xref: /freebsd/sys/geom/raid/tr_concat.c (revision ce3adf4362fcca6a43e500b2531f0038adbfbd21)
1 /*-
2  * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org>
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
29 
30 #include <sys/param.h>
31 #include <sys/bio.h>
32 #include <sys/endian.h>
33 #include <sys/kernel.h>
34 #include <sys/kobj.h>
35 #include <sys/lock.h>
36 #include <sys/malloc.h>
37 #include <sys/mutex.h>
38 #include <sys/systm.h>
39 #include <geom/geom.h>
40 #include "geom/raid/g_raid.h"
41 #include "g_raid_tr_if.h"
42 
43 static MALLOC_DEFINE(M_TR_CONCAT, "tr_concat_data", "GEOM_RAID CONCAT data");
44 
45 struct g_raid_tr_concat_object {
46 	struct g_raid_tr_object	 trso_base;
47 	int			 trso_starting;
48 	int			 trso_stopped;
49 };
50 
51 static g_raid_tr_taste_t g_raid_tr_taste_concat;
52 static g_raid_tr_event_t g_raid_tr_event_concat;
53 static g_raid_tr_start_t g_raid_tr_start_concat;
54 static g_raid_tr_stop_t g_raid_tr_stop_concat;
55 static g_raid_tr_iostart_t g_raid_tr_iostart_concat;
56 static g_raid_tr_iodone_t g_raid_tr_iodone_concat;
57 static g_raid_tr_kerneldump_t g_raid_tr_kerneldump_concat;
58 static g_raid_tr_free_t g_raid_tr_free_concat;
59 
60 static kobj_method_t g_raid_tr_concat_methods[] = {
61 	KOBJMETHOD(g_raid_tr_taste,	g_raid_tr_taste_concat),
62 	KOBJMETHOD(g_raid_tr_event,	g_raid_tr_event_concat),
63 	KOBJMETHOD(g_raid_tr_start,	g_raid_tr_start_concat),
64 	KOBJMETHOD(g_raid_tr_stop,	g_raid_tr_stop_concat),
65 	KOBJMETHOD(g_raid_tr_iostart,	g_raid_tr_iostart_concat),
66 	KOBJMETHOD(g_raid_tr_iodone,	g_raid_tr_iodone_concat),
67 	KOBJMETHOD(g_raid_tr_kerneldump,	g_raid_tr_kerneldump_concat),
68 	KOBJMETHOD(g_raid_tr_free,	g_raid_tr_free_concat),
69 	{ 0, 0 }
70 };
71 
72 static struct g_raid_tr_class g_raid_tr_concat_class = {
73 	"CONCAT",
74 	g_raid_tr_concat_methods,
75 	sizeof(struct g_raid_tr_concat_object),
76 	.trc_enable = 1,
77 	.trc_priority = 50
78 };
79 
80 static int
81 g_raid_tr_taste_concat(struct g_raid_tr_object *tr, struct g_raid_volume *volume)
82 {
83 	struct g_raid_tr_concat_object *trs;
84 
85 	trs = (struct g_raid_tr_concat_object *)tr;
86 	if (tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_SINGLE &&
87 	    tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_CONCAT &&
88 	    !(tr->tro_volume->v_disks_count == 1 &&
89 	      tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_UNKNOWN))
90 		return (G_RAID_TR_TASTE_FAIL);
91 	trs->trso_starting = 1;
92 	return (G_RAID_TR_TASTE_SUCCEED);
93 }
94 
95 static int
96 g_raid_tr_update_state_concat(struct g_raid_volume *vol)
97 {
98 	struct g_raid_tr_concat_object *trs;
99 	struct g_raid_softc *sc;
100 	off_t size;
101 	u_int s;
102 	int i, n, f;
103 
104 	sc = vol->v_softc;
105 	trs = (struct g_raid_tr_concat_object *)vol->v_tr;
106 	if (trs->trso_stopped)
107 		s = G_RAID_VOLUME_S_STOPPED;
108 	else if (trs->trso_starting)
109 		s = G_RAID_VOLUME_S_STARTING;
110 	else {
111 		n = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_ACTIVE);
112 		f = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_FAILED);
113 		if (n + f == vol->v_disks_count) {
114 			if (f == 0)
115 				s = G_RAID_VOLUME_S_OPTIMAL;
116 			else
117 				s = G_RAID_VOLUME_S_SUBOPTIMAL;
118 		} else
119 			s = G_RAID_VOLUME_S_BROKEN;
120 	}
121 	if (s != vol->v_state) {
122 
123 		/*
124 		 * Some metadata modules may not know CONCAT volume
125 		 * mediasize until all disks connected. Recalculate.
