 174 pjdlog_errno(LOG_ERR, "Unable to send control header"); 175 goto end; 176 } 177 178 /* 179 * Receive response. 180 */ 181 if (hast_proto_recv_hdr(res->hr_ctrl, &cnvin) < 0) { 182 error = errno; 183 pjdlog_errno(LOG_ERR, "Unable to receive control header"); 184 goto end; 185 } 186 187 error = nv_get_int16(cnvin, "error"); 188 if (error != 0) 189 goto end; 190 191 if ((str = nv_get_string(cnvin, "status")) == NULL) { 192 error = ENOENT; 193 pjdlog_errno(LOG_ERR, "Field 'status' is missing."); 194 goto end; 195 } 196 nv_add_string(nvout, str, "status%u", no); 197 nv_add_uint64(nvout, nv_get_uint64(cnvin, "dirty"), "dirty%u", no); 198 nv_add_uint32(nvout, nv_get_uint32(cnvin, "extentsize"), 199 "extentsize%u", no); 200 nv_add_uint32(nvout, nv_get_uint32(cnvin, "keepdirty"), 201 "keepdirty%u", no); 202 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_read"), 203 "stat_read%u", no); 204 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_write"), 205 "stat_write%u", no); 206 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_delete"), 207 "stat_delete%u", no); 208 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_flush"), 209 "stat_flush%u", no); 210 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_activemap_update"), 211 "stat_activemap_update%u", no); 212end: 213 if (cnvin != NULL) 214 nv_free(cnvin); 215 if (cnvout != NULL) 216 nv_free(cnvout); 217 if (error != 0) 218 nv_add_int16(nvout, error, "error"); 219} 220 221static void 222control_status(struct hastd_config *cfg, struct nv *nvout, 223 struct hast_resource *res, const char *name, unsigned int no) 224{ 225 226 assert(cfg != NULL); 227 assert(nvout != NULL); 228 assert(name != NULL); 229 230 /* Name is always needed. */ 231 nv_add_string(nvout, name, "resource%u", no); 232 233 if (res == NULL) { 234 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) { 235 if (strcmp(res->hr_name, name) == 0) 236 break; 237 } 238 if (res == NULL) { 239 nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no); 240 return; 241 } 242 } 243 assert(res != NULL); 244 nv_add_string(nvout, res->hr_provname, "provname%u", no); 245 nv_add_string(nvout, res->hr_localpath, "localpath%u", no); 246 nv_add_string(nvout, res->hr_remoteaddr, "remoteaddr%u", no); 247 if (res->hr_sourceaddr[0] != '\0') 248 nv_add_string(nvout, res->hr_sourceaddr, "sourceaddr%u", no); 249 switch (res->hr_replication) { 250 case HAST_REPLICATION_FULLSYNC: 251 nv_add_string(nvout, "fullsync", "replication%u", no); 252 break; 253 case HAST_REPLICATION_MEMSYNC: 254 nv_add_string(nvout, "memsync", "replication%u", no); 255 break; 256 case HAST_REPLICATION_ASYNC: 257 nv_add_string(nvout, "async", "replication%u", no); 258 break; 259 default: 260 nv_add_string(nvout, "unknown", "replication%u", no); 261 break; 262 } 263 nv_add_string(nvout, checksum_name(res->hr_checksum), 264 "checksum%u", no); 265 nv_add_string(nvout, compression_name(res->hr_compression), 266 "compression%u", no); 267 nv_add_string(nvout, role2str(res->hr_role), "role%u", no); 268 269 switch (res->hr_role) { 270 case HAST_ROLE_PRIMARY: 271 assert(res->hr_workerpid != 0); 272 /* FALLTHROUGH */ 273 case HAST_ROLE_SECONDARY: 274 if (res->hr_workerpid != 0) 275 break; 276 /* FALLTHROUGH */ 277 default: 278 return; 279 } 280 281 /* 282 * If we are here, it means that we have a worker process, which we 283 * want to ask some questions. 284 */ 285 control_status_worker(res, nvout, no); 286} 287 288void 289control_handle(struct hastd_config *cfg) 290{ 291 struct proto_conn *conn; 292 struct nv *nvin, *nvout; 293 unsigned int ii; 294 const char *str; 295 uint8_t cmd, role; 296 int error; 297 298 if (proto_accept(cfg->hc_controlconn, &conn) < 0) { 299 pjdlog_errno(LOG_ERR, "Unable to accept control connection"); 300 return; 301 } 302 303 cfg->hc_controlin = conn; 304 nvin = nvout = NULL; 305 role = HAST_ROLE_UNDEF; 306 307 if (hast_proto_recv_hdr(conn, &nvin) < 0) { 308 pjdlog_errno(LOG_ERR, "Unable to receive control header"); 309 nvin = NULL; 310 goto close; 311 } 312 313 /* Obtain command code. 0 means that nv_get_uint8() failed. */ 314 cmd = nv_get_uint8(nvin, "cmd"); 315 if (cmd == 0) { 316 pjdlog_error("Control header is missing 'cmd' field."); 317 error = EHAST_INVALID; 318 goto close; 319 } 320 321 /* Allocate outgoing nv structure. */ 322 nvout = nv_alloc(); 323 if (nvout == NULL) { 324 pjdlog_error("Unable to allocate header for control response."); 325 error = EHAST_NOMEMORY; 326 goto close; 327 } 328 329 error = 0; 330 331 str = nv_get_string(nvin, "resource0"); 332 if (str == NULL) { 333 pjdlog_error("Control header is missing 'resource0' field."); 334 error = EHAST_INVALID; 335 goto fail; 336 } 337 if (cmd == HASTCTL_CMD_SETROLE) { 338 role = nv_get_uint8(nvin, "role"); 339 switch (role) { 340 case HAST_ROLE_INIT: 341 case HAST_ROLE_PRIMARY: 342 case HAST_ROLE_SECONDARY: 343 break; 344 default: 345 pjdlog_error("Invalid role received (%hhu).", role); 346 error = EHAST_INVALID; 347 goto fail; 348 } 349 } 350 if (strcmp(str, "all") == 0) { 351 struct hast_resource *res; 352 353 /* All configured resources. */ 354 355 ii = 0; 356 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) { 357 switch (cmd) { 358 case HASTCTL_CMD_SETROLE: 359 control_set_role_common(cfg, nvout, role, res, 360 res->hr_name, ii++); 361 break; 362 case HASTCTL_CMD_STATUS: 363 control_status(cfg, nvout, res, res->hr_name, 364 ii++); 365 break; 366 default: 367 pjdlog_error("Invalid command received (%hhu).", 368 cmd); 369 error = EHAST_UNIMPLEMENTED; 370 goto fail; 371 } 372 } 373 } else { 374 /* Only selected resources. */ 375 376 for (ii = 0; ; ii++) { 377 str = nv_get_string(nvin, "resource%u", ii); 378 if (str == NULL) 379 break; 380 switch (cmd) { 381 case HASTCTL_CMD_SETROLE: 382 control_set_role_common(cfg, nvout, role, NULL, 383 str, ii); 384 break; 385 case HASTCTL_CMD_STATUS: 386 control_status(cfg, nvout, NULL, str, ii); 387 break; 388 default: 389 pjdlog_error("Invalid command received (%hhu).", 390 cmd); 391 error = EHAST_UNIMPLEMENTED; 392 goto fail; 393 } 394 } 395 } 396 if (nv_error(nvout) != 0) 397 goto close; 398fail: 399 if (error != 0) 400 nv_add_int16(nvout, error, "error"); 401 402 if (hast_proto_send(NULL, conn, nvout, NULL, 0) < 0) 403 pjdlog_errno(LOG_ERR, "Unable to send control response"); 404close: 405 if (nvin != NULL) 406 nv_free(nvin); 407 if (nvout != NULL) 408 nv_free(nvout); 409 proto_close(conn); 410 cfg->hc_controlin = NULL; 411} 412 413/* 414 * Thread handles control requests from the parent. 415 */ 416void * 417ctrl_thread(void *arg) 418{ 419 struct hast_resource *res = arg; 420 struct nv *nvin, *nvout; 421 uint8_t cmd; 422 423 for (;;) { 424 if (hast_proto_recv_hdr(res->hr_ctrl, &nvin) < 0) { 425 if (sigexit_received) 426 pthread_exit(NULL); 427 pjdlog_errno(LOG_ERR, 428 "Unable to receive control message"); 429 kill(getpid(), SIGTERM); 430 pthread_exit(NULL); 431 } 432 cmd = nv_get_uint8(nvin, "cmd"); 433 if (cmd == 0) { 434 pjdlog_error("Control message is missing 'cmd' field."); 435 nv_free(nvin); 436 continue; 437 } 438 nvout = nv_alloc(); 439 switch (cmd) { 440 case CONTROL_STATUS: 441 if (res->hr_remotein != NULL && 442 res->hr_remoteout != NULL) { 443 nv_add_string(nvout, "complete", "status"); 444 } else { 445 nv_add_string(nvout, "degraded", "status"); 446 } 447 nv_add_uint32(nvout, (uint32_t)res->hr_extentsize, 448 "extentsize"); 449 if (res->hr_role == HAST_ROLE_PRIMARY) { 450 nv_add_uint32(nvout, 451 (uint32_t)res->hr_keepdirty, "keepdirty"); 452 nv_add_uint64(nvout, 453 (uint64_t)(activemap_ndirty(res->hr_amp) * 454 res->hr_extentsize), "dirty"); 455 } else { 456 nv_add_uint32(nvout, (uint32_t)0, "keepdirty"); 457 nv_add_uint64(nvout, (uint64_t)0, "dirty"); 458 } 459 nv_add_uint64(nvout, res->hr_stat_read, "stat_read"); 460 nv_add_uint64(nvout, res->hr_stat_write, "stat_write"); 461 nv_add_uint64(nvout, res->hr_stat_delete, 462 "stat_delete"); 463 nv_add_uint64(nvout, res->hr_stat_flush, "stat_flush"); 464 nv_add_uint64(nvout, res->hr_stat_activemap_update, 465 "stat_activemap_update"); 466 nv_add_int16(nvout, 0, "error"); 467 break; 468 case CONTROL_RELOAD: 469 /* 470 * When parent receives SIGHUP and discovers that 471 * something related to us has changes, it sends reload 472 * message to us. 473 */ 474 assert(res->hr_role == HAST_ROLE_PRIMARY); 475 primary_config_reload(res, nvin); 476 nv_add_int16(nvout, 0, "error"); 477 break; 478 default: 479 nv_add_int16(nvout, EINVAL, "error"); 480 break; 481 } 482 nv_free(nvin); 483 if (nv_error(nvout) != 0) { 484 pjdlog_error("Unable to create answer on control message."); 485 nv_free(nvout); 486 continue; 487 } 488 if (hast_proto_send(NULL, res->hr_ctrl, nvout, NULL, 0) < 0) { 489 pjdlog_errno(LOG_ERR, 490 "Unable to send reply to control message"); 491 } 492 nv_free(nvout); 493 } 494 /* NOTREACHED */ 495 return (NULL); 496} 497