1/*-
2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3 *
4 *  Copyright (c) 2004, 2007 Lukas Ertl
5 *  Copyright (c) 2007, 2009 Ulf Lilleengen
6 *  All rights reserved.
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 AUTHOR 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 AUTHOR 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
31#include <sys/cdefs.h>
32__FBSDID("$FreeBSD$");
33
34#include <sys/param.h>
35#include <sys/bio.h>
36#include <sys/kernel.h>
37#include <sys/kthread.h>
38#include <sys/lock.h>
39#include <sys/malloc.h>
40#include <sys/module.h>
41#include <sys/mutex.h>
42#include <sys/sbuf.h>
43#include <sys/sysctl.h>
44#include <sys/systm.h>
45
46#include <geom/geom.h>
47#include <geom/geom_dbg.h>
48#include <geom/vinum/geom_vinum_var.h>
49#include <geom/vinum/geom_vinum.h>
50#include <geom/vinum/geom_vinum_raid5.h>
51
52SYSCTL_DECL(_kern_geom);
53static SYSCTL_NODE(_kern_geom, OID_AUTO, vinum, CTLFLAG_RW, 0,
54    "GEOM_VINUM stuff");
55u_int g_vinum_debug = 0;
56SYSCTL_UINT(_kern_geom_vinum, OID_AUTO, debug, CTLFLAG_RWTUN, &g_vinum_debug, 0,
57    "Debug level");
58
59static int	gv_create(struct g_geom *, struct gctl_req *);
60static void	gv_attach(struct gv_softc *, struct gctl_req *);
61static void	gv_detach(struct gv_softc *, struct gctl_req *);
62static void	gv_parityop(struct gv_softc *, struct gctl_req *);
63
64
65static void
66gv_orphan(struct g_consumer *cp)
67{
68	struct g_geom *gp;
69	struct gv_softc *sc;
70	struct gv_drive *d;
71
72	g_topology_assert();
73
74	KASSERT(cp != NULL, ("gv_orphan: null cp"));
75	gp = cp->geom;
76	KASSERT(gp != NULL, ("gv_orphan: null gp"));
77	sc = gp->softc;
78	KASSERT(sc != NULL, ("gv_orphan: null sc"));
79	d = cp->private;
80	KASSERT(d != NULL, ("gv_orphan: null d"));
81
82	g_trace(G_T_TOPOLOGY, "gv_orphan(%s)", gp->name);
83
84	gv_post_event(sc, GV_EVENT_DRIVE_LOST, d, NULL, 0, 0);
85}
86
87void
88gv_start(struct bio *bp)
89{
90	struct g_geom *gp;
91	struct gv_softc *sc;
92
93	gp = bp->bio_to->geom;
94	sc = gp->softc;
95
96	switch (bp->bio_cmd) {
97	case BIO_READ:
98	case BIO_WRITE:
99	case BIO_DELETE:
100		break;
101	case BIO_GETATTR:
102	default:
103		g_io_deliver(bp, EOPNOTSUPP);
104		return;
105	}
106	mtx_lock(&sc->bqueue_mtx);
107	bioq_disksort(sc->bqueue_down, bp);
108	wakeup(sc);
109	mtx_unlock(&sc->bqueue_mtx);
110}
111
112void
113gv_done(struct bio *bp)
114{
115	struct g_geom *gp;
116	struct gv_softc *sc;
117
118	KASSERT(bp != NULL, ("NULL bp"));
119
120	gp = bp->bio_from->geom;
121	sc = gp->softc;
122
123	mtx_lock(&sc->bqueue_mtx);
124	bioq_disksort(sc->bqueue_up, bp);
125	wakeup(sc);
126	mtx_unlock(&sc->bqueue_mtx);
127}
128
129int
130gv_access(struct g_provider *pp, int dr, int dw, int de)
131{
132	struct g_geom *gp;
133	struct gv_softc *sc;
134	struct gv_drive *d, *d2;
135	int error;
136
137	gp = pp->geom;
138	sc = gp->softc;
139	/*
140	 * We want to modify the read count with the write count in case we have
141	 * plexes in a RAID-5 organization.
