master
1/*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@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#ifndef _G_RAID3_H_
30#define _G_RAID3_H_
31
32#include <sys/endian.h>
33#include <sys/md5.h>
34
35#define G_RAID3_CLASS_NAME "RAID3"
36
37#define G_RAID3_MAGIC "GEOM::RAID3"
38/*
39 * Version history:
40 * 0 - Initial version number.
41 * 1 - Added 'round-robin reading' algorithm.
42 * 2 - Added 'verify reading' algorithm.
43 * 3 - Added md_genid field to metadata.
44 * 4 - Added md_provsize field to metadata.
45 * 5 - Added 'no failure synchronization' flag.
46 */
47#define G_RAID3_VERSION 5
48
49#define G_RAID3_DISK_FLAG_DIRTY 0x0000000000000001ULL
50#define G_RAID3_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL
51#define G_RAID3_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL
52#define G_RAID3_DISK_FLAG_HARDCODED 0x0000000000000008ULL
53#define G_RAID3_DISK_FLAG_BROKEN 0x0000000000000010ULL
54#define G_RAID3_DISK_FLAG_MASK (G_RAID3_DISK_FLAG_DIRTY | \
55 G_RAID3_DISK_FLAG_SYNCHRONIZING | \
56 G_RAID3_DISK_FLAG_FORCE_SYNC)
57
58#define G_RAID3_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL
59#define G_RAID3_DEVICE_FLAG_ROUND_ROBIN 0x0000000000000002ULL
60#define G_RAID3_DEVICE_FLAG_VERIFY 0x0000000000000004ULL
61#define G_RAID3_DEVICE_FLAG_NOFAILSYNC 0x0000000000000008ULL
62#define G_RAID3_DEVICE_FLAG_MASK (G_RAID3_DEVICE_FLAG_NOAUTOSYNC | \
63 G_RAID3_DEVICE_FLAG_ROUND_ROBIN | \
64 G_RAID3_DEVICE_FLAG_VERIFY | \
65 G_RAID3_DEVICE_FLAG_NOFAILSYNC)
66
67#ifdef _KERNEL
68extern u_int g_raid3_debug;
69
70#define G_RAID3_DEBUG(lvl, ...) \
71 _GEOM_DEBUG("GEOM_RAID3", g_raid3_debug, (lvl), NULL, __VA_ARGS__)
72#define G_RAID3_LOGREQ(lvl, bp, ...) \
73 _GEOM_DEBUG("GEOM_RAID3", g_raid3_debug, (lvl), (bp), __VA_ARGS__)
74
75#define G_RAID3_BIO_CFLAG_REGULAR 0x01
76#define G_RAID3_BIO_CFLAG_SYNC 0x02
77#define G_RAID3_BIO_CFLAG_PARITY 0x04
78#define G_RAID3_BIO_CFLAG_NODISK 0x08
79#define G_RAID3_BIO_CFLAG_REGSYNC 0x10
80#define G_RAID3_BIO_CFLAG_MASK (G_RAID3_BIO_CFLAG_REGULAR | \
81 G_RAID3_BIO_CFLAG_SYNC | \
82 G_RAID3_BIO_CFLAG_PARITY | \
83 G_RAID3_BIO_CFLAG_NODISK | \
84 G_RAID3_BIO_CFLAG_REGSYNC)
85
86#define G_RAID3_BIO_PFLAG_DEGRADED 0x01
87#define G_RAID3_BIO_PFLAG_NOPARITY 0x02
88#define G_RAID3_BIO_PFLAG_VERIFY 0x04
89#define G_RAID3_BIO_PFLAG_MASK (G_RAID3_BIO_PFLAG_DEGRADED | \
90 G_RAID3_BIO_PFLAG_NOPARITY | \
91 G_RAID3_BIO_PFLAG_VERIFY)
92
93/*
94 * Informations needed for synchronization.
95 */
96struct g_raid3_disk_sync {
97 struct g_consumer *ds_consumer; /* Consumer connected to our device. */
98 off_t ds_offset; /* Offset of next request to send. */
99 off_t ds_offset_done; /* Offset of already synchronized
100 region. */
101 off_t ds_resync; /* Resynchronize from this offset. */
102 u_int ds_syncid; /* Disk's synchronization ID. */
103 u_int ds_inflight; /* Number of in-flight sync requests. */
104 struct bio **ds_bios; /* BIOs for synchronization I/O. */
105};
106
107/*
108 * Informations needed for synchronization.
