Commit | Line | Data |
---|---|---|
9d09e663 N |
1 | /* |
2 | * Copyright (C) 2010-2011 Neil Brown | |
6e53636f | 3 | * Copyright (C) 2010-2017 Red Hat, Inc. All rights reserved. |
9d09e663 N |
4 | * |
5 | * This file is released under the GPL. | |
6 | */ | |
7 | ||
8 | #include <linux/slab.h> | |
056075c7 | 9 | #include <linux/module.h> |
9d09e663 N |
10 | |
11 | #include "md.h" | |
32737279 | 12 | #include "raid1.h" |
9d09e663 | 13 | #include "raid5.h" |
63f33b8d | 14 | #include "raid10.h" |
9d09e663 N |
15 | #include "bitmap.h" |
16 | ||
3e8dbb7f AK |
17 | #include <linux/device-mapper.h> |
18 | ||
9d09e663 | 19 | #define DM_MSG_PREFIX "raid" |
92c83d79 | 20 | #define MAX_RAID_DEVICES 253 /* md-raid kernel limit */ |
9d09e663 | 21 | |
40ba37e5 HM |
22 | /* |
23 | * Minimum sectors of free reshape space per raid device | |
24 | */ | |
25 | #define MIN_FREE_RESHAPE_SPACE to_sector(4*4096) | |
26 | ||
63c32ed4 HM |
27 | /* |
28 | * Minimum journal space 4 MiB in sectors. | |
29 | */ | |
30 | #define MIN_RAID456_JOURNAL_SPACE (4*2048) | |
31 | ||
48cf06bc HM |
32 | static bool devices_handle_discard_safely = false; |
33 | ||
9d09e663 | 34 | /* |
b12d437b JB |
35 | * The following flags are used by dm-raid.c to set up the array state. |
36 | * They must be cleared before md_run is called. | |
9d09e663 | 37 | */ |
43157840 | 38 | #define FirstUse 10 /* rdev flag */ |
9d09e663 N |
39 | |
40 | struct raid_dev { | |
41 | /* | |
42 | * Two DM devices, one to hold metadata and one to hold the | |
43157840 | 43 | * actual data/parity. The reason for this is to not confuse |
9d09e663 N |
44 | * ti->len and give more flexibility in altering size and |
45 | * characteristics. | |
46 | * | |
47 | * While it is possible for this device to be associated | |
48 | * with a different physical device than the data_dev, it | |
49 | * is intended for it to be the same. | |
50 | * |--------- Physical Device ---------| | |
51 | * |- meta_dev -|------ data_dev ------| | |
52 | */ | |
53 | struct dm_dev *meta_dev; | |
54 | struct dm_dev *data_dev; | |
3cb03002 | 55 | struct md_rdev rdev; |
9d09e663 N |
56 | }; |
57 | ||
58 | /* | |
4286325b | 59 | * Bits for establishing rs->ctr_flags |
702108d1 HM |
60 | * |
61 | * 1 = no flag value | |
62 | * 2 = flag with value | |
9d09e663 | 63 | */ |
4286325b MS |
64 | #define __CTR_FLAG_SYNC 0 /* 1 */ /* Not with raid0! */ |
65 | #define __CTR_FLAG_NOSYNC 1 /* 1 */ /* Not with raid0! */ | |
66 | #define __CTR_FLAG_REBUILD 2 /* 2 */ /* Not with raid0! */ | |
67 | #define __CTR_FLAG_DAEMON_SLEEP 3 /* 2 */ /* Not with raid0! */ | |
68 | #define __CTR_FLAG_MIN_RECOVERY_RATE 4 /* 2 */ /* Not with raid0! */ | |
69 | #define __CTR_FLAG_MAX_RECOVERY_RATE 5 /* 2 */ /* Not with raid0! */ | |
70 | #define __CTR_FLAG_MAX_WRITE_BEHIND 6 /* 2 */ /* Only with raid1! */ | |
71 | #define __CTR_FLAG_WRITE_MOSTLY 7 /* 2 */ /* Only with raid1! */ | |
72 | #define __CTR_FLAG_STRIPE_CACHE 8 /* 2 */ /* Only with raid4/5/6! */ | |
73 | #define __CTR_FLAG_REGION_SIZE 9 /* 2 */ /* Not with raid0! */ | |
74 | #define __CTR_FLAG_RAID10_COPIES 10 /* 2 */ /* Only with raid10 */ | |
75 | #define __CTR_FLAG_RAID10_FORMAT 11 /* 2 */ /* Only with raid10 */ | |
9b6e5423 | 76 | /* New for v1.9.0 */ |
d7ccc2e2 | 77 | #define __CTR_FLAG_DELTA_DISKS 12 /* 2 */ /* Only with reshapable raid1/4/5/6/10! */ |
4286325b MS |
78 | #define __CTR_FLAG_DATA_OFFSET 13 /* 2 */ /* Only with reshapable raid4/5/6/10! */ |
79 | #define __CTR_FLAG_RAID10_USE_NEAR_SETS 14 /* 2 */ /* Only with raid10! */ | |
80 | ||
63c32ed4 | 81 | /* New for v1.10.0 */ |
6e53636f HM |
82 | #define __CTR_FLAG_JOURNAL_DEV 15 /* 2 */ /* Only with raid4/5/6 (journal device)! */ |
83 | ||
84 | /* New for v1.11.1 */ | |
85 | #define __CTR_FLAG_JOURNAL_MODE 16 /* 2 */ /* Only with raid4/5/6 (journal mode)! */ | |
63c32ed4 | 86 | |
4286325b MS |
87 | /* |
88 | * Flags for rs->ctr_flags field. | |
89 | */ | |
90 | #define CTR_FLAG_SYNC (1 << __CTR_FLAG_SYNC) | |
91 | #define CTR_FLAG_NOSYNC (1 << __CTR_FLAG_NOSYNC) | |
92 | #define CTR_FLAG_REBUILD (1 << __CTR_FLAG_REBUILD) | |
93 | #define CTR_FLAG_DAEMON_SLEEP (1 << __CTR_FLAG_DAEMON_SLEEP) | |
94 | #define CTR_FLAG_MIN_RECOVERY_RATE (1 << __CTR_FLAG_MIN_RECOVERY_RATE) | |
95 | #define CTR_FLAG_MAX_RECOVERY_RATE (1 << __CTR_FLAG_MAX_RECOVERY_RATE) | |
96 | #define CTR_FLAG_MAX_WRITE_BEHIND (1 << __CTR_FLAG_MAX_WRITE_BEHIND) | |
97 | #define CTR_FLAG_WRITE_MOSTLY (1 << __CTR_FLAG_WRITE_MOSTLY) | |
98 | #define CTR_FLAG_STRIPE_CACHE (1 << __CTR_FLAG_STRIPE_CACHE) | |
99 | #define CTR_FLAG_REGION_SIZE (1 << __CTR_FLAG_REGION_SIZE) | |
100 | #define CTR_FLAG_RAID10_COPIES (1 << __CTR_FLAG_RAID10_COPIES) | |
101 | #define CTR_FLAG_RAID10_FORMAT (1 << __CTR_FLAG_RAID10_FORMAT) | |
102 | #define CTR_FLAG_DELTA_DISKS (1 << __CTR_FLAG_DELTA_DISKS) | |
103 | #define CTR_FLAG_DATA_OFFSET (1 << __CTR_FLAG_DATA_OFFSET) | |
104 | #define CTR_FLAG_RAID10_USE_NEAR_SETS (1 << __CTR_FLAG_RAID10_USE_NEAR_SETS) | |
63c32ed4 | 105 | #define CTR_FLAG_JOURNAL_DEV (1 << __CTR_FLAG_JOURNAL_DEV) |
6e53636f | 106 | #define CTR_FLAG_JOURNAL_MODE (1 << __CTR_FLAG_JOURNAL_MODE) |
63f33b8d | 107 | |
2664f3c9 MS |
108 | #define RESUME_STAY_FROZEN_FLAGS (CTR_FLAG_DELTA_DISKS | CTR_FLAG_DATA_OFFSET) |
109 | ||
f090279e HM |
110 | /* |
111 | * Definitions of various constructor flags to | |
112 | * be used in checks of valid / invalid flags | |
113 | * per raid level. | |
114 | */ | |
115 | /* Define all any sync flags */ | |
116 | #define CTR_FLAGS_ANY_SYNC (CTR_FLAG_SYNC | CTR_FLAG_NOSYNC) | |
117 | ||
118 | /* Define flags for options without argument (e.g. 'nosync') */ | |
33e53f06 HM |
119 | #define CTR_FLAG_OPTIONS_NO_ARGS (CTR_FLAGS_ANY_SYNC | \ |
120 | CTR_FLAG_RAID10_USE_NEAR_SETS) | |
f090279e HM |
121 | |
122 | /* Define flags for options with one argument (e.g. 'delta_disks +2') */ | |