142	 */
143	dr += dw;
144
145	LIST_FOREACH(d, &sc->drives, drive) {
146		if (d->consumer == NULL)
147			continue;
148		error = g_access(d->consumer, dr, dw, de);
149		if (error) {
150			LIST_FOREACH(d2, &sc->drives, drive) {
151				if (d == d2)
152					break;
153				g_access(d2->consumer, -dr, -dw, -de);
154			}
155			G_VINUM_DEBUG(0, "g_access '%s' failed: %d", d->name,
156			    error);
157			return (error);
158		}
159	}
160	return (0);
161}
162
163static void
164gv_init(struct g_class *mp)
165{
166	struct g_geom *gp;
167	struct gv_softc *sc;
168
169	g_trace(G_T_TOPOLOGY, "gv_init(%p)", mp);
170
171	gp = g_new_geomf(mp, "VINUM");
172	gp->spoiled = gv_orphan;
173	gp->orphan = gv_orphan;
174	gp->access = gv_access;
175	gp->start = gv_start;
176	gp->softc = g_malloc(sizeof(struct gv_softc), M_WAITOK | M_ZERO);
177	sc = gp->softc;
178	sc->geom = gp;
179	sc->bqueue_down = g_malloc(sizeof(struct bio_queue_head),
180	    M_WAITOK | M_ZERO);
181	sc->bqueue_up = g_malloc(sizeof(struct bio_queue_head),
182	    M_WAITOK | M_ZERO);
183	bioq_init(sc->bqueue_down);
184	bioq_init(sc->bqueue_up);
185	LIST_INIT(&sc->drives);
186	LIST_INIT(&sc->subdisks);
187	LIST_INIT(&sc->plexes);
188	LIST_INIT(&sc->volumes);
189	TAILQ_INIT(&sc->equeue);
190	mtx_init(&sc->config_mtx, "gv_config", NULL, MTX_DEF);
191	mtx_init(&sc->equeue_mtx, "gv_equeue", NULL, MTX_DEF);
192	mtx_init(&sc->bqueue_mtx, "gv_bqueue", NULL, MTX_DEF);
193	kproc_create(gv_worker, sc, &sc->worker, 0, 0, "gv_worker");
194}
195
196static int
197gv_unload(struct gctl_req *req, struct g_class *mp, struct g_geom *gp)
198{
199	struct gv_softc *sc;
200
201	g_trace(G_T_TOPOLOGY, "gv_unload(%p)", mp);
202
203	g_topology_assert();
204	sc = gp->softc;
205
206	if (sc != NULL) {
207		gv_worker_exit(sc);
208		gp->softc = NULL;
209		g_wither_geom(gp, ENXIO);
210	}
211
212	return (0);
213}
214
215/* Handle userland request of attaching object. */
216static void
217gv_attach(struct gv_softc *sc, struct gctl_req *req)
218{
219	struct gv_volume *v;
220	struct gv_plex *p;
221	struct gv_sd *s;
222	off_t *offset;
223	int *rename, type_child, type_parent;
224	char *child, *parent;
225
226	child = gctl_get_param(req, "child", NULL);
227	if (child == NULL) {
228		gctl_error(req, "no child given");
229		return;
230	}
231	parent = gctl_get_param(req, "parent", NULL);
232	if (parent == NULL) {
233		gctl_error(req, "no parent given");
234		return;
235	}
236	offset = gctl_get_paraml(req, "offset", sizeof(*offset));
237	if (offset == NULL) {
238		gctl_error(req, "no offset given");
239		return;
240	}
241	rename = gctl_get_paraml(req, "rename", sizeof(*rename));
242	if (rename == NULL) {
243		gctl_error(req, "no rename flag given");
244		return;
245	}
246
247	type_child = gv_object_type(sc, child);
248	type_parent = gv_object_type(sc, parent);
249
250	switch (type_child) {
251	case GV_TYPE_PLEX:
252		if (type_parent != GV_TYPE_VOL) {
253			gctl_error(req, "no such volume to attach to");
254			return;
255		}
256		v = gv_find_vol(sc, parent);
257		p = gv_find_plex(sc, child);
258		gv_post_event(sc, GV_EVENT_ATTACH_PLEX, p, v, *offset, *rename);
259		break;
260	case GV_TYPE_SD:
261		if (type_parent != GV_TYPE_PLEX) {
262			gctl_error(req, "no such plex to attach to");
263			return;
264		}
265		p = gv_find_plex(sc, parent);
266		s = gv_find_sd(sc, child);
267		gv_post_event(sc, GV_EVENT_ATTACH_SD, s, p, *offset, *rename);
268		break;
269	default:
270		gctl_error(req, "invalid child type");
271		break;
