xref: /freebsd/sys/geom/raid/tr_concat.c (revision 2008043f386721d58158e37e0d7e50df8095942d)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
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/geom_dbg.h>
41 #include "geom/raid/g_raid.h"
42 #include "g_raid_tr_if.h"
43 
44 static MALLOC_DEFINE(M_TR_CONCAT, "tr_concat_data", "GEOM_RAID CONCAT data");
45 
46 struct g_raid_tr_concat_object {
47 	struct g_raid_tr_object	 trso_base;
48 	int			 trso_starting;
49 	int			 trso_stopped;
50 };
51 
52 static g_raid_tr_taste_t g_raid_tr_taste_concat;
53 static g_raid_tr_event_t g_raid_tr_event_concat;
54 static g_raid_tr_start_t g_raid_tr_start_concat;
55 static g_raid_tr_stop_t g_raid_tr_stop_concat;
56 static g_raid_tr_iostart_t g_raid_tr_iostart_concat;
57 static g_raid_tr_iodone_t g_raid_tr_iodone_concat;
58 static g_raid_tr_kerneldump_t g_raid_tr_kerneldump_concat;
59 static g_raid_tr_free_t g_raid_tr_free_concat;
60 
61 static kobj_method_t g_raid_tr_concat_methods[] = {
62 	KOBJMETHOD(g_raid_tr_taste,	g_raid_tr_taste_concat),
63 	KOBJMETHOD(g_raid_tr_event,	g_raid_tr_event_concat),
64 	KOBJMETHOD(g_raid_tr_start,	g_raid_tr_start_concat),
65 	KOBJMETHOD(g_raid_tr_stop,	g_raid_tr_stop_concat),
66 	KOBJMETHOD(g_raid_tr_iostart,	g_raid_tr_iostart_concat),
67 	KOBJMETHOD(g_raid_tr_iodone,	g_raid_tr_iodone_concat),
68 	KOBJMETHOD(g_raid_tr_kerneldump,	g_raid_tr_kerneldump_concat),
69 	KOBJMETHOD(g_raid_tr_free,	g_raid_tr_free_concat),
70 	{ 0, 0 }
71 };
72 
73 static struct g_raid_tr_class g_raid_tr_concat_class = {
74 	"CONCAT",
75 	g_raid_tr_concat_methods,
76 	sizeof(struct g_raid_tr_concat_object),
77 	.trc_enable = 1,
78 	.trc_priority = 50,
79 	.trc_accept_unmapped = 1
80 };
81 
82 static int
83 g_raid_tr_taste_concat(struct g_raid_tr_object *tr, struct g_raid_volume *volume)
84 {
85 	struct g_raid_tr_concat_object *trs;
86 
87 	trs = (struct g_raid_tr_concat_object *)tr;
88 	if (tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_SINGLE &&
89 	    tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_CONCAT &&
90 	    !(tr->tro_volume->v_disks_count == 1 &&
91 	      tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_UNKNOWN))
92 		return (G_RAID_TR_TASTE_FAIL);
93 	trs->trso_starting = 1;
94 	return (G_RAID_TR_TASTE_SUCCEED);
95 }
96 
97 static int
98 g_raid_tr_update_state_concat(struct g_raid_volume *vol)
99 {
100 	struct g_raid_tr_concat_object *trs;
101 	struct g_raid_softc *sc;
102 	off_t size;
103 	u_int s;
104 	int i, n, f;
105 
106 	sc = vol->v_softc;
107 	trs = (struct g_raid_tr_concat_object *)vol->v_tr;
108 	if (trs->trso_stopped)
109 		s = G_RAID_VOLUME_S_STOPPED;
110 	else if (trs->trso_starting)
111 		s = G_RAID_VOLUME_S_STARTING;
112 	else {
113 		n = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_ACTIVE);
114 		f = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_FAILED);
115 		if (n + f == vol->v_disks_count) {
116 			if (f == 0)
117 				s = G_RAID_VOLUME_S_OPTIMAL;
118 			else
119 				s = G_RAID_VOLUME_S_SUBOPTIMAL;
120 		} else
121 			s = G_RAID_VOLUME_S_BROKEN;
122 	}
123 	if (s != vol->v_state) {
124 		/*
125 		 * Some metadata modules may not know CONCAT volume
126 		 * mediasize until all disks connected. Recalculate.