123 | #define CTR_FLAG_OPTIONS_ONE_ARG (CTR_FLAG_REBUILD | \ | |
124 | CTR_FLAG_WRITE_MOSTLY | \ | |
125 | CTR_FLAG_DAEMON_SLEEP | \ | |
126 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
127 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
128 | CTR_FLAG_MAX_WRITE_BEHIND | \ | |
129 | CTR_FLAG_STRIPE_CACHE | \ | |
130 | CTR_FLAG_REGION_SIZE | \ | |
131 | CTR_FLAG_RAID10_COPIES | \ | |
33e53f06 HM |
132 | CTR_FLAG_RAID10_FORMAT | \ |
133 | CTR_FLAG_DELTA_DISKS | \ | |
134 | CTR_FLAG_DATA_OFFSET) | |
f090279e | 135 | |
a30cbc0d HM |
136 | /* Valid options definitions per raid level... */ |
137 | ||
138 | /* "raid0" does only accept data offset */ | |
139 | #define RAID0_VALID_FLAGS (CTR_FLAG_DATA_OFFSET) | |
140 | ||
141 | /* "raid1" does not accept stripe cache, data offset, delta_disks or any raid10 options */ | |
142 | #define RAID1_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ | |
143 | CTR_FLAG_REBUILD | \ | |
144 | CTR_FLAG_WRITE_MOSTLY | \ | |
145 | CTR_FLAG_DAEMON_SLEEP | \ | |
146 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
147 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
148 | CTR_FLAG_MAX_WRITE_BEHIND | \ | |
149 | CTR_FLAG_REGION_SIZE | \ | |
7a7c330f | 150 | CTR_FLAG_DELTA_DISKS | \ |
a30cbc0d | 151 | CTR_FLAG_DATA_OFFSET) |
f090279e | 152 | |
a30cbc0d HM |
153 | /* "raid10" does not accept any raid1 or stripe cache options */ |
154 | #define RAID10_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ | |
155 | CTR_FLAG_REBUILD | \ | |
156 | CTR_FLAG_DAEMON_SLEEP | \ | |
157 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
158 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
159 | CTR_FLAG_REGION_SIZE | \ | |
f090279e | 160 | CTR_FLAG_RAID10_COPIES | \ |
33e53f06 HM |
161 | CTR_FLAG_RAID10_FORMAT | \ |
162 | CTR_FLAG_DELTA_DISKS | \ | |
a30cbc0d HM |
163 | CTR_FLAG_DATA_OFFSET | \ |
164 | CTR_FLAG_RAID10_USE_NEAR_SETS) | |
f090279e | 165 | |
f090279e HM |
166 | /* |
167 | * "raid4/5/6" do not accept any raid1 or raid10 specific options | |
168 | * | |
169 | * "raid6" does not accept "nosync", because it is not guaranteed | |
170 | * that both parity and q-syndrome are being written properly with | |
171 | * any writes | |
172 | */ | |
a30cbc0d HM |
173 | #define RAID45_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ |
174 | CTR_FLAG_REBUILD | \ | |
175 | CTR_FLAG_DAEMON_SLEEP | \ | |
176 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
177 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
a30cbc0d HM |
178 | CTR_FLAG_STRIPE_CACHE | \ |
179 | CTR_FLAG_REGION_SIZE | \ | |
180 | CTR_FLAG_DELTA_DISKS | \ | |
63c32ed4 | 181 | CTR_FLAG_DATA_OFFSET | \ |
6e53636f HM |
182 | CTR_FLAG_JOURNAL_DEV | \ |
183 | CTR_FLAG_JOURNAL_MODE) | |
a30cbc0d HM |
184 | |
185 | #define RAID6_VALID_FLAGS (CTR_FLAG_SYNC | \ | |
186 | CTR_FLAG_REBUILD | \ | |
187 | CTR_FLAG_DAEMON_SLEEP | \ | |
188 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
189 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
a30cbc0d HM |
190 | CTR_FLAG_STRIPE_CACHE | \ |
191 | CTR_FLAG_REGION_SIZE | \ | |
192 | CTR_FLAG_DELTA_DISKS | \ | |
63c32ed4 | 193 | CTR_FLAG_DATA_OFFSET | \ |
6e53636f HM |
194 | CTR_FLAG_JOURNAL_DEV | \ |
195 | CTR_FLAG_JOURNAL_MODE) | |
a30cbc0d | 196 | /* ...valid options definitions per raid level */ |
f090279e | 197 | |
ecbfb9f1 HM |
198 | /* |
199 | * Flags for rs->runtime_flags field | |
200 | * (RT_FLAG prefix meaning "runtime flag") | |
201 | * | |
202 | * These are all internal and used to define runtime state, | |
203 | * e.g. to prevent another resume from preresume processing | |
204 | * the raid set all over again. | |
205 | */ | |
40ba37e5 HM |
206 | #define RT_FLAG_RS_PRERESUMED 0 |
207 | #define RT_FLAG_RS_RESUMED 1 | |
208 | #define RT_FLAG_RS_BITMAP_LOADED 2 | |
209 | #define RT_FLAG_UPDATE_SBS 3 | |
9dbd1aa3 | 210 | #define RT_FLAG_RESHAPE_RS 4 |
ecbfb9f1 | 211 | |
d7ccc2e2 | 212 | /* Array elements of 64 bit needed for rebuild/failed disk bits */ |
33e53f06 HM |
213 | #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8) |
214 | ||
ecbfb9f1 HM |
215 | /* |
216 | * raid set level, layout and chunk sectors backup/restore | |
217 | */ | |
218 | struct rs_layout { | |
219 | int new_level; | |
220 | int new_layout; | |
221 | int new_chunk_sectors; | |
222 | }; | |
223 | ||
9d09e663 N |
224 | struct raid_set { |
225 | struct dm_target *ti; | |
226 | ||
34f8ac6d | 227 | uint32_t bitmap_loaded; |
9dbd1aa3 | 228 | uint32_t stripe_cache_entries; |
4286325b MS |
229 | unsigned long ctr_flags; |
230 | unsigned long runtime_flags; | |
ecbfb9f1 HM |
231 | |
232 | uint64_t rebuild_disks[DISKS_ARRAY_ELEMS]; | |
9d09e663 | 233 | |
33e53f06 HM |
234 | int raid_disks; |
235 | int delta_disks; | |
4763e543 | 236 | int data_offset; |
33e53f06 | 237 | int raid10_copies; |
4257e085 | 238 | int requested_bitmap_chunk_sectors; |
33e53f06 | 239 | |
fd01b88c | 240 | struct mddev md; |
9d09e663 N |
241 | struct raid_type *raid_type; |
242 | struct dm_target_callbacks callbacks; | |
243 | ||
63c32ed4 HM |
244 | /* Optional raid4/5/6 journal device */ |
245 | struct journal_dev { | |
246 | struct dm_dev *dev; | |
247 | struct md_rdev rdev; | |
6e53636f | 248 | int mode; |
63c32ed4 HM |
249 | } journal_dev; |
250 | ||
9d09e663 N |
251 | struct raid_dev dev[0]; |
252 | }; | |
253 | ||
9dbd1aa3 | 254 | static void rs_config_backup(struct raid_set *rs, struct rs_layout *l) |
ecbfb9f1 HM |
255 | { |
256 | struct mddev *mddev = &rs->md; | |
257 | ||
258 | l->new_level = mddev->new_level; | |
259 | l->new_layout = mddev->new_layout; | |
260 | l->new_chunk_sectors = mddev->new_chunk_sectors; | |
261 | } | |
262 | ||
9dbd1aa3 | 263 | static void rs_config_restore(struct raid_set *rs, struct rs_layout *l) |
ecbfb9f1 HM |
264 | { |
265 | struct mddev *mddev = &rs->md; | |
266 | ||
267 | mddev->new_level = l->new_level; | |