272	}
273}
274
275/* Handle userland request of detaching object. */
276static void
277gv_detach(struct gv_softc *sc, struct gctl_req *req)
278{
279	struct gv_plex *p;
280	struct gv_sd *s;
281	int *flags, type;
282	char *object;
283
284	object = gctl_get_param(req, "object", NULL);
285	if (object == NULL) {
286		gctl_error(req, "no argument given");
287		return;
288	}
289
290	flags = gctl_get_paraml(req, "flags", sizeof(*flags));
291	type = gv_object_type(sc, object);
292	switch (type) {
293	case GV_TYPE_PLEX:
294		p = gv_find_plex(sc, object);
295		gv_post_event(sc, GV_EVENT_DETACH_PLEX, p, NULL, *flags, 0);
296		break;
297	case GV_TYPE_SD:
298		s = gv_find_sd(sc, object);
299		gv_post_event(sc, GV_EVENT_DETACH_SD, s, NULL, *flags, 0);
300		break;
301	default:
302		gctl_error(req, "invalid object type");
303		break;
304	}
305}
306
307/* Handle userland requests for creating new objects. */
308static int
309gv_create(struct g_geom *gp, struct gctl_req *req)
310{
311	struct gv_softc *sc;
312	struct gv_drive *d, *d2;
313	struct gv_plex *p, *p2;
314	struct gv_sd *s, *s2;
315	struct gv_volume *v, *v2;
316	struct g_provider *pp;
317	int error, i, *drives, *flags, *plexes, *subdisks, *volumes;
318	char buf[20];
319
320	g_topology_assert();
321
322	sc = gp->softc;
323
324	/* Find out how many of each object have been passed in. */
325	volumes = gctl_get_paraml(req, "volumes", sizeof(*volumes));
326	plexes = gctl_get_paraml(req, "plexes", sizeof(*plexes));
327	subdisks = gctl_get_paraml(req, "subdisks", sizeof(*subdisks));
328	drives = gctl_get_paraml(req, "drives", sizeof(*drives));
329	if (volumes == NULL || plexes == NULL || subdisks == NULL ||
330	    drives == NULL) {
331		gctl_error(req, "number of objects not given");
332		return (-1);
333	}
334	flags = gctl_get_paraml(req, "flags", sizeof(*flags));
335	if (flags == NULL) {
336		gctl_error(req, "flags not given");
337		return (-1);
338	}
339
340	/* First, handle drive definitions ... */
341	for (i = 0; i < *drives; i++) {
342		snprintf(buf, sizeof(buf), "drive%d", i);
343		d2 = gctl_get_paraml(req, buf, sizeof(*d2));
344		if (d2 == NULL) {
345			gctl_error(req, "no drive definition given");
346			return (-1);
347		}
348		/*
349		 * Make sure that the device specified in the drive config is
350		 * an active GEOM provider.
351		 */
352		pp = g_provider_by_name(d2->device);
353		if (pp == NULL) {
354			gctl_error(req, "%s: device not found", d2->device);
355			goto error;
356		}
357		if (gv_find_drive(sc, d2->name) != NULL) {
358			/* Ignore error. */
359			if (*flags & GV_FLAG_F)
360				continue;
361			gctl_error(req, "drive '%s' already exists", d2->name);
362			goto error;
363		}
364		if (gv_find_drive_device(sc, d2->device) != NULL) {
365			gctl_error(req, "device '%s' already configured in "
366			    "gvinum", d2->device);
367			goto error;
368		}
369
370
371		d = g_malloc(sizeof(*d), M_WAITOK | M_ZERO);
372		bcopy(d2, d, sizeof(*d));
373
374		gv_post_event(sc, GV_EVENT_CREATE_DRIVE, d, NULL, 0, 0);
375	}
376
377	/* ... then volume definitions ... */
378	for (i = 0; i < *volumes; i++) {
379		error = 0;
380		snprintf(buf, sizeof(buf), "volume%d", i);
381		v2 = gctl_get_paraml(req, buf, sizeof(*v2));
382		if (v2 == NULL) {
383			gctl_error(req, "no volume definition given");
384			return (-1);
385		}
386		if (gv_find_vol(sc, v2->name) != NULL) {
387			/* Ignore error. */
388			if (*flags & GV_FLAG_F)
389				continue;
390			gctl_error(req, "volume '%s' already exists", v2->name);
