xref: /freebsd/sbin/hastd/primary.c (revision 774f94f14c92bf94afc21d8c8d7a1e8f2fdf5a48)
1 /*-
2  * Copyright (c) 2009 The FreeBSD Foundation
3  * All rights reserved.
4  *
5  * This software was developed by Pawel Jakub Dawidek under sponsorship from
6  * the FreeBSD Foundation.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/types.h>
34 #include <sys/time.h>
35 #include <sys/bio.h>
36 #include <sys/disk.h>
37 #include <sys/refcount.h>
38 #include <sys/stat.h>
39 
40 #include <geom/gate/g_gate.h>
41 
42 #include <assert.h>
43 #include <err.h>
44 #include <errno.h>
45 #include <fcntl.h>
46 #include <libgeom.h>
47 #include <pthread.h>
48 #include <stdint.h>
49 #include <stdio.h>
50 #include <string.h>
51 #include <sysexits.h>
52 #include <unistd.h>
53 
54 #include <activemap.h>
55 #include <nv.h>
56 #include <rangelock.h>
57 
58 #include "control.h"
59 #include "hast.h"
60 #include "hast_proto.h"
61 #include "hastd.h"
62 #include "metadata.h"
63 #include "proto.h"
64 #include "pjdlog.h"
65 #include "subr.h"
66 #include "synch.h"
67 
68 struct hio {
69 	/*
70 	 * Number of components we are still waiting for.
71 	 * When this field goes to 0, we can send the request back to the
72 	 * kernel. Each component has to decrease this counter by one
73 	 * even on failure.
74 	 */
75 	unsigned int		 hio_countdown;
76 	/*
77 	 * Each component has a place to store its own error.
78 	 * Once the request is handled by all components we can decide if the
79 	 * request overall is successful or not.
80 	 */
81 	int			*hio_errors;
82 	/*
83 	 * Structure used to comunicate with GEOM Gate class.
84 	 */
85 	struct g_gate_ctl_io	 hio_ggio;
86 	TAILQ_ENTRY(hio)	*hio_next;
87 };
88 #define	hio_free_next	hio_next[0]
89 #define	hio_done_next	hio_next[0]
90 
91 /*
92  * Free list holds unused structures. When free list is empty, we have to wait
93  * until some in-progress requests are freed.
94  */
95 static TAILQ_HEAD(, hio) hio_free_list;
96 static pthread_mutex_t hio_free_list_lock;
97 static pthread_cond_t hio_free_list_cond;
98 /*
99  * There is one send list for every component. One requests is placed on all
100  * send lists - each component gets the same request, but each component is
101  * responsible for managing his own send list.
102  */
103 static TAILQ_HEAD(, hio) *hio_send_list;
104 static pthread_mutex_t *hio_send_list_lock;
105 static pthread_cond_t *hio_send_list_cond;
106 /*
107  * There is one recv list for every component, although local components don't
108  * use recv lists as local requests are done synchronously.
109  */
110 static TAILQ_HEAD(, hio) *hio_recv_list;
111 static pthread_mutex_t *hio_recv_list_lock;
112 static pthread_cond_t *hio_recv_list_cond;
113 /*
114  * Request is placed on done list by the slowest component (the one that
115  * decreased hio_countdown from 1 to 0).
116  */
117 static TAILQ_HEAD(, hio) hio_done_list;
118 static pthread_mutex_t hio_done_list_lock;
119 static pthread_cond_t hio_done_list_cond;
120 /*
121  * Structure below are for interaction with sync thread.
122  */
123 static bool sync_inprogress;
124 static pthread_mutex_t sync_lock;
125 static pthread_cond_t sync_cond;
126 /*
127  * The lock below allows to synchornize access to remote connections.
128  */
129 static pthread_rwlock_t *hio_remote_lock;
130 static pthread_mutex_t hio_guard_lock;
131 static pthread_cond_t hio_guard_cond;
132 
133 /*
134  * Lock to synchronize metadata updates. Also synchronize access to
135  * hr_primary_localcnt and hr_primary_remotecnt fields.
136  */
137 static pthread_mutex_t metadata_lock;
138 
139 /*
140  * Maximum number of outstanding I/O requests.
141  */
142 #define	HAST_HIO_MAX	256
143 /*
144  * Number of components. At this point there are only two components: local
145  * and remote, but in the future it might be possible to use multiple local
146  * and remote components.
147  */
148 #define	HAST_NCOMPONENTS	2
149 /*
150  * Number of seconds to sleep before next reconnect try.
151  */
152 #define	RECONNECT_SLEEP		5
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)	do {				\
179 	mtx_lock(&hio_##name##_list_lock[(ncomp)]);			\
180 	while (((hio) = TAILQ_FIRST(&hio_##name##_list[(ncomp)])) == NULL) { \
181 		cv_wait(&hio_##name##_list_cond[(ncomp)],		\
182 		    &hio_##name##_list_lock[(ncomp)]);			\
183 	}								\
184 	TAILQ_REMOVE(&hio_##name##_list[(ncomp)], (hio),		\
185 	    hio_next[(ncomp)]);						\
186 	mtx_unlock(&hio_##name##_list_lock[(ncomp)]);			\
187 } while (0)
188 #define	QUEUE_TAKE2(hio, name)	do {					\
189 	mtx_lock(&hio_##name##_list_lock);				\
190 	while (((hio) = TAILQ_FIRST(&hio_##name##_list)) == NULL) {	\
191 		cv_wait(&hio_##name##_list_cond,			\
192 		    &hio_##name##_list_lock);				\
193 	}								\
194 	TAILQ_REMOVE(&hio_##name##_list, (hio), hio_##name##_next);	\
195 	mtx_unlock(&hio_##name##_list_lock);				\
196 } while (0)
197 
198 #define	SYNCREQ(hio)		do {					\
199 	(hio)->hio_ggio.gctl_unit = -1;					\
200 	(hio)->hio_ggio.gctl_seq = 1;					\
201 } while (0)
202 #define	ISSYNCREQ(hio)		((hio)->hio_ggio.gctl_unit == -1)
203 #define	SYNCREQDONE(hio)	do { (hio)->hio_ggio.gctl_unit = -2; } while (0)
204 #define	ISSYNCREQDONE(hio)	((hio)->hio_ggio.gctl_unit == -2)
205 
206 static struct hast_resource *gres;
207 
208 static pthread_mutex_t range_lock;
209 static struct rangelocks *range_regular;
210 static bool range_regular_wait;
211 static pthread_cond_t range_regular_cond;
212 static struct rangelocks *range_sync;
213 static bool range_sync_wait;
214 static pthread_cond_t range_sync_cond;
215 
216 static void *ggate_recv_thread(void *arg);
217 static void *local_send_thread(void *arg);
218 static void *remote_send_thread(void *arg);
219 static void *remote_recv_thread(void *arg);
220 static void *ggate_send_thread(void *arg);
221 static void *sync_thread(void *arg);
222 static void *guard_thread(void *arg);
223 
224 static void sighandler(int sig);
225 
226 static void
227 cleanup(struct hast_resource *res)
228 {
229 	int rerrno;
230 
231 	/* Remember errno. */
232 	rerrno = errno;
233 
234 	/*
235 	 * Close descriptor to /dev/hast/<name>
236 	 * to work-around race in the kernel.
237 	 */
238 	close(res->hr_localfd);
239 
240 	/* Destroy ggate provider if we created one. */
241 	if (res->hr_ggateunit >= 0) {
242 		struct g_gate_ctl_destroy ggiod;
243 
244 		ggiod.gctl_version = G_GATE_VERSION;
245 		ggiod.gctl_unit = res->hr_ggateunit;
246 		ggiod.gctl_force = 1;
247 		if (ioctl(res->hr_ggatefd, G_GATE_CMD_DESTROY, &ggiod) < 0) {
248 			pjdlog_warning("Unable to destroy hast/%s device",
249 			    res->hr_provname);
250 		}
251 		res->hr_ggateunit = -1;
252 	}
253 
254 	/* Restore errno. */
255 	errno = rerrno;
256 }
257 
258 static void
259 primary_exit(int exitcode, const char *fmt, ...)
260 {
261 	va_list ap;
262 
263 	assert(exitcode != EX_OK);
264 	va_start(ap, fmt);
265 	pjdlogv_errno(LOG_ERR, fmt, ap);
266 	va_end(ap);
267 	cleanup(gres);
268 	exit(exitcode);
269 }
270 
271 static void
272 primary_exitx(int exitcode, const char *fmt, ...)
273 {
274 	va_list ap;
275 
276 	va_start(ap, fmt);
277 	pjdlogv(exitcode == EX_OK ? LOG_INFO : LOG_ERR, fmt, ap);
278 	va_end(ap);
279 	cleanup(gres);
280 	exit(exitcode);
281 }
282 
283 static int
284 hast_activemap_flush(struct hast_resource *res)
285 {
286 	const unsigned char *buf;
287 	size_t size;
288 
289 	buf = activemap_bitmap(res->hr_amp, &size);
290 	assert(buf != NULL);
291 	assert((size % res->hr_local_sectorsize) == 0);
292 	if (pwrite(res->hr_localfd, buf, size, METADATA_SIZE) !=
293 	    (ssize_t)size) {
294 		KEEP_ERRNO(pjdlog_errno(LOG_ERR,
295 		    "Unable to flush activemap to disk"));
296 		return (-1);
297 	}
298 	return (0);
299 }
300 
301 static void
302 init_environment(struct hast_resource *res __unused)
303 {
304 	struct hio *hio;
305 	unsigned int ii, ncomps;
306 
307 	/*
308 	 * In the future it might be per-resource value.
