2 * Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
32 #include <sys/endian.h>
35 #define G_MIRROR_CLASS_NAME "MIRROR"
37 #define G_MIRROR_MAGIC "GEOM::MIRROR"
40 * 0 - Initial version number.
41 * 1 - Added 'prefer' balance algorithm.
42 * 2 - Added md_genid field to metadata.
43 * 3 - Added md_provsize field to metadata.
44 * 4 - Added 'no failure synchronization' flag.
46 #define G_MIRROR_VERSION 4
48 #define G_MIRROR_BALANCE_NONE 0
49 #define G_MIRROR_BALANCE_ROUND_ROBIN 1
50 #define G_MIRROR_BALANCE_LOAD 2
51 #define G_MIRROR_BALANCE_SPLIT 3
52 #define G_MIRROR_BALANCE_PREFER 4
53 #define G_MIRROR_BALANCE_MIN G_MIRROR_BALANCE_NONE
54 #define G_MIRROR_BALANCE_MAX G_MIRROR_BALANCE_PREFER
56 #define G_MIRROR_DISK_FLAG_DIRTY 0x0000000000000001ULL
57 #define G_MIRROR_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL
58 #define G_MIRROR_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL
59 #define G_MIRROR_DISK_FLAG_INACTIVE 0x0000000000000008ULL
60 #define G_MIRROR_DISK_FLAG_HARDCODED 0x0000000000000010ULL
61 #define G_MIRROR_DISK_FLAG_BROKEN 0x0000000000000020ULL
62 #define G_MIRROR_DISK_FLAG_CANDELETE 0x0000000000000040ULL
63 #define G_MIRROR_DISK_FLAG_MASK (G_MIRROR_DISK_FLAG_DIRTY | \
64 G_MIRROR_DISK_FLAG_SYNCHRONIZING | \
65 G_MIRROR_DISK_FLAG_FORCE_SYNC | \
66 G_MIRROR_DISK_FLAG_INACTIVE | \
67 G_MIRROR_DISK_FLAG_CANDELETE)
69 #define G_MIRROR_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL
70 #define G_MIRROR_DEVICE_FLAG_NOFAILSYNC 0x0000000000000002ULL
71 #define G_MIRROR_DEVICE_FLAG_MASK (G_MIRROR_DEVICE_FLAG_NOAUTOSYNC | \
72 G_MIRROR_DEVICE_FLAG_NOFAILSYNC)
75 extern u_int g_mirror_debug;
77 #define G_MIRROR_DEBUG(lvl, ...) do { \
78 if (g_mirror_debug >= (lvl)) { \
79 printf("GEOM_MIRROR"); \
80 if (g_mirror_debug > 0) \
81 printf("[%u]", lvl); \
83 printf(__VA_ARGS__); \
87 #define G_MIRROR_LOGREQ(lvl, bp, ...) do { \
88 if (g_mirror_debug >= (lvl)) { \
89 printf("GEOM_MIRROR"); \
90 if (g_mirror_debug > 0) \
91 printf("[%u]", lvl); \
93 printf(__VA_ARGS__); \
100 #define G_MIRROR_BIO_FLAG_REGULAR 0x01
101 #define G_MIRROR_BIO_FLAG_SYNC 0x02
104 * Informations needed for synchronization.
106 struct g_mirror_disk_sync {
107 struct g_consumer *ds_consumer; /* Consumer connected to our mirror. */
108 off_t ds_offset; /* Offset of next request to send. */
109 off_t ds_offset_done; /* Offset of already synchronized
111 u_int ds_syncid; /* Disk's synchronization ID. */
112 u_int ds_inflight; /* Number of in-flight sync requests. */
113 struct bio **ds_bios; /* BIOs for synchronization I/O. */
117 * Informations needed for synchronization.