268 | mddev->new_layout = l->new_layout; | |
269 | mddev->new_chunk_sectors = l->new_chunk_sectors; | |
270 | } | |
271 | ||
33e53f06 HM |
272 | /* raid10 algorithms (i.e. formats) */ |
273 | #define ALGORITHM_RAID10_DEFAULT 0 | |
274 | #define ALGORITHM_RAID10_NEAR 1 | |
275 | #define ALGORITHM_RAID10_OFFSET 2 | |
276 | #define ALGORITHM_RAID10_FAR 3 | |
277 | ||
9d09e663 N |
278 | /* Supported raid types and properties. */ |
279 | static struct raid_type { | |
280 | const char *name; /* RAID algorithm. */ | |
281 | const char *descr; /* Descriptor text for logging. */ | |
094f394d HM |
282 | const unsigned int parity_devs; /* # of parity devices. */ |
283 | const unsigned int minimal_devs;/* minimal # of devices in set. */ | |
284 | const unsigned int level; /* RAID level. */ | |
285 | const unsigned int algorithm; /* RAID algorithm. */ | |
9d09e663 | 286 | } raid_types[] = { |
43157840 MS |
287 | {"raid0", "raid0 (striping)", 0, 2, 0, 0 /* NONE */}, |
288 | {"raid1", "raid1 (mirroring)", 0, 2, 1, 0 /* NONE */}, | |
289 | {"raid10_far", "raid10 far (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_FAR}, | |
33e53f06 | 290 | {"raid10_offset", "raid10 offset (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_OFFSET}, |
43157840 MS |
291 | {"raid10_near", "raid10 near (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_NEAR}, |
292 | {"raid10", "raid10 (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_DEFAULT}, | |
b052b07c | 293 | {"raid4", "raid4 (dedicated first parity disk)", 1, 2, 5, ALGORITHM_PARITY_0}, /* raid4 layout = raid5_0 */ |
43157840 MS |
294 | {"raid5_n", "raid5 (dedicated last parity disk)", 1, 2, 5, ALGORITHM_PARITY_N}, |
295 | {"raid5_ls", "raid5 (left symmetric)", 1, 2, 5, ALGORITHM_LEFT_SYMMETRIC}, | |
296 | {"raid5_rs", "raid5 (right symmetric)", 1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC}, | |
297 | {"raid5_la", "raid5 (left asymmetric)", 1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC}, | |
298 | {"raid5_ra", "raid5 (right asymmetric)", 1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC}, | |
299 | {"raid6_zr", "raid6 (zero restart)", 2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART}, | |
300 | {"raid6_nr", "raid6 (N restart)", 2, 4, 6, ALGORITHM_ROTATING_N_RESTART}, | |
301 | {"raid6_nc", "raid6 (N continue)", 2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}, | |
302 | {"raid6_n_6", "raid6 (dedicated parity/Q n/6)", 2, 4, 6, ALGORITHM_PARITY_N_6}, | |
303 | {"raid6_ls_6", "raid6 (left symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_SYMMETRIC_6}, | |
304 | {"raid6_rs_6", "raid6 (right symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_SYMMETRIC_6}, | |
305 | {"raid6_la_6", "raid6 (left asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_ASYMMETRIC_6}, | |
306 | {"raid6_ra_6", "raid6 (right asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_ASYMMETRIC_6} | |
9d09e663 N |
307 | }; |
308 | ||
92c83d79 | 309 | /* True, if @v is in inclusive range [@min, @max] */ |
bb91a63f | 310 | static bool __within_range(long v, long min, long max) |
92c83d79 HM |
311 | { |
312 | return v >= min && v <= max; | |
313 | } | |
314 | ||
702108d1 HM |
315 | /* All table line arguments are defined here */ |
316 | static struct arg_name_flag { | |
4286325b | 317 | const unsigned long flag; |
702108d1 | 318 | const char *name; |
e6ca5e1a | 319 | } __arg_name_flags[] = { |
702108d1 HM |
320 | { CTR_FLAG_SYNC, "sync"}, |
321 | { CTR_FLAG_NOSYNC, "nosync"}, | |
322 | { CTR_FLAG_REBUILD, "rebuild"}, | |
323 | { CTR_FLAG_DAEMON_SLEEP, "daemon_sleep"}, | |
324 | { CTR_FLAG_MIN_RECOVERY_RATE, "min_recovery_rate"}, | |
325 | { CTR_FLAG_MAX_RECOVERY_RATE, "max_recovery_rate"}, | |
326 | { CTR_FLAG_MAX_WRITE_BEHIND, "max_write_behind"}, | |
65359ee6 | 327 | { CTR_FLAG_WRITE_MOSTLY, "write_mostly"}, |
702108d1 HM |
328 | { CTR_FLAG_STRIPE_CACHE, "stripe_cache"}, |
329 | { CTR_FLAG_REGION_SIZE, "region_size"}, | |
330 | { CTR_FLAG_RAID10_COPIES, "raid10_copies"}, | |
331 | { CTR_FLAG_RAID10_FORMAT, "raid10_format"}, | |
4763e543 HM |
332 | { CTR_FLAG_DATA_OFFSET, "data_offset"}, |
333 | { CTR_FLAG_DELTA_DISKS, "delta_disks"}, | |
334 | { CTR_FLAG_RAID10_USE_NEAR_SETS, "raid10_use_near_sets"}, | |
63c32ed4 | 335 | { CTR_FLAG_JOURNAL_DEV, "journal_dev" }, |
6e53636f | 336 | { CTR_FLAG_JOURNAL_MODE, "journal_mode" }, |
702108d1 HM |
337 | }; |
338 | ||
339 | /* Return argument name string for given @flag */ | |
3fa6cf38 | 340 | static const char *dm_raid_arg_name_by_flag(const uint32_t flag) |
702108d1 HM |
341 | { |
342 | if (hweight32(flag) == 1) { | |
e6ca5e1a | 343 | struct arg_name_flag *anf = __arg_name_flags + ARRAY_SIZE(__arg_name_flags); |
702108d1 | 344 | |
e6ca5e1a | 345 | while (anf-- > __arg_name_flags) |
4286325b | 346 | if (flag & anf->flag) |
702108d1 HM |
347 | return anf->name; |
348 | ||
349 | } else | |
350 | DMERR("%s called with more than one flag!", __func__); | |
351 | ||
352 | return NULL; | |
353 | } | |
354 | ||
6e53636f HM |
355 | /* Define correlation of raid456 journal cache modes and dm-raid target line parameters */ |
356 | static struct { | |
357 | const int mode; | |
358 | const char *param; | |
359 | } _raid456_journal_mode[] = { | |
360 | { R5C_JOURNAL_MODE_WRITE_THROUGH , "writethrough" }, | |
361 | { R5C_JOURNAL_MODE_WRITE_BACK , "writeback" } | |
362 | }; | |
363 | ||
364 | /* Return MD raid4/5/6 journal mode for dm @journal_mode one */ | |
365 | static int dm_raid_journal_mode_to_md(const char *mode) | |
366 | { | |
367 | int m = ARRAY_SIZE(_raid456_journal_mode); | |
368 | ||
369 | while (m--) | |
370 | if (!strcasecmp(mode, _raid456_journal_mode[m].param)) | |
371 | return _raid456_journal_mode[m].mode; | |
372 | ||
373 | return -EINVAL; | |
374 | } | |
375 | ||
376 | /* Return dm-raid raid4/5/6 journal mode string for @mode */ | |
377 | static const char *md_journal_mode_to_dm_raid(const int mode) | |