309 	 */
310 	ncomps = HAST_NCOMPONENTS;
311 
312 	/*
313 	 * Allocate memory needed by lists.
314 	 */
315 	hio_send_list = malloc(sizeof(hio_send_list[0]) * ncomps);
316 	if (hio_send_list == NULL) {
317 		primary_exitx(EX_TEMPFAIL,
318 		    "Unable to allocate %zu bytes of memory for send lists.",
319 		    sizeof(hio_send_list[0]) * ncomps);
320 	}
321 	hio_send_list_lock = malloc(sizeof(hio_send_list_lock[0]) * ncomps);
322 	if (hio_send_list_lock == NULL) {
323 		primary_exitx(EX_TEMPFAIL,
324 		    "Unable to allocate %zu bytes of memory for send list locks.",
325 		    sizeof(hio_send_list_lock[0]) * ncomps);
326 	}
327 	hio_send_list_cond = malloc(sizeof(hio_send_list_cond[0]) * ncomps);
328 	if (hio_send_list_cond == NULL) {
329 		primary_exitx(EX_TEMPFAIL,
330 		    "Unable to allocate %zu bytes of memory for send list condition variables.",
331 		    sizeof(hio_send_list_cond[0]) * ncomps);
332 	}
333 	hio_recv_list = malloc(sizeof(hio_recv_list[0]) * ncomps);
334 	if (hio_recv_list == NULL) {
335 		primary_exitx(EX_TEMPFAIL,
336 		    "Unable to allocate %zu bytes of memory for recv lists.",
337 		    sizeof(hio_recv_list[0]) * ncomps);
338 	}
339 	hio_recv_list_lock = malloc(sizeof(hio_recv_list_lock[0]) * ncomps);
340 	if (hio_recv_list_lock == NULL) {
341 		primary_exitx(EX_TEMPFAIL,
342 		    "Unable to allocate %zu bytes of memory for recv list locks.",
343 		    sizeof(hio_recv_list_lock[0]) * ncomps);
344 	}
345 	hio_recv_list_cond = malloc(sizeof(hio_recv_list_cond[0]) * ncomps);
346 	if (hio_recv_list_cond == NULL) {
347 		primary_exitx(EX_TEMPFAIL,
348 		    "Unable to allocate %zu bytes of memory for recv list condition variables.",
349 		    sizeof(hio_recv_list_cond[0]) * ncomps);
350 	}
351 	hio_remote_lock = malloc(sizeof(hio_remote_lock[0]) * ncomps);
352 	if (hio_remote_lock == NULL) {
353 		primary_exitx(EX_TEMPFAIL,
354 		    "Unable to allocate %zu bytes of memory for remote connections locks.",
355 		    sizeof(hio_remote_lock[0]) * ncomps);
356 	}
357 
358 	/*
359 	 * Initialize lists, their locks and theirs condition variables.
360 	 */
361 	TAILQ_INIT(&hio_free_list);
362 	mtx_init(&hio_free_list_lock);
363 	cv_init(&hio_free_list_cond);
364 	for (ii = 0; ii < HAST_NCOMPONENTS; ii++) {
365 		TAILQ_INIT(&hio_send_list[ii]);
366 		mtx_init(&hio_send_list_lock[ii]);
367 		cv_init(&hio_send_list_cond[ii]);
368 		TAILQ_INIT(&hio_recv_list[ii]);
369 		mtx_init(&hio_recv_list_lock[ii]);
370 		cv_init(&hio_recv_list_cond[ii]);
371 		rw_init(&hio_remote_lock[ii]);
372 	}
373 	TAILQ_INIT(&hio_done_list);
374 	mtx_init(&hio_done_list_lock);
375 	cv_init(&hio_done_list_cond);
376 	mtx_init(&hio_guard_lock);
377 	cv_init(&hio_guard_cond);
378 	mtx_init(&metadata_lock);
379 
380 	/*
381 	 * Allocate requests pool and initialize requests.
382 	 */
383 	for (ii = 0; ii < HAST_HIO_MAX; ii++) {
384 		hio = malloc(sizeof(*hio));
385 		if (hio == NULL) {
386 			primary_exitx(EX_TEMPFAIL,
387 			    "Unable to allocate %zu bytes of memory for hio request.",
388 			    sizeof(*hio));
389 		}
390 		hio->hio_countdown = 0;
391 		hio->hio_errors = malloc(sizeof(hio->hio_errors[0]) * ncomps);
392 		if (hio->hio_errors == NULL) {
393 			primary_exitx(EX_TEMPFAIL,
394 			    "Unable allocate %zu bytes of memory for hio errors.",
395 			    sizeof(hio->hio_errors[0]) * ncomps);
396 		}
397 		hio->hio_next = malloc(sizeof(hio->hio_next[0]) * ncomps);
398 		if (hio->hio_next == NULL) {
399 			primary_exitx(EX_TEMPFAIL,
400 			    "Unable allocate %zu bytes of memory for hio_next field.",
401 			    sizeof(hio->hio_next[0]) * ncomps);
402 		}
403 		hio->hio_ggio.gctl_version = G_GATE_VERSION;
404 		hio->hio_ggio.gctl_data = malloc(MAXPHYS);
405 		if (hio->hio_ggio.gctl_data == NULL) {
406 			primary_exitx(EX_TEMPFAIL,
407 			    "Unable to allocate %zu bytes of memory for gctl_data.",
408 			    MAXPHYS);
409 		}
410 		hio->hio_ggio.gctl_length = MAXPHYS;
411 		hio->hio_ggio.gctl_error = 0;
412 		TAILQ_INSERT_HEAD(&hio_free_list, hio, hio_free_next);
413 	}
414 
415 	/*
416 	 * Turn on signals handling.
417 	 */
418 	signal(SIGINT, sighandler);
419 	signal(SIGTERM, sighandler);
420 }
421 
422 static void
423 init_local(struct hast_resource *res)
424 {
425 	unsigned char *buf;
426 	size_t mapsize;
427 
428 	if (metadata_read(res, true) < 0)
429 		exit(EX_NOINPUT);
430 	mtx_init(&res->hr_amp_lock);
431 	if (activemap_init(&res->hr_amp, res->hr_datasize, res->hr_extentsize,
432 	    res->hr_local_sectorsize, res->hr_keepdirty) < 0) {
433 		primary_exit(EX_TEMPFAIL, "Unable to create activemap");
434 	}
435 	mtx_init(&range_lock);
436 	cv_init(&range_regular_cond);
437 	if (rangelock_init(&range_regular) < 0)
438 		primary_exit(EX_TEMPFAIL, "Unable to create regular range lock");
439 	cv_init(&range_sync_cond);
440 	if (rangelock_init(&range_sync) < 0)
441 		primary_exit(EX_TEMPFAIL, "Unable to create sync range lock");
442 	mapsize = activemap_ondisk_size(res->hr_amp);
443 	buf = calloc(1, mapsize);
444 	if (buf == NULL) {
445 		primary_exitx(EX_TEMPFAIL,
446 		    "Unable to allocate buffer for activemap.");
447 	}
448 	if (pread(res->hr_localfd, buf, mapsize, METADATA_SIZE) !=
449 	    (ssize_t)mapsize) {
450 		primary_exit(EX_NOINPUT, "Unable to read activemap");
451 	}
452 	activemap_copyin(res->hr_amp, buf, mapsize);
453 	free(buf);
454 	if (res->hr_resuid != 0)
455 		return;
456 	/*
457 	 * We're using provider for the first time, so we have to generate
458 	 * resource unique identifier and initialize local and remote counts.
459 	 */
460 	arc4random_buf(&res->hr_resuid, sizeof(res->hr_resuid));
461 	res->hr_primary_localcnt = 1;
462 	res->hr_primary_remotecnt = 0;
463 	if (metadata_write(res) < 0)
464 		exit(EX_NOINPUT);
465 }
466 
467 static bool
468 init_remote(struct hast_resource *res, struct proto_conn **inp,
469     struct proto_conn **outp)
470 {
471 	struct proto_conn *in, *out;
472 	struct nv *nvout, *nvin;
473 	const unsigned char *token;
474 	unsigned char *map;
475 	const char *errmsg;
476 	int32_t extentsize;
477 	int64_t datasize;
478 	uint32_t mapsize;
479 	size_t size;
480 
481 	assert((inp == NULL && outp == NULL) || (inp != NULL && outp != NULL));
482 
483 	in = out = NULL;
484 
485 	/* Prepare outgoing connection with remote node. */
486 	if (proto_client(res->hr_remoteaddr, &out) < 0) {
487 		primary_exit(EX_TEMPFAIL, "Unable to create connection to %s",
488 		    res->hr_remoteaddr);
489 	}
490 	/* Try to connect, but accept failure. */
491 	if (proto_connect(out) < 0) {
492 		pjdlog_errno(LOG_WARNING, "Unable to connect to %s",
493 		    res->hr_remoteaddr);
494 		goto close;
495 	}
496 	/* Error in setting timeout is not critical, but why should it fail? */
497 	if (proto_timeout(out, res->hr_timeout) < 0)
498 		pjdlog_errno(LOG_WARNING, "Unable to set connection timeout");
499 	/*
500 	 * First handshake step.