119 struct g_mirror_device_sync {
120 struct g_geom *ds_geom; /* Synchronization geom. */
121 u_int ds_ndisks; /* Number of disks in SYNCHRONIZING
125 #define G_MIRROR_DISK_STATE_NONE 0
126 #define G_MIRROR_DISK_STATE_NEW 1
127 #define G_MIRROR_DISK_STATE_ACTIVE 2
128 #define G_MIRROR_DISK_STATE_STALE 3
129 #define G_MIRROR_DISK_STATE_SYNCHRONIZING 4
130 #define G_MIRROR_DISK_STATE_DISCONNECTED 5
131 #define G_MIRROR_DISK_STATE_DESTROY 6
132 struct g_mirror_disk {
133 uint32_t d_id; /* Disk ID. */
134 struct g_consumer *d_consumer; /* Consumer. */
135 struct g_mirror_softc *d_softc; /* Back-pointer to softc. */
136 int d_state; /* Disk state. */
137 u_int d_priority; /* Disk priority. */
138 u_int load; /* Averaged queue length */
139 off_t d_last_offset; /* Last read offset */
140 uint64_t d_flags; /* Additional flags. */
141 u_int d_genid; /* Disk's generation ID. */
142 struct g_mirror_disk_sync d_sync;/* Sync information. */
143 LIST_ENTRY(g_mirror_disk) d_next;
145 #define d_name d_consumer->provider->name
147 #define G_MIRROR_EVENT_DONTWAIT 0x1
148 #define G_MIRROR_EVENT_WAIT 0x2
149 #define G_MIRROR_EVENT_DEVICE 0x4
150 #define G_MIRROR_EVENT_DONE 0x8
151 struct g_mirror_event {
152 struct g_mirror_disk *e_disk;
156 TAILQ_ENTRY(g_mirror_event) e_next;
159 #define G_MIRROR_DEVICE_FLAG_DESTROY 0x0100000000000000ULL
160 #define G_MIRROR_DEVICE_FLAG_WAIT 0x0200000000000000ULL
161 #define G_MIRROR_DEVICE_FLAG_DESTROYING 0x0400000000000000ULL
162 #define G_MIRROR_DEVICE_FLAG_TASTING 0x0800000000000000ULL
163 #define G_MIRROR_DEVICE_FLAG_WIPE 0x1000000000000000ULL
165 #define G_MIRROR_DEVICE_STATE_STARTING 0
166 #define G_MIRROR_DEVICE_STATE_RUNNING 1
168 /* Bump syncid on first write. */
169 #define G_MIRROR_BUMP_SYNCID 0x1
170 /* Bump genid immediately. */
171 #define G_MIRROR_BUMP_GENID 0x2
172 struct g_mirror_softc {
173 u_int sc_state; /* Device state. */
174 uint32_t sc_slice; /* Slice size. */
175 uint8_t sc_balance; /* Balance algorithm. */
176 uint64_t sc_mediasize; /* Device size. */
177 uint32_t sc_sectorsize; /* Sector size. */
178 uint64_t sc_flags; /* Additional flags. */
180 struct g_geom *sc_geom;
181 struct g_provider *sc_provider;
183 uint32_t sc_id; /* Mirror unique ID. */
186 struct bio_queue_head sc_queue;
187 struct mtx sc_queue_mtx;
188 struct proc *sc_worker;
189 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due
192 struct bio_queue_head sc_inflight; /* In-flight regular write
194 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due
195 collision with regular
198 LIST_HEAD(, g_mirror_disk) sc_disks;
199 u_int sc_ndisks; /* Number of disks. */
200 struct g_mirror_disk *sc_hint;
202 u_int sc_genid; /* Generation ID. */
203 u_int sc_syncid; /* Synchronization ID. */
205 struct g_mirror_device_sync sc_sync;
206 int sc_idle; /* DIRTY flags removed. */
207 time_t sc_last_write;
210 TAILQ_HEAD(, g_mirror_event) sc_events;
211 struct mtx sc_events_mtx;
213 struct callout sc_callout;
215 struct root_hold_token *sc_rootmount;
217 struct mtx sc_done_mtx;
219 #define sc_name sc_geom->name
221 u_int g_mirror_ndisks(struct g_mirror_softc *sc, int state);
222 #define G_MIRROR_DESTROY_SOFT 0
223 #define G_MIRROR_DESTROY_DELAYED 1
224 #define G_MIRROR_DESTROY_HARD 2
225 int g_mirror_destroy(struct g_mirror_softc *sc, int how);
226 int g_mirror_event_send(void *arg, int state, int flags);
227 struct g_mirror_metadata;
228 int g_mirror_add_disk(struct g_mirror_softc *sc, struct g_provider *pp,
229 struct g_mirror_metadata *md);
230 int g_mirror_read_metadata(struct g_consumer *cp, struct g_mirror_metadata *md);
231 void g_mirror_fill_metadata(struct g_mirror_softc *sc,
232 struct g_mirror_disk *disk, struct g_mirror_metadata *md);