378 | { | |
379 | int m = ARRAY_SIZE(_raid456_journal_mode); | |
380 | ||
381 | while (m--) | |
382 | if (mode == _raid456_journal_mode[m].mode) | |
383 | return _raid456_journal_mode[m].param; | |
384 | ||
385 | return "unknown"; | |
386 | } | |
387 | ||
33e53f06 | 388 | /* |
d7ccc2e2 HM |
389 | * Bool helpers to test for various raid levels of a raid set. |
390 | * It's level as reported by the superblock rather than | |
33e53f06 HM |
391 | * the requested raid_type passed to the constructor. |
392 | */ | |
393 | /* Return true, if raid set in @rs is raid0 */ | |
394 | static bool rs_is_raid0(struct raid_set *rs) | |
395 | { | |
396 | return !rs->md.level; | |
397 | } | |
398 | ||
9dbd1aa3 HM |
399 | /* Return true, if raid set in @rs is raid1 */ |
400 | static bool rs_is_raid1(struct raid_set *rs) | |
401 | { | |
402 | return rs->md.level == 1; | |
403 | } | |
404 | ||
33e53f06 HM |
405 | /* Return true, if raid set in @rs is raid10 */ |
406 | static bool rs_is_raid10(struct raid_set *rs) | |
407 | { | |
408 | return rs->md.level == 10; | |
409 | } | |
410 | ||
4dff2f1e HM |
411 | /* Return true, if raid set in @rs is level 6 */ |
412 | static bool rs_is_raid6(struct raid_set *rs) | |
413 | { | |
414 | return rs->md.level == 6; | |
415 | } | |
416 | ||
40ba37e5 HM |
417 | /* Return true, if raid set in @rs is level 4, 5 or 6 */ |
418 | static bool rs_is_raid456(struct raid_set *rs) | |
419 | { | |
420 | return __within_range(rs->md.level, 4, 6); | |
421 | } | |
422 | ||
423 | /* Return true, if raid set in @rs is reshapable */ | |
d7ccc2e2 | 424 | static bool __is_raid10_far(int layout); |
40ba37e5 HM |
425 | static bool rs_is_reshapable(struct raid_set *rs) |
426 | { | |
427 | return rs_is_raid456(rs) || | |
428 | (rs_is_raid10(rs) && !__is_raid10_far(rs->md.new_layout)); | |
429 | } | |
430 | ||
9dbd1aa3 HM |
431 | /* Return true, if raid set in @rs is recovering */ |
432 | static bool rs_is_recovering(struct raid_set *rs) | |
433 | { | |
50c4feb9 | 434 | return rs->md.recovery_cp < rs->md.dev_sectors; |
9dbd1aa3 HM |
435 | } |
436 | ||
437 | /* Return true, if raid set in @rs is reshaping */ | |
438 | static bool rs_is_reshaping(struct raid_set *rs) | |
439 | { | |
9dbd1aa3 HM |
440 | return rs->md.reshape_position != MaxSector; |
441 | } | |
442 | ||
f090279e | 443 | /* |
d7ccc2e2 | 444 | * bool helpers to test for various raid levels of a raid type @rt |
f090279e HM |
445 | */ |
446 | ||
447 | /* Return true, if raid type in @rt is raid0 */ | |
448 | static bool rt_is_raid0(struct raid_type *rt) | |
449 | { | |
450 | return !rt->level; | |
451 | } | |
452 | ||
453 | /* Return true, if raid type in @rt is raid1 */ | |
454 | static bool rt_is_raid1(struct raid_type *rt) | |
455 | { | |
456 | return rt->level == 1; | |
457 | } | |
458 | ||
459 | /* Return true, if raid type in @rt is raid10 */ | |
460 | static bool rt_is_raid10(struct raid_type *rt) | |
461 | { | |
462 | return rt->level == 10; | |
463 | } | |
464 | ||
465 | /* Return true, if raid type in @rt is raid4/5 */ | |
466 | static bool rt_is_raid45(struct raid_type *rt) | |
467 | { | |
bb91a63f | 468 | return __within_range(rt->level, 4, 5); |
f090279e HM |
469 | } |
470 | ||
471 | /* Return true, if raid type in @rt is raid6 */ | |
472 | static bool rt_is_raid6(struct raid_type *rt) | |
473 | { | |
474 | return rt->level == 6; | |
475 | } | |
676fa5ad HM |
476 | |
477 | /* Return true, if raid type in @rt is raid4/5/6 */ | |
478 | static bool rt_is_raid456(struct raid_type *rt) | |
479 | { | |
bb91a63f | 480 | return __within_range(rt->level, 4, 6); |
676fa5ad | 481 | } |
f090279e HM |
482 | /* END: raid level bools */ |
483 | ||
a30cbc0d HM |
484 | /* Return valid ctr flags for the raid level of @rs */ |
485 | static unsigned long __valid_flags(struct raid_set *rs) | |
f090279e HM |
486 | { |
487 | if (rt_is_raid0(rs->raid_type)) | |
a30cbc0d | 488 | return RAID0_VALID_FLAGS; |
f090279e | 489 | else if (rt_is_raid1(rs->raid_type)) |
a30cbc0d | 490 | return RAID1_VALID_FLAGS; |
f090279e | 491 | else if (rt_is_raid10(rs->raid_type)) |
a30cbc0d | 492 | return RAID10_VALID_FLAGS; |
f090279e | 493 | else if (rt_is_raid45(rs->raid_type)) |
a30cbc0d | 494 | return RAID45_VALID_FLAGS; |
f090279e | 495 | else if (rt_is_raid6(rs->raid_type)) |
a30cbc0d | 496 | return RAID6_VALID_FLAGS; |
f090279e | 497 | |
d7ccc2e2 | 498 | return 0; |
f090279e HM |
499 | } |
500 | ||
501 | /* | |
a30cbc0d | 502 | * Check for valid flags set on @rs |
f090279e HM |
503 | * |
504 | * Has to be called after parsing of the ctr flags! | |
505 | */ | |
a30cbc0d | 506 | static int rs_check_for_valid_flags(struct raid_set *rs) |
f090279e | 507 | { |
a30cbc0d | 508 | if (rs->ctr_flags & ~__valid_flags(rs)) { |
4286325b | 509 | rs->ti->error = "Invalid flags combination"; |
bd83a4c4 MS |
510 | return -EINVAL; |
511 | } | |
f090279e HM |
512 | |
513 | return 0; | |
514 | } | |
515 | ||
33e53f06 HM |
516 | /* MD raid10 bit definitions and helpers */ |
517 | #define RAID10_OFFSET (1 << 16) /* stripes with data copies area adjacent on devices */ | |
518 | #define RAID10_BROCKEN_USE_FAR_SETS (1 << 17) /* Broken in raid10.c: use sets instead of whole stripe rotation */ | |
519 | #define RAID10_USE_FAR_SETS (1 << 18) /* Use sets instead of whole stripe rotation */ | |
520 | #define RAID10_FAR_COPIES_SHIFT 8 /* raid10 # far copies shift (2nd byte of layout) */ | |
521 | ||
522 | /* Return md raid10 near copies for @layout */ | |
e6ca5e1a | 523 | static unsigned int __raid10_near_copies(int layout) |
33e53f06 HM |
524 | { |
525 | return layout & 0xFF; | |
526 | } | |
527 | ||
528 | /* Return md raid10 far copies for @layout */ | |
e6ca5e1a | 529 | static unsigned int __raid10_far_copies(int layout) |
33e53f06 | 530 | { |
e6ca5e1a | 531 | return __raid10_near_copies(layout >> RAID10_FAR_COPIES_SHIFT); |
33e53f06 HM |
532 | } |
