2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
34 #include <sys/endian.h>
37 #define G_MIRROR_CLASS_NAME "MIRROR"
39 #define G_MIRROR_MAGIC "GEOM::MIRROR"
42 * 0 - Initial version number.
43 * 1 - Added 'prefer' balance algorithm.
44 * 2 - Added md_genid field to metadata.
45 * 3 - Added md_provsize field to metadata.
46 * 4 - Added 'no failure synchronization' flag.
48 #define G_MIRROR_VERSION 4
50 #define G_MIRROR_BALANCE_NONE 0
51 #define G_MIRROR_BALANCE_ROUND_ROBIN 1
52 #define G_MIRROR_BALANCE_LOAD 2
53 #define G_MIRROR_BALANCE_SPLIT 3
54 #define G_MIRROR_BALANCE_PREFER 4
55 #define G_MIRROR_BALANCE_MIN G_MIRROR_BALANCE_NONE
56 #define G_MIRROR_BALANCE_MAX G_MIRROR_BALANCE_PREFER
58 #define G_MIRROR_DISK_FLAG_DIRTY 0x0000000000000001ULL
59 #define G_MIRROR_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL
60 #define G_MIRROR_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL
61 #define G_MIRROR_DISK_FLAG_INACTIVE 0x0000000000000008ULL
62 #define G_MIRROR_DISK_FLAG_HARDCODED 0x0000000000000010ULL
63 #define G_MIRROR_DISK_FLAG_BROKEN 0x0000000000000020ULL
64 #define G_MIRROR_DISK_FLAG_CANDELETE 0x0000000000000040ULL
65 #define G_MIRROR_DISK_FLAG_MASK (G_MIRROR_DISK_FLAG_DIRTY | \
66 G_MIRROR_DISK_FLAG_SYNCHRONIZING | \
67 G_MIRROR_DISK_FLAG_FORCE_SYNC | \
68 G_MIRROR_DISK_FLAG_INACTIVE | \
69 G_MIRROR_DISK_FLAG_CANDELETE)
71 #define G_MIRROR_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL
72 #define G_MIRROR_DEVICE_FLAG_NOFAILSYNC 0x0000000000000002ULL
73 #define G_MIRROR_DEVICE_FLAG_MASK (G_MIRROR_DEVICE_FLAG_NOAUTOSYNC | \
74 G_MIRROR_DEVICE_FLAG_NOFAILSYNC)
77 extern int g_mirror_debug;
79 #define G_MIRROR_DEBUG(lvl, ...) do { \
80 if (g_mirror_debug >= (lvl)) { \
81 printf("GEOM_MIRROR"); \
82 if (g_mirror_debug > 0) \
83 printf("[%u]", lvl); \
85 printf(__VA_ARGS__); \
89 #define G_MIRROR_LOGREQ(lvl, bp, ...) do { \
90 if (g_mirror_debug >= (lvl)) { \
91 printf("GEOM_MIRROR"); \
92 if (g_mirror_debug > 0) \
93 printf("[%u]", lvl); \
95 printf(__VA_ARGS__); \
102 #define G_MIRROR_BIO_FLAG_REGULAR 0x01
103 #define G_MIRROR_BIO_FLAG_SYNC 0x02
106 * Informations needed for synchronization.
108 struct g_mirror_disk_sync {
109 struct g_consumer *ds_consumer; /* Consumer connected to our mirror. */
110 off_t ds_offset; /* Offset of next request to send. */
111 off_t ds_offset_done; /* Offset of already synchronized
113 time_t ds_update_ts; /* Time of last metadata update. */
114 u_int ds_syncid; /* Disk's synchronization ID. */
115 u_int ds_inflight; /* Number of in-flight sync requests. */
116 struct bio **ds_bios; /* BIOs for synchronization I/O. */
120 * Informations needed for synchronization.
