sbin/hastd/primary.c

   1 /*-
   2  * Copyright (c) 2009 The FreeBSD Foundation
   3  * Copyright (c) 2010-2011 Pawel Jakub Dawidek <pawel@dawidek.net>
   4  * All rights reserved.
   5  *
   6  * This software was developed by Pawel Jakub Dawidek under sponsorship from
   7  * the FreeBSD Foundation.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
  19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
  22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  28  * SUCH DAMAGE.
  29  */
  30
  31 #include <sys/cdefs.h>
  32 __FBSDID("$FreeBSD$");
  33
  34 #include <sys/types.h>
  35 #include <sys/time.h>
  36 #include <sys/bio.h>
  37 #include <sys/disk.h>
  38 #include <sys/refcount.h>
  39 #include <sys/stat.h>
  40
  41 #include <geom/gate/g_gate.h>
  42
  43 #include <err.h>
  44 #include <errno.h>
  45 #include <fcntl.h>
  46 #include <libgeom.h>
  47 #include <pthread.h>
  48 #include <signal.h>
  49 #include <stdint.h>
  50 #include <stdio.h>
  51 #include <string.h>
  52 #include <sysexits.h>
  53 #include <unistd.h>
  54
  55 #include <activemap.h>
  56 #include <nv.h>
  57 #include <rangelock.h>
  58
  59 #include "control.h"
  60 #include "event.h"
  61 #include "hast.h"
  62 #include "hast_proto.h"
  63 #include "hastd.h"
  64 #include "hooks.h"
  65 #include "metadata.h"
  66 #include "proto.h"
  67 #include "pjdlog.h"
  68 #include "subr.h"
  69 #include "synch.h"
  70
  71 /* The is only one remote component for now. */
  72 #define ISREMOTE(no)    ((no) == 1)
  73
  74 struct hio {
  75         /*
  76          * Number of components we are still waiting for.
  77          * When this field goes to 0, we can send the request back to the
  78          * kernel. Each component has to decrease this counter by one
  79          * even on failure.
  80          */
  81         unsigned int             hio_countdown;
  82         /*
  83          * Each component has a place to store its own error.
  84          * Once the request is handled by all components we can decide if the
  85          * request overall is successful or not.
  86          */
  87         int                     *hio_errors;
  88         /*
  89          * Structure used to communicate with GEOM Gate class.
  90          */
  91         struct g_gate_ctl_io     hio_ggio;
  92         TAILQ_ENTRY(hio)        *hio_next;
  93 };
  94 #define hio_free_next   hio_next[0]
  95 #define hio_done_next   hio_next[0]
  96
  97 /*
  98  * Free list holds unused structures. When free list is empty, we have to wait
  99  * until some in-progress requests are freed.
 100  */
 101 static TAILQ_HEAD(, hio) hio_free_list;
 102 static pthread_mutex_t hio_free_list_lock;
 103 static pthread_cond_t hio_free_list_cond;
 104 /*
 105  * There is one send list for every component. One requests is placed on all
 106  * send lists - each component gets the same request, but each component is
 107  * responsible for managing his own send list.
 108  */
 109 static TAILQ_HEAD(, hio) *hio_send_list;
 110 static pthread_mutex_t *hio_send_list_lock;
 111 static pthread_cond_t *hio_send_list_cond;
 112 /*
 113  * There is one recv list for every component, although local components don't
 114  * use recv lists as local requests are done synchronously.
 115  */
 116 static TAILQ_HEAD(, hio) *hio_recv_list;
 117 static pthread_mutex_t *hio_recv_list_lock;
 118 static pthread_cond_t *hio_recv_list_cond;
 119 /*
 120  * Request is placed on done list by the slowest component (the one that
 121  * decreased hio_countdown from 1 to 0).
 122  */
 123 static TAILQ_HEAD(, hio) hio_done_list;
 124 static pthread_mutex_t hio_done_list_lock;
 125 static pthread_cond_t hio_done_list_cond;
 126 /*
 127  * Structure below are for interaction with sync thread.
 128  */
 129 static bool sync_inprogress;
 130 static pthread_mutex_t sync_lock;
 131 static pthread_cond_t sync_cond;
 132 /*
 133  * The lock below allows to synchornize access to remote connections.
 134  */
 135 static pthread_rwlock_t *hio_remote_lock;
 136
 137 /*
 138  * Lock to synchronize metadata updates. Also synchronize access to
 139  * hr_primary_localcnt and hr_primary_remotecnt fields.
 140  */
 141 static pthread_mutex_t metadata_lock;
 142
 143 /*
 144  * Maximum number of outstanding I/O requests.
 145  */
 146 #define HAST_HIO_MAX    256
 147 /*
 148  * Number of components. At this point there are only two components: local
 149  * and remote, but in the future it might be possible to use multiple local
 150  * and remote components.
 151  */
 152 #define HAST_NCOMPONENTS        2
 153
 154 #define ISCONNECTED(res, no)    \
 155         ((res)->hr_remotein != NULL && (res)->hr_remoteout != NULL)
 156
 157 #define QUEUE_INSERT1(hio, name, ncomp) do {                            \
 158         bool _wakeup;                                                   \
 159                                                                         \
 160         mtx_lock(&hio_##name##_list_lock[(ncomp)]);                     \
 161         _wakeup = TAILQ_EMPTY(&hio_##name##_list[(ncomp)]);             \
 162         TAILQ_INSERT_TAIL(&hio_##name##_list[(ncomp)], (hio),           \
 163             hio_next[(ncomp)]);                                         \
 164         mtx_unlock(&hio_##name##_list_lock[ncomp]);                     \
 165         if (_wakeup)                                                    \
 166                 cv_signal(&hio_##name##_list_cond[(ncomp)]);            \
 167 } while (0)
 168 #define QUEUE_INSERT2(hio, name)        do {                            \
 169         bool _wakeup;                                                   \
 170                                                                         \
 171         mtx_lock(&hio_##name##_list_lock);                              \
 172         _wakeup = TAILQ_EMPTY(&hio_##name##_list);                      \
 173         TAILQ_INSERT_TAIL(&hio_##name##_list, (hio), hio_##name##_next);\
 174         mtx_unlock(&hio_##name##_list_lock);                            \
 175         if (_wakeup)                                                    \
 176                 cv_signal(&hio_##name##_list_cond);                     \
 177 } while (0)
 178 #define QUEUE_TAKE1(hio, name, ncomp, timeout)  do {                    \
 179         bool _last;                                                     \
 180                                                                         \
 181         mtx_lock(&hio_##name##_list_lock[(ncomp)]);                     \
 182         _last = false;                                                  \
 183         while (((hio) = TAILQ_FIRST(&hio_##name##_list[(ncomp)])) == NULL && !_last) { \
 184                 cv_timedwait(&hio_##name##_list_cond[(ncomp)],          \
 185                     &hio_##name##_list_lock[(ncomp)], (timeout));       \
 186                 if ((timeout) != 0)                                     \
 187                         _last = true;                                   \
 188         }                                                               \
 189         if (hio != NULL) {                                              \
 190                 TAILQ_REMOVE(&hio_##name##_list[(ncomp)], (hio),        \
 191                     hio_next[(ncomp)]);                                 \
 192         }                                                               \
 193         mtx_unlock(&hio_##name##_list_lock[(ncomp)]);                   \
 194 } while (0)
 195 #define QUEUE_TAKE2(hio, name)  do {                                    \
 196         mtx_lock(&hio_##name##_list_lock);                              \
 197         while (((hio) = TAILQ_FIRST(&hio_##name##_list)) == NULL) {     \
 198                 cv_wait(&hio_##name##_list_cond,                        \
 199                     &hio_##name##_list_lock);                           \
 200         }                                                               \
 201         TAILQ_REMOVE(&hio_##name##_list, (hio), hio_##name##_next);     \
 202         mtx_unlock(&hio_##name##_list_lock);                            \
 203 } while (0)
 204
 205 #define SYNCREQ(hio)            do {                                    \
 206         (hio)->hio_ggio.gctl_unit = -1;                                 \
 207         (hio)->hio_ggio.gctl_seq = 1;                                   \
 208 } while (0)
 209 #define ISSYNCREQ(hio)          ((hio)->hio_ggio.gctl_unit == -1)
 210 #define SYNCREQDONE(hio)        do { (hio)->hio_ggio.gctl_unit = -2; } while (0)
 211 #define ISSYNCREQDONE(hio)      ((hio)->hio_ggio.gctl_unit == -2)
 212
 213 static struct hast_resource *gres;
 214
 215 static pthread_mutex_t range_lock;
 216 static struct rangelocks *range_regular;
 217 static bool range_regular_wait;
 218 static pthread_cond_t range_regular_cond;
 219 static struct rangelocks *range_sync;
 220 static bool range_sync_wait;
 221 static pthread_cond_t range_sync_cond;
 222 static bool fullystarted;
 223
 224 static void *ggate_recv_thread(void *arg);
 225 static void *local_send_thread(void *arg);
 226 static void *remote_send_thread(void *arg);
 227 static void *remote_recv_thread(void *arg);
 228 static void *ggate_send_thread(void *arg);
 229 static void *sync_thread(void *arg);
 230 static void *guard_thread(void *arg);
 231
 232 static void
 233 cleanup(struct hast_resource *res)
 234 {
 235         int rerrno;
 236
 237         /* Remember errno. */
 238         rerrno = errno;
 239
 240         /* Destroy ggate provider if we created one. */
 241         if (res->hr_ggateunit >= 0) {
 242                 struct g_gate_ctl_destroy ggiod;
 243
 244                 bzero(&ggiod, sizeof(ggiod));
 245                 ggiod.gctl_version = G_GATE_VERSION;
 246                 ggiod.gctl_unit = res->hr_ggateunit;
 247                 ggiod.gctl_force = 1;
 248                 if (ioctl(res->hr_ggatefd, G_GATE_CMD_DESTROY, &ggiod) < 0) {
 249                         pjdlog_errno(LOG_WARNING,
 250                             "Unable to destroy hast/%s device",
 251                             res->hr_provname);
 252                 }
 253                 res->hr_ggateunit = -1;
 254         }
 255
 256         /* Restore errno. */
 257         errno = rerrno;
 258 }
 259
 260 static __dead2 void
 261 primary_exit(int exitcode, const char *fmt, ...)
 262 {
 263         va_list ap;
 264
 265         PJDLOG_ASSERT(exitcode != EX_OK);
 266         va_start(ap, fmt);
 267         pjdlogv_errno(LOG_ERR, fmt, ap);
 268         va_end(ap);
 269         cleanup(gres);
 270         exit(exitcode);
 271 }
 272
 273 static __dead2 void
 274 primary_exitx(int exitcode, const char *fmt, ...)
 275 {
 276         va_list ap;
 277
 278         va_start(ap, fmt);
 279         pjdlogv(exitcode == EX_OK ? LOG_INFO : LOG_ERR, fmt, ap);
 280         va_end(ap);
 281         cleanup(gres);
 282         exit(exitcode);
 283 }
 284
 285 static int
 286 hast_activemap_flush(struct hast_resource *res)
 287 {
 288         const unsigned char *buf;
 289         size_t size;
 290
 291         buf = activemap_bitmap(res->hr_amp, &size);
 292         PJDLOG_ASSERT(buf != NULL);
 293         PJDLOG_ASSERT((size % res->hr_local_sectorsize) == 0);
 294         if (pwrite(res->hr_localfd, buf, size, METADATA_SIZE) !=
 295             (ssize_t)size) {
 296                 KEEP_ERRNO(pjdlog_errno(LOG_ERR,
 297                     "Unable to flush activemap to disk"));
 298                 return (-1);
 299         }
 300         return (0);
 301 }
 302
 303 static bool
 304 real_remote(const struct hast_resource *res)
 305 {
 306
 307         return (strcmp(res->hr_remoteaddr, "none") != 0);
 308 }
 309
 310 static void
 311 init_environment(struct hast_resource *res __unused)
 312 {
 313         struct hio *hio;
 314         unsigned int ii, ncomps;
 315
 316         /*
 317          * In the future it might be per-resource value.