533 | ||
534 | /* Return true if md raid10 offset for @layout */ | |
d7ccc2e2 | 535 | static bool __is_raid10_offset(int layout) |
33e53f06 | 536 | { |
d7ccc2e2 | 537 | return !!(layout & RAID10_OFFSET); |
33e53f06 HM |
538 | } |
539 | ||
540 | /* Return true if md raid10 near for @layout */ | |
d7ccc2e2 | 541 | static bool __is_raid10_near(int layout) |
33e53f06 | 542 | { |
e6ca5e1a | 543 | return !__is_raid10_offset(layout) && __raid10_near_copies(layout) > 1; |
33e53f06 HM |
544 | } |
545 | ||
546 | /* Return true if md raid10 far for @layout */ | |
d7ccc2e2 | 547 | static bool __is_raid10_far(int layout) |
33e53f06 | 548 | { |
e6ca5e1a | 549 | return !__is_raid10_offset(layout) && __raid10_far_copies(layout) > 1; |
33e53f06 HM |
550 | } |
551 | ||
552 | /* Return md raid10 layout string for @layout */ | |
553 | static const char *raid10_md_layout_to_format(int layout) | |
fe5d2f4a JB |
554 | { |
555 | /* | |
33e53f06 HM |
556 | * Bit 16 stands for "offset" |
557 | * (i.e. adjacent stripes hold copies) | |
558 | * | |
fe5d2f4a JB |
559 | * Refer to MD's raid10.c for details |
560 | */ | |
e6ca5e1a | 561 | if (__is_raid10_offset(layout)) |
fe5d2f4a JB |
562 | return "offset"; |
563 | ||
e6ca5e1a | 564 | if (__raid10_near_copies(layout) > 1) |
fe5d2f4a JB |
565 | return "near"; |
566 | ||
e6ca5e1a | 567 | WARN_ON(__raid10_far_copies(layout) < 2); |
33e53f06 | 568 | |
fe5d2f4a JB |
569 | return "far"; |
570 | } | |
571 | ||
33e53f06 | 572 | /* Return md raid10 algorithm for @name */ |
68c1c4d5 | 573 | static int raid10_name_to_format(const char *name) |
33e53f06 HM |
574 | { |
575 | if (!strcasecmp(name, "near")) | |
576 | return ALGORITHM_RAID10_NEAR; | |
577 | else if (!strcasecmp(name, "offset")) | |
578 | return ALGORITHM_RAID10_OFFSET; | |
579 | else if (!strcasecmp(name, "far")) | |
580 | return ALGORITHM_RAID10_FAR; | |
581 | ||
582 | return -EINVAL; | |
583 | } | |
584 | ||
33e53f06 HM |
585 | /* Return md raid10 copies for @layout */ |
586 | static unsigned int raid10_md_layout_to_copies(int layout) | |
63f33b8d | 587 | { |
d7ccc2e2 | 588 | return max(__raid10_near_copies(layout), __raid10_far_copies(layout)); |
63f33b8d JB |
589 | } |
590 | ||
33e53f06 HM |
591 | /* Return md raid10 format id for @format string */ |
592 | static int raid10_format_to_md_layout(struct raid_set *rs, | |
593 | unsigned int algorithm, | |
594 | unsigned int copies) | |
63f33b8d | 595 | { |
33e53f06 | 596 | unsigned int n = 1, f = 1, r = 0; |
fe5d2f4a | 597 | |
33e53f06 HM |
598 | /* |
599 | * MD resilienece flaw: | |
600 | * | |
601 | * enabling use_far_sets for far/offset formats causes copies | |
602 | * to be colocated on the same devs together with their origins! | |
603 | * | |
604 | * -> disable it for now in the definition above | |
605 | */ | |
606 | if (algorithm == ALGORITHM_RAID10_DEFAULT || | |
607 | algorithm == ALGORITHM_RAID10_NEAR) | |
fe5d2f4a | 608 | n = copies; |
33e53f06 HM |
609 | |
610 | else if (algorithm == ALGORITHM_RAID10_OFFSET) { | |
611 | f = copies; | |
612 | r = RAID10_OFFSET; | |
4286325b | 613 | if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) |
33e53f06 HM |
614 | r |= RAID10_USE_FAR_SETS; |
615 | ||
616 | } else if (algorithm == ALGORITHM_RAID10_FAR) { | |
fe5d2f4a | 617 | f = copies; |
33e53f06 | 618 | r = !RAID10_OFFSET; |
4286325b | 619 | if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) |
33e53f06 | 620 | r |= RAID10_USE_FAR_SETS; |
fe5d2f4a | 621 | |
33e53f06 HM |
622 | } else |
623 | return -EINVAL; | |
624 | ||
625 | return r | (f << RAID10_FAR_COPIES_SHIFT) | n; | |
626 | } | |
627 | /* END: MD raid10 bit definitions and helpers */ | |
fe5d2f4a | 628 | |
33e53f06 | 629 | /* Check for any of the raid10 algorithms */ |
d7ccc2e2 | 630 | static bool __got_raid10(struct raid_type *rtp, const int layout) |
33e53f06 HM |
631 | { |
632 | if (rtp->level == 10) { | |
633 | switch (rtp->algorithm) { | |
634 | case ALGORITHM_RAID10_DEFAULT: | |
635 | case ALGORITHM_RAID10_NEAR: | |
e6ca5e1a | 636 | return __is_raid10_near(layout); |
33e53f06 | 637 | case ALGORITHM_RAID10_OFFSET: |
e6ca5e1a | 638 | return __is_raid10_offset(layout); |
33e53f06 | 639 | case ALGORITHM_RAID10_FAR: |
e6ca5e1a | 640 | return __is_raid10_far(layout); |
33e53f06 HM |
641 | default: |
642 | break; | |
643 | } | |
644 | } | |
fe5d2f4a | 645 | |
d7ccc2e2 | 646 | return false; |
63f33b8d JB |
647 | } |
648 | ||
33e53f06 | 649 | /* Return raid_type for @name */ |
92c83d79 | 650 | static struct raid_type *get_raid_type(const char *name) |
9d09e663 | 651 | { |
33e53f06 | 652 | struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); |
9d09e663 | 653 | |
33e53f06 HM |
654 | while (rtp-- > raid_types) |
655 | if (!strcasecmp(rtp->name, name)) | |
656 | return rtp; | |
9d09e663 N |
657 | |
658 | return NULL; | |
659 | } | |
660 | ||
33e53f06 HM |
661 | /* Return raid_type for @name based derived from @level and @layout */ |
662 | static struct raid_type *get_raid_type_by_ll(const int level, const int layout) | |
663 | { | |
664 | struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); | |
665 | ||
666 | while (rtp-- > raid_types) { | |
667 | /* RAID10 special checks based on @layout flags/properties */ | |
668 | if (rtp->level == level && | |
e6ca5e1a | 669 | (__got_raid10(rtp, layout) || rtp->algorithm == layout)) |
33e53f06 HM |
670 | return rtp; |
671 | } | |
672 | ||
673 | return NULL; | |
674 | } | |
675 | ||
9dbd1aa3 HM |
676 | /* |
677 | * Conditionally change bdev capacity of @rs | |
678 | * in case of a disk add/remove reshape | |
679 | */ | |
680 | static void rs_set_capacity(struct raid_set *rs) | |
681 | { | |
682 | struct mddev *mddev = &rs->md; | |
fbe6365b | 683 | struct md_rdev *rdev; |
0095dbc9 | 684 | struct gendisk *gendisk = dm_disk(dm_table_get_md(rs->ti->table)); |
9dbd1aa3 | 685 | |
fbe6365b HM |
686 | /* |