126 		 */
127 		if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT &&
128 		    G_RAID_VOLUME_S_ALIVE(s) &&
129 		    !G_RAID_VOLUME_S_ALIVE(vol->v_state)) {
130 			size = 0;
131 			for (i = 0; i < vol->v_disks_count; i++) {
132 				if (vol->v_subdisks[i].sd_state !=
133 				    G_RAID_SUBDISK_S_NONE)
134 					size += vol->v_subdisks[i].sd_size;
135 			}
136 			vol->v_mediasize = size;
137 		}
138 
139 		g_raid_event_send(vol, G_RAID_VOLUME_S_ALIVE(s) ?
140 		    G_RAID_VOLUME_E_UP : G_RAID_VOLUME_E_DOWN,
141 		    G_RAID_EVENT_VOLUME);
142 		g_raid_change_volume_state(vol, s);
143 		if (!trs->trso_starting && !trs->trso_stopped)
144 			g_raid_write_metadata(sc, vol, NULL, NULL);
145 	}
146 	return (0);
147 }
148 
149 static int
150 g_raid_tr_event_concat(struct g_raid_tr_object *tr,
151     struct g_raid_subdisk *sd, u_int event)
152 {
153 	struct g_raid_tr_concat_object *trs;
154 	struct g_raid_softc *sc;
155 	struct g_raid_volume *vol;
156 	int state;
157 
158 	trs = (struct g_raid_tr_concat_object *)tr;
159 	vol = tr->tro_volume;
160 	sc = vol->v_softc;
161 
162 	state = sd->sd_state;
163 	if (state != G_RAID_SUBDISK_S_NONE &&
164 	    state != G_RAID_SUBDISK_S_FAILED &&
165 	    state != G_RAID_SUBDISK_S_ACTIVE) {
166 		G_RAID_DEBUG1(1, sc,
167 		    "Promote subdisk %s:%d from %s to ACTIVE.",
168 		    vol->v_name, sd->sd_pos,
169 		    g_raid_subdisk_state2str(sd->sd_state));
170 		g_raid_change_subdisk_state(sd, G_RAID_SUBDISK_S_ACTIVE);
171 	}
172 	if (state != sd->sd_state &&
173 	    !trs->trso_starting && !trs->trso_stopped)
174 		g_raid_write_metadata(sc, vol, sd, NULL);
175 	g_raid_tr_update_state_concat(vol);
176 	return (0);
177 }
178 
179 static int
180 g_raid_tr_start_concat(struct g_raid_tr_object *tr)
181 {
182 	struct g_raid_tr_concat_object *trs;
183 	struct g_raid_volume *vol;
184 
185 	trs = (struct g_raid_tr_concat_object *)tr;
186 	vol = tr->tro_volume;
187 	trs->trso_starting = 0;
188 	g_raid_tr_update_state_concat(vol);
189 	return (0);
190 }
191 
192 static int
193 g_raid_tr_stop_concat(struct g_raid_tr_object *tr)
194 {
195 	struct g_raid_tr_concat_object *trs;
196 	struct g_raid_volume *vol;
197 
198 	trs = (struct g_raid_tr_concat_object *)tr;
199 	vol = tr->tro_volume;
200 	trs->trso_starting = 0;
201 	trs->trso_stopped = 1;
202 	g_raid_tr_update_state_concat(vol);
203 	return (0);
204 }
205 
206 static void
207 g_raid_tr_iostart_concat(struct g_raid_tr_object *tr, struct bio *bp)
208 {
209 	struct g_raid_volume *vol;
210 	struct g_raid_subdisk *sd;
211 	struct bio_queue_head queue;
212 	struct bio *cbp;
213 	char *addr;
214 	off_t offset, length, remain;
215 	u_int no;
216 
217 	vol = tr->tro_volume;
218 	if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL &&
219 	    vol->v_state != G_RAID_VOLUME_S_SUBOPTIMAL) {
220 		g_raid_iodone(bp, EIO);
221 		return;
222 	}
223 	if (bp->bio_cmd == BIO_FLUSH) {
224 		g_raid_tr_flush_common(tr, bp);
225 		return;
226 	}
227 
228 	offset = bp->bio_offset;
229 	remain = bp->bio_length;
230 	addr = bp->bio_data;
231 	no = 0;
232 	while (no < vol->v_disks_count &&
233 	    offset >= vol->v_subdisks[no].sd_size) {