 318          */
 319         ncomps = HAST_NCOMPONENTS;
 320
 321         /*
 322          * Allocate memory needed by lists.
 323          */
 324         hio_send_list = malloc(sizeof(hio_send_list[0]) * ncomps);
 325         if (hio_send_list == NULL) {
 326                 primary_exitx(EX_TEMPFAIL,
 327                     "Unable to allocate %zu bytes of memory for send lists.",
 328                     sizeof(hio_send_list[0]) * ncomps);
 329         }
 330         hio_send_list_lock = malloc(sizeof(hio_send_list_lock[0]) * ncomps);
 331         if (hio_send_list_lock == NULL) {
 332                 primary_exitx(EX_TEMPFAIL,
 333                     "Unable to allocate %zu bytes of memory for send list locks.",
 334                     sizeof(hio_send_list_lock[0]) * ncomps);
 335         }
 336         hio_send_list_cond = malloc(sizeof(hio_send_list_cond[0]) * ncomps);
 337         if (hio_send_list_cond == NULL) {
 338                 primary_exitx(EX_TEMPFAIL,
 339                     "Unable to allocate %zu bytes of memory for send list condition variables.",
 340                     sizeof(hio_send_list_cond[0]) * ncomps);
 341         }
 342         hio_recv_list = malloc(sizeof(hio_recv_list[0]) * ncomps);
 343         if (hio_recv_list == NULL) {
 344                 primary_exitx(EX_TEMPFAIL,
 345                     "Unable to allocate %zu bytes of memory for recv lists.",
 346                     sizeof(hio_recv_list[0]) * ncomps);
 347         }
 348         hio_recv_list_lock = malloc(sizeof(hio_recv_list_lock[0]) * ncomps);
 349         if (hio_recv_list_lock == NULL) {
 350                 primary_exitx(EX_TEMPFAIL,
 351                     "Unable to allocate %zu bytes of memory for recv list locks.",
 352                     sizeof(hio_recv_list_lock[0]) * ncomps);
 353         }
 354         hio_recv_list_cond = malloc(sizeof(hio_recv_list_cond[0]) * ncomps);
 355         if (hio_recv_list_cond == NULL) {
 356                 primary_exitx(EX_TEMPFAIL,
 357                     "Unable to allocate %zu bytes of memory for recv list condition variables.",
 358                     sizeof(hio_recv_list_cond[0]) * ncomps);
 359         }
 360         hio_remote_lock = malloc(sizeof(hio_remote_lock[0]) * ncomps);
 361         if (hio_remote_lock == NULL) {
 362                 primary_exitx(EX_TEMPFAIL,
 363                     "Unable to allocate %zu bytes of memory for remote connections locks.",
 364                     sizeof(hio_remote_lock[0]) * ncomps);
 365         }
 366
 367         /*
 368          * Initialize lists, their locks and theirs condition variables.
 369          */
 370         TAILQ_INIT(&hio_free_list);
 371         mtx_init(&hio_free_list_lock);
 372         cv_init(&hio_free_list_cond);
 373         for (ii = 0; ii < HAST_NCOMPONENTS; ii++) {
 374                 TAILQ_INIT(&hio_send_list[ii]);
 375                 mtx_init(&hio_send_list_lock[ii]);
 376                 cv_init(&hio_send_list_cond[ii]);
 377                 TAILQ_INIT(&hio_recv_list[ii]);
 378                 mtx_init(&hio_recv_list_lock[ii]);
 379                 cv_init(&hio_recv_list_cond[ii]);
 380                 rw_init(&hio_remote_lock[ii]);
 381         }
 382         TAILQ_INIT(&hio_done_list);
 383         mtx_init(&hio_done_list_lock);
 384         cv_init(&hio_done_list_cond);
 385         mtx_init(&metadata_lock);
 386
 387         /*
 388          * Allocate requests pool and initialize requests.
 389          */
 390         for (ii = 0; ii < HAST_HIO_MAX; ii++) {
 391                 hio = malloc(sizeof(*hio));
 392                 if (hio == NULL) {
 393                         primary_exitx(EX_TEMPFAIL,
 394                             "Unable to allocate %zu bytes of memory for hio request.",
 395                             sizeof(*hio));
 396                 }
 397                 hio->hio_countdown = 0;
 398                 hio->hio_errors = malloc(sizeof(hio->hio_errors[0]) * ncomps);
 399                 if (hio->hio_errors == NULL) {
 400                         primary_exitx(EX_TEMPFAIL,
 401                             "Unable allocate %zu bytes of memory for hio errors.",
 402                             sizeof(hio->hio_errors[0]) * ncomps);
 403                 }
 404                 hio->hio_next = malloc(sizeof(hio->hio_next[0]) * ncomps);
 405                 if (hio->hio_next == NULL) {
 406                         primary_exitx(EX_TEMPFAIL,
 407                             "Unable allocate %zu bytes of memory for hio_next field.",
 408                             sizeof(hio->hio_next[0]) * ncomps);
 409                 }
 410                 hio->hio_ggio.gctl_version = G_GATE_VERSION;
 411                 hio->hio_ggio.gctl_data = malloc(MAXPHYS);
 412                 if (hio->hio_ggio.gctl_data == NULL) {
 413                         primary_exitx(EX_TEMPFAIL,
 414                             "Unable to allocate %zu bytes of memory for gctl_data.",
 415                             MAXPHYS);
 416                 }
 417                 hio->hio_ggio.gctl_length = MAXPHYS;
 418                 hio->hio_ggio.gctl_error = 0;
 419                 TAILQ_INSERT_HEAD(&hio_free_list, hio, hio_free_next);
 420         }
 421 }
 422
 423 static bool
 424 init_resuid(struct hast_resource *res)
 425 {
 426
 427         mtx_lock(&metadata_lock);
 428         if (res->hr_resuid != 0) {
 429                 mtx_unlock(&metadata_lock);
 430                 return (false);
 431         } else {
 432                 /* Initialize unique resource identifier. */
 433                 arc4random_buf(&res->hr_resuid, sizeof(res->hr_resuid));
 434                 mtx_unlock(&metadata_lock);
 435                 if (metadata_write(res) < 0)
 436                         exit(EX_NOINPUT);
 437                 return (true);
 438         }
 439 }
 440
 441 static void
 442 init_local(struct hast_resource *res)
 443 {
 444         unsigned char *buf;
 445         size_t mapsize;
 446
 447         if (metadata_read(res, true) < 0)
 448                 exit(EX_NOINPUT);
 449         mtx_init(&res->hr_amp_lock);
 450         if (activemap_init(&res->hr_amp, res->hr_datasize, res->hr_extentsize,
 451             res->hr_local_sectorsize, res->hr_keepdirty) < 0) {
 452                 primary_exit(EX_TEMPFAIL, "Unable to create activemap");
 453         }
 454         mtx_init(&range_lock);
 455         cv_init(&range_regular_cond);
 456         if (rangelock_init(&range_regular) < 0)
 457                 primary_exit(EX_TEMPFAIL, "Unable to create regular range lock");
 458         cv_init(&range_sync_cond);
 459         if (rangelock_init(&range_sync) < 0)
 460                 primary_exit(EX_TEMPFAIL, "Unable to create sync range lock");
 461         mapsize = activemap_ondisk_size(res->hr_amp);
 462         buf = calloc(1, mapsize);
 463         if (buf == NULL) {
 464                 primary_exitx(EX_TEMPFAIL,
 465                     "Unable to allocate buffer for activemap.");
 466         }
 467         if (pread(res->hr_localfd, buf, mapsize, METADATA_SIZE) !=
 468             (ssize_t)mapsize) {
 469                 primary_exit(EX_NOINPUT, "Unable to read activemap");
 470         }
 471         activemap_copyin(res->hr_amp, buf, mapsize);
 472         free(buf);
 473         if (res->hr_resuid != 0)
 474                 return;
 475         /*
 476          * We're using provider for the first time. Initialize local and remote
 477          * counters. We don't initialize resuid here, as we want to do it just
 478          * in time. The reason for this is that we want to inform secondary
 479          * that there were no writes yet, so there is no need to synchronize
 480          * anything.
 481          */
 482         res->hr_primary_localcnt = 0;
 483         res->hr_primary_remotecnt = 0;
 484         if (metadata_write(res) < 0)
 485                 exit(EX_NOINPUT);
 486 }
 487
 488 static int
 489 primary_connect(struct hast_resource *res, struct proto_conn **connp)
 490 {
 491         struct proto_conn *conn;
 492         int16_t val;
 493
 494         val = 1;
 495         if (proto_send(res->hr_conn, &val, sizeof(val)) < 0) {
 496                 primary_exit(EX_TEMPFAIL,
 497                     "Unable to send connection request to parent");
 498         }
 499         if (proto_recv(res->hr_conn, &val, sizeof(val)) < 0) {
 500                 primary_exit(EX_TEMPFAIL,
 501                     "Unable to receive reply to connection request from parent");
 502         }
 503         if (val != 0) {
 504                 errno = val;
 505                 pjdlog_errno(LOG_WARNING, "Unable to connect to %s",
 506                     res->hr_remoteaddr);
 507                 return (-1);
 508         }
 509         if (proto_connection_recv(res->hr_conn, true, &conn) < 0) {
 510                 primary_exit(EX_TEMPFAIL,
 511                     "Unable to receive connection from parent");
 512         }
 513         if (proto_connect_wait(conn, res->hr_timeout) < 0) {
 514                 pjdlog_errno(LOG_WARNING, "Unable to connect to %s",
 515                     res->hr_remoteaddr);
 516                 proto_close(conn);
 517                 return (-1);
 518         }
 519         /* Error in setting timeout is not critical, but why should it fail? */
 520         if (proto_timeout(conn, res->hr_timeout) < 0)
 521                 pjdlog_errno(LOG_WARNING, "Unable to set connection timeout");
 522
 523         *connp = conn;
 524
 525         return (0);
 526 }
 527
 528 static int
 529 init_remote(struct hast_resource *res, struct proto_conn **inp,
 530     struct proto_conn **outp)
 531 {
 532         struct proto_conn *in, *out;
 533         struct nv *nvout, *nvin;
 534         const unsigned char *token;
 535         unsigned char *map;
 536         const char *errmsg;
 537         int32_t extentsize;
 538         int64_t datasize;
 539         uint32_t mapsize;
 540         size_t size;
 541         int error;
 542
 543         PJDLOG_ASSERT((inp == NULL && outp == NULL) || (inp != NULL && outp != NULL));
 544         PJDLOG_ASSERT(real_remote(res));
 545
 546         in = out = NULL;
 547         errmsg = NULL;
 548
 549         if (primary_connect(res, &out) == -1)
 550                 return (ECONNREFUSED);
 551
 552         error = ECONNABORTED;
 553
 554         /*
 555          * First handshake step.
 556          * Setup outgoing connection with remote node.