687 | * raid10 sets rdev->sector to the device size, which | |
688 | * is unintended in case of out-of-place reshaping | |
689 | */ | |
690 | rdev_for_each(rdev, mddev) | |
63c32ed4 HM |
691 | if (!test_bit(Journal, &rdev->flags)) |
692 | rdev->sectors = mddev->dev_sectors; | |
fbe6365b | 693 | |
0095dbc9 HM |
694 | set_capacity(gendisk, mddev->array_sectors); |
695 | revalidate_disk(gendisk); | |
9dbd1aa3 HM |
696 | } |
697 | ||
3a1c1ef2 HM |
698 | /* |
699 | * Set the mddev properties in @rs to the current | |
700 | * ones retrieved from the freshest superblock | |
701 | */ | |
702 | static void rs_set_cur(struct raid_set *rs) | |
703 | { | |
704 | struct mddev *mddev = &rs->md; | |
705 | ||
706 | mddev->new_level = mddev->level; | |
707 | mddev->new_layout = mddev->layout; | |
708 | mddev->new_chunk_sectors = mddev->chunk_sectors; | |
709 | } | |
710 | ||
33e53f06 HM |
711 | /* |
712 | * Set the mddev properties in @rs to the new | |
713 | * ones requested by the ctr | |
714 | */ | |
715 | static void rs_set_new(struct raid_set *rs) | |
716 | { | |
717 | struct mddev *mddev = &rs->md; | |
718 | ||
719 | mddev->level = mddev->new_level; | |
720 | mddev->layout = mddev->new_layout; | |
721 | mddev->chunk_sectors = mddev->new_chunk_sectors; | |
3a1c1ef2 | 722 | mddev->raid_disks = rs->raid_disks; |
33e53f06 HM |
723 | mddev->delta_disks = 0; |
724 | } | |
725 | ||
bfcee0e3 | 726 | static struct raid_set *raid_set_alloc(struct dm_target *ti, struct raid_type *raid_type, |
094f394d | 727 | unsigned int raid_devs) |
9d09e663 | 728 | { |
094f394d | 729 | unsigned int i; |
9d09e663 | 730 | struct raid_set *rs; |
9d09e663 | 731 | |
bd83a4c4 MS |
732 | if (raid_devs <= raid_type->parity_devs) { |
733 | ti->error = "Insufficient number of devices"; | |
734 | return ERR_PTR(-EINVAL); | |
735 | } | |
9d09e663 | 736 | |
9d09e663 | 737 | rs = kzalloc(sizeof(*rs) + raid_devs * sizeof(rs->dev[0]), GFP_KERNEL); |
bd83a4c4 MS |
738 | if (!rs) { |
739 | ti->error = "Cannot allocate raid context"; | |
740 | return ERR_PTR(-ENOMEM); | |
741 | } | |
9d09e663 N |
742 | |
743 | mddev_init(&rs->md); | |
744 | ||
33e53f06 HM |
745 | rs->raid_disks = raid_devs; |
746 | rs->delta_disks = 0; | |
747 | ||
9d09e663 N |
748 | rs->ti = ti; |
749 | rs->raid_type = raid_type; | |
9dbd1aa3 | 750 | rs->stripe_cache_entries = 256; |
9d09e663 N |
751 | rs->md.raid_disks = raid_devs; |
752 | rs->md.level = raid_type->level; | |
753 | rs->md.new_level = rs->md.level; | |
9d09e663 N |
754 | rs->md.layout = raid_type->algorithm; |
755 | rs->md.new_layout = rs->md.layout; | |
756 | rs->md.delta_disks = 0; | |
4dff2f1e | 757 | rs->md.recovery_cp = MaxSector; |
9d09e663 N |
758 | |
759 | for (i = 0; i < raid_devs; i++) | |
760 | md_rdev_init(&rs->dev[i].rdev); | |
761 | ||
762 | /* | |
763 | * Remaining items to be initialized by further RAID params: | |
764 | * rs->md.persistent | |
765 | * rs->md.external | |
766 | * rs->md.chunk_sectors | |
767 | * rs->md.new_chunk_sectors | |
c039c332 | 768 | * rs->md.dev_sectors |
9d09e663 N |
769 | */ |
770 | ||
771 | return rs; | |
772 | } | |
773 | ||
bfcee0e3 | 774 | static void raid_set_free(struct raid_set *rs) |
9d09e663 N |
775 | { |
776 | int i; | |
777 | ||
63c32ed4 HM |
778 | if (rs->journal_dev.dev) { |
779 | md_rdev_clear(&rs->journal_dev.rdev); | |
780 | dm_put_device(rs->ti, rs->journal_dev.dev); | |
781 | } | |
782 | ||
ffeeac75 | 783 | for (i = 0; i < rs->raid_disks; i++) { |
b12d437b JB |
784 | if (rs->dev[i].meta_dev) |
785 | dm_put_device(rs->ti, rs->dev[i].meta_dev); | |
545c8795 | 786 | md_rdev_clear(&rs->dev[i].rdev); |
9d09e663 N |
787 | if (rs->dev[i].data_dev) |
788 | dm_put_device(rs->ti, rs->dev[i].data_dev); | |
b12d437b | 789 | } |
9d09e663 N |
790 | |
791 | kfree(rs); | |
792 | } | |
793 | ||
794 | /* | |
795 | * For every device we have two words | |
796 | * <meta_dev>: meta device name or '-' if missing | |
797 | * <data_dev>: data device name or '-' if missing | |
798 | * | |
b12d437b JB |
799 | * The following are permitted: |
800 | * - - | |
801 | * - <data_dev> | |
802 | * <meta_dev> <data_dev> | |
803 | * | |
804 | * The following is not allowed: | |
805 | * <meta_dev> - | |
806 | * | |
807 | * This code parses those words. If there is a failure, | |
bfcee0e3 | 808 | * the caller must use raid_set_free() to unwind the operations. |
9d09e663 | 809 | */ |
702108d1 | 810 | static int parse_dev_params(struct raid_set *rs, struct dm_arg_set *as) |
9d09e663 N |
811 | { |
812 | int i; | |
813 | int rebuild = 0; | |
814 | int metadata_available = 0; | |
73c6f239 | 815 | int r = 0; |
92c83d79 | 816 | const char *arg; |
9d09e663 | 817 | |
92c83d79 HM |
818 | /* Put off the number of raid devices argument to get to dev pairs */ |
819 | arg = dm_shift_arg(as); | |
820 | if (!arg) | |
821 | return -EINVAL; | |
822 | ||
ffeeac75 | 823 | for (i = 0; i < rs->raid_disks; i++) { |
9d09e663 N |
824 | rs->dev[i].rdev.raid_disk = i; |
825 | ||
826 | rs->dev[i].meta_dev = NULL; | |
827 | rs->dev[i].data_dev = NULL; | |
828 | ||
829 | /* | |
63c32ed4 HM |
830 | * There are no offsets initially. |
831 | * Out of place reshape will set them accordingly. | |
9d09e663 N |
832 | */ |
833 | rs->dev[i].rdev.data_offset = 0; | |
63c32ed4 | 834 | rs->dev[i].rdev.new_data_offset = 0; |
9d09e663 N |
835 | rs->dev[i].rdev.mddev = &rs->md; |
836 | ||
92c83d79 HM |
837 | arg = dm_shift_arg(as); |
838 | if (!arg) | |
839 | return -EINVAL; | |
840 | ||
841 | if (strcmp(arg, "-")) { | |
bd83a4c4 MS |
842 | r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), |
843 | &rs->dev[i].meta_dev); | |
844 | if (r) { | |
845 | rs->ti->error = "RAID metadata device lookup failure"; | |
846 | return r; | |
847 | } | |
b12d437b JB |
848 | |
849 | rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL); | |
bd83a4c4 MS |
850 | if (!rs->dev[i].rdev.sb_page) { |
851 | rs->ti->error = "Failed to allocate superblock page"; | |