109 */
110struct g_raid3_device_sync {
111 struct g_geom *ds_geom; /* Synchronization geom. */
112};
113
114#define G_RAID3_DISK_STATE_NODISK 0
115#define G_RAID3_DISK_STATE_NONE 1
116#define G_RAID3_DISK_STATE_NEW 2
117#define G_RAID3_DISK_STATE_ACTIVE 3
118#define G_RAID3_DISK_STATE_STALE 4
119#define G_RAID3_DISK_STATE_SYNCHRONIZING 5
120#define G_RAID3_DISK_STATE_DISCONNECTED 6
121#define G_RAID3_DISK_STATE_DESTROY 7
122struct g_raid3_disk {
123 u_int d_no; /* Disk number. */
124 struct g_consumer *d_consumer; /* Consumer. */
125 struct g_raid3_softc *d_softc; /* Back-pointer to softc. */
126 int d_state; /* Disk state. */
127 uint64_t d_flags; /* Additional flags. */
128 u_int d_genid; /* Disk's generation ID. */
129 struct g_raid3_disk_sync d_sync; /* Sync information. */
130 LIST_ENTRY(g_raid3_disk) d_next;
131};
132#define d_name d_consumer->provider->name
133
134#define G_RAID3_EVENT_DONTWAIT 0x1
135#define G_RAID3_EVENT_WAIT 0x2
136#define G_RAID3_EVENT_DEVICE 0x4
137#define G_RAID3_EVENT_DONE 0x8
138struct g_raid3_event {
139 struct g_raid3_disk *e_disk;
140 int e_state;
141 int e_flags;
142 int e_error;
143 TAILQ_ENTRY(g_raid3_event) e_next;
144};
145
146#define G_RAID3_DEVICE_FLAG_DESTROY 0x0100000000000000ULL
147#define G_RAID3_DEVICE_FLAG_WAIT 0x0200000000000000ULL
148#define G_RAID3_DEVICE_FLAG_DESTROYING 0x0400000000000000ULL
149
150#define G_RAID3_DEVICE_STATE_STARTING 0
151#define G_RAID3_DEVICE_STATE_DEGRADED 1
152#define G_RAID3_DEVICE_STATE_COMPLETE 2
153
154/* Bump syncid on first write. */
155#define G_RAID3_BUMP_SYNCID 0x1
156/* Bump genid immediately. */
157#define G_RAID3_BUMP_GENID 0x2
158
159enum g_raid3_zones {
160 G_RAID3_ZONE_64K,
161 G_RAID3_ZONE_16K,
162 G_RAID3_ZONE_4K,
163 G_RAID3_NUM_ZONES
164};
165
166static __inline enum g_raid3_zones
167g_raid3_zone(size_t nbytes) {
168 if (nbytes > 65536)
169 return (G_RAID3_NUM_ZONES);
170 else if (nbytes > 16384)
171 return (G_RAID3_ZONE_64K);
172 else if (nbytes > 4096)
173 return (G_RAID3_ZONE_16K);
174 else
175 return (G_RAID3_ZONE_4K);
176};
177
178struct g_raid3_softc {
179 u_int sc_state; /* Device state. */
180 uint64_t sc_mediasize; /* Device size. */
181 uint32_t sc_sectorsize; /* Sector size. */
182 uint64_t sc_flags; /* Additional flags. */
183
184 struct g_geom *sc_geom;
185 struct g_provider *sc_provider;
186
187 uint32_t sc_id; /* Device unique ID. */
188
189 struct sx sc_lock;
190 struct bio_queue_head sc_queue;
191 struct mtx sc_queue_mtx;
192 struct proc *sc_worker;
193 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due
194 collision with sync
195 requests. */
196 struct bio_queue_head sc_inflight; /* In-flight regular write
197 requests. */
198 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due
199 collision with regular
200 requests. */
201
202 struct g_raid3_disk *sc_disks;
203 u_int sc_ndisks; /* Number of disks. */
204 u_int sc_round_robin;
205 struct g_raid3_disk *sc_syncdisk;
206
207 struct g_raid3_zone {
208 uma_zone_t sz_zone;
209 size_t sz_inuse;
210 size_t sz_max;
211 u_int sz_requested;
212 u_int sz_failed;
213 } sc_zones[G_RAID3_NUM_ZONES];
214
215 u_int sc_genid; /* Generation ID. */
216 u_int sc_syncid; /* Synchronization ID. */
217 int sc_bump_id;
218 struct g_raid3_device_sync sc_sync;
219 int sc_idle; /* DIRTY flags removed. */
220 time_t sc_last_write;
221 u_int sc_writes;
222 u_int sc_refcnt; /* Number of softc references. */
223
224 TAILQ_HEAD(, g_raid3_event) sc_events;
225 struct mtx sc_events_mtx;
226 struct g_raid3_event *sc_timeout_event;
227
228 struct callout sc_callout;
229
230 struct root_hold_token *sc_rootmount;
231};
232#define sc_name sc_geom->name
233
234const char *g_raid3_get_diskname(struct g_raid3_disk *disk);
235u_int g_raid3_ndisks(struct g_raid3_softc *sc, int state);