 557          */
 558         nvout = nv_alloc();
 559         nv_add_string(nvout, res->hr_name, "resource");
 560         if (nv_error(nvout) != 0) {
 561                 pjdlog_common(LOG_WARNING, 0, nv_error(nvout),
 562                     "Unable to allocate header for connection with %s",
 563                     res->hr_remoteaddr);
 564                 nv_free(nvout);
 565                 goto close;
 566         }
 567         if (hast_proto_send(res, out, nvout, NULL, 0) < 0) {
 568                 pjdlog_errno(LOG_WARNING,
 569                     "Unable to send handshake header to %s",
 570                     res->hr_remoteaddr);
 571                 nv_free(nvout);
 572                 goto close;
 573         }
 574         nv_free(nvout);
 575         if (hast_proto_recv_hdr(out, &nvin) < 0) {
 576                 pjdlog_errno(LOG_WARNING,
 577                     "Unable to receive handshake header from %s",
 578                     res->hr_remoteaddr);
 579                 goto close;
 580         }
 581         errmsg = nv_get_string(nvin, "errmsg");
 582         if (errmsg != NULL) {
 583                 pjdlog_warning("%s", errmsg);
 584                 if (nv_exists(nvin, "wait"))
 585                         error = EBUSY;
 586                 nv_free(nvin);
 587                 goto close;
 588         }
 589         token = nv_get_uint8_array(nvin, &size, "token");
 590         if (token == NULL) {
 591                 pjdlog_warning("Handshake header from %s has no 'token' field.",
 592                     res->hr_remoteaddr);
 593                 nv_free(nvin);
 594                 goto close;
 595         }
 596         if (size != sizeof(res->hr_token)) {
 597                 pjdlog_warning("Handshake header from %s contains 'token' of wrong size (got %zu, expected %zu).",
 598                     res->hr_remoteaddr, size, sizeof(res->hr_token));
 599                 nv_free(nvin);
 600                 goto close;
 601         }
 602         bcopy(token, res->hr_token, sizeof(res->hr_token));
 603         nv_free(nvin);
 604
 605         /*
 606          * Second handshake step.
 607          * Setup incoming connection with remote node.
 608          */
 609         if (primary_connect(res, &in) == -1)
 610                 goto close;
 611
 612         nvout = nv_alloc();
 613         nv_add_string(nvout, res->hr_name, "resource");
 614         nv_add_uint8_array(nvout, res->hr_token, sizeof(res->hr_token),
 615             "token");
 616         if (res->hr_resuid == 0) {
 617                 /*
 618                  * The resuid field was not yet initialized.
 619                  * Because we do synchronization inside init_resuid(), it is
 620                  * possible that someone already initialized it, the function
 621                  * will return false then, but if we successfully initialized
 622                  * it, we will get true. True means that there were no writes
 623                  * to this resource yet and we want to inform secondary that
 624                  * synchronization is not needed by sending "virgin" argument.
 625                  */
 626                 if (init_resuid(res))
 627                         nv_add_int8(nvout, 1, "virgin");
 628         }
 629         nv_add_uint64(nvout, res->hr_resuid, "resuid");
 630         nv_add_uint64(nvout, res->hr_primary_localcnt, "localcnt");
 631         nv_add_uint64(nvout, res->hr_primary_remotecnt, "remotecnt");
 632         if (nv_error(nvout) != 0) {
 633                 pjdlog_common(LOG_WARNING, 0, nv_error(nvout),
 634                     "Unable to allocate header for connection with %s",
 635                     res->hr_remoteaddr);
 636                 nv_free(nvout);
 637                 goto close;
 638         }
 639         if (hast_proto_send(res, in, nvout, NULL, 0) < 0) {
 640                 pjdlog_errno(LOG_WARNING,
 641                     "Unable to send handshake header to %s",
 642                     res->hr_remoteaddr);
 643                 nv_free(nvout);
 644                 goto close;
 645         }
 646         nv_free(nvout);
 647         if (hast_proto_recv_hdr(out, &nvin) < 0) {
 648                 pjdlog_errno(LOG_WARNING,
 649                     "Unable to receive handshake header from %s",
 650                     res->hr_remoteaddr);
 651                 goto close;
 652         }
 653         errmsg = nv_get_string(nvin, "errmsg");
 654         if (errmsg != NULL) {
 655                 pjdlog_warning("%s", errmsg);
 656                 nv_free(nvin);
 657                 goto close;
 658         }
 659         datasize = nv_get_int64(nvin, "datasize");
 660         if (datasize != res->hr_datasize) {
 661                 pjdlog_warning("Data size differs between nodes (local=%jd, remote=%jd).",
 662                     (intmax_t)res->hr_datasize, (intmax_t)datasize);
 663                 nv_free(nvin);
 664                 goto close;
 665         }
 666         extentsize = nv_get_int32(nvin, "extentsize");
 667         if (extentsize != res->hr_extentsize) {
 668                 pjdlog_warning("Extent size differs between nodes (local=%zd, remote=%zd).",
 669                     (ssize_t)res->hr_extentsize, (ssize_t)extentsize);
 670                 nv_free(nvin);
 671                 goto close;
 672         }
 673         res->hr_secondary_localcnt = nv_get_uint64(nvin, "localcnt");
 674         res->hr_secondary_remotecnt = nv_get_uint64(nvin, "remotecnt");
 675         res->hr_syncsrc = nv_get_uint8(nvin, "syncsrc");
 676         if (nv_exists(nvin, "virgin")) {
 677                 /*
 678                  * Secondary was reinitialized, bump localcnt if it is 0 as
 679                  * only we have the data.
 680                  */
 681                 PJDLOG_ASSERT(res->hr_syncsrc == HAST_SYNCSRC_PRIMARY);
 682                 PJDLOG_ASSERT(res->hr_secondary_localcnt == 0);
 683
 684                 if (res->hr_primary_localcnt == 0) {
 685                         PJDLOG_ASSERT(res->hr_secondary_remotecnt == 0);
 686
 687                         mtx_lock(&metadata_lock);
 688                         res->hr_primary_localcnt++;
 689                         pjdlog_debug(1, "Increasing localcnt to %ju.",
 690                             (uintmax_t)res->hr_primary_localcnt);
 691                         (void)metadata_write(res);
 692                         mtx_unlock(&metadata_lock);
 693                 }
 694         }
 695         map = NULL;
 696         mapsize = nv_get_uint32(nvin, "mapsize");
 697         if (mapsize > 0) {
 698                 map = malloc(mapsize);
 699                 if (map == NULL) {
 700                         pjdlog_error("Unable to allocate memory for remote activemap (mapsize=%ju).",
 701                             (uintmax_t)mapsize);
 702                         nv_free(nvin);
 703                         goto close;
 704                 }
 705                 /*
 706                  * Remote node have some dirty extents on its own, lets
 707                  * download its activemap.
 708                  */
 709                 if (hast_proto_recv_data(res, out, nvin, map,
 710                     mapsize) < 0) {
 711                         pjdlog_errno(LOG_ERR,
 712                             "Unable to receive remote activemap");
 713                         nv_free(nvin);
 714                         free(map);
 715                         goto close;
 716                 }
 717                 /*
 718                  * Merge local and remote bitmaps.
 719                  */
 720                 activemap_merge(res->hr_amp, map, mapsize);
 721                 free(map);
 722                 /*
 723                  * Now that we merged bitmaps from both nodes, flush it to the
 724                  * disk before we start to synchronize.
 725                  */
 726                 (void)hast_activemap_flush(res);
 727         }
 728         nv_free(nvin);
 729         /* Setup directions. */
 730         if (proto_send(out, NULL, 0) == -1)
 731                 pjdlog_errno(LOG_WARNING, "Unable to set connection direction");
 732         if (proto_recv(in, NULL, 0) == -1)
 733                 pjdlog_errno(LOG_WARNING, "Unable to set connection direction");
 734         pjdlog_info("Connected to %s.", res->hr_remoteaddr);
 735         if (inp != NULL && outp != NULL) {
 736                 *inp = in;
 737                 *outp = out;
 738         } else {
 739                 res->hr_remotein = in;
 740                 res->hr_remoteout = out;
 741         }
 742         event_send(res, EVENT_CONNECT);
 743         return (0);
 744 close:
 745         if (errmsg != NULL && strcmp(errmsg, "Split-brain condition!") == 0)
 746                 event_send(res, EVENT_SPLITBRAIN);
 747         proto_close(out);
 748         if (in != NULL)
 749                 proto_close(in);
 750         return (error);
 751 }
 752
 753 static void
 754 sync_start(void)
 755 {
 756
 757         mtx_lock(&sync_lock);
 758         sync_inprogress = true;
 759         mtx_unlock(&sync_lock);
 760         cv_signal(&sync_cond);
 761 }
 762
 763 static void
 764 sync_stop(void)
 765 {
 766
 767         mtx_lock(&sync_lock);
 768         if (sync_inprogress)
 769                 sync_inprogress = false;
 770         mtx_unlock(&sync_lock);
 771 }
 772
 773 static void
 774 init_ggate(struct hast_resource *res)
 775 {
 776         struct g_gate_ctl_create ggiocreate;
 777         struct g_gate_ctl_cancel ggiocancel;
 778
 779         /*
 780          * We communicate with ggate via /dev/ggctl. Open it.
 781          */
 782         res->hr_ggatefd = open("/dev/" G_GATE_CTL_NAME, O_RDWR);
 783         if (res->hr_ggatefd < 0)
 784                 primary_exit(EX_OSFILE, "Unable to open /dev/" G_GATE_CTL_NAME);
 785         /*
 786          * Create provider before trying to connect, as connection failure
 787          * is not critical, but may take some time.
 788          */
 789         bzero(&ggiocreate, sizeof(ggiocreate));
 790         ggiocreate.gctl_version = G_GATE_VERSION;
 791         ggiocreate.gctl_mediasize = res->hr_datasize;
 792         ggiocreate.gctl_sectorsize = res->hr_local_sectorsize;
 793         ggiocreate.gctl_flags = 0;
 794         ggiocreate.gctl_maxcount = 0;
 795         ggiocreate.gctl_timeout = 0;
 796         ggiocreate.gctl_unit = G_GATE_NAME_GIVEN;
 797         snprintf(ggiocreate.gctl_name, sizeof(ggiocreate.gctl_name), "hast/%s",
 798             res->hr_provname);
 799         if (ioctl(res->hr_ggatefd, G_GATE_CMD_CREATE, &ggiocreate) == 0) {
 800                 pjdlog_info("Device hast/%s created.", res->hr_provname);
 801                 res->hr_ggateunit = ggiocreate.gctl_unit;
 802                 return;
 803         }
 804         if (errno != EEXIST) {
 805                 primary_exit(EX_OSERR, "Unable to create hast/%s device",
 806                     res->hr_provname);
 807         }
 808         pjdlog_debug(1,
 809             "Device hast/%s already exists, we will try to take it over.",
 810             res->hr_provname);
 811         /*
 812          * If we received EEXIST, we assume that the process who created the
 813          * provider died and didn't clean up. In that case we will start from
 814          * where he left of.
 815          */
 816         bzero(&ggiocancel, sizeof(ggiocancel));
 817         ggiocancel.gctl_version = G_GATE_VERSION;
 818         ggiocancel.gctl_unit = G_GATE_NAME_GIVEN;
 819         snprintf(ggiocancel.gctl_name, sizeof(ggiocancel.gctl_name), "hast/%s",
 820             res->hr_provname);
 821         if (ioctl(res->hr_ggatefd, G_GATE_CMD_CANCEL, &ggiocancel) == 0) {
 822                 pjdlog_info("Device hast/%s recovered.", res->hr_provname);
 823                 res->hr_ggateunit = ggiocancel.gctl_unit;
 824                 return;
 825         }
 826         primary_exit(EX_OSERR, "Unable to take over hast/%s device",
 827             res->hr_provname);
 828 }
 829
 830 void
 831 hastd_primary(struct hast_resource *res)
 832 {
 833         pthread_t td;
 834         pid_t pid;
 835         int error, mode, debuglevel;
 836
 837         /*
 838          * Create communication channel for sending control commands from
 839          * parent to child.