501 	 * Setup outgoing connection with remote node.
502 	 */
503 	nvout = nv_alloc();
504 	nv_add_string(nvout, res->hr_name, "resource");
505 	if (nv_error(nvout) != 0) {
506 		pjdlog_common(LOG_WARNING, 0, nv_error(nvout),
507 		    "Unable to allocate header for connection with %s",
508 		    res->hr_remoteaddr);
509 		nv_free(nvout);
510 		goto close;
511 	}
512 	if (hast_proto_send(res, out, nvout, NULL, 0) < 0) {
513 		pjdlog_errno(LOG_WARNING,
514 		    "Unable to send handshake header to %s",
515 		    res->hr_remoteaddr);
516 		nv_free(nvout);
517 		goto close;
518 	}
519 	nv_free(nvout);
520 	if (hast_proto_recv_hdr(out, &nvin) < 0) {
521 		pjdlog_errno(LOG_WARNING,
522 		    "Unable to receive handshake header from %s",
523 		    res->hr_remoteaddr);
524 		goto close;
525 	}
526 	errmsg = nv_get_string(nvin, "errmsg");
527 	if (errmsg != NULL) {
528 		pjdlog_warning("%s", errmsg);
529 		nv_free(nvin);
530 		goto close;
531 	}
532 	token = nv_get_uint8_array(nvin, &size, "token");
533 	if (token == NULL) {
534 		pjdlog_warning("Handshake header from %s has no 'token' field.",
535 		    res->hr_remoteaddr);
536 		nv_free(nvin);
537 		goto close;
538 	}
539 	if (size != sizeof(res->hr_token)) {
540 		pjdlog_warning("Handshake header from %s contains 'token' of wrong size (got %zu, expected %zu).",
541 		    res->hr_remoteaddr, size, sizeof(res->hr_token));
542 		nv_free(nvin);
543 		goto close;
544 	}
545 	bcopy(token, res->hr_token, sizeof(res->hr_token));
546 	nv_free(nvin);
547 
548 	/*
549 	 * Second handshake step.
550 	 * Setup incoming connection with remote node.
551 	 */
552 	if (proto_client(res->hr_remoteaddr, &in) < 0) {
553 		pjdlog_errno(LOG_WARNING, "Unable to create connection to %s",
554 		    res->hr_remoteaddr);
555 	}
556 	/* Try to connect, but accept failure. */
557 	if (proto_connect(in) < 0) {
558 		pjdlog_errno(LOG_WARNING, "Unable to connect to %s",
559 		    res->hr_remoteaddr);
560 		goto close;
561 	}
562 	/* Error in setting timeout is not critical, but why should it fail? */
563 	if (proto_timeout(in, res->hr_timeout) < 0)
564 		pjdlog_errno(LOG_WARNING, "Unable to set connection timeout");
565 	nvout = nv_alloc();
566 	nv_add_string(nvout, res->hr_name, "resource");
567 	nv_add_uint8_array(nvout, res->hr_token, sizeof(res->hr_token),
568 	    "token");
569 	nv_add_uint64(nvout, res->hr_resuid, "resuid");
570 	nv_add_uint64(nvout, res->hr_primary_localcnt, "localcnt");
571 	nv_add_uint64(nvout, res->hr_primary_remotecnt, "remotecnt");
572 	if (nv_error(nvout) != 0) {
573 		pjdlog_common(LOG_WARNING, 0, nv_error(nvout),
574 		    "Unable to allocate header for connection with %s",
575 		    res->hr_remoteaddr);
576 		nv_free(nvout);
577 		goto close;
578 	}
579 	if (hast_proto_send(res, in, nvout, NULL, 0) < 0) {
580 		pjdlog_errno(LOG_WARNING,
581 		    "Unable to send handshake header to %s",
582 		    res->hr_remoteaddr);
583 		nv_free(nvout);
584 		goto close;
585 	}
586 	nv_free(nvout);
587 	if (hast_proto_recv_hdr(out, &nvin) < 0) {
588 		pjdlog_errno(LOG_WARNING,
589 		    "Unable to receive handshake header from %s",
590 		    res->hr_remoteaddr);
591 		goto close;
592 	}
593 	errmsg = nv_get_string(nvin, "errmsg");
594 	if (errmsg != NULL) {
595 		pjdlog_warning("%s", errmsg);
596 		nv_free(nvin);
597 		goto close;
598 	}
599 	datasize = nv_get_int64(nvin, "datasize");
600 	if (datasize != res->hr_datasize) {
601 		pjdlog_warning("Data size differs between nodes (local=%jd, remote=%jd).",
602 		    (intmax_t)res->hr_datasize, (intmax_t)datasize);
603 		nv_free(nvin);
604 		goto close;
605 	}
606 	extentsize = nv_get_int32(nvin, "extentsize");
607 	if (extentsize != res->hr_extentsize) {
608 		pjdlog_warning("Extent size differs between nodes (local=%zd, remote=%zd).",
609 		    (ssize_t)res->hr_extentsize, (ssize_t)extentsize);
610 		nv_free(nvin);
611 		goto close;
612 	}
613 	res->hr_secondary_localcnt = nv_get_uint64(nvin, "localcnt");
614 	res->hr_secondary_remotecnt = nv_get_uint64(nvin, "remotecnt");
615 	res->hr_syncsrc = nv_get_uint8(nvin, "syncsrc");
616 	map = NULL;
617 	mapsize = nv_get_uint32(nvin, "mapsize");
618 	if (mapsize > 0) {
619 		map = malloc(mapsize);
620 		if (map == NULL) {
621 			pjdlog_error("Unable to allocate memory for remote activemap (mapsize=%ju).",
622 			    (uintmax_t)mapsize);
623 			nv_free(nvin);
624 			goto close;
625 		}
626 		/*
627 		 * Remote node have some dirty extents on its own, lets
628 		 * download its activemap.
629 		 */
630 		if (hast_proto_recv_data(res, out, nvin, map,
631 		    mapsize) < 0) {
632 			pjdlog_errno(LOG_ERR,
633 			    "Unable to receive remote activemap");
634 			nv_free(nvin);
635 			free(map);
636 			goto close;
637 		}
638 		/*
639 		 * Merge local and remote bitmaps.
640 		 */
641 		activemap_merge(res->hr_amp, map, mapsize);
642 		free(map);
643 		/*
644 		 * Now that we merged bitmaps from both nodes, flush it to the
645 		 * disk before we start to synchronize.
646 		 */
647 		(void)hast_activemap_flush(res);
648 	}
649 	pjdlog_info("Connected to %s.", res->hr_remoteaddr);
650 	if (inp != NULL && outp != NULL) {
651 		*inp = in;
652 		*outp = out;
653 	} else {
654 		res->hr_remotein = in;
655 		res->hr_remoteout = out;
656 	}
657 	return (true);
658 close:
659 	proto_close(out);
660 	if (in != NULL)
661 		proto_close(in);
662 	return (false);
663 }
664 
665 static void
666 sync_start(void)
667 {
668 
669 	mtx_lock(&sync_lock);
670 	sync_inprogress = true;
671 	mtx_unlock(&sync_lock);
672 	cv_signal(&sync_cond);
673 }
674 
675 static void
676 init_ggate(struct hast_resource *res)
677 {
678 	struct g_gate_ctl_create ggiocreate;
679 	struct g_gate_ctl_cancel ggiocancel;
680 
681 	/*
682 	 * We communicate with ggate via /dev/ggctl. Open it.
683 	 */
684 	res->hr_ggatefd = open("/dev/" G_GATE_CTL_NAME, O_RDWR);
685 	if (res->hr_ggatefd < 0)
686 		primary_exit(EX_OSFILE, "Unable to open /dev/" G_GATE_CTL_NAME);
687 	/*
688 	 * Create provider before trying to connect, as connection failure
689 	 * is not critical, but may take some time.