236#define G_RAID3_DESTROY_SOFT 0
237#define G_RAID3_DESTROY_DELAYED 1
238#define G_RAID3_DESTROY_HARD 2
239int g_raid3_destroy(struct g_raid3_softc *sc, int how);
240int g_raid3_event_send(void *arg, int state, int flags);
241struct g_raid3_metadata;
242int g_raid3_add_disk(struct g_raid3_softc *sc, struct g_provider *pp,
243 struct g_raid3_metadata *md);
244int g_raid3_read_metadata(struct g_consumer *cp, struct g_raid3_metadata *md);
245void g_raid3_fill_metadata(struct g_raid3_disk *disk,
246 struct g_raid3_metadata *md);
247int g_raid3_clear_metadata(struct g_raid3_disk *disk);
248void g_raid3_update_metadata(struct g_raid3_disk *disk);
249
250g_ctl_req_t g_raid3_config;
251#endif /* _KERNEL */
252
253struct g_raid3_metadata {
254 char md_magic[16]; /* Magic value. */
255 uint32_t md_version; /* Version number. */
256 char md_name[16]; /* Device name. */
257 uint32_t md_id; /* Device unique ID. */
258 uint16_t md_no; /* Component number. */
259 uint16_t md_all; /* Number of disks in device. */
260 uint32_t md_genid; /* Generation ID. */
261 uint32_t md_syncid; /* Synchronization ID. */
262 uint64_t md_mediasize; /* Size of whole device. */
263 uint32_t md_sectorsize; /* Sector size. */
264 uint64_t md_sync_offset; /* Synchronized offset. */
265 uint64_t md_mflags; /* Additional device flags. */
266 uint64_t md_dflags; /* Additional disk flags. */
267 char md_provider[16]; /* Hardcoded provider. */
268 uint64_t md_provsize; /* Provider's size. */
269 u_char md_hash[16]; /* MD5 hash. */
270};
271static __inline void
272raid3_metadata_encode(struct g_raid3_metadata *md, u_char *data)
273{
274 MD5_CTX ctx;
275
276 bcopy(md->md_magic, data, 16);
277 le32enc(data + 16, md->md_version);
278 bcopy(md->md_name, data + 20, 16);
279 le32enc(data + 36, md->md_id);
280 le16enc(data + 40, md->md_no);
281 le16enc(data + 42, md->md_all);
282 le32enc(data + 44, md->md_genid);
283 le32enc(data + 48, md->md_syncid);
284 le64enc(data + 52, md->md_mediasize);
285 le32enc(data + 60, md->md_sectorsize);
286 le64enc(data + 64, md->md_sync_offset);
287 le64enc(data + 72, md->md_mflags);
288 le64enc(data + 80, md->md_dflags);
289 bcopy(md->md_provider, data + 88, 16);
290 le64enc(data + 104, md->md_provsize);
291 MD5Init(&ctx);
292 MD5Update(&ctx, data, 112);
293 MD5Final(md->md_hash, &ctx);
294 bcopy(md->md_hash, data + 112, 16);
295}
296static __inline int
297raid3_metadata_decode_v0v1v2(const u_char *data, struct g_raid3_metadata *md)
298{
299 MD5_CTX ctx;
300
301 bcopy(data + 20, md->md_name, 16);
302 md->md_id = le32dec(data + 36);
303 md->md_no = le16dec(data + 40);
304 md->md_all = le16dec(data + 42);
305 md->md_syncid = le32dec(data + 44);
306 md->md_mediasize = le64dec(data + 48);
307 md->md_sectorsize = le32dec(data + 56);
308 md->md_sync_offset = le64dec(data + 60);
309 md->md_mflags = le64dec(data + 68);
310 md->md_dflags = le64dec(data + 76);
311 bcopy(data + 84, md->md_provider, 16);
312 bcopy(data + 100, md->md_hash, 16);
313 MD5Init(&ctx);
314 MD5Update(&ctx, data, 100);
315 MD5Final(md->md_hash, &ctx);
316 if (bcmp(md->md_hash, data + 100, 16) != 0)
317 return (EINVAL);
318
319 /* New fields. */
320 md->md_genid = 0;
321 md->md_provsize = 0;
322
323 return (0);
324}
325static __inline int
326raid3_metadata_decode_v3(const u_char *data, struct g_raid3_metadata *md)
327{
328 MD5_CTX ctx;
329
330 bcopy(data + 20, md->md_name, 16);
331 md->md_id = le32dec(data + 36);
332 md->md_no = le16dec(data + 40);
333 md->md_all = le16dec(data + 42);
334 md->md_genid = le32dec(data + 44);
335 md->md_syncid = le32dec(data + 48);
336 md->md_mediasize = le64dec(data + 52);
337 md->md_sectorsize = le32dec(data + 60);
338 md->md_sync_offset = le64dec(data + 64);
339 md->md_mflags = le64dec(data + 72);