 840          */
 841         if (proto_client(NULL, "socketpair://", &res->hr_ctrl) < 0) {
 842                 /* TODO: There's no need for this to be fatal error. */
 843                 KEEP_ERRNO((void)pidfile_remove(pfh));
 844                 pjdlog_exit(EX_OSERR,
 845                     "Unable to create control sockets between parent and child");
 846         }
 847         /*
 848          * Create communication channel for sending events from child to parent.
 849          */
 850         if (proto_client(NULL, "socketpair://", &res->hr_event) < 0) {
 851                 /* TODO: There's no need for this to be fatal error. */
 852                 KEEP_ERRNO((void)pidfile_remove(pfh));
 853                 pjdlog_exit(EX_OSERR,
 854                     "Unable to create event sockets between child and parent");
 855         }
 856         /*
 857          * Create communication channel for sending connection requests from
 858          * child to parent.
 859          */
 860         if (proto_client(NULL, "socketpair://", &res->hr_conn) < 0) {
 861                 /* TODO: There's no need for this to be fatal error. */
 862                 KEEP_ERRNO((void)pidfile_remove(pfh));
 863                 pjdlog_exit(EX_OSERR,
 864                     "Unable to create connection sockets between child and parent");
 865         }
 866
 867         pid = fork();
 868         if (pid < 0) {
 869                 /* TODO: There's no need for this to be fatal error. */
 870                 KEEP_ERRNO((void)pidfile_remove(pfh));
 871                 pjdlog_exit(EX_TEMPFAIL, "Unable to fork");
 872         }
 873
 874         if (pid > 0) {
 875                 /* This is parent. */
 876                 /* Declare that we are receiver. */
 877                 proto_recv(res->hr_event, NULL, 0);
 878                 proto_recv(res->hr_conn, NULL, 0);
 879                 /* Declare that we are sender. */
 880                 proto_send(res->hr_ctrl, NULL, 0);
 881                 res->hr_workerpid = pid;
 882                 return;
 883         }
 884
 885         gres = res;
 886         mode = pjdlog_mode_get();
 887         debuglevel = pjdlog_debug_get();
 888
 889         /* Declare that we are sender. */
 890         proto_send(res->hr_event, NULL, 0);
 891         proto_send(res->hr_conn, NULL, 0);
 892         /* Declare that we are receiver. */
 893         proto_recv(res->hr_ctrl, NULL, 0);
 894         descriptors_cleanup(res);
 895
 896         descriptors_assert(res, mode);
 897
 898         pjdlog_init(mode);
 899         pjdlog_debug_set(debuglevel);
 900         pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
 901         setproctitle("%s (%s)", res->hr_name, role2str(res->hr_role));
 902
 903         init_local(res);
 904         init_ggate(res);
 905         init_environment(res);
 906
 907         if (drop_privs(res) != 0) {
 908                 cleanup(res);
 909                 exit(EX_CONFIG);
 910         }
 911         pjdlog_info("Privileges successfully dropped.");
 912
 913         /*
 914          * Create the guard thread first, so we can handle signals from the
 915          * very begining.
 916          */
 917         error = pthread_create(&td, NULL, guard_thread, res);
 918         PJDLOG_ASSERT(error == 0);
 919         /*
 920          * Create the control thread before sending any event to the parent,
 921          * as we can deadlock when parent sends control request to worker,
 922          * but worker has no control thread started yet, so parent waits.
 923          * In the meantime worker sends an event to the parent, but parent
 924          * is unable to handle the event, because it waits for control
 925          * request response.
 926          */
 927         error = pthread_create(&td, NULL, ctrl_thread, res);
 928         PJDLOG_ASSERT(error == 0);
 929         if (real_remote(res)) {
 930                 error = init_remote(res, NULL, NULL);
 931                 if (error == 0) {
 932                         sync_start();
 933                 } else if (error == EBUSY) {
 934                         time_t start = time(NULL);
 935
 936                         pjdlog_warning("Waiting for remote node to become %s for %ds.",
 937                             role2str(HAST_ROLE_SECONDARY),
 938                             res->hr_timeout);
 939                         for (;;) {
 940                                 sleep(1);
 941                                 error = init_remote(res, NULL, NULL);
 942                                 if (error != EBUSY)
 943                                         break;
 944                                 if (time(NULL) > start + res->hr_timeout)
 945                                         break;
 946                         }
 947                         if (error == EBUSY) {
 948                                 pjdlog_warning("Remote node is still %s, starting anyway.",
 949                                     role2str(HAST_ROLE_PRIMARY));
 950                         }
 951                 }
 952         }
 953         error = pthread_create(&td, NULL, ggate_recv_thread, res);
 954         PJDLOG_ASSERT(error == 0);
 955         error = pthread_create(&td, NULL, local_send_thread, res);
 956         PJDLOG_ASSERT(error == 0);
 957         error = pthread_create(&td, NULL, remote_send_thread, res);
 958         PJDLOG_ASSERT(error == 0);
 959         error = pthread_create(&td, NULL, remote_recv_thread, res);
 960         PJDLOG_ASSERT(error == 0);
 961         error = pthread_create(&td, NULL, ggate_send_thread, res);
 962         PJDLOG_ASSERT(error == 0);
 963         fullystarted = true;
 964         (void)sync_thread(res);
 965 }
 966
 967 static void
 968 reqlog(int loglevel, int debuglevel, struct g_gate_ctl_io *ggio, const char *fmt, ...)
 969 {
 970         char msg[1024];
 971         va_list ap;
 972         int len;
 973
 974         va_start(ap, fmt);
 975         len = vsnprintf(msg, sizeof(msg), fmt, ap);
 976         va_end(ap);
 977         if ((size_t)len < sizeof(msg)) {
 978                 switch (ggio->gctl_cmd) {
 979                 case BIO_READ:
 980                         (void)snprintf(msg + len, sizeof(msg) - len,
 981                             "READ(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
 982                             (uintmax_t)ggio->gctl_length);
 983                         break;
 984                 case BIO_DELETE:
 985                         (void)snprintf(msg + len, sizeof(msg) - len,
 986                             "DELETE(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
 987                             (uintmax_t)ggio->gctl_length);
 988                         break;
 989                 case BIO_FLUSH:
 990                         (void)snprintf(msg + len, sizeof(msg) - len, "FLUSH.");
 991                         break;
 992                 case BIO_WRITE:
 993                         (void)snprintf(msg + len, sizeof(msg) - len,
 994                             "WRITE(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
 995                             (uintmax_t)ggio->gctl_length);
 996                         break;
 997                 default:
 998                         (void)snprintf(msg + len, sizeof(msg) - len,
 999                             "UNKNOWN(%u).", (unsigned int)ggio->gctl_cmd);
1000                         break;
1001                 }
1002         }
1003         pjdlog_common(loglevel, debuglevel, -1, "%s", msg);
1004 }
1005
1006 static void
1007 remote_close(struct hast_resource *res, int ncomp)
1008 {
1009
1010         rw_wlock(&hio_remote_lock[ncomp]);
1011         /*
1012          * A race is possible between dropping rlock and acquiring wlock -
1013          * another thread can close connection in-between.
1014          */
1015         if (!ISCONNECTED(res, ncomp)) {
1016                 PJDLOG_ASSERT(res->hr_remotein == NULL);
1017                 PJDLOG_ASSERT(res->hr_remoteout == NULL);
1018                 rw_unlock(&hio_remote_lock[ncomp]);
1019                 return;
1020         }
1021
1022         PJDLOG_ASSERT(res->hr_remotein != NULL);
1023         PJDLOG_ASSERT(res->hr_remoteout != NULL);
1024
1025         pjdlog_debug(2, "Closing incoming connection to %s.",
1026             res->hr_remoteaddr);
1027         proto_close(res->hr_remotein);
1028         res->hr_remotein = NULL;
1029         pjdlog_debug(2, "Closing outgoing connection to %s.",
1030             res->hr_remoteaddr);
1031         proto_close(res->hr_remoteout);
1032         res->hr_remoteout = NULL;
1033
1034         rw_unlock(&hio_remote_lock[ncomp]);
1035
1036         pjdlog_warning("Disconnected from %s.", res->hr_remoteaddr);
1037
1038         /*
1039          * Stop synchronization if in-progress.
1040          */
1041         sync_stop();
1042
1043         event_send(res, EVENT_DISCONNECT);
1044 }
1045
1046 /*
1047  * Thread receives ggate I/O requests from the kernel and passes them to
1048  * appropriate threads:
1049  * WRITE - always goes to both local_send and remote_send threads
1050  * READ (when the block is up-to-date on local component) -
1051  *      only local_send thread
1052  * READ (when the block isn't up-to-date on local component) -
1053  *      only remote_send thread
1054  * DELETE - always goes to both local_send and remote_send threads
1055  * FLUSH - always goes to both local_send and remote_send threads
1056  */
1057 static void *
1058 ggate_recv_thread(void *arg)
1059 {
1060         struct hast_resource *res = arg;
1061         struct g_gate_ctl_io *ggio;
1062         struct hio *hio;
1063         unsigned int ii, ncomp, ncomps;
1064         int error;
1065
1066         ncomps = HAST_NCOMPONENTS;
1067
1068         for (;;) {
1069                 pjdlog_debug(2, "ggate_recv: Taking free request.");
1070                 QUEUE_TAKE2(hio, free);
1071                 pjdlog_debug(2, "ggate_recv: (%p) Got free request.", hio);
1072                 ggio = &hio->hio_ggio;
1073                 ggio->gctl_unit = res->hr_ggateunit;
1074                 ggio->gctl_length = MAXPHYS;
1075                 ggio->gctl_error = 0;
1076                 pjdlog_debug(2,
1077                     "ggate_recv: (%p) Waiting for request from the kernel.",
1078                     hio);
1079                 if (ioctl(res->hr_ggatefd, G_GATE_CMD_START, ggio) < 0) {
1080                         if (sigexit_received)
1081                                 pthread_exit(NULL);
1082                         primary_exit(EX_OSERR, "G_GATE_CMD_START failed");
1083                 }
1084                 error = ggio->gctl_error;
1085                 switch (error) {
1086                 case 0:
1087                         break;
1088                 case ECANCELED:
1089                         /* Exit gracefully. */
1090                         if (!sigexit_received) {
1091                                 pjdlog_debug(2,
1092                                     "ggate_recv: (%p) Received cancel from the kernel.",
1093                                     hio);
1094                                 pjdlog_info("Received cancel from the kernel, exiting.");
1095                         }
1096                         pthread_exit(NULL);
1097                 case ENOMEM:
1098                         /*
1099                          * Buffer too small? Impossible, we allocate MAXPHYS
1100                          * bytes - request can't be bigger than that.
1101                          */
1102                         /* FALLTHROUGH */
1103                 case ENXIO:
1104                 default:
1105                         primary_exitx(EX_OSERR, "G_GATE_CMD_START failed: %s.",
1106                             strerror(error));
1107                 }
1108                 for (ii = 0; ii < ncomps; ii++)
1109                         hio->hio_errors[ii] = EINVAL;
1110                 reqlog(LOG_DEBUG, 2, ggio,
1111                     "ggate_recv: (%p) Request received from the kernel: ",
1112                     hio);
1113                 /*
1114                  * Inform all components about new write request.