234 		offset -= vol->v_subdisks[no].sd_size;
235 		no++;
236 	}
237 	KASSERT(no < vol->v_disks_count,
238 	    ("Request starts after volume end (%ju)", bp->bio_offset));
239 	bioq_init(&queue);
240 	do {
241 		sd = &vol->v_subdisks[no];
242 		length = MIN(sd->sd_size - offset, remain);
243 		cbp = g_clone_bio(bp);
244 		if (cbp == NULL)
245 			goto failure;
246 		cbp->bio_offset = offset;
247 		cbp->bio_data = addr;
248 		cbp->bio_length = length;
249 		cbp->bio_caller1 = sd;
250 		bioq_insert_tail(&queue, cbp);
251 		remain -= length;
252 		if (bp->bio_cmd != BIO_DELETE)
253 			addr += length;
254 		offset = 0;
255 		no++;
256 		KASSERT(no < vol->v_disks_count || remain == 0,
257 		    ("Request ends after volume end (%ju, %ju)",
258 			bp->bio_offset, bp->bio_length));
259 	} while (remain > 0);
260 	for (cbp = bioq_first(&queue); cbp != NULL;
261 	    cbp = bioq_first(&queue)) {
262 		bioq_remove(&queue, cbp);
263 		sd = cbp->bio_caller1;
264 		cbp->bio_caller1 = NULL;
265 		g_raid_subdisk_iostart(sd, cbp);
266 	}
267 	return;
268 failure:
269 	for (cbp = bioq_first(&queue); cbp != NULL;
270 	    cbp = bioq_first(&queue)) {
271 		bioq_remove(&queue, cbp);
272 		g_destroy_bio(cbp);
273 	}
274 	if (bp->bio_error == 0)
275 		bp->bio_error = ENOMEM;
276 	g_raid_iodone(bp, bp->bio_error);
277 }
278 
279 static int
280 g_raid_tr_kerneldump_concat(struct g_raid_tr_object *tr,
281     void *virtual, vm_offset_t physical, off_t boffset, size_t blength)
282 {
283 	struct g_raid_volume *vol;
284 	struct g_raid_subdisk *sd;
285 	char *addr;
286 	off_t offset, length, remain;
287 	int error, no;
288 
289 	vol = tr->tro_volume;
290 	if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
291 		return (ENXIO);
292 
293 	offset = boffset;
294 	remain = blength;
295 	addr = virtual;
296 	no = 0;
297 	while (no < vol->v_disks_count &&
298 	    offset >= vol->v_subdisks[no].sd_size) {
299 		offset -= vol->v_subdisks[no].sd_size;
300 		no++;
301 	}
302 	KASSERT(no < vol->v_disks_count,
303 	    ("Request starts after volume end (%ju)", boffset));
304 	do {
305 		sd = &vol->v_subdisks[no];
306 		length = MIN(sd->sd_size - offset, remain);
307 		error = g_raid_subdisk_kerneldump(&vol->v_subdisks[no],
308 		    addr, 0, offset, length);
309 		if (error != 0)
310 			return (error);
311 		remain -= length;
312 		addr += length;
313 		offset = 0;
314 		no++;
315 		KASSERT(no < vol->v_disks_count || remain == 0,
316 		    ("Request ends after volume end (%ju, %zu)",
317 			boffset, blength));
318 	} while (remain > 0);
319 	return (0);
320 }
321 
322 static void
323 g_raid_tr_iodone_concat(struct g_raid_tr_object *tr,
324     struct g_raid_subdisk *sd,struct bio *bp)
325 {
326 	struct bio *pbp;
327 
328 	pbp = bp->bio_parent;
329 	if (pbp->bio_error == 0)
330 		pbp->bio_error = bp->bio_error;
331 	g_destroy_bio(bp);
332 	pbp->bio_inbed++;
333 	if (pbp->bio_children == pbp->bio_inbed) {
334 		pbp->bio_completed = pbp->bio_length;
335 		g_raid_iodone(pbp, bp->bio_error);
336 	}
337 }
338 
339 static int
340 g_raid_tr_free_concat(struct g_raid_tr_object *tr)
341 {
342 
343 	return (0);
344 }
345 
346 G_RAID_TR_DECLARE(concat, "CONCAT");
347