391			goto error;
392		}
393
394		v = g_malloc(sizeof(*v), M_WAITOK | M_ZERO);
395		bcopy(v2, v, sizeof(*v));
396
397		gv_post_event(sc, GV_EVENT_CREATE_VOLUME, v, NULL, 0, 0);
398	}
399
400	/* ... then plex definitions ... */
401	for (i = 0; i < *plexes; i++) {
402		error = 0;
403		snprintf(buf, sizeof(buf), "plex%d", i);
404		p2 = gctl_get_paraml(req, buf, sizeof(*p2));
405		if (p2 == NULL) {
406			gctl_error(req, "no plex definition given");
407			return (-1);
408		}
409		if (gv_find_plex(sc, p2->name) != NULL) {
410			/* Ignore error. */
411			if (*flags & GV_FLAG_F)
412				continue;
413			gctl_error(req, "plex '%s' already exists", p2->name);
414			goto error;
415		}
416
417		p = g_malloc(sizeof(*p), M_WAITOK | M_ZERO);
418		bcopy(p2, p, sizeof(*p));
419
420		gv_post_event(sc, GV_EVENT_CREATE_PLEX, p, NULL, 0, 0);
421	}
422
423	/* ... and, finally, subdisk definitions. */
424	for (i = 0; i < *subdisks; i++) {
425		error = 0;
426		snprintf(buf, sizeof(buf), "sd%d", i);
427		s2 = gctl_get_paraml(req, buf, sizeof(*s2));
428		if (s2 == NULL) {
429			gctl_error(req, "no subdisk definition given");
430			return (-1);
431		}
432		if (gv_find_sd(sc, s2->name) != NULL) {
433			/* Ignore error. */
434			if (*flags & GV_FLAG_F)
435				continue;
436			gctl_error(req, "sd '%s' already exists", s2->name);
437			goto error;
438		}
439
440		s = g_malloc(sizeof(*s), M_WAITOK | M_ZERO);
441		bcopy(s2, s, sizeof(*s));
442
443		gv_post_event(sc, GV_EVENT_CREATE_SD, s, NULL, 0, 0);
444	}
445
446error:
447	gv_post_event(sc, GV_EVENT_SETUP_OBJECTS, sc, NULL, 0, 0);
448	gv_post_event(sc, GV_EVENT_SAVE_CONFIG, sc, NULL, 0, 0);
449
450	return (0);
451}
452
453static void
454gv_config(struct gctl_req *req, struct g_class *mp, char const *verb)
455{
456	struct g_geom *gp;
457	struct gv_softc *sc;
458	struct sbuf *sb;
459	char *comment;
460
461	g_topology_assert();
462
463	gp = LIST_FIRST(&mp->geom);
464	sc = gp->softc;
465
466	if (!strcmp(verb, "attach")) {
467		gv_attach(sc, req);
468
469	} else if (!strcmp(verb, "concat")) {
470		gv_concat(gp, req);
471
472	} else if (!strcmp(verb, "detach")) {
473		gv_detach(sc, req);
474
475	} else if (!strcmp(verb, "list")) {
476		gv_list(gp, req);
477
478	/* Save our configuration back to disk. */
479	} else if (!strcmp(verb, "saveconfig")) {
480		gv_post_event(sc, GV_EVENT_SAVE_CONFIG, sc, NULL, 0, 0);
481
482	/* Return configuration in string form. */
483	} else if (!strcmp(verb, "getconfig")) {
484		comment = gctl_get_param(req, "comment", NULL);
485		if (comment == NULL) {
486			gctl_error(req, "no comment parameter given");
487			return;
488		}
489		sb = sbuf_new(NULL, NULL, GV_CFG_LEN, SBUF_FIXEDLEN);
490		gv_format_config(sc, sb, 0, comment);
491		sbuf_finish(sb);
492		gctl_set_param(req, "config", sbuf_data(sb), sbuf_len(sb) + 1);
493		sbuf_delete(sb);
494
495	} else if (!strcmp(verb, "create")) {
496		gv_create(gp, req);
497
498	} else if (!strcmp(verb, "mirror")) {
499		gv_mirror(gp, req);
500
501	} else if (!strcmp(verb, "move")) {
502		gv_move(gp, req);
503
504	} else if (!strcmp(verb, "raid5")) {
505		gv_raid5(gp, req);
506
507	} else if (!strcmp(verb, "rebuildparity") ||
508	    !strcmp(verb, "checkparity")) {
509		gv_parityop(sc, req);
510
511	} else if (!strcmp(verb, "remove")) {
512		gv_remove(gp, req);
513
514	} else if (!strcmp(verb, "rename")) {
515		gv_rename(gp, req);
516
517	} else if (!strcmp(verb, "resetconfig")) {