1115                  * For read request prefer local component unless the given
1116                  * range is out-of-date, then use remote component.
1117                  */
1118                 switch (ggio->gctl_cmd) {
1119                 case BIO_READ:
1120                         res->hr_stat_read++;
1121                         pjdlog_debug(2,
1122                             "ggate_recv: (%p) Moving request to the send queue.",
1123                             hio);
1124                         refcount_init(&hio->hio_countdown, 1);
1125                         mtx_lock(&metadata_lock);
1126                         if (res->hr_syncsrc == HAST_SYNCSRC_UNDEF ||
1127                             res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1128                                 /*
1129                                  * This range is up-to-date on local component,
1130                                  * so handle request locally.
1131                                  */
1132                                  /* Local component is 0 for now. */
1133                                 ncomp = 0;
1134                         } else /* if (res->hr_syncsrc ==
1135                             HAST_SYNCSRC_SECONDARY) */ {
1136                                 PJDLOG_ASSERT(res->hr_syncsrc ==
1137                                     HAST_SYNCSRC_SECONDARY);
1138                                 /*
1139                                  * This range is out-of-date on local component,
1140                                  * so send request to the remote node.
1141                                  */
1142                                  /* Remote component is 1 for now. */
1143                                 ncomp = 1;
1144                         }
1145                         mtx_unlock(&metadata_lock);
1146                         QUEUE_INSERT1(hio, send, ncomp);
1147                         break;
1148                 case BIO_WRITE:
1149                         res->hr_stat_write++;
1150                         if (res->hr_resuid == 0) {
1151                                 /*
1152                                  * This is first write, initialize localcnt and
1153                                  * resuid.
1154                                  */
1155                                 res->hr_primary_localcnt = 1;
1156                                 (void)init_resuid(res);
1157                         }
1158                         for (;;) {
1159                                 mtx_lock(&range_lock);
1160                                 if (rangelock_islocked(range_sync,
1161                                     ggio->gctl_offset, ggio->gctl_length)) {
1162                                         pjdlog_debug(2,
1163                                             "regular: Range offset=%jd length=%zu locked.",
1164                                             (intmax_t)ggio->gctl_offset,
1165                                             (size_t)ggio->gctl_length);
1166                                         range_regular_wait = true;
1167                                         cv_wait(&range_regular_cond, &range_lock);
1168                                         range_regular_wait = false;
1169                                         mtx_unlock(&range_lock);
1170                                         continue;
1171                                 }
1172                                 if (rangelock_add(range_regular,
1173                                     ggio->gctl_offset, ggio->gctl_length) < 0) {
1174                                         mtx_unlock(&range_lock);
1175                                         pjdlog_debug(2,
1176                                             "regular: Range offset=%jd length=%zu is already locked, waiting.",
1177                                             (intmax_t)ggio->gctl_offset,
1178                                             (size_t)ggio->gctl_length);
1179                                         sleep(1);
1180                                         continue;
1181                                 }
1182                                 mtx_unlock(&range_lock);
1183                                 break;
1184                         }
1185                         mtx_lock(&res->hr_amp_lock);
1186                         if (activemap_write_start(res->hr_amp,
1187                             ggio->gctl_offset, ggio->gctl_length)) {
1188                                 res->hr_stat_activemap_update++;
1189                                 (void)hast_activemap_flush(res);
1190                         }
1191                         mtx_unlock(&res->hr_amp_lock);
1192                         /* FALLTHROUGH */
1193                 case BIO_DELETE:
1194                 case BIO_FLUSH:
1195                         switch (ggio->gctl_cmd) {
1196                         case BIO_DELETE:
1197                                 res->hr_stat_delete++;
1198                                 break;
1199                         case BIO_FLUSH:
1200                                 res->hr_stat_flush++;
1201                                 break;
1202                         }
1203                         pjdlog_debug(2,
1204                             "ggate_recv: (%p) Moving request to the send queues.",
1205                             hio);
1206                         refcount_init(&hio->hio_countdown, ncomps);
1207                         for (ii = 0; ii < ncomps; ii++)
1208                                 QUEUE_INSERT1(hio, send, ii);
1209                         break;
1210                 }
1211         }
1212         /* NOTREACHED */
1213         return (NULL);
1214 }
1215
1216 /*
1217  * Thread reads from or writes to local component.
1218  * If local read fails, it redirects it to remote_send thread.
1219  */
1220 static void *
1221 local_send_thread(void *arg)
1222 {
1223         struct hast_resource *res = arg;
1224         struct g_gate_ctl_io *ggio;
1225         struct hio *hio;
1226         unsigned int ncomp, rncomp;
1227         ssize_t ret;
1228
1229         /* Local component is 0 for now. */
1230         ncomp = 0;
1231         /* Remote component is 1 for now. */
1232         rncomp = 1;
1233
1234         for (;;) {
1235                 pjdlog_debug(2, "local_send: Taking request.");
1236                 QUEUE_TAKE1(hio, send, ncomp, 0);
1237                 pjdlog_debug(2, "local_send: (%p) Got request.", hio);
1238                 ggio = &hio->hio_ggio;
1239                 switch (ggio->gctl_cmd) {
1240                 case BIO_READ:
1241                         ret = pread(res->hr_localfd, ggio->gctl_data,
1242                             ggio->gctl_length,
1243                             ggio->gctl_offset + res->hr_localoff);
1244                         if (ret == ggio->gctl_length)
1245                                 hio->hio_errors[ncomp] = 0;
1246                         else if (!ISSYNCREQ(hio)) {
1247                                 /*
1248                                  * If READ failed, try to read from remote node.
1249                                  */
1250                                 if (ret < 0) {
1251                                         reqlog(LOG_WARNING, 0, ggio,
1252                                             "Local request failed (%s), trying remote node. ",
1253                                             strerror(errno));
1254                                 } else if (ret != ggio->gctl_length) {
1255                                         reqlog(LOG_WARNING, 0, ggio,
1256                                             "Local request failed (%zd != %jd), trying remote node. ",
1257                                             ret, (intmax_t)ggio->gctl_length);
1258                                 }
1259                                 QUEUE_INSERT1(hio, send, rncomp);
1260                                 continue;
1261                         }
1262                         break;
1263                 case BIO_WRITE:
1264                         ret = pwrite(res->hr_localfd, ggio->gctl_data,
1265                             ggio->gctl_length,
1266                             ggio->gctl_offset + res->hr_localoff);
1267                         if (ret < 0) {
1268                                 hio->hio_errors[ncomp] = errno;
1269                                 reqlog(LOG_WARNING, 0, ggio,
1270                                     "Local request failed (%s): ",
1271                                     strerror(errno));
1272                         } else if (ret != ggio->gctl_length) {
1273                                 hio->hio_errors[ncomp] = EIO;
1274                                 reqlog(LOG_WARNING, 0, ggio,
1275                                     "Local request failed (%zd != %jd): ",
1276                                     ret, (intmax_t)ggio->gctl_length);
1277                         } else {
1278                                 hio->hio_errors[ncomp] = 0;
1279                         }
1280                         break;
1281                 case BIO_DELETE:
1282                         ret = g_delete(res->hr_localfd,
1283                             ggio->gctl_offset + res->hr_localoff,
1284                             ggio->gctl_length);
1285                         if (ret < 0) {
1286                                 hio->hio_errors[ncomp] = errno;
1287                                 reqlog(LOG_WARNING, 0, ggio,
1288                                     "Local request failed (%s): ",
1289                                     strerror(errno));
1290                         } else {
1291                                 hio->hio_errors[ncomp] = 0;
1292                         }
1293                         break;
1294                 case BIO_FLUSH:
1295                         ret = g_flush(res->hr_localfd);
1296                         if (ret < 0) {
1297                                 hio->hio_errors[ncomp] = errno;
1298                                 reqlog(LOG_WARNING, 0, ggio,
1299                                     "Local request failed (%s): ",
1300                                     strerror(errno));
1301                         } else {
1302                                 hio->hio_errors[ncomp] = 0;
1303                         }
1304                         break;
1305                 }
1306                 if (refcount_release(&hio->hio_countdown)) {
1307                         if (ISSYNCREQ(hio)) {
1308                                 mtx_lock(&sync_lock);
1309                                 SYNCREQDONE(hio);
1310                                 mtx_unlock(&sync_lock);
1311                                 cv_signal(&sync_cond);
1312                         } else {
1313                                 pjdlog_debug(2,
1314                                     "local_send: (%p) Moving request to the done queue.",
1315                                     hio);
1316                                 QUEUE_INSERT2(hio, done);
1317                         }
1318                 }
1319         }
1320         /* NOTREACHED */
1321         return (NULL);
1322 }
1323
1324 static void
1325 keepalive_send(struct hast_resource *res, unsigned int ncomp)
1326 {
1327         struct nv *nv;
1328
1329         rw_rlock(&hio_remote_lock[ncomp]);
1330
1331         if (!ISCONNECTED(res, ncomp)) {
1332                 rw_unlock(&hio_remote_lock[ncomp]);
1333                 return;
1334         }
1335
1336         PJDLOG_ASSERT(res->hr_remotein != NULL);
1337         PJDLOG_ASSERT(res->hr_remoteout != NULL);
1338
1339         nv = nv_alloc();
1340         nv_add_uint8(nv, HIO_KEEPALIVE, "cmd");
1341         if (nv_error(nv) != 0) {
1342                 rw_unlock(&hio_remote_lock[ncomp]);
1343                 nv_free(nv);
1344                 pjdlog_debug(1,
1345                     "keepalive_send: Unable to prepare header to send.");
1346                 return;
1347         }
1348         if (hast_proto_send(res, res->hr_remoteout, nv, NULL, 0) < 0) {
1349                 rw_unlock(&hio_remote_lock[ncomp]);
1350                 pjdlog_common(LOG_DEBUG, 1, errno,
1351                     "keepalive_send: Unable to send request");
1352                 nv_free(nv);
1353                 remote_close(res, ncomp);
1354                 return;
1355         }
1356
1357         rw_unlock(&hio_remote_lock[ncomp]);
1358         nv_free(nv);
1359         pjdlog_debug(2, "keepalive_send: Request sent.");
1360 }
1361
1362 /*
1363  * Thread sends request to secondary node.