690 	 */
691 	ggiocreate.gctl_version = G_GATE_VERSION;
692 	ggiocreate.gctl_mediasize = res->hr_datasize;
693 	ggiocreate.gctl_sectorsize = res->hr_local_sectorsize;
694 	ggiocreate.gctl_flags = 0;
695 	ggiocreate.gctl_maxcount = G_GATE_MAX_QUEUE_SIZE;
696 	ggiocreate.gctl_timeout = 0;
697 	ggiocreate.gctl_unit = G_GATE_NAME_GIVEN;
698 	snprintf(ggiocreate.gctl_name, sizeof(ggiocreate.gctl_name), "hast/%s",
699 	    res->hr_provname);
700 	bzero(ggiocreate.gctl_info, sizeof(ggiocreate.gctl_info));
701 	if (ioctl(res->hr_ggatefd, G_GATE_CMD_CREATE, &ggiocreate) == 0) {
702 		pjdlog_info("Device hast/%s created.", res->hr_provname);
703 		res->hr_ggateunit = ggiocreate.gctl_unit;
704 		return;
705 	}
706 	if (errno != EEXIST) {
707 		primary_exit(EX_OSERR, "Unable to create hast/%s device",
708 		    res->hr_provname);
709 	}
710 	pjdlog_debug(1,
711 	    "Device hast/%s already exists, we will try to take it over.",
712 	    res->hr_provname);
713 	/*
714 	 * If we received EEXIST, we assume that the process who created the
715 	 * provider died and didn't clean up. In that case we will start from
716 	 * where he left of.
717 	 */
718 	ggiocancel.gctl_version = G_GATE_VERSION;
719 	ggiocancel.gctl_unit = G_GATE_NAME_GIVEN;
720 	snprintf(ggiocancel.gctl_name, sizeof(ggiocancel.gctl_name), "hast/%s",
721 	    res->hr_provname);
722 	if (ioctl(res->hr_ggatefd, G_GATE_CMD_CANCEL, &ggiocancel) == 0) {
723 		pjdlog_info("Device hast/%s recovered.", res->hr_provname);
724 		res->hr_ggateunit = ggiocancel.gctl_unit;
725 		return;
726 	}
727 	primary_exit(EX_OSERR, "Unable to take over hast/%s device",
728 	    res->hr_provname);
729 }
730 
731 void
732 hastd_primary(struct hast_resource *res)
733 {
734 	pthread_t td;
735 	pid_t pid;
736 	int error;
737 
738 	gres = res;
739 
740 	/*
741 	 * Create communication channel between parent and child.
742 	 */
743 	if (proto_client("socketpair://", &res->hr_ctrl) < 0) {
744 		KEEP_ERRNO((void)pidfile_remove(pfh));
745 		primary_exit(EX_OSERR,
746 		    "Unable to create control sockets between parent and child");
747 	}
748 
749 	pid = fork();
750 	if (pid < 0) {
751 		KEEP_ERRNO((void)pidfile_remove(pfh));
752 		primary_exit(EX_TEMPFAIL, "Unable to fork");
753 	}
754 
755 	if (pid > 0) {
756 		/* This is parent. */
757 		res->hr_workerpid = pid;
758 		return;
759 	}
760 	(void)pidfile_close(pfh);
761 
762 	setproctitle("%s (primary)", res->hr_name);
763 
764 	init_local(res);
765 	if (init_remote(res, NULL, NULL))
766 		sync_start();
767 	init_ggate(res);
768 	init_environment(res);
769 	error = pthread_create(&td, NULL, ggate_recv_thread, res);
770 	assert(error == 0);
771 	error = pthread_create(&td, NULL, local_send_thread, res);
772 	assert(error == 0);
773 	error = pthread_create(&td, NULL, remote_send_thread, res);
774 	assert(error == 0);
775 	error = pthread_create(&td, NULL, remote_recv_thread, res);
776 	assert(error == 0);
777 	error = pthread_create(&td, NULL, ggate_send_thread, res);
778 	assert(error == 0);
779 	error = pthread_create(&td, NULL, sync_thread, res);
780 	assert(error == 0);
781 	error = pthread_create(&td, NULL, ctrl_thread, res);
782 	assert(error == 0);
783 	(void)guard_thread(res);
784 }
785 
786 static void
787 reqlog(int loglevel, int debuglevel, struct g_gate_ctl_io *ggio, const char *fmt, ...)
788 {
789 	char msg[1024];
790 	va_list ap;
791 	int len;
792 
793 	va_start(ap, fmt);
794 	len = vsnprintf(msg, sizeof(msg), fmt, ap);
795 	va_end(ap);
796 	if ((size_t)len < sizeof(msg)) {
797 		switch (ggio->gctl_cmd) {
798 		case BIO_READ:
799 			(void)snprintf(msg + len, sizeof(msg) - len,
800 			    "READ(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
801 			    (uintmax_t)ggio->gctl_length);
802 			break;
803 		case BIO_DELETE:
804 			(void)snprintf(msg + len, sizeof(msg) - len,
805 			    "DELETE(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
806 			    (uintmax_t)ggio->gctl_length);
807 			break;
808 		case BIO_FLUSH:
809 			(void)snprintf(msg + len, sizeof(msg) - len, "FLUSH.");
810 			break;
811 		case BIO_WRITE:
812 			(void)snprintf(msg + len, sizeof(msg) - len,
813 			    "WRITE(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
814 			    (uintmax_t)ggio->gctl_length);
815 			break;
816 		default:
817 			(void)snprintf(msg + len, sizeof(msg) - len,
818 			    "UNKNOWN(%u).", (unsigned int)ggio->gctl_cmd);
819 			break;
820 		}
821 	}
822 	pjdlog_common(loglevel, debuglevel, -1, "%s", msg);
823 }
824 
825 static void
826 remote_close(struct hast_resource *res, int ncomp)
827 {
828 
829 	rw_wlock(&hio_remote_lock[ncomp]);
830 	/*
831 	 * A race is possible between dropping rlock and acquiring wlock -
832 	 * another thread can close connection in-between.
833 	 */
834 	if (!ISCONNECTED(res, ncomp)) {
835 		assert(res->hr_remotein == NULL);
836 		assert(res->hr_remoteout == NULL);
837 		rw_unlock(&hio_remote_lock[ncomp]);
838 		return;
839 	}
840 
841 	assert(res->hr_remotein != NULL);
842 	assert(res->hr_remoteout != NULL);
843 
844 	pjdlog_debug(2, "Closing old incoming connection to %s.",
845 	    res->hr_remoteaddr);
846 	proto_close(res->hr_remotein);
847 	res->hr_remotein = NULL;
848 	pjdlog_debug(2, "Closing old outgoing connection to %s.",
849 	    res->hr_remoteaddr);
850 	proto_close(res->hr_remoteout);
851 	res->hr_remoteout = NULL;
852 
853 	rw_unlock(&hio_remote_lock[ncomp]);
854 
855 	/*
856 	 * Stop synchronization if in-progress.
857 	 */
858 	mtx_lock(&sync_lock);
859 	if (sync_inprogress)
860 		sync_inprogress = false;
861 	mtx_unlock(&sync_lock);
862 
863 	/*
864 	 * Wake up guard thread, so it can immediately start reconnect.
865 	 */
866 	mtx_lock(&hio_guard_lock);
867 	cv_signal(&hio_guard_cond);
868 	mtx_unlock(&hio_guard_lock);
869 }
870 
871 /*
872  * Thread receives ggate I/O requests from the kernel and passes them to
873  * appropriate threads:
874  * WRITE - always goes to both local_send and remote_send threads
875  * READ (when the block is up-to-date on local component) -
876  *	only local_send thread
877  * READ (when the block isn't up-to-date on local component) -
878  *	only remote_send thread
879  * DELETE - always goes to both local_send and remote_send threads
880  * FLUSH - always goes to both local_send and remote_send threads
881  */
882 static void *
883 ggate_recv_thread(void *arg)
884 {
885 	struct hast_resource *res = arg;
886 	struct g_gate_ctl_io *ggio;
887 	struct hio *hio;
888 	unsigned int ii, ncomp, ncomps;
889 	int error;
890 
891 	ncomps = HAST_NCOMPONENTS;
892 
893 	for (;;) {
894 		pjdlog_debug(2, "ggate_recv: Taking free request.");
895 		QUEUE_TAKE2(hio, free);
896 		pjdlog_debug(2, "ggate_recv: (%p) Got free request.", hio);
897 		ggio = &hio->hio_ggio;
898 		ggio->gctl_unit = res->hr_ggateunit;
899 		ggio->gctl_length = MAXPHYS;
900 		ggio->gctl_error = 0;
901 		pjdlog_debug(2,
902 		    "ggate_recv: (%p) Waiting for request from the kernel.",
903 		    hio);
904 		if (ioctl(res->hr_ggatefd, G_GATE_CMD_START, ggio) < 0) {
905 			if (sigexit_received)
906 				pthread_exit(NULL);
907 			primary_exit(EX_OSERR, "G_GATE_CMD_START failed");
908 		}
909 		error = ggio->gctl_error;
910 		switch (error) {
911 		case 0:
912 			break;
913 		case ECANCELED:
914 			/* Exit gracefully. */
915 			if (!sigexit_received) {
916 				pjdlog_debug(2,
917 				    "ggate_recv: (%p) Received cancel from the kernel.",
918 				    hio);
919 				pjdlog_info("Received cancel from the kernel, exiting.");
920 			}
921 			pthread_exit(NULL);
922 		case ENOMEM:
923 			/*
924 			 * Buffer too small? Impossible, we allocate MAXPHYS
925 			 * bytes - request can't be bigger than that.