340 md->md_dflags = le64dec(data + 80);
341 bcopy(data + 88, md->md_provider, 16);
342 bcopy(data + 104, md->md_hash, 16);
343 MD5Init(&ctx);
344 MD5Update(&ctx, data, 104);
345 MD5Final(md->md_hash, &ctx);
346 if (bcmp(md->md_hash, data + 104, 16) != 0)
347 return (EINVAL);
348
349 /* New fields. */
350 md->md_provsize = 0;
351
352 return (0);
353}
354static __inline int
355raid3_metadata_decode_v4v5(const u_char *data, struct g_raid3_metadata *md)
356{
357 MD5_CTX ctx;
358
359 bcopy(data + 20, md->md_name, 16);
360 md->md_id = le32dec(data + 36);
361 md->md_no = le16dec(data + 40);
362 md->md_all = le16dec(data + 42);
363 md->md_genid = le32dec(data + 44);
364 md->md_syncid = le32dec(data + 48);
365 md->md_mediasize = le64dec(data + 52);
366 md->md_sectorsize = le32dec(data + 60);
367 md->md_sync_offset = le64dec(data + 64);
368 md->md_mflags = le64dec(data + 72);
369 md->md_dflags = le64dec(data + 80);
370 bcopy(data + 88, md->md_provider, 16);
371 md->md_provsize = le64dec(data + 104);
372 bcopy(data + 112, md->md_hash, 16);
373 MD5Init(&ctx);
374 MD5Update(&ctx, data, 112);
375 MD5Final(md->md_hash, &ctx);
376 if (bcmp(md->md_hash, data + 112, 16) != 0)
377 return (EINVAL);
378 return (0);
379}
380static __inline int
381raid3_metadata_decode(const u_char *data, struct g_raid3_metadata *md)
382{
383 int error;
384
385 bcopy(data, md->md_magic, 16);
386 md->md_version = le32dec(data + 16);
387 switch (md->md_version) {
388 case 0:
389 case 1:
390 case 2:
391 error = raid3_metadata_decode_v0v1v2(data, md);
392 break;
393 case 3:
394 error = raid3_metadata_decode_v3(data, md);
395 break;
396 case 4:
397 case 5:
398 error = raid3_metadata_decode_v4v5(data, md);
399 break;
400 default:
401 error = EINVAL;
402 break;
403 }
404 return (error);
405}
406
407static __inline void
408raid3_metadata_dump(const struct g_raid3_metadata *md)
409{
410 static const char hex[] = "0123456789abcdef";
411 char hash[16 * 2 + 1];
412 u_int i;
413
414 printf(" magic: %s\n", md->md_magic);
415 printf(" version: %u\n", (u_int)md->md_version);
416 printf(" name: %s\n", md->md_name);
417 printf(" id: %u\n", (u_int)md->md_id);
418 printf(" no: %u\n", (u_int)md->md_no);
419 printf(" all: %u\n", (u_int)md->md_all);
420 printf(" genid: %u\n", (u_int)md->md_genid);
421 printf(" syncid: %u\n", (u_int)md->md_syncid);
422 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize);
423 printf("sectorsize: %u\n", (u_int)md->md_sectorsize);
424 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset);
425 printf(" mflags:");
426 if (md->md_mflags == 0)
427 printf(" NONE");
428 else {
429 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0)
430 printf(" NOAUTOSYNC");
431 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0)
432 printf(" ROUND-ROBIN");
433 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_VERIFY) != 0)
434 printf(" VERIFY");
435 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) != 0)
436 printf(" NOFAILSYNC");
437 }
438 printf("\n");
439 printf(" dflags:");
440 if (md->md_dflags == 0)
441 printf(" NONE");
442 else {
443 if ((md->md_dflags & G_RAID3_DISK_FLAG_DIRTY) != 0)
444 printf(" DIRTY");
445 if ((md->md_dflags & G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0)
446 printf(" SYNCHRONIZING");
447 if ((md->md_dflags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0)
448 printf(" FORCE_SYNC");
449 }
450 printf("\n");
451 printf("hcprovider: %s\n", md->md_provider);
452 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize);
453 bzero(hash, sizeof(hash));
454 for (i = 0; i < 16; i++) {
455 hash[i * 2] = hex[md->md_hash[i] >> 4];
456 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f];
457 }
458 printf(" MD5 hash: %s\n", hash);
459}
460#endif /* !_G_RAID3_H_ */