518		gv_post_event(sc, GV_EVENT_RESET_CONFIG, sc, NULL, 0, 0);
519
520	} else if (!strcmp(verb, "start")) {
521		gv_start_obj(gp, req);
522
523	} else if (!strcmp(verb, "stripe")) {
524		gv_stripe(gp, req);
525
526	} else if (!strcmp(verb, "setstate")) {
527		gv_setstate(gp, req);
528	} else
529		gctl_error(req, "Unknown verb parameter");
530}
531
532static void
533gv_parityop(struct gv_softc *sc, struct gctl_req *req)
534{
535	struct gv_plex *p;
536	int *flags, *rebuild, type;
537	char *plex;
538
539	plex = gctl_get_param(req, "plex", NULL);
540	if (plex == NULL) {
541		gctl_error(req, "no plex given");
542		return;
543	}
544
545	flags = gctl_get_paraml(req, "flags", sizeof(*flags));
546	if (flags == NULL) {
547		gctl_error(req, "no flags given");
548		return;
549	}
550
551	rebuild = gctl_get_paraml(req, "rebuild", sizeof(*rebuild));
552	if (rebuild == NULL) {
553		gctl_error(req, "no operation given");
554		return;
555	}
556
557	type = gv_object_type(sc, plex);
558	if (type != GV_TYPE_PLEX) {
559		gctl_error(req, "'%s' is not a plex", plex);
560		return;
561	}
562	p = gv_find_plex(sc, plex);
563
564	if (p->state != GV_PLEX_UP) {
565		gctl_error(req, "plex %s is not completely accessible",
566		    p->name);
567		return;
568	}
569
570	if (p->org != GV_PLEX_RAID5) {
571		gctl_error(req, "plex %s is not a RAID5 plex", p->name);
572		return;
573	}
574
575	/* Put it in the event queue. */
576	/* XXX: The state of the plex might have changed when this event is
577	 * picked up ... We should perhaps check this afterwards. */
578	if (*rebuild)
579		gv_post_event(sc, GV_EVENT_PARITY_REBUILD, p, NULL, 0, 0);
580	else
581		gv_post_event(sc, GV_EVENT_PARITY_CHECK, p, NULL, 0, 0);
582}
583
584
585static struct g_geom *
586gv_taste(struct g_class *mp, struct g_provider *pp, int flags __unused)
587{
588	struct g_geom *gp;
589	struct g_consumer *cp;
590	struct gv_softc *sc;
591	struct gv_hdr vhdr;
592	int error;
593
594 	g_topology_assert();
595	g_trace(G_T_TOPOLOGY, "gv_taste(%s, %s)", mp->name, pp->name);
596
597	gp = LIST_FIRST(&mp->geom);
598	if (gp == NULL) {
599		G_VINUM_DEBUG(0, "error: tasting, but not initialized?");
600		return (NULL);
601	}
602	sc = gp->softc;
603
604	cp = g_new_consumer(gp);
605	if (g_attach(cp, pp) != 0) {
606		g_destroy_consumer(cp);
607		return (NULL);
608	}
609	if (g_access(cp, 1, 0, 0) != 0) {
610		g_detach(cp);
611		g_destroy_consumer(cp);
612		return (NULL);
613	}
614	g_topology_unlock();
615
616	error = gv_read_header(cp, &vhdr);
617
618	g_topology_lock();
619	g_access(cp, -1, 0, 0);
620	g_detach(cp);
621	g_destroy_consumer(cp);
622
623	/* Check if what we've been given is a valid vinum drive. */
624	if (!error)
625		gv_post_event(sc, GV_EVENT_DRIVE_TASTED, pp, NULL, 0, 0);
626
627	return (NULL);
628}
629
630void
631gv_worker(void *arg)
632{
633	struct g_provider *pp;
634	struct gv_softc *sc;
635	struct gv_event *ev;
636	struct gv_volume *v;
637	struct gv_plex *p;
638	struct gv_sd *s;
639	struct gv_drive *d;
640	struct bio *bp;
641	int newstate, flags, err, rename;
642	char *newname;
643	off_t offset;
644
645	sc = arg;
646	KASSERT(sc != NULL, ("NULL sc"));
647	for (;;) {
648		/* Look at the events first... */
649		ev = gv_get_event(sc);
650		if (ev != NULL) {
651			gv_remove_event(sc, ev);
652
653			switch (ev->type) {
654			case GV_EVENT_DRIVE_TASTED:
655				G_VINUM_DEBUG(2, "event 'drive tasted'");
656				pp = ev->arg1;
657				gv_drive_tasted(sc, pp);
658				break;
659
660			case GV_EVENT_DRIVE_LOST:
661				G_VINUM_DEBUG(2, "event 'drive lost'");
662				d = ev->arg1;
663				gv_drive_lost(sc, d);
664				break;
665
666			case GV_EVENT_CREATE_DRIVE:
667				G_VINUM_DEBUG(2, "event 'create drive'");
668				d = ev->arg1;
669				gv_create_drive(sc, d);
670				break;
671
672			case GV_EVENT_CREATE_VOLUME:
673				G_VINUM_DEBUG(2, "event 'create volume'");
674				v = ev->arg1;
675				gv_create_volume(sc, v);
676				break;
677
678			case GV_EVENT_CREATE_PLEX:
679				G_VINUM_DEBUG(2, "event 'create plex'");
680				p = ev->arg1;
681				gv_create_plex(sc, p);
682				break;
683
684			case GV_EVENT_CREATE_SD:
685				G_VINUM_DEBUG(2, "event 'create sd'");
686				s = ev->arg1;
687				gv_create_sd(sc, s);
688				break;
689
690			case GV_EVENT_RM_DRIVE:
691				G_VINUM_DEBUG(2, "event 'remove drive'");
692				d = ev->arg1;
693				flags = ev->arg3;
694				gv_rm_drive(sc, d, flags);
695				/*gv_setup_objects(sc);*/
696				break;
697
698			case GV_EVENT_RM_VOLUME:
699				G_VINUM_DEBUG(2, "event 'remove volume'");
700				v = ev->arg1;
701				gv_rm_vol(sc, v);
702				/*gv_setup_objects(sc);*/
703				break;
704
705			case GV_EVENT_RM_PLEX:
706				G_VINUM_DEBUG(2, "event 'remove plex'");
707				p = ev->arg1;
708				gv_rm_plex(sc, p);
709				/*gv_setup_objects(sc);*/
710				break;
711
712			case GV_EVENT_RM_SD:
713				G_VINUM_DEBUG(2, "event 'remove sd'");
714				s = ev->arg1;
715				gv_rm_sd(sc, s);
716				/*gv_setup_objects(sc);*/
717				break;
718
719			case GV_EVENT_SAVE_CONFIG:
720				G_VINUM_DEBUG(2, "event 'save config'");
721				gv_save_config(sc);
722				break;
723
724			case GV_EVENT_SET_SD_STATE:
725				G_VINUM_DEBUG(2, "event 'setstate sd'");
726				s = ev->arg1;
727				newstate = ev->arg3;
728				flags = ev->arg4;
729				err = gv_set_sd_state(s, newstate, flags);
730				if (err)
731					G_VINUM_DEBUG(0, "error setting subdisk"
732					    " state: error code %d", err);
733				break;
734
735			case GV_EVENT_SET_DRIVE_STATE:
736				G_VINUM_DEBUG(2, "event 'setstate drive'");
737				d = ev->arg1;
738				newstate = ev->arg3;
739				flags = ev->arg4;
740				err = gv_set_drive_state(d, newstate, flags);
741				if (err)
742					G_VINUM_DEBUG(0, "error setting drive "
743					    "state: error code %d", err);
744				break;
745
746			case GV_EVENT_SET_VOL_STATE:
747				G_VINUM_DEBUG(2, "event 'setstate volume'");
748				v = ev->arg1;
749				newstate = ev->arg3;
750				flags = ev->arg4;
751				err = gv_set_vol_state(v, newstate, flags);
752				if (err)
753					G_VINUM_DEBUG(0, "error setting volume "
754					    "state: error code %d", err);
755				break;
756
757			case GV_EVENT_SET_PLEX_STATE:
758				G_VINUM_DEBUG(2, "event 'setstate plex'");
759				p = ev->arg1;
760				newstate = ev->arg3;
761				flags = ev->arg4;
762				err = gv_set_plex_state(p, newstate, flags);
763				if (err)
764					G_VINUM_DEBUG(0, "error setting plex "
765					    "state: error code %d", err);
766				break;
767
768			case GV_EVENT_SETUP_OBJECTS:
769				G_VINUM_DEBUG(2, "event 'setup objects'");
770				gv_setup_objects(sc);
771				break;
772
773			case GV_EVENT_RESET_CONFIG:
774				G_VINUM_DEBUG(2, "event 'resetconfig'");
775				err = gv_resetconfig(sc);
776				if (err)
777					G_VINUM_DEBUG(0, "error resetting "
778					    "config: error code %d", err);
779				break;
780
781			case GV_EVENT_PARITY_REBUILD:
782				/*
783				 * Start the rebuild. The gv_plex_done will
784				 * handle issuing of the remaining rebuild bio's
785				 * until it's finished.