926 			 */
927 			/* FALLTHROUGH */
928 		case ENXIO:
929 		default:
930 			primary_exitx(EX_OSERR, "G_GATE_CMD_START failed: %s.",
931 			    strerror(error));
932 		}
933 		for (ii = 0; ii < ncomps; ii++)
934 			hio->hio_errors[ii] = EINVAL;
935 		reqlog(LOG_DEBUG, 2, ggio,
936 		    "ggate_recv: (%p) Request received from the kernel: ",
937 		    hio);
938 		/*
939 		 * Inform all components about new write request.
940 		 * For read request prefer local component unless the given
941 		 * range is out-of-date, then use remote component.
942 		 */
943 		switch (ggio->gctl_cmd) {
944 		case BIO_READ:
945 			pjdlog_debug(2,
946 			    "ggate_recv: (%p) Moving request to the send queue.",
947 			    hio);
948 			refcount_init(&hio->hio_countdown, 1);
949 			mtx_lock(&metadata_lock);
950 			if (res->hr_syncsrc == HAST_SYNCSRC_UNDEF ||
951 			    res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
952 				/*
953 				 * This range is up-to-date on local component,
954 				 * so handle request locally.
955 				 */
956 				 /* Local component is 0 for now. */
957 				ncomp = 0;
958 			} else /* if (res->hr_syncsrc ==
959 			    HAST_SYNCSRC_SECONDARY) */ {
960 				assert(res->hr_syncsrc ==
961 				    HAST_SYNCSRC_SECONDARY);
962 				/*
963 				 * This range is out-of-date on local component,
964 				 * so send request to the remote node.
965 				 */
966 				 /* Remote component is 1 for now. */
967 				ncomp = 1;
968 			}
969 			mtx_unlock(&metadata_lock);
970 			QUEUE_INSERT1(hio, send, ncomp);
971 			break;
972 		case BIO_WRITE:
973 			for (;;) {
974 				mtx_lock(&range_lock);
975 				if (rangelock_islocked(range_sync,
976 				    ggio->gctl_offset, ggio->gctl_length)) {
977 					pjdlog_debug(2,
978 					    "regular: Range offset=%jd length=%zu locked.",
979 					    (intmax_t)ggio->gctl_offset,
980 					    (size_t)ggio->gctl_length);
981 					range_regular_wait = true;
982 					cv_wait(&range_regular_cond, &range_lock);
983 					range_regular_wait = false;
984 					mtx_unlock(&range_lock);
985 					continue;
986 				}
987 				if (rangelock_add(range_regular,
988 				    ggio->gctl_offset, ggio->gctl_length) < 0) {
989 					mtx_unlock(&range_lock);
990 					pjdlog_debug(2,
991 					    "regular: Range offset=%jd length=%zu is already locked, waiting.",
992 					    (intmax_t)ggio->gctl_offset,
993 					    (size_t)ggio->gctl_length);
994 					sleep(1);
995 					continue;
996 				}
997 				mtx_unlock(&range_lock);
998 				break;
999 			}
1000 			mtx_lock(&res->hr_amp_lock);
1001 			if (activemap_write_start(res->hr_amp,
1002 			    ggio->gctl_offset, ggio->gctl_length)) {
1003 				(void)hast_activemap_flush(res);
1004 			}
1005 			mtx_unlock(&res->hr_amp_lock);
1006 			/* FALLTHROUGH */
1007 		case BIO_DELETE:
1008 		case BIO_FLUSH:
1009 			pjdlog_debug(2,
1010 			    "ggate_recv: (%p) Moving request to the send queues.",
1011 			    hio);
1012 			refcount_init(&hio->hio_countdown, ncomps);
1013 			for (ii = 0; ii < ncomps; ii++)
1014 				QUEUE_INSERT1(hio, send, ii);
1015 			break;
1016 		}
1017 	}
1018 	/* NOTREACHED */
1019 	return (NULL);
1020 }
1021 
1022 /*
1023  * Thread reads from or writes to local component.
1024  * If local read fails, it redirects it to remote_send thread.
1025  */
1026 static void *
1027 local_send_thread(void *arg)
1028 {
1029 	struct hast_resource *res = arg;
1030 	struct g_gate_ctl_io *ggio;
1031 	struct hio *hio;
1032 	unsigned int ncomp, rncomp;
1033 	ssize_t ret;
1034 
1035 	/* Local component is 0 for now. */
1036 	ncomp = 0;
1037 	/* Remote component is 1 for now. */
1038 	rncomp = 1;
1039 
1040 	for (;;) {
1041 		pjdlog_debug(2, "local_send: Taking request.");
1042 		QUEUE_TAKE1(hio, send, ncomp);
1043 		pjdlog_debug(2, "local_send: (%p) Got request.", hio);
1044 		ggio = &hio->hio_ggio;
1045 		switch (ggio->gctl_cmd) {
1046 		case BIO_READ:
1047 			ret = pread(res->hr_localfd, ggio->gctl_data,
1048 			    ggio->gctl_length,
1049 			    ggio->gctl_offset + res->hr_localoff);
1050 			if (ret == ggio->gctl_length)
1051 				hio->hio_errors[ncomp] = 0;
1052 			else {
1053 				/*
1054 				 * If READ failed, try to read from remote node.
1055 				 */
1056 				QUEUE_INSERT1(hio, send, rncomp);
1057 				continue;
1058 			}
1059 			break;
1060 		case BIO_WRITE:
1061 			ret = pwrite(res->hr_localfd, ggio->gctl_data,
1062 			    ggio->gctl_length,
1063 			    ggio->gctl_offset + res->hr_localoff);
1064 			if (ret < 0)
1065 				hio->hio_errors[ncomp] = errno;
1066 			else if (ret != ggio->gctl_length)
1067 				hio->hio_errors[ncomp] = EIO;
1068 			else
1069 				hio->hio_errors[ncomp] = 0;
1070 			break;
1071 		case BIO_DELETE:
1072 			ret = g_delete(res->hr_localfd,
1073 			    ggio->gctl_offset + res->hr_localoff,
1074 			    ggio->gctl_length);
1075 			if (ret < 0)
1076 				hio->hio_errors[ncomp] = errno;
1077 			else
1078 				hio->hio_errors[ncomp] = 0;
1079 			break;
1080 		case BIO_FLUSH:
1081 			ret = g_flush(res->hr_localfd);
1082 			if (ret < 0)
1083 				hio->hio_errors[ncomp] = errno;
1084 			else
1085 				hio->hio_errors[ncomp] = 0;
1086 			break;
1087 		}
1088 		if (refcount_release(&hio->hio_countdown)) {
1089 			if (ISSYNCREQ(hio)) {
1090 				mtx_lock(&sync_lock);
1091 				SYNCREQDONE(hio);
1092 				mtx_unlock(&sync_lock);
1093 				cv_signal(&sync_cond);
1094 			} else {
1095 				pjdlog_debug(2,
1096 				    "local_send: (%p) Moving request to the done queue.",
1097 				    hio);
1098 				QUEUE_INSERT2(hio, done);
1099 			}
1100 		}
1101 	}
1102 	/* NOTREACHED */
1103 	return (NULL);
1104 }
1105 
1106 /*
1107  * Thread sends request to secondary node.
1108  */
1109 static void *
1110 remote_send_thread(void *arg)
1111 {
1112 	struct hast_resource *res = arg;
1113 	struct g_gate_ctl_io *ggio;
1114 	struct hio *hio;
1115 	struct nv *nv;
1116 	unsigned int ncomp;
1117 	bool wakeup;
1118 	uint64_t offset, length;
1119 	uint8_t cmd;
1120 	void *data;
1121 
1122 	/* Remote component is 1 for now. */
1123 	ncomp = 1;
1124 
1125 	for (;;) {
1126 		pjdlog_debug(2, "remote_send: Taking request.");
1127 		QUEUE_TAKE1(hio, send, ncomp);
1128 		pjdlog_debug(2, "remote_send: (%p) Got request.", hio);
1129 		ggio = &hio->hio_ggio;
1130 		switch (ggio->gctl_cmd) {
1131 		case BIO_READ:
1132 			cmd = HIO_READ;
1133 			data = NULL;
1134 			offset = ggio->gctl_offset;
1135 			length = ggio->gctl_length;
1136 			break;
1137 		case BIO_WRITE:
1138 			cmd = HIO_WRITE;
1139 			data = ggio->gctl_data;
1140 			offset = ggio->gctl_offset;
1141 			length = ggio->gctl_length;
1142 			break;
1143 		case BIO_DELETE:
1144 			cmd = HIO_DELETE;
1145 			data = NULL;
1146 			offset = ggio->gctl_offset;
1147 			length = ggio->gctl_length;
1148 			break;
1149 		case BIO_FLUSH:
1150 			cmd = HIO_FLUSH;
1151 			data = NULL;
1152 			offset = 0;
1153 			length = 0;
1154 			break;
1155 		default:
1156 			assert(!"invalid condition");
1157 			abort();
1158 		}
1159 		nv = nv_alloc();
1160 		nv_add_uint8(nv, cmd, "cmd");
1161 		nv_add_uint64(nv, (uint64_t)ggio->gctl_seq, "seq");
1162 		nv_add_uint64(nv, offset, "offset");
1163 		nv_add_uint64(nv, length, "length");
1164 		if (nv_error(nv) != 0) {
1165 			hio->hio_errors[ncomp] = nv_error(nv);
1166 			pjdlog_debug(2,
1167 			    "remote_send: (%p) Unable to prepare header to send.",
1168 			    hio);
1169 			reqlog(LOG_ERR, 0, ggio,
1170 			    "Unable to prepare header to send (%s): ",
1171 			    strerror(nv_error(nv)));
1172 			/* Move failed request immediately to the done queue. */
1173 			goto done_queue;
1174 		}
1175 		pjdlog_debug(2,
1176 		    "remote_send: (%p) Moving request to the recv queue.",
1177 		    hio);
1178 		/*
1179 		 * Protect connection from disappearing.