122 struct g_mirror_device_sync {
123 struct g_geom *ds_geom; /* Synchronization geom. */
124 u_int ds_ndisks; /* Number of disks in SYNCHRONIZING
128 #define G_MIRROR_DISK_STATE_NONE 0
129 #define G_MIRROR_DISK_STATE_NEW 1
130 #define G_MIRROR_DISK_STATE_ACTIVE 2
131 #define G_MIRROR_DISK_STATE_STALE 3
132 #define G_MIRROR_DISK_STATE_SYNCHRONIZING 4
133 #define G_MIRROR_DISK_STATE_DISCONNECTED 5
134 #define G_MIRROR_DISK_STATE_DESTROY 6
135 struct g_mirror_disk {
136 uint32_t d_id; /* Disk ID. */
137 struct g_consumer *d_consumer; /* Consumer. */
138 struct g_mirror_softc *d_softc; /* Back-pointer to softc. */
139 int d_state; /* Disk state. */
140 u_int d_priority; /* Disk priority. */
141 u_int load; /* Averaged queue length */
142 off_t d_last_offset; /* Last read offset */
143 uint64_t d_flags; /* Additional flags. */
144 u_int d_genid; /* Disk's generation ID. */
145 struct g_mirror_disk_sync d_sync;/* Sync information. */
146 LIST_ENTRY(g_mirror_disk) d_next;
148 #define d_name d_consumer->provider->name
150 #define G_MIRROR_EVENT_DONTWAIT 0x1
151 #define G_MIRROR_EVENT_WAIT 0x2
152 #define G_MIRROR_EVENT_DEVICE 0x4
153 #define G_MIRROR_EVENT_DONE 0x8
154 struct g_mirror_event {
155 struct g_mirror_disk *e_disk;
159 TAILQ_ENTRY(g_mirror_event) e_next;
162 #define G_MIRROR_DEVICE_FLAG_DESTROY 0x0100000000000000ULL
163 #define G_MIRROR_DEVICE_FLAG_DRAIN 0x0200000000000000ULL
164 #define G_MIRROR_DEVICE_FLAG_CLOSEWAIT 0x0400000000000000ULL
165 #define G_MIRROR_DEVICE_FLAG_TASTING 0x0800000000000000ULL
166 #define G_MIRROR_DEVICE_FLAG_WIPE 0x1000000000000000ULL
168 #define G_MIRROR_DEVICE_STATE_STARTING 0
169 #define G_MIRROR_DEVICE_STATE_RUNNING 1
171 #define G_MIRROR_TYPE_MANUAL 0
172 #define G_MIRROR_TYPE_AUTOMATIC 1
174 /* Bump syncid on first write. */
175 #define G_MIRROR_BUMP_SYNCID 0x1
176 /* Bump genid immediately. */
177 #define G_MIRROR_BUMP_GENID 0x2
178 /* Bump syncid immediately. */
179 #define G_MIRROR_BUMP_SYNCID_NOW 0x4
180 struct g_mirror_softc {
181 u_int sc_type; /* Device type (manual/automatic). */
182 u_int sc_state; /* Device state. */
183 uint32_t sc_slice; /* Slice size. */
184 uint8_t sc_balance; /* Balance algorithm. */
185 uint64_t sc_mediasize; /* Device size. */
186 uint32_t sc_sectorsize; /* Sector size. */
187 uint64_t sc_flags; /* Additional flags. */
189 struct g_geom *sc_geom;
190 struct g_provider *sc_provider;
191 int sc_provider_open;
193 uint32_t sc_id; /* Mirror unique ID. */
196 struct bio_queue sc_queue;
197 struct mtx sc_queue_mtx;
198 struct proc *sc_worker;
199 struct bio_queue sc_inflight; /* In-flight regular write requests. */
200 struct bio_queue sc_regular_delayed; /* Delayed I/O requests due to
201 collision with sync requests. */
202 struct bio_queue sc_sync_delayed; /* Delayed sync requests due to
203 collision with regular requests. */
205 LIST_HEAD(, g_mirror_disk) sc_disks;
206 u_int sc_ndisks; /* Number of disks. */
207 struct g_mirror_disk *sc_hint;
209 u_int sc_genid; /* Generation ID. */
210 u_int sc_syncid; /* Synchronization ID. */
212 struct g_mirror_device_sync sc_sync;
213 int sc_idle; /* DIRTY flags removed. */
214 time_t sc_last_write;
216 u_int sc_refcnt; /* Number of softc references */
218 TAILQ_HEAD(, g_mirror_event) sc_events;
219 struct mtx sc_events_mtx;
221 struct callout sc_callout;
223 struct root_hold_token *sc_rootmount;
225 struct mtx sc_done_mtx;
227 #define sc_name sc_geom->name
229 struct g_mirror_metadata;
231 u_int g_mirror_ndisks(struct g_mirror_softc *sc, int state);
232 struct g_geom * g_mirror_create(struct g_class *mp,
233 const struct g_mirror_metadata *md, u_int type);
234 #define G_MIRROR_DESTROY_SOFT 0
235 #define G_MIRROR_DESTROY_DELAYED 1
236 #define G_MIRROR_DESTROY_HARD 2
237 int g_mirror_destroy(struct g_mirror_softc *sc, int how);
238 int g_mirror_event_send(void *arg, int state, int flags);