1364  */
1365 static void *
1366 remote_send_thread(void *arg)
1367 {
1368         struct hast_resource *res = arg;
1369         struct g_gate_ctl_io *ggio;
1370         time_t lastcheck, now;
1371         struct hio *hio;
1372         struct nv *nv;
1373         unsigned int ncomp;
1374         bool wakeup;
1375         uint64_t offset, length;
1376         uint8_t cmd;
1377         void *data;
1378
1379         /* Remote component is 1 for now. */
1380         ncomp = 1;
1381         lastcheck = time(NULL);
1382
1383         for (;;) {
1384                 pjdlog_debug(2, "remote_send: Taking request.");
1385                 QUEUE_TAKE1(hio, send, ncomp, HAST_KEEPALIVE);
1386                 if (hio == NULL) {
1387                         now = time(NULL);
1388                         if (lastcheck + HAST_KEEPALIVE <= now) {
1389                                 keepalive_send(res, ncomp);
1390                                 lastcheck = now;
1391                         }
1392                         continue;
1393                 }
1394                 pjdlog_debug(2, "remote_send: (%p) Got request.", hio);
1395                 ggio = &hio->hio_ggio;
1396                 switch (ggio->gctl_cmd) {
1397                 case BIO_READ:
1398                         cmd = HIO_READ;
1399                         data = NULL;
1400                         offset = ggio->gctl_offset;
1401                         length = ggio->gctl_length;
1402                         break;
1403                 case BIO_WRITE:
1404                         cmd = HIO_WRITE;
1405                         data = ggio->gctl_data;
1406                         offset = ggio->gctl_offset;
1407                         length = ggio->gctl_length;
1408                         break;
1409                 case BIO_DELETE:
1410                         cmd = HIO_DELETE;
1411                         data = NULL;
1412                         offset = ggio->gctl_offset;
1413                         length = ggio->gctl_length;
1414                         break;
1415                 case BIO_FLUSH:
1416                         cmd = HIO_FLUSH;
1417                         data = NULL;
1418                         offset = 0;
1419                         length = 0;
1420                         break;
1421                 default:
1422                         PJDLOG_ASSERT(!"invalid condition");
1423                         abort();
1424                 }
1425                 nv = nv_alloc();
1426                 nv_add_uint8(nv, cmd, "cmd");
1427                 nv_add_uint64(nv, (uint64_t)ggio->gctl_seq, "seq");
1428                 nv_add_uint64(nv, offset, "offset");
1429                 nv_add_uint64(nv, length, "length");
1430                 if (nv_error(nv) != 0) {
1431                         hio->hio_errors[ncomp] = nv_error(nv);
1432                         pjdlog_debug(2,
1433                             "remote_send: (%p) Unable to prepare header to send.",
1434                             hio);
1435                         reqlog(LOG_ERR, 0, ggio,
1436                             "Unable to prepare header to send (%s): ",
1437                             strerror(nv_error(nv)));
1438                         /* Move failed request immediately to the done queue. */
1439                         goto done_queue;
1440                 }
1441                 pjdlog_debug(2,
1442                     "remote_send: (%p) Moving request to the recv queue.",
1443                     hio);
1444                 /*
1445                  * Protect connection from disappearing.
1446                  */
1447                 rw_rlock(&hio_remote_lock[ncomp]);
1448                 if (!ISCONNECTED(res, ncomp)) {
1449                         rw_unlock(&hio_remote_lock[ncomp]);
1450                         hio->hio_errors[ncomp] = ENOTCONN;
1451                         goto done_queue;
1452                 }
1453                 /*
1454                  * Move the request to recv queue before sending it, because
1455                  * in different order we can get reply before we move request
1456                  * to recv queue.
1457                  */
1458                 mtx_lock(&hio_recv_list_lock[ncomp]);
1459                 wakeup = TAILQ_EMPTY(&hio_recv_list[ncomp]);
1460                 TAILQ_INSERT_TAIL(&hio_recv_list[ncomp], hio, hio_next[ncomp]);
1461                 mtx_unlock(&hio_recv_list_lock[ncomp]);
1462                 if (hast_proto_send(res, res->hr_remoteout, nv, data,
1463                     data != NULL ? length : 0) < 0) {
1464                         hio->hio_errors[ncomp] = errno;
1465                         rw_unlock(&hio_remote_lock[ncomp]);
1466                         pjdlog_debug(2,
1467                             "remote_send: (%p) Unable to send request.", hio);
1468                         reqlog(LOG_ERR, 0, ggio,
1469                             "Unable to send request (%s): ",
1470                             strerror(hio->hio_errors[ncomp]));
1471                         remote_close(res, ncomp);
1472                         /*
1473                          * Take request back from the receive queue and move
1474                          * it immediately to the done queue.
1475                          */
1476                         mtx_lock(&hio_recv_list_lock[ncomp]);
1477                         TAILQ_REMOVE(&hio_recv_list[ncomp], hio, hio_next[ncomp]);
1478                         mtx_unlock(&hio_recv_list_lock[ncomp]);
1479                         goto done_queue;
1480                 }
1481                 rw_unlock(&hio_remote_lock[ncomp]);
1482                 nv_free(nv);
1483                 if (wakeup)
1484                         cv_signal(&hio_recv_list_cond[ncomp]);
1485                 continue;
1486 done_queue:
1487                 nv_free(nv);
1488                 if (ISSYNCREQ(hio)) {
1489                         if (!refcount_release(&hio->hio_countdown))
1490                                 continue;
1491                         mtx_lock(&sync_lock);
1492                         SYNCREQDONE(hio);
1493                         mtx_unlock(&sync_lock);
1494                         cv_signal(&sync_cond);
1495                         continue;
1496                 }
1497                 if (ggio->gctl_cmd == BIO_WRITE) {
1498                         mtx_lock(&res->hr_amp_lock);
1499                         if (activemap_need_sync(res->hr_amp, ggio->gctl_offset,
1500                             ggio->gctl_length)) {
1501                                 (void)hast_activemap_flush(res);
1502                         }
1503                         mtx_unlock(&res->hr_amp_lock);
1504                 }
1505                 if (!refcount_release(&hio->hio_countdown))
1506                         continue;
1507                 pjdlog_debug(2,
1508                     "remote_send: (%p) Moving request to the done queue.",
1509                     hio);
1510                 QUEUE_INSERT2(hio, done);
1511         }
1512         /* NOTREACHED */
1513         return (NULL);
1514 }
1515
1516 /*
1517  * Thread receives answer from secondary node and passes it to ggate_send
1518  * thread.
1519  */
1520 static void *
1521 remote_recv_thread(void *arg)
1522 {
1523         struct hast_resource *res = arg;
1524         struct g_gate_ctl_io *ggio;
1525         struct hio *hio;
1526         struct nv *nv;
1527         unsigned int ncomp;
1528         uint64_t seq;
1529         int error;
1530
1531         /* Remote component is 1 for now. */
1532         ncomp = 1;
1533
1534         for (;;) {
1535                 /* Wait until there is anything to receive. */
1536                 mtx_lock(&hio_recv_list_lock[ncomp]);
1537                 while (TAILQ_EMPTY(&hio_recv_list[ncomp])) {
1538                         pjdlog_debug(2, "remote_recv: No requests, waiting.");
1539                         cv_wait(&hio_recv_list_cond[ncomp],
1540                             &hio_recv_list_lock[ncomp]);
1541                 }
1542                 mtx_unlock(&hio_recv_list_lock[ncomp]);
1543                 rw_rlock(&hio_remote_lock[ncomp]);
1544                 if (!ISCONNECTED(res, ncomp)) {
1545                         rw_unlock(&hio_remote_lock[ncomp]);
1546                         /*
1547                          * Connection is dead, so move all pending requests to
1548                          * the done queue (one-by-one).
1549                          */
1550                         mtx_lock(&hio_recv_list_lock[ncomp]);
1551                         hio = TAILQ_FIRST(&hio_recv_list[ncomp]);
1552                         PJDLOG_ASSERT(hio != NULL);
1553                         TAILQ_REMOVE(&hio_recv_list[ncomp], hio,
1554                             hio_next[ncomp]);
1555                         mtx_unlock(&hio_recv_list_lock[ncomp]);
1556                         goto done_queue;
1557                 }
1558                 if (hast_proto_recv_hdr(res->hr_remotein, &nv) < 0) {
1559                         pjdlog_errno(LOG_ERR,
1560                             "Unable to receive reply header");
1561                         rw_unlock(&hio_remote_lock[ncomp]);
1562                         remote_close(res, ncomp);
1563                         continue;
1564                 }
1565                 rw_unlock(&hio_remote_lock[ncomp]);
1566                 seq = nv_get_uint64(nv, "seq");
1567                 if (seq == 0) {
1568                         pjdlog_error("Header contains no 'seq' field.");
1569                         nv_free(nv);
1570                         continue;
1571                 }
1572                 mtx_lock(&hio_recv_list_lock[ncomp]);
1573                 TAILQ_FOREACH(hio, &hio_recv_list[ncomp], hio_next[ncomp]) {
1574                         if (hio->hio_ggio.gctl_seq == seq) {
1575                                 TAILQ_REMOVE(&hio_recv_list[ncomp], hio,
1576                                     hio_next[ncomp]);
1577                                 break;
1578                         }
1579                 }
1580                 mtx_unlock(&hio_recv_list_lock[ncomp]);
1581                 if (hio == NULL) {
1582                         pjdlog_error("Found no request matching received 'seq' field (%ju).",
1583                             (uintmax_t)seq);
1584                         nv_free(nv);
1585                         continue;
1586                 }
1587                 error = nv_get_int16(nv, "error");
1588                 if (error != 0) {
1589                         /* Request failed on remote side. */
1590                         hio->hio_errors[ncomp] = error;
1591                         reqlog(LOG_WARNING, 0, &hio->hio_ggio,
1592                             "Remote request failed (%s): ", strerror(error));
1593                         nv_free(nv);
1594                         goto done_queue;
1595                 }
1596                 ggio = &hio->hio_ggio;
1597                 switch (ggio->gctl_cmd) {
1598                 case BIO_READ:
1599                         rw_rlock(&hio_remote_lock[ncomp]);
1600                         if (!ISCONNECTED(res, ncomp)) {
1601                                 rw_unlock(&hio_remote_lock[ncomp]);
1602                                 nv_free(nv);
1603                                 goto done_queue;
1604                         }
1605                         if (hast_proto_recv_data(res, res->hr_remotein, nv,
1606                             ggio->gctl_data, ggio->gctl_length) < 0) {
1607                                 hio->hio_errors[ncomp] = errno;
1608                                 pjdlog_errno(LOG_ERR,
1609                                     "Unable to receive reply data");
1610                                 rw_unlock(&hio_remote_lock[ncomp]);
1611                                 nv_free(nv);
1612                                 remote_close(res, ncomp);
1613                                 goto done_queue;
1614                         }
1615                         rw_unlock(&hio_remote_lock[ncomp]);
1616                         break;
1617                 case BIO_WRITE:
1618                 case BIO_DELETE:
1619                 case BIO_FLUSH:
1620                         break;
1621                 default:
1622                         PJDLOG_ASSERT(!"invalid condition");
1623                         abort();
1624                 }
1625                 hio->hio_errors[ncomp] = 0;
1626                 nv_free(nv);
1627 done_queue:
1628                 if (refcount_release(&hio->hio_countdown)) {
1629                         if (ISSYNCREQ(hio)) {
1630                                 mtx_lock(&sync_lock);
1631                                 SYNCREQDONE(hio);
1632                                 mtx_unlock(&sync_lock);
1633                                 cv_signal(&sync_cond);
1634                         } else {
1635                                 pjdlog_debug(2,
1636                                     "remote_recv: (%p) Moving request to the done queue.",
1637                                     hio);
1638                                 QUEUE_INSERT2(hio, done);
1639                         }
1640                 }
1641         }
1642         /* NOTREACHED */
1643         return (NULL);
1644 }
1645
1646 /*
1647  * Thread sends answer to the kernel.
1648  */
1649 static void *
1650 ggate_send_thread(void *arg)
1651 {
1652         struct hast_resource *res = arg;
1653         struct g_gate_ctl_io *ggio;
1654         struct hio *hio;
1655         unsigned int ii, ncomp, ncomps;
1656
1657         ncomps = HAST_NCOMPONENTS;
1658
1659         for (;;) {
1660                 pjdlog_debug(2, "ggate_send: Taking request.");
1661                 QUEUE_TAKE2(hio, done);
1662                 pjdlog_debug(2, "ggate_send: (%p) Got request.", hio);
1663                 ggio = &hio->hio_ggio;
1664                 for (ii = 0; ii < ncomps; ii++) {
1665                         if (hio->hio_errors[ii] == 0) {
1666                                 /*
1667                                  * One successful request is enough to declare
1668                                  * success.