1180 		 */
1181 		rw_rlock(&hio_remote_lock[ncomp]);
1182 		if (!ISCONNECTED(res, ncomp)) {
1183 			rw_unlock(&hio_remote_lock[ncomp]);
1184 			hio->hio_errors[ncomp] = ENOTCONN;
1185 			goto done_queue;
1186 		}
1187 		/*
1188 		 * Move the request to recv queue before sending it, because
1189 		 * in different order we can get reply before we move request
1190 		 * to recv queue.
1191 		 */
1192 		mtx_lock(&hio_recv_list_lock[ncomp]);
1193 		wakeup = TAILQ_EMPTY(&hio_recv_list[ncomp]);
1194 		TAILQ_INSERT_TAIL(&hio_recv_list[ncomp], hio, hio_next[ncomp]);
1195 		mtx_unlock(&hio_recv_list_lock[ncomp]);
1196 		if (hast_proto_send(res, res->hr_remoteout, nv, data,
1197 		    data != NULL ? length : 0) < 0) {
1198 			hio->hio_errors[ncomp] = errno;
1199 			rw_unlock(&hio_remote_lock[ncomp]);
1200 			remote_close(res, ncomp);
1201 			pjdlog_debug(2,
1202 			    "remote_send: (%p) Unable to send request.", hio);
1203 			reqlog(LOG_ERR, 0, ggio,
1204 			    "Unable to send request (%s): ",
1205 			    strerror(hio->hio_errors[ncomp]));
1206 			/*
1207 			 * Take request back from the receive queue and move
1208 			 * it immediately to the done queue.
1209 			 */
1210 			mtx_lock(&hio_recv_list_lock[ncomp]);
1211 			TAILQ_REMOVE(&hio_recv_list[ncomp], hio, hio_next[ncomp]);
1212 			mtx_unlock(&hio_recv_list_lock[ncomp]);
1213 			goto done_queue;
1214 		}
1215 		rw_unlock(&hio_remote_lock[ncomp]);
1216 		nv_free(nv);
1217 		if (wakeup)
1218 			cv_signal(&hio_recv_list_cond[ncomp]);
1219 		continue;
1220 done_queue:
1221 		nv_free(nv);
1222 		if (ISSYNCREQ(hio)) {
1223 			if (!refcount_release(&hio->hio_countdown))
1224 				continue;
1225 			mtx_lock(&sync_lock);
1226 			SYNCREQDONE(hio);
1227 			mtx_unlock(&sync_lock);
1228 			cv_signal(&sync_cond);
1229 			continue;
1230 		}
1231 		if (ggio->gctl_cmd == BIO_WRITE) {
1232 			mtx_lock(&res->hr_amp_lock);
1233 			if (activemap_need_sync(res->hr_amp, ggio->gctl_offset,
1234 			    ggio->gctl_length)) {
1235 				(void)hast_activemap_flush(res);
1236 			}
1237 			mtx_unlock(&res->hr_amp_lock);
1238 		}
1239 		if (!refcount_release(&hio->hio_countdown))
1240 			continue;
1241 		pjdlog_debug(2,
1242 		    "remote_send: (%p) Moving request to the done queue.",
1243 		    hio);
1244 		QUEUE_INSERT2(hio, done);
1245 	}
1246 	/* NOTREACHED */
1247 	return (NULL);
1248 }
1249 
1250 /*
1251  * Thread receives answer from secondary node and passes it to ggate_send
1252  * thread.
1253  */
1254 static void *
1255 remote_recv_thread(void *arg)
1256 {
1257 	struct hast_resource *res = arg;
1258 	struct g_gate_ctl_io *ggio;
1259 	struct hio *hio;
1260 	struct nv *nv;
1261 	unsigned int ncomp;
1262 	uint64_t seq;
1263 	int error;
1264 
1265 	/* Remote component is 1 for now. */
1266 	ncomp = 1;
1267 
1268 	for (;;) {
1269 		/* Wait until there is anything to receive. */
1270 		mtx_lock(&hio_recv_list_lock[ncomp]);
1271 		while (TAILQ_EMPTY(&hio_recv_list[ncomp])) {
1272 			pjdlog_debug(2, "remote_recv: No requests, waiting.");
1273 			cv_wait(&hio_recv_list_cond[ncomp],
1274 			    &hio_recv_list_lock[ncomp]);
1275 		}
1276 		mtx_unlock(&hio_recv_list_lock[ncomp]);
1277 		rw_rlock(&hio_remote_lock[ncomp]);
1278 		if (!ISCONNECTED(res, ncomp)) {
1279 			rw_unlock(&hio_remote_lock[ncomp]);
1280 			/*
1281 			 * Connection is dead, so move all pending requests to
1282 			 * the done queue (one-by-one).
1283 			 */
1284 			mtx_lock(&hio_recv_list_lock[ncomp]);
1285 			hio = TAILQ_FIRST(&hio_recv_list[ncomp]);
1286 			assert(hio != NULL);
1287 			TAILQ_REMOVE(&hio_recv_list[ncomp], hio,
1288 			    hio_next[ncomp]);
1289 			mtx_unlock(&hio_recv_list_lock[ncomp]);
1290 			goto done_queue;
1291 		}
1292 		if (hast_proto_recv_hdr(res->hr_remotein, &nv) < 0) {
1293 			pjdlog_errno(LOG_ERR,
1294 			    "Unable to receive reply header");
1295 			rw_unlock(&hio_remote_lock[ncomp]);
1296 			remote_close(res, ncomp);
1297 			continue;
1298 		}
1299 		rw_unlock(&hio_remote_lock[ncomp]);
1300 		seq = nv_get_uint64(nv, "seq");
1301 		if (seq == 0) {
1302 			pjdlog_error("Header contains no 'seq' field.");
1303 			nv_free(nv);
1304 			continue;
1305 		}
1306 		mtx_lock(&hio_recv_list_lock[ncomp]);
1307 		TAILQ_FOREACH(hio, &hio_recv_list[ncomp], hio_next[ncomp]) {
1308 			if (hio->hio_ggio.gctl_seq == seq) {
1309 				TAILQ_REMOVE(&hio_recv_list[ncomp], hio,
1310 				    hio_next[ncomp]);
1311 				break;
1312 			}
1313 		}
1314 		mtx_unlock(&hio_recv_list_lock[ncomp]);
1315 		if (hio == NULL) {
1316 			pjdlog_error("Found no request matching received 'seq' field (%ju).",
1317 			    (uintmax_t)seq);
1318 			nv_free(nv);
1319 			continue;
1320 		}
1321 		error = nv_get_int16(nv, "error");
1322 		if (error != 0) {
1323 			/* Request failed on remote side. */
1324 			hio->hio_errors[ncomp] = 0;
1325 			nv_free(nv);
1326 			goto done_queue;
1327 		}
1328 		ggio = &hio->hio_ggio;
1329 		switch (ggio->gctl_cmd) {
1330 		case BIO_READ:
1331 			rw_rlock(&hio_remote_lock[ncomp]);
1332 			if (!ISCONNECTED(res, ncomp)) {
1333 				rw_unlock(&hio_remote_lock[ncomp]);
1334 				nv_free(nv);
1335 				goto done_queue;
1336 			}
1337 			if (hast_proto_recv_data(res, res->hr_remotein, nv,
1338 			    ggio->gctl_data, ggio->gctl_length) < 0) {
1339 				hio->hio_errors[ncomp] = errno;
1340 				pjdlog_errno(LOG_ERR,
1341 				    "Unable to receive reply data");
1342 				rw_unlock(&hio_remote_lock[ncomp]);
1343 				nv_free(nv);
1344 				remote_close(res, ncomp);
1345 				goto done_queue;
1346 			}
1347 			rw_unlock(&hio_remote_lock[ncomp]);
1348 			break;
1349 		case BIO_WRITE:
1350 		case BIO_DELETE:
1351 		case BIO_FLUSH:
1352 			break;
1353 		default:
1354 			assert(!"invalid condition");
1355 			abort();
1356 		}
1357 		hio->hio_errors[ncomp] = 0;
1358 		nv_free(nv);
1359 done_queue:
1360 		if (refcount_release(&hio->hio_countdown)) {
1361 			if (ISSYNCREQ(hio)) {
1362 				mtx_lock(&sync_lock);
1363 				SYNCREQDONE(hio);
1364 				mtx_unlock(&sync_lock);
1365 				cv_signal(&sync_cond);
1366 			} else {
1367 				pjdlog_debug(2,
1368 				    "remote_recv: (%p) Moving request to the done queue.",
1369 				    hio);
1370 				QUEUE_INSERT2(hio, done);
1371 			}
1372 		}
1373 	}
1374 	/* NOTREACHED */
1375 	return (NULL);
1376 }
1377 
1378 /*
1379  * Thread sends answer to the kernel.