852 | return -ENOMEM; | |
853 | } | |
9d09e663 N |
854 | } |
855 | ||
92c83d79 HM |
856 | arg = dm_shift_arg(as); |
857 | if (!arg) | |
858 | return -EINVAL; | |
859 | ||
860 | if (!strcmp(arg, "-")) { | |
9d09e663 | 861 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags) && |
bd83a4c4 MS |
862 | (!rs->dev[i].rdev.recovery_offset)) { |
863 | rs->ti->error = "Drive designated for rebuild not specified"; | |
864 | return -EINVAL; | |
865 | } | |
9d09e663 | 866 | |
bd83a4c4 MS |
867 | if (rs->dev[i].meta_dev) { |
868 | rs->ti->error = "No data device supplied with metadata device"; | |
869 | return -EINVAL; | |
870 | } | |
b12d437b | 871 | |
9d09e663 N |
872 | continue; |
873 | } | |
874 | ||
bd83a4c4 MS |
875 | r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), |
876 | &rs->dev[i].data_dev); | |
877 | if (r) { | |
878 | rs->ti->error = "RAID device lookup failure"; | |
879 | return r; | |
880 | } | |
9d09e663 | 881 | |
b12d437b JB |
882 | if (rs->dev[i].meta_dev) { |
883 | metadata_available = 1; | |
884 | rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev; | |
885 | } | |
9d09e663 | 886 | rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev; |
3a1c1ef2 | 887 | list_add_tail(&rs->dev[i].rdev.same_set, &rs->md.disks); |
9d09e663 N |
888 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags)) |
889 | rebuild++; | |
890 | } | |
891 | ||
63c32ed4 HM |
892 | if (rs->journal_dev.dev) |
893 | list_add_tail(&rs->journal_dev.rdev.same_set, &rs->md.disks); | |
894 | ||
9d09e663 N |
895 | if (metadata_available) { |
896 | rs->md.external = 0; | |
897 | rs->md.persistent = 1; | |
898 | rs->md.major_version = 2; | |
899 | } else if (rebuild && !rs->md.recovery_cp) { | |
900 | /* | |
901 | * Without metadata, we will not be able to tell if the array | |
902 | * is in-sync or not - we must assume it is not. Therefore, | |
903 | * it is impossible to rebuild a drive. | |
904 | * | |
905 | * Even if there is metadata, the on-disk information may | |
906 | * indicate that the array is not in-sync and it will then | |
907 | * fail at that time. | |
908 | * | |
909 | * User could specify 'nosync' option if desperate. | |
910 | */ | |
bd83a4c4 MS |
911 | rs->ti->error = "Unable to rebuild drive while array is not in-sync"; |
912 | return -EINVAL; | |
9d09e663 N |
913 | } |
914 | ||
915 | return 0; | |
916 | } | |
917 | ||
c1084561 JB |
918 | /* |
919 | * validate_region_size | |
920 | * @rs | |
921 | * @region_size: region size in sectors. If 0, pick a size (4MiB default). | |
922 | * | |
923 | * Set rs->md.bitmap_info.chunksize (which really refers to 'region size'). | |
924 | * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap. | |
925 | * | |
926 | * Returns: 0 on success, -EINVAL on failure. | |
927 | */ | |
928 | static int validate_region_size(struct raid_set *rs, unsigned long region_size) | |
929 | { | |
930 | unsigned long min_region_size = rs->ti->len / (1 << 21); | |
931 | ||
9e7d9367 HM |
932 | if (rs_is_raid0(rs)) |
933 | return 0; | |
934 | ||
c1084561 JB |
935 | if (!region_size) { |
936 | /* | |
43157840 | 937 | * Choose a reasonable default. All figures in sectors. |
c1084561 JB |
938 | */ |
939 | if (min_region_size > (1 << 13)) { | |
3a0f9aae | 940 | /* If not a power of 2, make it the next power of 2 */ |
042745ee | 941 | region_size = roundup_pow_of_two(min_region_size); |
c1084561 JB |
942 | DMINFO("Choosing default region size of %lu sectors", |
943 | region_size); | |
c1084561 JB |
944 | } else { |
945 | DMINFO("Choosing default region size of 4MiB"); | |
946 | region_size = 1 << 13; /* sectors */ | |
947 | } | |
948 | } else { | |
949 | /* | |
950 | * Validate user-supplied value. | |
951 | */ | |
bd83a4c4 MS |
952 | if (region_size > rs->ti->len) { |
953 | rs->ti->error = "Supplied region size is too large"; | |
954 | return -EINVAL; | |
955 | } | |
c1084561 JB |
956 | |
957 | if (region_size < min_region_size) { | |
958 | DMERR("Supplied region_size (%lu sectors) below minimum (%lu)", | |
959 | region_size, min_region_size); | |
bd83a4c4 MS |
960 | rs->ti->error = "Supplied region size is too small"; |
961 | return -EINVAL; | |
c1084561 JB |
962 | } |
963 | ||
bd83a4c4 MS |
964 | if (!is_power_of_2(region_size)) { |
965 | rs->ti->error = "Region size is not a power of 2"; | |
966 | return -EINVAL; | |
967 | } | |
c1084561 | 968 | |
bd83a4c4 MS |
969 | if (region_size < rs->md.chunk_sectors) { |
970 | rs->ti->error = "Region size is smaller than the chunk size"; | |
971 | return -EINVAL; | |
972 | } | |
c1084561 JB |
973 | } |
974 | ||
975 | /* | |
976 | * Convert sectors to bytes. | |
977 | */ | |
89d3d9a1 | 978 | rs->md.bitmap_info.chunksize = to_bytes(region_size); |
c1084561 JB |
979 | |
980 | return 0; | |
981 | } | |
982 | ||
eb649123 | 983 | /* |
55ebbb59 | 984 | * validate_raid_redundancy |
eb649123 JB |
985 | * @rs |
986 | * | |
55ebbb59 JB |
987 | * Determine if there are enough devices in the array that haven't |
988 | * failed (or are being rebuilt) to form a usable array. | |
eb649123 JB |
989 | * |
990 | * Returns: 0 on success, -EINVAL on failure. | |
991 | */ | |
55ebbb59 | 992 | static int validate_raid_redundancy(struct raid_set *rs) |
eb649123 | 993 | { |
094f394d HM |
994 | unsigned int i, rebuild_cnt = 0; |
995 | unsigned int rebuilds_per_group = 0, copies; | |
996 | unsigned int group_size, last_group_start; | |
eb649123 | 997 | |
eb649123 | 998 | for (i = 0; i < rs->md.raid_disks; i++) |
55ebbb59 JB |
999 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags) || |
1000 | !rs->dev[i].rdev.sb_page) | |
eb649123 JB |
1001 | rebuild_cnt++; |
1002 | ||
1003 | switch (rs->raid_type->level) { | |
9e7d9367 HM |
1004 | case 0: |
1005 | break; | |
eb649123 JB |
1006 | case 1: |
1007 | if (rebuild_cnt >= rs->md.raid_disks) | |
1008 | goto too_many; | |
1009 | break; | |
1010 | case 4: | |
1011 | case 5: | |