786				 */
787				G_VINUM_DEBUG(2, "event 'rebuild'");
788				p = ev->arg1;
789				if (p->state != GV_PLEX_UP) {
790					G_VINUM_DEBUG(0, "plex %s is not "
791					    "completely accessible", p->name);
792					break;
793				}
794				if (p->flags & GV_PLEX_SYNCING ||
795				    p->flags & GV_PLEX_REBUILDING ||
796				    p->flags & GV_PLEX_GROWING) {
797					G_VINUM_DEBUG(0, "plex %s is busy with "
798					    "syncing or parity build", p->name);
799					break;
800				}
801				p->synced = 0;
802				p->flags |= GV_PLEX_REBUILDING;
803				g_topology_assert_not();
804				g_topology_lock();
805				err = gv_access(p->vol_sc->provider, 1, 1, 0);
806				if (err) {
807					G_VINUM_DEBUG(0, "unable to access "
808					    "provider");
809					break;
810				}
811				g_topology_unlock();
812				gv_parity_request(p, GV_BIO_CHECK |
813				    GV_BIO_PARITY, 0);
814				break;
815
816			case GV_EVENT_PARITY_CHECK:
817				/* Start parity check. */
818				G_VINUM_DEBUG(2, "event 'check'");
819				p = ev->arg1;
820				if (p->state != GV_PLEX_UP) {
821					G_VINUM_DEBUG(0, "plex %s is not "
822					    "completely accessible", p->name);
823					break;
824				}
825				if (p->flags & GV_PLEX_SYNCING ||
826				    p->flags & GV_PLEX_REBUILDING ||
827				    p->flags & GV_PLEX_GROWING) {
828					G_VINUM_DEBUG(0, "plex %s is busy with "
829					    "syncing or parity build", p->name);
830					break;
831				}
832				p->synced = 0;
833				g_topology_assert_not();
834				g_topology_lock();
835				err = gv_access(p->vol_sc->provider, 1, 1, 0);
836				if (err) {
837					G_VINUM_DEBUG(0, "unable to access "
838					    "provider");
839					break;
840				}
841				g_topology_unlock();
842				gv_parity_request(p, GV_BIO_CHECK, 0);
843				break;
844
845			case GV_EVENT_START_PLEX:
846				G_VINUM_DEBUG(2, "event 'start' plex");
847				p = ev->arg1;
848				gv_start_plex(p);
849				break;
850
851			case GV_EVENT_START_VOLUME:
852				G_VINUM_DEBUG(2, "event 'start' volume");
853				v = ev->arg1;
854				gv_start_vol(v);
855				break;
856
857			case GV_EVENT_ATTACH_PLEX:
858				G_VINUM_DEBUG(2, "event 'attach' plex");
859				p = ev->arg1;
860				v = ev->arg2;
861				rename = ev->arg4;
862				err = gv_attach_plex(p, v, rename);
863				if (err)
864					G_VINUM_DEBUG(0, "error attaching %s to"
865					    " %s: error code %d", p->name,
866					    v->name, err);
867				break;
868
869			case GV_EVENT_ATTACH_SD:
870				G_VINUM_DEBUG(2, "event 'attach' sd");
871				s = ev->arg1;
872				p = ev->arg2;
873				offset = ev->arg3;
874				rename = ev->arg4;
875				err = gv_attach_sd(s, p, offset, rename);
876				if (err)
877					G_VINUM_DEBUG(0, "error attaching %s to"
878					    " %s: error code %d", s->name,
879					    p->name, err);
880				break;
881
882			case GV_EVENT_DETACH_PLEX:
883				G_VINUM_DEBUG(2, "event 'detach' plex");
884				p = ev->arg1;
885				flags = ev->arg3;
886				err = gv_detach_plex(p, flags);
887				if (err)
888					G_VINUM_DEBUG(0, "error detaching %s: "
889					    "error code %d", p->name, err);
890				break;
891
892			case GV_EVENT_DETACH_SD:
893				G_VINUM_DEBUG(2, "event 'detach' sd");
894				s = ev->arg1;
895				flags = ev->arg3;
896				err = gv_detach_sd(s, flags);
897				if (err)
898					G_VINUM_DEBUG(0, "error detaching %s: "
899					    "error code %d", s->name, err);
900				break;
901
902			case GV_EVENT_RENAME_VOL:
903				G_VINUM_DEBUG(2, "event 'rename' volume");
904				v = ev->arg1;
905				newname = ev->arg2;
906				flags = ev->arg3;
907				err = gv_rename_vol(sc, v, newname, flags);
908				if (err)
909					G_VINUM_DEBUG(0, "error renaming %s to "
910					    "%s: error code %d", v->name,
911					    newname, err);
912				g_free(newname);
913				/* Destroy and recreate the provider if we can. */
914				if (gv_provider_is_open(v->provider)) {
915					G_VINUM_DEBUG(0, "unable to rename "