233 void g_mirror_update_metadata(struct g_mirror_disk *disk);
235 g_ctl_req_t g_mirror_config;
238 struct g_mirror_metadata {
239 char md_magic[16]; /* Magic value. */
240 uint32_t md_version; /* Version number. */
241 char md_name[16]; /* Mirror name. */
242 uint32_t md_mid; /* Mirror unique ID. */
243 uint32_t md_did; /* Disk unique ID. */
244 uint8_t md_all; /* Number of disks in mirror. */
245 uint32_t md_genid; /* Generation ID. */
246 uint32_t md_syncid; /* Synchronization ID. */
247 uint8_t md_priority; /* Disk priority. */
248 uint32_t md_slice; /* Slice size. */
249 uint8_t md_balance; /* Balance type. */
250 uint64_t md_mediasize; /* Size of the smallest
252 uint32_t md_sectorsize; /* Sector size. */
253 uint64_t md_sync_offset; /* Synchronized offset. */
254 uint64_t md_mflags; /* Additional mirror flags. */
255 uint64_t md_dflags; /* Additional disk flags. */
256 char md_provider[16]; /* Hardcoded provider. */
257 uint64_t md_provsize; /* Provider's size. */
258 u_char md_hash[16]; /* MD5 hash. */
261 mirror_metadata_encode(struct g_mirror_metadata *md, u_char *data)
265 bcopy(md->md_magic, data, 16);
266 le32enc(data + 16, md->md_version);
267 bcopy(md->md_name, data + 20, 16);
268 le32enc(data + 36, md->md_mid);
269 le32enc(data + 40, md->md_did);
270 *(data + 44) = md->md_all;
271 le32enc(data + 45, md->md_genid);
272 le32enc(data + 49, md->md_syncid);
273 *(data + 53) = md->md_priority;
274 le32enc(data + 54, md->md_slice);
275 *(data + 58) = md->md_balance;
276 le64enc(data + 59, md->md_mediasize);
277 le32enc(data + 67, md->md_sectorsize);
278 le64enc(data + 71, md->md_sync_offset);
279 le64enc(data + 79, md->md_mflags);
280 le64enc(data + 87, md->md_dflags);
281 bcopy(md->md_provider, data + 95, 16);
282 le64enc(data + 111, md->md_provsize);
284 MD5Update(&ctx, data, 119);
285 MD5Final(md->md_hash, &ctx);
286 bcopy(md->md_hash, data + 119, 16);
289 mirror_metadata_decode_v0v1(const u_char *data, struct g_mirror_metadata *md)
293 bcopy(data + 20, md->md_name, 16);
294 md->md_mid = le32dec(data + 36);
295 md->md_did = le32dec(data + 40);
296 md->md_all = *(data + 44);
297 md->md_syncid = le32dec(data + 45);
298 md->md_priority = *(data + 49);
299 md->md_slice = le32dec(data + 50);
300 md->md_balance = *(data + 54);
301 md->md_mediasize = le64dec(data + 55);
302 md->md_sectorsize = le32dec(data + 63);
303 md->md_sync_offset = le64dec(data + 67);
304 md->md_mflags = le64dec(data + 75);
305 md->md_dflags = le64dec(data + 83);
306 bcopy(data + 91, md->md_provider, 16);
307 bcopy(data + 107, md->md_hash, 16);
309 MD5Update(&ctx, data, 107);
310 MD5Final(md->md_hash, &ctx);
311 if (bcmp(md->md_hash, data + 107, 16) != 0)
321 mirror_metadata_decode_v2(const u_char *data, struct g_mirror_metadata *md)
325 bcopy(data + 20, md->md_name, 16);
326 md->md_mid = le32dec(data + 36);
327 md->md_did = le32dec(data + 40);
328 md->md_all = *(data + 44);
329 md->md_genid = le32dec(data + 45);
330 md->md_syncid = le32dec(data + 49);
331 md->md_priority = *(data + 53);
332 md->md_slice = le32dec(data + 54);
333 md->md_balance = *(data + 58);
334 md->md_mediasize = le64dec(data + 59);
335 md->md_sectorsize = le32dec(data + 67);
336 md->md_sync_offset = le64dec(data + 71);
337 md->md_mflags = le64dec(data + 79);
338 md->md_dflags = le64dec(data + 87);
339 bcopy(data + 95, md->md_provider, 16);
340 bcopy(data + 111, md->md_hash, 16);
342 MD5Update(&ctx, data, 111);
343 MD5Final(md->md_hash, &ctx);
344 if (bcmp(md->md_hash, data + 111, 16) != 0)
353 mirror_metadata_decode_v3v4(const u_char *data, struct g_mirror_metadata *md)
357 bcopy(data + 20, md->md_name, 16);
358 md->md_mid = le32dec(data + 36);