127 		 */
128 		if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT &&
129 		    G_RAID_VOLUME_S_ALIVE(s) &&
130 		    !G_RAID_VOLUME_S_ALIVE(vol->v_state)) {
131 			size = 0;
132 			for (i = 0; i < vol->v_disks_count; i++) {
133 				if (vol->v_subdisks[i].sd_state !=
134 				    G_RAID_SUBDISK_S_NONE)
135 					size += vol->v_subdisks[i].sd_size;
136 			}
137 			vol->v_mediasize = size;
138 		}
139 
140 		g_raid_event_send(vol, G_RAID_VOLUME_S_ALIVE(s) ?
141 		    G_RAID_VOLUME_E_UP : G_RAID_VOLUME_E_DOWN,
142 		    G_RAID_EVENT_VOLUME);
143 		g_raid_change_volume_state(vol, s);
144 		if (!trs->trso_starting && !trs->trso_stopped)
145 			g_raid_write_metadata(sc, vol, NULL, NULL);
146 	}
147 	return (0);
148 }
149 
150 static int
151 g_raid_tr_event_concat(struct g_raid_tr_object *tr,
152     struct g_raid_subdisk *sd, u_int event)
153 {
154 	struct g_raid_tr_concat_object *trs;
155 	struct g_raid_softc *sc;
156 	struct g_raid_volume *vol;
157 	int state;
158 
159 	trs = (struct g_raid_tr_concat_object *)tr;
160 	vol = tr->tro_volume;
161 	sc = vol->v_softc;
162 
163 	state = sd->sd_state;
164 	if (state != G_RAID_SUBDISK_S_NONE &&
165 	    state != G_RAID_SUBDISK_S_FAILED &&
166 	    state != G_RAID_SUBDISK_S_ACTIVE) {
167 		G_RAID_DEBUG1(1, sc,
168 		    "Promote subdisk %s:%d from %s to ACTIVE.",
169 		    vol->v_name, sd->sd_pos,
170 		    g_raid_subdisk_state2str(sd->sd_state));
171 		g_raid_change_subdisk_state(sd, G_RAID_SUBDISK_S_ACTIVE);
172 	}
173 	if (state != sd->sd_state &&
174 	    !trs->trso_starting && !trs->trso_stopped)
175 		g_raid_write_metadata(sc, vol, sd, NULL);
176 	g_raid_tr_update_state_concat(vol);
177 	return (0);
178 }
179 
180 static int
181 g_raid_tr_start_concat(struct g_raid_tr_object *tr)
182 {
183 	struct g_raid_tr_concat_object *trs;
184 	struct g_raid_volume *vol;
185 
186 	trs = (struct g_raid_tr_concat_object *)tr;
187 	vol = tr->tro_volume;
188 	trs->trso_starting = 0;
189 	g_raid_tr_update_state_concat(vol);
190 	return (0);
191 }
192 
193 static int
194 g_raid_tr_stop_concat(struct g_raid_tr_object *tr)
195 {
196 	struct g_raid_tr_concat_object *trs;
197 	struct g_raid_volume *vol;
198 
199 	trs = (struct g_raid_tr_concat_object *)tr;
200 	vol = tr->tro_volume;
201 	trs->trso_starting = 0;
202 	trs->trso_stopped = 1;
203 	g_raid_tr_update_state_concat(vol);
204 	return (0);
205 }
206 
207 static void
208 g_raid_tr_iostart_concat(struct g_raid_tr_object *tr, struct bio *bp)
209 {
210 	struct g_raid_volume *vol;
211 	struct g_raid_subdisk *sd;
212 	struct bio_queue_head queue;
213 	struct bio *cbp;
214 	char *addr;
215 	off_t offset, length, remain;
216 	u_int no;
217 
218 	vol = tr->tro_volume;
219 	if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL &&
220 	    vol->v_state != G_RAID_VOLUME_S_SUBOPTIMAL) {
221 		g_raid_iodone(bp, EIO);
222 		return;
223 	}
224 	if (bp->bio_cmd == BIO_FLUSH || bp->bio_cmd == BIO_SPEEDUP) {
225 		g_raid_tr_flush_common(tr, bp);
226 		return;
227 	}
228 
229 	offset = bp->bio_offset;
230 	remain = bp->bio_length;
231 	if ((bp->bio_flags & BIO_UNMAPPED) != 0)
232 		addr = NULL;
233 	else
234 		addr = bp->bio_data;
235 	no = 0;