916					    "provider to %s: provider in use",
917					    v->name);
918					break;
919				}
920				g_topology_lock();
921				g_wither_provider(v->provider, ENOENT);
922				g_topology_unlock();
923				v->provider = NULL;
924				gv_post_event(sc, GV_EVENT_SETUP_OBJECTS, sc,
925				    NULL, 0, 0);
926				break;
927
928			case GV_EVENT_RENAME_PLEX:
929				G_VINUM_DEBUG(2, "event 'rename' plex");
930				p = ev->arg1;
931				newname = ev->arg2;
932				flags = ev->arg3;
933				err = gv_rename_plex(sc, p, newname, flags);
934				if (err)
935					G_VINUM_DEBUG(0, "error renaming %s to "
936					    "%s: error code %d", p->name,
937					    newname, err);
938				g_free(newname);
939				break;
940
941			case GV_EVENT_RENAME_SD:
942				G_VINUM_DEBUG(2, "event 'rename' sd");
943				s = ev->arg1;
944				newname = ev->arg2;
945				flags = ev->arg3;
946				err = gv_rename_sd(sc, s, newname, flags);
947				if (err)
948					G_VINUM_DEBUG(0, "error renaming %s to "
949					    "%s: error code %d", s->name,
950					    newname, err);
951				g_free(newname);
952				break;
953
954			case GV_EVENT_RENAME_DRIVE:
955				G_VINUM_DEBUG(2, "event 'rename' drive");
956				d = ev->arg1;
957				newname = ev->arg2;
958				flags = ev->arg3;
959				err = gv_rename_drive(sc, d, newname, flags);
960				if (err)
961					G_VINUM_DEBUG(0, "error renaming %s to "
962					    "%s: error code %d", d->name,
963					    newname, err);
964				g_free(newname);
965				break;
966
967			case GV_EVENT_MOVE_SD:
968				G_VINUM_DEBUG(2, "event 'move' sd");
969				s = ev->arg1;
970				d = ev->arg2;
971				flags = ev->arg3;
972				err = gv_move_sd(sc, s, d, flags);
973				if (err)
974					G_VINUM_DEBUG(0, "error moving %s to "
975					    "%s: error code %d", s->name,
976					    d->name, err);
977				break;
978
979			case GV_EVENT_THREAD_EXIT:
980				G_VINUM_DEBUG(2, "event 'thread exit'");
981				g_free(ev);
982				mtx_lock(&sc->equeue_mtx);
983				mtx_lock(&sc->bqueue_mtx);
984				gv_cleanup(sc);
985				mtx_destroy(&sc->bqueue_mtx);
986				mtx_destroy(&sc->equeue_mtx);
987				g_free(sc->bqueue_down);
988				g_free(sc->bqueue_up);
989				g_free(sc);
990				kproc_exit(0);
991				/* NOTREACHED */
992
993			default:
994				G_VINUM_DEBUG(1, "unknown event %d", ev->type);
995			}
996
997			g_free(ev);
998			continue;
999		}
1000
1001		/* ... then do I/O processing. */
1002		mtx_lock(&sc->bqueue_mtx);
1003		/* First do new requests. */
1004		bp = bioq_takefirst(sc->bqueue_down);
1005		if (bp != NULL) {
1006			mtx_unlock(&sc->bqueue_mtx);
1007			/* A bio that interfered with another bio. */
1008			if (bp->bio_pflags & GV_BIO_ONHOLD) {
1009				s = bp->bio_caller1;
1010				p = s->plex_sc;
1011				/* Is it still locked out? */
1012				if (gv_stripe_active(p, bp)) {
1013					/* Park the bio on the waiting queue. */
1014					bioq_disksort(p->wqueue, bp);
1015				} else {
1016					bp->bio_pflags &= ~GV_BIO_ONHOLD;
1017					g_io_request(bp, s->drive_sc->consumer);
1018				}
1019			/* A special request requireing special handling. */
1020			} else if (bp->bio_pflags & GV_BIO_INTERNAL) {
1021				p = bp->bio_caller1;
1022				gv_plex_start(p, bp);
1023			} else {
1024				gv_volume_start(sc, bp);
1025			}
1026			mtx_lock(&sc->bqueue_mtx);
1027		}
1028		/* Then do completed requests. */
1029		bp = bioq_takefirst(sc->bqueue_up);
1030		if (bp == NULL) {
1031			msleep(sc, &sc->bqueue_mtx, PRIBIO, "-", hz/10);
1032			mtx_unlock(&sc->bqueue_mtx);
1033			continue;
1034		}
1035		mtx_unlock(&sc->bqueue_mtx);
1036		gv_bio_done(sc, bp);
1037	}
1038}
1039
1040#define	VINUM_CLASS_NAME "VINUM"
1041
1042static struct g_class g_vinum_class	= {
1043	.name = VINUM_CLASS_NAME,
1044	.version = G_VERSION,
1045	.init = gv_init,
1046	.taste = gv_taste,
1047	.ctlreq = gv_config,
1048	.destroy_geom = gv_unload,
1049};
1050
1051DECLARE_GEOM_CLASS(g_vinum_class, g_vinum);
1052MODULE_VERSION(geom_vinum, 0);
1053