1669                                  */
1670                                 ggio->gctl_error = 0;
1671                                 break;
1672                         }
1673                 }
1674                 if (ii == ncomps) {
1675                         /*
1676                          * None of the requests were successful.
1677                          * Use the error from local component except the
1678                          * case when we did only remote request.
1679                          */
1680                         if (ggio->gctl_cmd == BIO_READ &&
1681                             res->hr_syncsrc == HAST_SYNCSRC_SECONDARY)
1682                                 ggio->gctl_error = hio->hio_errors[1];
1683                         else
1684                                 ggio->gctl_error = hio->hio_errors[0];
1685                 }
1686                 if (ggio->gctl_error == 0 && ggio->gctl_cmd == BIO_WRITE) {
1687                         mtx_lock(&res->hr_amp_lock);
1688                         activemap_write_complete(res->hr_amp,
1689                             ggio->gctl_offset, ggio->gctl_length);
1690                         mtx_unlock(&res->hr_amp_lock);
1691                 }
1692                 if (ggio->gctl_cmd == BIO_WRITE) {
1693                         /*
1694                          * Unlock range we locked.
1695                          */
1696                         mtx_lock(&range_lock);
1697                         rangelock_del(range_regular, ggio->gctl_offset,
1698                             ggio->gctl_length);
1699                         if (range_sync_wait)
1700                                 cv_signal(&range_sync_cond);
1701                         mtx_unlock(&range_lock);
1702                         /*
1703                          * Bump local count if this is first write after
1704                          * connection failure with remote node.
1705                          */
1706                         ncomp = 1;
1707                         rw_rlock(&hio_remote_lock[ncomp]);
1708                         if (!ISCONNECTED(res, ncomp)) {
1709                                 mtx_lock(&metadata_lock);
1710                                 if (res->hr_primary_localcnt ==
1711                                     res->hr_secondary_remotecnt) {
1712                                         res->hr_primary_localcnt++;
1713                                         pjdlog_debug(1,
1714                                             "Increasing localcnt to %ju.",
1715                                             (uintmax_t)res->hr_primary_localcnt);
1716                                         (void)metadata_write(res);
1717                                 }
1718                                 mtx_unlock(&metadata_lock);
1719                         }
1720                         rw_unlock(&hio_remote_lock[ncomp]);
1721                 }
1722                 if (ioctl(res->hr_ggatefd, G_GATE_CMD_DONE, ggio) < 0)
1723                         primary_exit(EX_OSERR, "G_GATE_CMD_DONE failed");
1724                 pjdlog_debug(2,
1725                     "ggate_send: (%p) Moving request to the free queue.", hio);
1726                 QUEUE_INSERT2(hio, free);
1727         }
1728         /* NOTREACHED */
1729         return (NULL);
1730 }
1731
1732 /*
1733  * Thread synchronize local and remote components.
1734  */
1735 static void *
1736 sync_thread(void *arg __unused)
1737 {
1738         struct hast_resource *res = arg;
1739         struct hio *hio;
1740         struct g_gate_ctl_io *ggio;
1741         struct timeval tstart, tend, tdiff;
1742         unsigned int ii, ncomp, ncomps;
1743         off_t offset, length, synced;
1744         bool dorewind;
1745         int syncext;
1746
1747         ncomps = HAST_NCOMPONENTS;
1748         dorewind = true;
1749         synced = 0;
1750         offset = -1;
1751
1752         for (;;) {
1753                 mtx_lock(&sync_lock);
1754                 if (offset >= 0 && !sync_inprogress) {
1755                         gettimeofday(&tend, NULL);
1756                         timersub(&tend, &tstart, &tdiff);
1757                         pjdlog_info("Synchronization interrupted after %#.0T. "
1758                             "%NB synchronized so far.", &tdiff,
1759                             (intmax_t)synced);
1760                         event_send(res, EVENT_SYNCINTR);
1761                 }
1762                 while (!sync_inprogress) {
1763                         dorewind = true;
1764                         synced = 0;
1765                         cv_wait(&sync_cond, &sync_lock);
1766                 }
1767                 mtx_unlock(&sync_lock);
1768                 /*
1769                  * Obtain offset at which we should synchronize.
1770                  * Rewind synchronization if needed.
1771                  */
1772                 mtx_lock(&res->hr_amp_lock);
1773                 if (dorewind)
1774                         activemap_sync_rewind(res->hr_amp);
1775                 offset = activemap_sync_offset(res->hr_amp, &length, &syncext);
1776                 if (syncext != -1) {
1777                         /*
1778                          * We synchronized entire syncext extent, we can mark
1779                          * it as clean now.
1780                          */
1781                         if (activemap_extent_complete(res->hr_amp, syncext))
1782                                 (void)hast_activemap_flush(res);
1783                 }
1784                 mtx_unlock(&res->hr_amp_lock);
1785                 if (dorewind) {
1786                         dorewind = false;
1787                         if (offset < 0)
1788                                 pjdlog_info("Nodes are in sync.");
1789                         else {
1790                                 pjdlog_info("Synchronization started. %NB to go.",
1791                                     (intmax_t)(res->hr_extentsize *
1792                                     activemap_ndirty(res->hr_amp)));
1793                                 event_send(res, EVENT_SYNCSTART);
1794                                 gettimeofday(&tstart, NULL);
1795                         }
1796                 }
1797                 if (offset < 0) {
1798                         sync_stop();
1799                         pjdlog_debug(1, "Nothing to synchronize.");
1800                         /*
1801                          * Synchronization complete, make both localcnt and
1802                          * remotecnt equal.
1803                          */
1804                         ncomp = 1;
1805                         rw_rlock(&hio_remote_lock[ncomp]);
1806                         if (ISCONNECTED(res, ncomp)) {
1807                                 if (synced > 0) {
1808                                         int64_t bps;
1809
1810                                         gettimeofday(&tend, NULL);
1811                                         timersub(&tend, &tstart, &tdiff);
1812                                         bps = (int64_t)((double)synced /
1813                                             ((double)tdiff.tv_sec +
1814                                             (double)tdiff.tv_usec / 1000000));
1815                                         pjdlog_info("Synchronization complete. "
1816                                             "%NB synchronized in %#.0lT (%NB/sec).",
1817                                             (intmax_t)synced, &tdiff,
1818                                             (intmax_t)bps);
1819                                         event_send(res, EVENT_SYNCDONE);
1820                                 }
1821                                 mtx_lock(&metadata_lock);
1822                                 res->hr_syncsrc = HAST_SYNCSRC_UNDEF;
1823                                 res->hr_primary_localcnt =
1824                                     res->hr_secondary_remotecnt;
1825                                 res->hr_primary_remotecnt =
1826                                     res->hr_secondary_localcnt;
1827                                 pjdlog_debug(1,
1828                                     "Setting localcnt to %ju and remotecnt to %ju.",
1829                                     (uintmax_t)res->hr_primary_localcnt,
1830                                     (uintmax_t)res->hr_primary_remotecnt);
1831                                 (void)metadata_write(res);
1832                                 mtx_unlock(&metadata_lock);
1833                         }
1834                         rw_unlock(&hio_remote_lock[ncomp]);
1835                         continue;
1836                 }
1837                 pjdlog_debug(2, "sync: Taking free request.");
1838                 QUEUE_TAKE2(hio, free);
1839                 pjdlog_debug(2, "sync: (%p) Got free request.", hio);
1840                 /*
1841                  * Lock the range we are going to synchronize. We don't want
1842                  * race where someone writes between our read and write.
1843                  */
1844                 for (;;) {
1845                         mtx_lock(&range_lock);
1846                         if (rangelock_islocked(range_regular, offset, length)) {
1847                                 pjdlog_debug(2,
1848                                     "sync: Range offset=%jd length=%jd locked.",
1849                                     (intmax_t)offset, (intmax_t)length);
1850                                 range_sync_wait = true;
1851                                 cv_wait(&range_sync_cond, &range_lock);
1852                                 range_sync_wait = false;
1853                                 mtx_unlock(&range_lock);
1854                                 continue;
1855                         }
1856                         if (rangelock_add(range_sync, offset, length) < 0) {
1857                                 mtx_unlock(&range_lock);
1858                                 pjdlog_debug(2,
1859                                     "sync: Range offset=%jd length=%jd is already locked, waiting.",
1860                                     (intmax_t)offset, (intmax_t)length);
1861                                 sleep(1);
1862                                 continue;
1863                         }
1864                         mtx_unlock(&range_lock);
1865                         break;
1866                 }
1867                 /*
1868                  * First read the data from synchronization source.
1869                  */
1870                 SYNCREQ(hio);
1871                 ggio = &hio->hio_ggio;
1872                 ggio->gctl_cmd = BIO_READ;
1873                 ggio->gctl_offset = offset;
1874                 ggio->gctl_length = length;
1875                 ggio->gctl_error = 0;
1876                 for (ii = 0; ii < ncomps; ii++)
1877                         hio->hio_errors[ii] = EINVAL;
1878                 reqlog(LOG_DEBUG, 2, ggio, "sync: (%p) Sending sync request: ",
1879                     hio);
1880                 pjdlog_debug(2, "sync: (%p) Moving request to the send queue.",
1881                     hio);
1882                 mtx_lock(&metadata_lock);
1883                 if (res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1884                         /*
1885                          * This range is up-to-date on local component,
1886                          * so handle request locally.
1887                          */
1888                          /* Local component is 0 for now. */
1889                         ncomp = 0;
1890                 } else /* if (res->hr_syncsrc == HAST_SYNCSRC_SECONDARY) */ {
1891                         PJDLOG_ASSERT(res->hr_syncsrc == HAST_SYNCSRC_SECONDARY);
1892                         /*
1893                          * This range is out-of-date on local component,
1894                          * so send request to the remote node.
1895                          */
1896                          /* Remote component is 1 for now. */
1897                         ncomp = 1;
1898                 }
1899                 mtx_unlock(&metadata_lock);
1900                 refcount_init(&hio->hio_countdown, 1);
1901                 QUEUE_INSERT1(hio, send, ncomp);
1902
1903                 /*
1904                  * Let's wait for READ to finish.
1905                  */
1906                 mtx_lock(&sync_lock);
1907                 while (!ISSYNCREQDONE(hio))
1908                         cv_wait(&sync_cond, &sync_lock);
1909                 mtx_unlock(&sync_lock);
1910
1911                 if (hio->hio_errors[ncomp] != 0) {
1912                         pjdlog_error("Unable to read synchronization data: %s.",
1913                             strerror(hio->hio_errors[ncomp]));
1914                         goto free_queue;
1915                 }
1916
1917                 /*
1918                  * We read the data from synchronization source, now write it
1919                  * to synchronization target.
1920                  */
1921                 SYNCREQ(hio);
1922                 ggio->gctl_cmd = BIO_WRITE;
1923                 for (ii = 0; ii < ncomps; ii++)
1924                         hio->hio_errors[ii] = EINVAL;
1925                 reqlog(LOG_DEBUG, 2, ggio, "sync: (%p) Sending sync request: ",
1926                     hio);
1927                 pjdlog_debug(2, "sync: (%p) Moving request to the send queue.",
1928                     hio);
1929                 mtx_lock(&metadata_lock);
1930                 if (res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1931                         /*
1932                          * This range is up-to-date on local component,
1933                          * so we update remote component.
1934                          */
1935                          /* Remote component is 1 for now. */
1936                         ncomp = 1;
1937                 } else /* if (res->hr_syncsrc == HAST_SYNCSRC_SECONDARY) */ {
1938                         PJDLOG_ASSERT(res->hr_syncsrc == HAST_SYNCSRC_SECONDARY);
1939                         /*
1940                          * This range is out-of-date on local component,
1941                          * so we update it.