239 struct g_mirror_metadata;
240 int g_mirror_add_disk(struct g_mirror_softc *sc, struct g_provider *pp,
241 struct g_mirror_metadata *md);
242 int g_mirror_read_metadata(struct g_consumer *cp, struct g_mirror_metadata *md);
243 void g_mirror_fill_metadata(struct g_mirror_softc *sc,
244 struct g_mirror_disk *disk, struct g_mirror_metadata *md);
245 void g_mirror_update_metadata(struct g_mirror_disk *disk);
247 g_ctl_req_t g_mirror_config;
250 struct g_mirror_metadata {
251 char md_magic[16]; /* Magic value. */
252 uint32_t md_version; /* Version number. */
253 char md_name[16]; /* Mirror name. */
254 uint32_t md_mid; /* Mirror unique ID. */
255 uint32_t md_did; /* Disk unique ID. */
256 uint8_t md_all; /* Number of disks in mirror. */
257 uint32_t md_genid; /* Generation ID. */
258 uint32_t md_syncid; /* Synchronization ID. */
259 uint8_t md_priority; /* Disk priority. */
260 uint32_t md_slice; /* Slice size. */
261 uint8_t md_balance; /* Balance type. */
262 uint64_t md_mediasize; /* Size of the smallest
264 uint32_t md_sectorsize; /* Sector size. */
265 uint64_t md_sync_offset; /* Synchronized offset. */
266 uint64_t md_mflags; /* Additional mirror flags. */
267 uint64_t md_dflags; /* Additional disk flags. */
268 char md_provider[16]; /* Hardcoded provider. */
269 uint64_t md_provsize; /* Provider's size. */
270 u_char md_hash[16]; /* MD5 hash. */
273 mirror_metadata_encode(struct g_mirror_metadata *md, u_char *data)
277 bcopy(md->md_magic, data, 16);
278 le32enc(data + 16, md->md_version);
279 bcopy(md->md_name, data + 20, 16);
280 le32enc(data + 36, md->md_mid);
281 le32enc(data + 40, md->md_did);
282 *(data + 44) = md->md_all;
283 le32enc(data + 45, md->md_genid);
284 le32enc(data + 49, md->md_syncid);
285 *(data + 53) = md->md_priority;
286 le32enc(data + 54, md->md_slice);
287 *(data + 58) = md->md_balance;
288 le64enc(data + 59, md->md_mediasize);
289 le32enc(data + 67, md->md_sectorsize);
290 le64enc(data + 71, md->md_sync_offset);
291 le64enc(data + 79, md->md_mflags);
292 le64enc(data + 87, md->md_dflags);
293 bcopy(md->md_provider, data + 95, 16);
294 le64enc(data + 111, md->md_provsize);
296 MD5Update(&ctx, data, 119);
297 MD5Final(md->md_hash, &ctx);
298 bcopy(md->md_hash, data + 119, 16);
301 mirror_metadata_decode_v0v1(const u_char *data, struct g_mirror_metadata *md)
305 bcopy(data + 20, md->md_name, 16);
306 md->md_mid = le32dec(data + 36);
307 md->md_did = le32dec(data + 40);
308 md->md_all = *(data + 44);
309 md->md_syncid = le32dec(data + 45);
310 md->md_priority = *(data + 49);
311 md->md_slice = le32dec(data + 50);
312 md->md_balance = *(data + 54);
313 md->md_mediasize = le64dec(data + 55);
314 md->md_sectorsize = le32dec(data + 63);
315 md->md_sync_offset = le64dec(data + 67);
316 md->md_mflags = le64dec(data + 75);
317 md->md_dflags = le64dec(data + 83);
318 bcopy(data + 91, md->md_provider, 16);
319 bcopy(data + 107, md->md_hash, 16);
321 MD5Update(&ctx, data, 107);
322 MD5Final(md->md_hash, &ctx);
323 if (bcmp(md->md_hash, data + 107, 16) != 0)
333 mirror_metadata_decode_v2(const u_char *data, struct g_mirror_metadata *md)
337 bcopy(data + 20, md->md_name, 16);
338 md->md_mid = le32dec(data + 36);
339 md->md_did = le32dec(data + 40);
340 md->md_all = *(data + 44);
341 md->md_genid = le32dec(data + 45);
342 md->md_syncid = le32dec(data + 49);
343 md->md_priority = *(data + 53);
344 md->md_slice = le32dec(data + 54);
345 md->md_balance = *(data + 58);
346 md->md_mediasize = le64dec(data + 59);
347 md->md_sectorsize = le32dec(data + 67);
348 md->md_sync_offset = le64dec(data + 71);
349 md->md_mflags = le64dec(data + 79);
350 md->md_dflags = le64dec(data + 87);
351 bcopy(data + 95, md->md_provider, 16);
352 bcopy(data + 111, md->md_hash, 16);
354 MD5Update(&ctx, data, 111);
355 MD5Final(md->md_hash, &ctx);
356 if (bcmp(md->md_hash, data + 111, 16) != 0)