236 	while (no < vol->v_disks_count &&
237 	    offset >= vol->v_subdisks[no].sd_size) {
238 		offset -= vol->v_subdisks[no].sd_size;
239 		no++;
240 	}
241 	if (no >= vol->v_disks_count) {
242 		g_raid_iodone(bp, EIO);
243 		return;
244 	}
245 	bioq_init(&queue);
246 	do {
247 		sd = &vol->v_subdisks[no];
248 		length = MIN(sd->sd_size - offset, remain);
249 		cbp = g_clone_bio(bp);
250 		if (cbp == NULL)
251 			goto failure;
252 		cbp->bio_offset = offset;
253 		cbp->bio_length = length;
254 		if ((bp->bio_flags & BIO_UNMAPPED) != 0 &&
255 		    bp->bio_cmd != BIO_DELETE) {
256 			cbp->bio_ma_offset += (uintptr_t)addr;
257 			cbp->bio_ma += cbp->bio_ma_offset / PAGE_SIZE;
258 			cbp->bio_ma_offset %= PAGE_SIZE;
259 			cbp->bio_ma_n = round_page(cbp->bio_ma_offset +
260 			    cbp->bio_length) / PAGE_SIZE;
261 		} else
262 			cbp->bio_data = addr;
263 		cbp->bio_caller1 = sd;
264 		bioq_insert_tail(&queue, cbp);
265 		remain -= length;
266 		if (bp->bio_cmd != BIO_DELETE)
267 			addr += length;
268 		offset = 0;
269 		no++;
270 	} while (remain > 0 && no < vol->v_disks_count);
271 	bp->bio_completed = bp->bio_length - remain;
272 	while ((cbp = bioq_takefirst(&queue)) != NULL) {
273 		sd = cbp->bio_caller1;
274 		cbp->bio_caller1 = NULL;
275 		g_raid_subdisk_iostart(sd, cbp);
276 	}
277 	return;
278 failure:
279 	while ((cbp = bioq_takefirst(&queue)) != NULL)
280 		g_destroy_bio(cbp);
281 	if (bp->bio_error == 0)
282 		bp->bio_error = ENOMEM;
283 	g_raid_iodone(bp, bp->bio_error);
284 }
285 
286 static int
287 g_raid_tr_kerneldump_concat(struct g_raid_tr_object *tr, void *virtual,
288     off_t boffset, size_t blength)
289 {
290 	struct g_raid_volume *vol;
291 	struct g_raid_subdisk *sd;
292 	char *addr;
293 	off_t offset, length, remain;
294 	int error, no;
295 
296 	vol = tr->tro_volume;
297 	if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
298 		return (ENXIO);
299 
300 	offset = boffset;
301 	remain = blength;
302 	addr = virtual;
303 	no = 0;
304 	while (no < vol->v_disks_count &&
305 	    offset >= vol->v_subdisks[no].sd_size) {
306 		offset -= vol->v_subdisks[no].sd_size;
307 		no++;
308 	}
309 	if (no >= vol->v_disks_count)
310 		return (EIO);
311 	do {
312 		sd = &vol->v_subdisks[no];
313 		length = MIN(sd->sd_size - offset, remain);
314 		error = g_raid_subdisk_kerneldump(&vol->v_subdisks[no],
315 		    addr, offset, length);
316 		if (error != 0)
317 			return (error);
318 		remain -= length;
319 		addr += length;
320 		offset = 0;
321 		no++;
322 	} while (remain > 0 && no < vol->v_disks_count);
323 	if (remain > 0)
324 		return (EIO);
325 	return (0);
326 }
327 
328 static void
329 g_raid_tr_iodone_concat(struct g_raid_tr_object *tr,
330     struct g_raid_subdisk *sd,struct bio *bp)
331 {
332 	struct bio *pbp;
333 
334 	pbp = bp->bio_parent;
335 	if (pbp->bio_error == 0)
336 		pbp->bio_error = bp->bio_error;
337 	g_destroy_bio(bp);
338 	pbp->bio_inbed++;
339 	if (pbp->bio_children == pbp->bio_inbed) {
340 		g_raid_iodone(pbp, pbp->bio_error);
341 	}
342 }
343 
344 static int
345 g_raid_tr_free_concat(struct g_raid_tr_object *tr)
346 {
347 
348 	return (0);
349 }
350 
351 G_RAID_TR_DECLARE(concat, "CONCAT");
352