1380  */
1381 static void *
1382 ggate_send_thread(void *arg)
1383 {
1384 	struct hast_resource *res = arg;
1385 	struct g_gate_ctl_io *ggio;
1386 	struct hio *hio;
1387 	unsigned int ii, ncomp, ncomps;
1388 
1389 	ncomps = HAST_NCOMPONENTS;
1390 
1391 	for (;;) {
1392 		pjdlog_debug(2, "ggate_send: Taking request.");
1393 		QUEUE_TAKE2(hio, done);
1394 		pjdlog_debug(2, "ggate_send: (%p) Got request.", hio);
1395 		ggio = &hio->hio_ggio;
1396 		for (ii = 0; ii < ncomps; ii++) {
1397 			if (hio->hio_errors[ii] == 0) {
1398 				/*
1399 				 * One successful request is enough to declare
1400 				 * success.
1401 				 */
1402 				ggio->gctl_error = 0;
1403 				break;
1404 			}
1405 		}
1406 		if (ii == ncomps) {
1407 			/*
1408 			 * None of the requests were successful.
1409 			 * Use first error.
1410 			 */
1411 			ggio->gctl_error = hio->hio_errors[0];
1412 		}
1413 		if (ggio->gctl_error == 0 && ggio->gctl_cmd == BIO_WRITE) {
1414 			mtx_lock(&res->hr_amp_lock);
1415 			activemap_write_complete(res->hr_amp,
1416 			    ggio->gctl_offset, ggio->gctl_length);
1417 			mtx_unlock(&res->hr_amp_lock);
1418 		}
1419 		if (ggio->gctl_cmd == BIO_WRITE) {
1420 			/*
1421 			 * Unlock range we locked.
1422 			 */
1423 			mtx_lock(&range_lock);
1424 			rangelock_del(range_regular, ggio->gctl_offset,
1425 			    ggio->gctl_length);
1426 			if (range_sync_wait)
1427 				cv_signal(&range_sync_cond);
1428 			mtx_unlock(&range_lock);
1429 			/*
1430 			 * Bump local count if this is first write after
1431 			 * connection failure with remote node.
1432 			 */
1433 			ncomp = 1;
1434 			rw_rlock(&hio_remote_lock[ncomp]);
1435 			if (!ISCONNECTED(res, ncomp)) {
1436 				mtx_lock(&metadata_lock);
1437 				if (res->hr_primary_localcnt ==
1438 				    res->hr_secondary_remotecnt) {
1439 					res->hr_primary_localcnt++;
1440 					pjdlog_debug(1,
1441 					    "Increasing localcnt to %ju.",
1442 					    (uintmax_t)res->hr_primary_localcnt);
1443 					(void)metadata_write(res);
1444 				}
1445 				mtx_unlock(&metadata_lock);
1446 			}
1447 			rw_unlock(&hio_remote_lock[ncomp]);
1448 		}
1449 		if (ioctl(res->hr_ggatefd, G_GATE_CMD_DONE, ggio) < 0)
1450 			primary_exit(EX_OSERR, "G_GATE_CMD_DONE failed");
1451 		pjdlog_debug(2,
1452 		    "ggate_send: (%p) Moving request to the free queue.", hio);
1453 		QUEUE_INSERT2(hio, free);
1454 	}
1455 	/* NOTREACHED */
1456 	return (NULL);
1457 }
1458 
1459 /*
1460  * Thread synchronize local and remote components.
1461  */
1462 static void *
1463 sync_thread(void *arg __unused)
1464 {
1465 	struct hast_resource *res = arg;
1466 	struct hio *hio;
1467 	struct g_gate_ctl_io *ggio;
1468 	unsigned int ii, ncomp, ncomps;
1469 	off_t offset, length, synced;
1470 	bool dorewind;
1471 	int syncext;
1472 
1473 	ncomps = HAST_NCOMPONENTS;
1474 	dorewind = true;
1475 	synced = 0;
1476 
1477 	for (;;) {
1478 		mtx_lock(&sync_lock);
1479 		while (!sync_inprogress) {
1480 			dorewind = true;
1481 			synced = 0;
1482 			cv_wait(&sync_cond, &sync_lock);
1483 		}
1484 		mtx_unlock(&sync_lock);
1485 		/*
1486 		 * Obtain offset at which we should synchronize.
1487 		 * Rewind synchronization if needed.
1488 		 */
1489 		mtx_lock(&res->hr_amp_lock);
1490 		if (dorewind)
1491 			activemap_sync_rewind(res->hr_amp);
1492 		offset = activemap_sync_offset(res->hr_amp, &length, &syncext);
1493 		if (syncext != -1) {
1494 			/*
1495 			 * We synchronized entire syncext extent, we can mark
1496 			 * it as clean now.
1497 			 */
1498 			if (activemap_extent_complete(res->hr_amp, syncext))
1499 				(void)hast_activemap_flush(res);
1500 		}
1501 		mtx_unlock(&res->hr_amp_lock);
1502 		if (dorewind) {
1503 			dorewind = false;
1504 			if (offset < 0)
1505 				pjdlog_info("Nodes are in sync.");
1506 			else {
1507 				pjdlog_info("Synchronization started. %ju bytes to go.",
1508 				    (uintmax_t)(res->hr_extentsize *
1509 				    activemap_ndirty(res->hr_amp)));
1510 			}
1511 		}
1512 		if (offset < 0) {
1513 			mtx_lock(&sync_lock);
1514 			sync_inprogress = false;
1515 			mtx_unlock(&sync_lock);
1516 			pjdlog_debug(1, "Nothing to synchronize.");
1517 			/*
1518 			 * Synchronization complete, make both localcnt and
1519 			 * remotecnt equal.
1520 			 */
1521 			ncomp = 1;
1522 			rw_rlock(&hio_remote_lock[ncomp]);
1523 			if (ISCONNECTED(res, ncomp)) {
1524 				if (synced > 0) {
1525 					pjdlog_info("Synchronization complete. "
1526 					    "%jd bytes synchronized.",
1527 					    (intmax_t)synced);
1528 				}
1529 				mtx_lock(&metadata_lock);
1530 				res->hr_syncsrc = HAST_SYNCSRC_UNDEF;
1531 				res->hr_primary_localcnt =
1532 				    res->hr_secondary_localcnt;
1533 				res->hr_primary_remotecnt =
1534 				    res->hr_secondary_remotecnt;
1535 				pjdlog_debug(1,
1536 				    "Setting localcnt to %ju and remotecnt to %ju.",
1537 				    (uintmax_t)res->hr_primary_localcnt,
1538 				    (uintmax_t)res->hr_secondary_localcnt);
1539 				(void)metadata_write(res);
1540 				mtx_unlock(&metadata_lock);
1541 			} else if (synced > 0) {
1542 				pjdlog_info("Synchronization interrupted. "
1543 				    "%jd bytes synchronized so far.",
1544 				    (intmax_t)synced);
1545 			}
1546 			rw_unlock(&hio_remote_lock[ncomp]);
1547 			continue;
1548 		}
1549 		pjdlog_debug(2, "sync: Taking free request.");
1550 		QUEUE_TAKE2(hio, free);
1551 		pjdlog_debug(2, "sync: (%p) Got free request.", hio);
1552 		/*
1553 		 * Lock the range we are going to synchronize. We don't want
1554 		 * race where someone writes between our read and write.
1555 		 */
1556 		for (;;) {
1557 			mtx_lock(&range_lock);
1558 			if (rangelock_islocked(range_regular, offset, length)) {
1559 				pjdlog_debug(2,
1560 				    "sync: Range offset=%jd length=%jd locked.",
1561 				    (intmax_t)offset, (intmax_t)length);
1562 				range_sync_wait = true;
1563 				cv_wait(&range_sync_cond, &range_lock);
1564 				range_sync_wait = false;
1565 				mtx_unlock(&range_lock);
1566 				continue;
1567 			}
1568 			if (rangelock_add(range_sync, offset, length) < 0) {
1569 				mtx_unlock(&range_lock);
1570 				pjdlog_debug(2,
1571 				    "sync: Range offset=%jd length=%jd is already locked, waiting.",
1572 				    (intmax_t)offset, (intmax_t)length);
1573 				sleep(1);
1574 				continue;
1575 			}
1576 			mtx_unlock(&range_lock);
1577 			break;
1578 		}
1579 		/*
1580 		 * First read the data from synchronization source.