365 mirror_metadata_decode_v3v4(const u_char *data, struct g_mirror_metadata *md)
369 bcopy(data + 20, md->md_name, 16);
370 md->md_mid = le32dec(data + 36);
371 md->md_did = le32dec(data + 40);
372 md->md_all = *(data + 44);
373 md->md_genid = le32dec(data + 45);
374 md->md_syncid = le32dec(data + 49);
375 md->md_priority = *(data + 53);
376 md->md_slice = le32dec(data + 54);
377 md->md_balance = *(data + 58);
378 md->md_mediasize = le64dec(data + 59);
379 md->md_sectorsize = le32dec(data + 67);
380 md->md_sync_offset = le64dec(data + 71);
381 md->md_mflags = le64dec(data + 79);
382 md->md_dflags = le64dec(data + 87);
383 bcopy(data + 95, md->md_provider, 16);
384 md->md_provsize = le64dec(data + 111);
385 bcopy(data + 119, md->md_hash, 16);
387 MD5Update(&ctx, data, 119);
388 MD5Final(md->md_hash, &ctx);
389 if (bcmp(md->md_hash, data + 119, 16) != 0)
394 mirror_metadata_decode(const u_char *data, struct g_mirror_metadata *md)
398 bcopy(data, md->md_magic, 16);
399 md->md_version = le32dec(data + 16);
400 switch (md->md_version) {
403 error = mirror_metadata_decode_v0v1(data, md);
406 error = mirror_metadata_decode_v2(data, md);
410 error = mirror_metadata_decode_v3v4(data, md);
419 static __inline const char *
420 balance_name(u_int balance)
422 static const char *algorithms[] = {
423 [G_MIRROR_BALANCE_NONE] = "none",
424 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin",
425 [G_MIRROR_BALANCE_LOAD] = "load",
426 [G_MIRROR_BALANCE_SPLIT] = "split",
427 [G_MIRROR_BALANCE_PREFER] = "prefer",
428 [G_MIRROR_BALANCE_MAX + 1] = "unknown"
431 if (balance > G_MIRROR_BALANCE_MAX)
432 balance = G_MIRROR_BALANCE_MAX + 1;
434 return (algorithms[balance]);
438 balance_id(const char *name)
440 static const char *algorithms[] = {
441 [G_MIRROR_BALANCE_NONE] = "none",
442 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin",
443 [G_MIRROR_BALANCE_LOAD] = "load",
444 [G_MIRROR_BALANCE_SPLIT] = "split",
445 [G_MIRROR_BALANCE_PREFER] = "prefer"
449 for (n = G_MIRROR_BALANCE_MIN; n <= G_MIRROR_BALANCE_MAX; n++) {
450 if (strcmp(name, algorithms[n]) == 0)
457 mirror_metadata_dump(const struct g_mirror_metadata *md)
459 static const char hex[] = "0123456789abcdef";
460 char hash[16 * 2 + 1];
463 printf(" magic: %s\n", md->md_magic);
464 printf(" version: %u\n", (u_int)md->md_version);
465 printf(" name: %s\n", md->md_name);
466 printf(" mid: %u\n", (u_int)md->md_mid);
467 printf(" did: %u\n", (u_int)md->md_did);
468 printf(" all: %u\n", (u_int)md->md_all);
469 printf(" genid: %u\n", (u_int)md->md_genid);
470 printf(" syncid: %u\n", (u_int)md->md_syncid);
471 printf(" priority: %u\n", (u_int)md->md_priority);
472 printf(" slice: %u\n", (u_int)md->md_slice);
473 printf(" balance: %s\n", balance_name((u_int)md->md_balance));
474 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize);
475 printf("sectorsize: %u\n", (u_int)md->md_sectorsize);
476 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset);
478 if (md->md_mflags == 0)
481 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOFAILSYNC) != 0)
482 printf(" NOFAILSYNC");
483 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) != 0)
484 printf(" NOAUTOSYNC");
488 if (md->md_dflags == 0)
491 if ((md->md_dflags & G_MIRROR_DISK_FLAG_DIRTY) != 0)
493 if ((md->md_dflags & G_MIRROR_DISK_FLAG_SYNCHRONIZING) != 0)
494 printf(" SYNCHRONIZING");
495 if ((md->md_dflags & G_MIRROR_DISK_FLAG_FORCE_SYNC) != 0)
496 printf(" FORCE_SYNC");
497 if ((md->md_dflags & G_MIRROR_DISK_FLAG_INACTIVE) != 0)
501 printf("hcprovider: %s\n", md->md_provider);
502 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize);
503 bzero(hash, sizeof(hash));
504 for (i = 0; i < 16; i++) {
505 hash[i * 2] = hex[md->md_hash[i] >> 4];
506 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f];
508 printf(" MD5 hash: %s\n", hash);
510 #endif /* !_G_MIRROR_H_ */