359 md->md_did = le32dec(data + 40);
360 md->md_all = *(data + 44);
361 md->md_genid = le32dec(data + 45);
362 md->md_syncid = le32dec(data + 49);
363 md->md_priority = *(data + 53);
364 md->md_slice = le32dec(data + 54);
365 md->md_balance = *(data + 58);
366 md->md_mediasize = le64dec(data + 59);
367 md->md_sectorsize = le32dec(data + 67);
368 md->md_sync_offset = le64dec(data + 71);
369 md->md_mflags = le64dec(data + 79);
370 md->md_dflags = le64dec(data + 87);
371 bcopy(data + 95, md->md_provider, 16);
372 md->md_provsize = le64dec(data + 111);
373 bcopy(data + 119, md->md_hash, 16);
375 MD5Update(&ctx, data, 119);
376 MD5Final(md->md_hash, &ctx);
377 if (bcmp(md->md_hash, data + 119, 16) != 0)
382 mirror_metadata_decode(const u_char *data, struct g_mirror_metadata *md)
386 bcopy(data, md->md_magic, 16);
387 md->md_version = le32dec(data + 16);
388 switch (md->md_version) {
391 error = mirror_metadata_decode_v0v1(data, md);
394 error = mirror_metadata_decode_v2(data, md);
398 error = mirror_metadata_decode_v3v4(data, md);
407 static __inline const char *
408 balance_name(u_int balance)
410 static const char *algorithms[] = {
411 [G_MIRROR_BALANCE_NONE] = "none",
412 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin",
413 [G_MIRROR_BALANCE_LOAD] = "load",
414 [G_MIRROR_BALANCE_SPLIT] = "split",
415 [G_MIRROR_BALANCE_PREFER] = "prefer",
416 [G_MIRROR_BALANCE_MAX + 1] = "unknown"
419 if (balance > G_MIRROR_BALANCE_MAX)
420 balance = G_MIRROR_BALANCE_MAX + 1;
422 return (algorithms[balance]);
426 balance_id(const char *name)
428 static const char *algorithms[] = {
429 [G_MIRROR_BALANCE_NONE] = "none",
430 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin",
431 [G_MIRROR_BALANCE_LOAD] = "load",
432 [G_MIRROR_BALANCE_SPLIT] = "split",
433 [G_MIRROR_BALANCE_PREFER] = "prefer"
437 for (n = G_MIRROR_BALANCE_MIN; n <= G_MIRROR_BALANCE_MAX; n++) {
438 if (strcmp(name, algorithms[n]) == 0)
445 mirror_metadata_dump(const struct g_mirror_metadata *md)
447 static const char hex[] = "0123456789abcdef";
448 char hash[16 * 2 + 1];
451 printf(" magic: %s\n", md->md_magic);
452 printf(" version: %u\n", (u_int)md->md_version);
453 printf(" name: %s\n", md->md_name);
454 printf(" mid: %u\n", (u_int)md->md_mid);
455 printf(" did: %u\n", (u_int)md->md_did);
456 printf(" all: %u\n", (u_int)md->md_all);
457 printf(" genid: %u\n", (u_int)md->md_genid);
458 printf(" syncid: %u\n", (u_int)md->md_syncid);
459 printf(" priority: %u\n", (u_int)md->md_priority);
460 printf(" slice: %u\n", (u_int)md->md_slice);
461 printf(" balance: %s\n", balance_name((u_int)md->md_balance));
462 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize);
463 printf("sectorsize: %u\n", (u_int)md->md_sectorsize);
464 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset);
466 if (md->md_mflags == 0)
469 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOFAILSYNC) != 0)
470 printf(" NOFAILSYNC");
471 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) != 0)
472 printf(" NOAUTOSYNC");
476 if (md->md_dflags == 0)
479 if ((md->md_dflags & G_MIRROR_DISK_FLAG_DIRTY) != 0)
481 if ((md->md_dflags & G_MIRROR_DISK_FLAG_SYNCHRONIZING) != 0)
482 printf(" SYNCHRONIZING");
483 if ((md->md_dflags & G_MIRROR_DISK_FLAG_FORCE_SYNC) != 0)
484 printf(" FORCE_SYNC");
485 if ((md->md_dflags & G_MIRROR_DISK_FLAG_INACTIVE) != 0)
489 printf("hcprovider: %s\n", md->md_provider);
490 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize);
491 bzero(hash, sizeof(hash));
492 for (i = 0; i < 16; i++) {
493 hash[i * 2] = hex[md->md_hash[i] >> 4];
494 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f];
496 printf(" MD5 hash: %s\n", hash);
498 #endif /* !_G_MIRROR_H_ */