1942                          */
1943                          /* Local component is 0 for now. */
1944                         ncomp = 0;
1945                 }
1946                 mtx_unlock(&metadata_lock);
1947
1948                 pjdlog_debug(2, "sync: (%p) Moving request to the send queues.",
1949                     hio);
1950                 refcount_init(&hio->hio_countdown, 1);
1951                 QUEUE_INSERT1(hio, send, ncomp);
1952
1953                 /*
1954                  * Let's wait for WRITE to finish.
1955                  */
1956                 mtx_lock(&sync_lock);
1957                 while (!ISSYNCREQDONE(hio))
1958                         cv_wait(&sync_cond, &sync_lock);
1959                 mtx_unlock(&sync_lock);
1960
1961                 if (hio->hio_errors[ncomp] != 0) {
1962                         pjdlog_error("Unable to write synchronization data: %s.",
1963                             strerror(hio->hio_errors[ncomp]));
1964                         goto free_queue;
1965                 }
1966
1967                 synced += length;
1968 free_queue:
1969                 mtx_lock(&range_lock);
1970                 rangelock_del(range_sync, offset, length);
1971                 if (range_regular_wait)
1972                         cv_signal(&range_regular_cond);
1973                 mtx_unlock(&range_lock);
1974                 pjdlog_debug(2, "sync: (%p) Moving request to the free queue.",
1975                     hio);
1976                 QUEUE_INSERT2(hio, free);
1977         }
1978         /* NOTREACHED */
1979         return (NULL);
1980 }
1981
1982 void
1983 primary_config_reload(struct hast_resource *res, struct nv *nv)
1984 {
1985         unsigned int ii, ncomps;
1986         int modified, vint;
1987         const char *vstr;
1988
1989         pjdlog_info("Reloading configuration...");
1990
1991         PJDLOG_ASSERT(res->hr_role == HAST_ROLE_PRIMARY);
1992         PJDLOG_ASSERT(gres == res);
1993         nv_assert(nv, "remoteaddr");
1994         nv_assert(nv, "sourceaddr");
1995         nv_assert(nv, "replication");
1996         nv_assert(nv, "checksum");
1997         nv_assert(nv, "compression");
1998         nv_assert(nv, "timeout");
1999         nv_assert(nv, "exec");
2000
2001         ncomps = HAST_NCOMPONENTS;
2002
2003 #define MODIFIED_REMOTEADDR     0x01
2004 #define MODIFIED_SOURCEADDR     0x02
2005 #define MODIFIED_REPLICATION    0x04
2006 #define MODIFIED_CHECKSUM       0x08
2007 #define MODIFIED_COMPRESSION    0x10
2008 #define MODIFIED_TIMEOUT        0x20
2009 #define MODIFIED_EXEC           0x40
2010         modified = 0;
2011
2012         vstr = nv_get_string(nv, "remoteaddr");
2013         if (strcmp(gres->hr_remoteaddr, vstr) != 0) {
2014                 /*
2015                  * Don't copy res->hr_remoteaddr to gres just yet.
2016                  * We want remote_close() to log disconnect from the old
2017                  * addresses, not from the new ones.
2018                  */
2019                 modified |= MODIFIED_REMOTEADDR;
2020         }
2021         vstr = nv_get_string(nv, "sourceaddr");
2022         if (strcmp(gres->hr_sourceaddr, vstr) != 0) {
2023                 strlcpy(gres->hr_sourceaddr, vstr, sizeof(gres->hr_sourceaddr));
2024                 modified |= MODIFIED_SOURCEADDR;
2025         }
2026         vint = nv_get_int32(nv, "replication");
2027         if (gres->hr_replication != vint) {
2028                 gres->hr_replication = vint;
2029                 modified |= MODIFIED_REPLICATION;
2030         }
2031         vint = nv_get_int32(nv, "checksum");
2032         if (gres->hr_checksum != vint) {
2033                 gres->hr_checksum = vint;
2034                 modified |= MODIFIED_CHECKSUM;
2035         }
2036         vint = nv_get_int32(nv, "compression");
2037         if (gres->hr_compression != vint) {
2038                 gres->hr_compression = vint;
2039                 modified |= MODIFIED_COMPRESSION;
2040         }
2041         vint = nv_get_int32(nv, "timeout");
2042         if (gres->hr_timeout != vint) {
2043                 gres->hr_timeout = vint;
2044                 modified |= MODIFIED_TIMEOUT;
2045         }
2046         vstr = nv_get_string(nv, "exec");
2047         if (strcmp(gres->hr_exec, vstr) != 0) {
2048                 strlcpy(gres->hr_exec, vstr, sizeof(gres->hr_exec));
2049                 modified |= MODIFIED_EXEC;
2050         }
2051
2052         /*
2053          * Change timeout for connected sockets.
2054          * Don't bother if we need to reconnect.
2055          */
2056         if ((modified & MODIFIED_TIMEOUT) != 0 &&
2057             (modified & (MODIFIED_REMOTEADDR | MODIFIED_SOURCEADDR |
2058             MODIFIED_REPLICATION)) == 0) {
2059                 for (ii = 0; ii < ncomps; ii++) {
2060                         if (!ISREMOTE(ii))
2061                                 continue;
2062                         rw_rlock(&hio_remote_lock[ii]);
2063                         if (!ISCONNECTED(gres, ii)) {
2064                                 rw_unlock(&hio_remote_lock[ii]);
2065                                 continue;
2066                         }
2067                         rw_unlock(&hio_remote_lock[ii]);
2068                         if (proto_timeout(gres->hr_remotein,
2069                             gres->hr_timeout) < 0) {
2070                                 pjdlog_errno(LOG_WARNING,
2071                                     "Unable to set connection timeout");
2072                         }
2073                         if (proto_timeout(gres->hr_remoteout,
2074                             gres->hr_timeout) < 0) {
2075                                 pjdlog_errno(LOG_WARNING,
2076                                     "Unable to set connection timeout");
2077                         }
2078                 }
2079         }
2080         if ((modified & (MODIFIED_REMOTEADDR | MODIFIED_SOURCEADDR |
2081             MODIFIED_REPLICATION)) != 0) {
2082                 for (ii = 0; ii < ncomps; ii++) {
2083                         if (!ISREMOTE(ii))
2084                                 continue;
2085                         remote_close(gres, ii);
2086                 }
2087                 if (modified & MODIFIED_REMOTEADDR) {
2088                         vstr = nv_get_string(nv, "remoteaddr");
2089                         strlcpy(gres->hr_remoteaddr, vstr,
2090                             sizeof(gres->hr_remoteaddr));
2091                 }
2092         }
2093 #undef  MODIFIED_REMOTEADDR
2094 #undef  MODIFIED_SOURCEADDR
2095 #undef  MODIFIED_REPLICATION
2096 #undef  MODIFIED_CHECKSUM
2097 #undef  MODIFIED_COMPRESSION
2098 #undef  MODIFIED_TIMEOUT
2099 #undef  MODIFIED_EXEC
2100
2101         pjdlog_info("Configuration reloaded successfully.");
2102 }
2103
2104 static void
2105 guard_one(struct hast_resource *res, unsigned int ncomp)
2106 {
2107         struct proto_conn *in, *out;
2108
2109         if (!ISREMOTE(ncomp))
2110                 return;
2111
2112         rw_rlock(&hio_remote_lock[ncomp]);
2113
2114         if (!real_remote(res)) {
2115                 rw_unlock(&hio_remote_lock[ncomp]);
2116                 return;
2117         }
2118
2119         if (ISCONNECTED(res, ncomp)) {
2120                 PJDLOG_ASSERT(res->hr_remotein != NULL);
2121                 PJDLOG_ASSERT(res->hr_remoteout != NULL);
2122                 rw_unlock(&hio_remote_lock[ncomp]);
2123                 pjdlog_debug(2, "remote_guard: Connection to %s is ok.",
2124                     res->hr_remoteaddr);
2125                 return;
2126         }
2127
2128         PJDLOG_ASSERT(res->hr_remotein == NULL);
2129         PJDLOG_ASSERT(res->hr_remoteout == NULL);
2130         /*
2131          * Upgrade the lock. It doesn't have to be atomic as no other thread
2132          * can change connection status from disconnected to connected.
2133          */
2134         rw_unlock(&hio_remote_lock[ncomp]);
2135         pjdlog_debug(2, "remote_guard: Reconnecting to %s.",
2136             res->hr_remoteaddr);
2137         in = out = NULL;
2138         if (init_remote(res, &in, &out) == 0) {
2139                 rw_wlock(&hio_remote_lock[ncomp]);
2140                 PJDLOG_ASSERT(res->hr_remotein == NULL);
2141                 PJDLOG_ASSERT(res->hr_remoteout == NULL);
2142                 PJDLOG_ASSERT(in != NULL && out != NULL);
2143                 res->hr_remotein = in;
2144                 res->hr_remoteout = out;
2145                 rw_unlock(&hio_remote_lock[ncomp]);
2146                 pjdlog_info("Successfully reconnected to %s.",
2147                     res->hr_remoteaddr);
2148                 sync_start();
2149         } else {
2150                 /* Both connections should be NULL. */
2151                 PJDLOG_ASSERT(res->hr_remotein == NULL);
2152                 PJDLOG_ASSERT(res->hr_remoteout == NULL);
2153                 PJDLOG_ASSERT(in == NULL && out == NULL);
2154                 pjdlog_debug(2, "remote_guard: Reconnect to %s failed.",
2155                     res->hr_remoteaddr);
2156         }
2157 }
2158
2159 /*
2160  * Thread guards remote connections and reconnects when needed, handles
2161  * signals, etc.
2162  */
2163 static void *
2164 guard_thread(void *arg)
2165 {
2166         struct hast_resource *res = arg;
2167         unsigned int ii, ncomps;
2168         struct timespec timeout;
2169         time_t lastcheck, now;
2170         sigset_t mask;
2171         int signo;
2172
2173         ncomps = HAST_NCOMPONENTS;
2174         lastcheck = time(NULL);
2175
2176         PJDLOG_VERIFY(sigemptyset(&mask) == 0);
2177         PJDLOG_VERIFY(sigaddset(&mask, SIGINT) == 0);
2178         PJDLOG_VERIFY(sigaddset(&mask, SIGTERM) == 0);
2179
2180         timeout.tv_sec = HAST_KEEPALIVE;
2181         timeout.tv_nsec = 0;
2182         signo = -1;
2183
2184         for (;;) {
2185                 switch (signo) {
2186                 case SIGINT:
2187                 case SIGTERM:
2188                         sigexit_received = true;
2189                         primary_exitx(EX_OK,
2190                             "Termination signal received, exiting.");
2191                         break;
2192                 default:
2193                         break;
2194                 }
2195
2196                 /*
2197                  * Don't check connections until we fully started,
2198                  * as we may still be looping, waiting for remote node
2199                  * to switch from primary to secondary.
2200                  */
2201                 if (fullystarted) {
2202                         pjdlog_debug(2, "remote_guard: Checking connections.");
2203                         now = time(NULL);
2204                         if (lastcheck + HAST_KEEPALIVE <= now) {
2205                                 for (ii = 0; ii < ncomps; ii++)
2206                                         guard_one(res, ii);
2207                                 lastcheck = now;
2208                         }
2209                 }
2210                 signo = sigtimedwait(&mask, NULL, &timeout);
2211         }
2212         /* NOTREACHED */
2213         return (NULL);
2214 }