1581 		 */
1582 		SYNCREQ(hio);
1583 		ggio = &hio->hio_ggio;
1584 		ggio->gctl_cmd = BIO_READ;
1585 		ggio->gctl_offset = offset;
1586 		ggio->gctl_length = length;
1587 		ggio->gctl_error = 0;
1588 		for (ii = 0; ii < ncomps; ii++)
1589 			hio->hio_errors[ii] = EINVAL;
1590 		reqlog(LOG_DEBUG, 2, ggio, "sync: (%p) Sending sync request: ",
1591 		    hio);
1592 		pjdlog_debug(2, "sync: (%p) Moving request to the send queue.",
1593 		    hio);
1594 		mtx_lock(&metadata_lock);
1595 		if (res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1596 			/*
1597 			 * This range is up-to-date on local component,
1598 			 * so handle request locally.
1599 			 */
1600 			 /* Local component is 0 for now. */
1601 			ncomp = 0;
1602 		} else /* if (res->hr_syncsrc == HAST_SYNCSRC_SECONDARY) */ {
1603 			assert(res->hr_syncsrc == HAST_SYNCSRC_SECONDARY);
1604 			/*
1605 			 * This range is out-of-date on local component,
1606 			 * so send request to the remote node.
1607 			 */
1608 			 /* Remote component is 1 for now. */
1609 			ncomp = 1;
1610 		}
1611 		mtx_unlock(&metadata_lock);
1612 		refcount_init(&hio->hio_countdown, 1);
1613 		QUEUE_INSERT1(hio, send, ncomp);
1614 
1615 		/*
1616 		 * Let's wait for READ to finish.
1617 		 */
1618 		mtx_lock(&sync_lock);
1619 		while (!ISSYNCREQDONE(hio))
1620 			cv_wait(&sync_cond, &sync_lock);
1621 		mtx_unlock(&sync_lock);
1622 
1623 		if (hio->hio_errors[ncomp] != 0) {
1624 			pjdlog_error("Unable to read synchronization data: %s.",
1625 			    strerror(hio->hio_errors[ncomp]));
1626 			goto free_queue;
1627 		}
1628 
1629 		/*
1630 		 * We read the data from synchronization source, now write it
1631 		 * to synchronization target.
1632 		 */
1633 		SYNCREQ(hio);
1634 		ggio->gctl_cmd = BIO_WRITE;
1635 		for (ii = 0; ii < ncomps; ii++)
1636 			hio->hio_errors[ii] = EINVAL;
1637 		reqlog(LOG_DEBUG, 2, ggio, "sync: (%p) Sending sync request: ",
1638 		    hio);
1639 		pjdlog_debug(2, "sync: (%p) Moving request to the send queue.",
1640 		    hio);
1641 		mtx_lock(&metadata_lock);
1642 		if (res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1643 			/*
1644 			 * This range is up-to-date on local component,
1645 			 * so we update remote component.
1646 			 */
1647 			 /* Remote component is 1 for now. */
1648 			ncomp = 1;
1649 		} else /* if (res->hr_syncsrc == HAST_SYNCSRC_SECONDARY) */ {
1650 			assert(res->hr_syncsrc == HAST_SYNCSRC_SECONDARY);
1651 			/*
1652 			 * This range is out-of-date on local component,
1653 			 * so we update it.
1654 			 */
1655 			 /* Local component is 0 for now. */
1656 			ncomp = 0;
1657 		}
1658 		mtx_unlock(&metadata_lock);
1659 
1660 		pjdlog_debug(2, "sync: (%p) Moving request to the send queues.",
1661 		    hio);
1662 		refcount_init(&hio->hio_countdown, 1);
1663 		QUEUE_INSERT1(hio, send, ncomp);
1664 
1665 		/*
1666 		 * Let's wait for WRITE to finish.
1667 		 */
1668 		mtx_lock(&sync_lock);
1669 		while (!ISSYNCREQDONE(hio))
1670 			cv_wait(&sync_cond, &sync_lock);
1671 		mtx_unlock(&sync_lock);
1672 
1673 		if (hio->hio_errors[ncomp] != 0) {
1674 			pjdlog_error("Unable to write synchronization data: %s.",
1675 			    strerror(hio->hio_errors[ncomp]));
1676 			goto free_queue;
1677 		}
1678 free_queue:
1679 		mtx_lock(&range_lock);
1680 		rangelock_del(range_sync, offset, length);
1681 		if (range_regular_wait)
1682 			cv_signal(&range_regular_cond);
1683 		mtx_unlock(&range_lock);
1684 
1685 		synced += length;
1686 
1687 		pjdlog_debug(2, "sync: (%p) Moving request to the free queue.",
1688 		    hio);
1689 		QUEUE_INSERT2(hio, free);
1690 	}
1691 	/* NOTREACHED */
1692 	return (NULL);
1693 }
1694 
1695 static void
1696 sighandler(int sig)
1697 {
1698 	bool unlock;
1699 
1700 	switch (sig) {
1701 	case SIGINT:
1702 	case SIGTERM:
1703 		sigexit_received = true;
1704 		break;
1705 	default:
1706 		assert(!"invalid condition");
1707 	}
1708 	/*
1709 	 * XXX: Racy, but if we cannot obtain hio_guard_lock here, we don't
1710 	 * want to risk deadlock.
1711 	 */
1712 	unlock = mtx_trylock(&hio_guard_lock);
1713 	cv_signal(&hio_guard_cond);
1714 	if (unlock)
1715 		mtx_unlock(&hio_guard_lock);
1716 }
1717 
1718 /*
1719  * Thread guards remote connections and reconnects when needed, handles
1720  * signals, etc.
1721  */
1722 static void *
1723 guard_thread(void *arg)
1724 {
1725 	struct hast_resource *res = arg;
1726 	struct proto_conn *in, *out;
1727 	unsigned int ii, ncomps;
1728 	int timeout;
1729 
1730 	ncomps = HAST_NCOMPONENTS;
1731 	/* The is only one remote component for now. */
1732 #define	ISREMOTE(no)	((no) == 1)
1733 
1734 	for (;;) {
1735 		if (sigexit_received) {
1736 			primary_exitx(EX_OK,
1737 			    "Termination signal received, exiting.");
1738 		}
1739 		/*
1740 		 * If all the connection will be fine, we will sleep until
1741 		 * someone wakes us up.
1742 		 * If any of the connections will be broken and we won't be
1743 		 * able to connect, we will sleep only for RECONNECT_SLEEP
1744 		 * seconds so we can retry soon.
1745 		 */
1746 		timeout = 0;
1747 		pjdlog_debug(2, "remote_guard: Checking connections.");
1748 		mtx_lock(&hio_guard_lock);
1749 		for (ii = 0; ii < ncomps; ii++) {
1750 			if (!ISREMOTE(ii))
1751 				continue;
1752 			rw_rlock(&hio_remote_lock[ii]);
1753 			if (ISCONNECTED(res, ii)) {
1754 				assert(res->hr_remotein != NULL);
1755 				assert(res->hr_remoteout != NULL);
1756 				rw_unlock(&hio_remote_lock[ii]);
1757 				pjdlog_debug(2,
1758 				    "remote_guard: Connection to %s is ok.",
1759 				    res->hr_remoteaddr);
1760 			} else {
1761 				assert(res->hr_remotein == NULL);
1762 				assert(res->hr_remoteout == NULL);
1763 				/*
1764 				 * Upgrade the lock. It doesn't have to be
1765 				 * atomic as no other thread can change
1766 				 * connection status from disconnected to
1767 				 * connected.
1768 				 */
1769 				rw_unlock(&hio_remote_lock[ii]);
1770 				pjdlog_debug(2,
1771 				    "remote_guard: Reconnecting to %s.",
1772 				    res->hr_remoteaddr);
1773 				in = out = NULL;
1774 				if (init_remote(res, &in, &out)) {
1775 					rw_wlock(&hio_remote_lock[ii]);
1776 					assert(res->hr_remotein == NULL);
1777 					assert(res->hr_remoteout == NULL);
1778 					assert(in != NULL && out != NULL);
1779 					res->hr_remotein = in;
1780 					res->hr_remoteout = out;
1781 					rw_unlock(&hio_remote_lock[ii]);
1782 					pjdlog_info("Successfully reconnected to %s.",
1783 					    res->hr_remoteaddr);
1784 					sync_start();
1785 				} else {
1786 					/* Both connections should be NULL. */
1787 					assert(res->hr_remotein == NULL);
1788 					assert(res->hr_remoteout == NULL);
1789 					assert(in == NULL && out == NULL);
1790 					pjdlog_debug(2,
1791 					    "remote_guard: Reconnect to %s failed.",
1792 					    res->hr_remoteaddr);
1793 					timeout = RECONNECT_SLEEP;
1794 				}
1795 			}
1796 		}
1797 		(void)cv_timedwait(&hio_guard_cond, &hio_guard_lock, timeout);
1798 		mtx_unlock(&hio_guard_lock);
1799 	}
1800 #undef	ISREMOTE
1801 	/* NOTREACHED */
1802 	return (NULL);
1803 }
1804