1012 | case 6: | |
1013 | if (rebuild_cnt > rs->raid_type->parity_devs) | |
1014 | goto too_many; | |
1015 | break; | |
1016 | case 10: | |
9dbd1aa3 | 1017 | copies = raid10_md_layout_to_copies(rs->md.new_layout); |
4ec1e369 JB |
1018 | if (rebuild_cnt < copies) |
1019 | break; | |
1020 | ||
1021 | /* | |
1022 | * It is possible to have a higher rebuild count for RAID10, | |
1023 | * as long as the failed devices occur in different mirror | |
1024 | * groups (i.e. different stripes). | |
1025 | * | |
4ec1e369 JB |
1026 | * When checking "near" format, make sure no adjacent devices |
1027 | * have failed beyond what can be handled. In addition to the | |
1028 | * simple case where the number of devices is a multiple of the | |
1029 | * number of copies, we must also handle cases where the number | |
1030 | * of devices is not a multiple of the number of copies. | |
43157840 MS |
1031 | * E.g. dev1 dev2 dev3 dev4 dev5 |
1032 | * A A B B C | |
1033 | * C D D E E | |
4ec1e369 | 1034 | */ |
9dbd1aa3 | 1035 | if (__is_raid10_near(rs->md.new_layout)) { |
ffeeac75 | 1036 | for (i = 0; i < rs->md.raid_disks; i++) { |
fe5d2f4a JB |
1037 | if (!(i % copies)) |
1038 | rebuilds_per_group = 0; | |
9dbd1aa3 | 1039 | if ((!rs->dev[i].rdev.sb_page || |
40ba37e5 | 1040 | !test_bit(In_sync, &rs->dev[i].rdev.flags)) && |
fe5d2f4a JB |
1041 | (++rebuilds_per_group >= copies)) |
1042 | goto too_many; | |
1043 | } | |
1044 | break; | |
1045 | } | |
1046 | ||
1047 | /* | |
1048 | * When checking "far" and "offset" formats, we need to ensure | |
1049 | * that the device that holds its copy is not also dead or | |
1050 | * being rebuilt. (Note that "far" and "offset" formats only | |
1051 | * support two copies right now. These formats also only ever | |
1052 | * use the 'use_far_sets' variant.) | |
1053 | * | |
1054 | * This check is somewhat complicated by the need to account | |
43157840 | 1055 | * for arrays that are not a multiple of (far) copies. This |
fe5d2f4a JB |
1056 | * results in the need to treat the last (potentially larger) |
1057 | * set differently. | |
1058 | */ | |
1059 | group_size = (rs->md.raid_disks / copies); | |
1060 | last_group_start = (rs->md.raid_disks / group_size) - 1; | |
1061 | last_group_start *= group_size; | |
1062 | for (i = 0; i < rs->md.raid_disks; i++) { | |
1063 | if (!(i % copies) && !(i > last_group_start)) | |
55ebbb59 | 1064 | rebuilds_per_group = 0; |
fe5d2f4a JB |
1065 | if ((!rs->dev[i].rdev.sb_page || |
1066 | !test_bit(In_sync, &rs->dev[i].rdev.flags)) && | |
4ec1e369 | 1067 | (++rebuilds_per_group >= copies)) |
fe5d2f4a | 1068 | goto too_many; |
4ec1e369 JB |
1069 | } |
1070 | break; | |
eb649123 | 1071 | default: |
55ebbb59 JB |
1072 | if (rebuild_cnt) |
1073 | return -EINVAL; | |
eb649123 JB |
1074 | } |
1075 | ||
1076 | return 0; | |
1077 | ||
1078 | too_many: | |
eb649123 JB |
1079 | return -EINVAL; |
1080 | } | |
1081 | ||
9d09e663 N |
1082 | /* |
1083 | * Possible arguments are... | |
9d09e663 N |
1084 | * <chunk_size> [optional_args] |
1085 | * | |
32737279 JB |
1086 | * Argument definitions |
1087 | * <chunk_size> The number of sectors per disk that | |
43157840 | 1088 | * will form the "stripe" |
32737279 | 1089 | * [[no]sync] Force or prevent recovery of the |
43157840 | 1090 | * entire array |
9d09e663 | 1091 | * [rebuild <idx>] Rebuild the drive indicated by the index |
32737279 | 1092 | * [daemon_sleep <ms>] Time between bitmap daemon work to |
43157840 | 1093 | * clear bits |
9d09e663 N |
1094 | * [min_recovery_rate <kB/sec/disk>] Throttle RAID initialization |
1095 | * [max_recovery_rate <kB/sec/disk>] Throttle RAID initialization | |
46bed2b5 | 1096 | * [write_mostly <idx>] Indicate a write mostly drive via index |
9d09e663 N |
1097 | * [max_write_behind <sectors>] See '-write-behind=' (man mdadm) |
1098 | * [stripe_cache <sectors>] Stripe cache size for higher RAIDs | |
43157840 | 1099 | * [region_size <sectors>] Defines granularity of bitmap |
63c32ed4 HM |
1100 | * [journal_dev <dev>] raid4/5/6 journaling deviice |
1101 | * (i.e. write hole closing log) | |
63f33b8d JB |
1102 | * |
1103 | * RAID10-only options: | |
43157840 | 1104 | * [raid10_copies <# copies>] Number of copies. (Default: 2) |
fe5d2f4a | 1105 | * [raid10_format <near|far|offset>] Layout algorithm. (Default: near) |
9d09e663 | 1106 | */ |
92c83d79 | 1107 | static int parse_raid_params(struct raid_set *rs, struct dm_arg_set *as, |
094f394d | 1108 | unsigned int num_raid_params) |
9d09e663 | 1109 | { |
9dbd1aa3 | 1110 | int value, raid10_format = ALGORITHM_RAID10_DEFAULT; |
094f394d HM |
1111 | unsigned int raid10_copies = 2; |
1112 | unsigned int i, write_mostly = 0; | |
1113 | unsigned int region_size = 0; | |
542f9038 | 1114 | sector_t max_io_len; |
92c83d79 | 1115 | const char *arg, *key; |
702108d1 | 1116 | struct raid_dev *rd; |
33e53f06 | 1117 | struct raid_type *rt = rs->raid_type; |
92c83d79 HM |
1118 | |
1119 | arg = dm_shift_arg(as); | |
1120 | num_raid_params--; /* Account for chunk_size argument */ | |
1121 | ||
9dbd1aa3 | 1122 | if (kstrtoint(arg, 10, &value) < 0) { |
bd83a4c4 MS |
1123 | rs->ti->error = "Bad numerical argument given for chunk_size"; |
1124 | return -EINVAL; | |
1125 | } | |
9d09e663 N |
1126 | |
1127 | /* | |
1128 | * First, parse the in-order required arguments | |
32737279 | 1129 | * "chunk_size" is the only argument of this type. |
9d09e663 | 1130 | */ |
33e53f06 | 1131 | if (rt_is_raid1(rt)) { |
32737279 JB |
1132 | if (value) |
1133 | DMERR("Ignoring chunk size parameter for RAID 1"); | |
1134 | value = 0; | |
bd83a4c4 MS |
1135 | } else if (!is_power_of_2(value)) { |
1136 | rs->ti->error = "Chunk size must be a power of 2"; | |
1137 | return -EINVAL; | |
1138 | } else if (value < 8) { | |
1139 | rs->ti->error = "Chunk size value is too small"; | |
1140 | return -EINVAL; | |
1141 | } | |
9d09e663 N |
1142 | |
1143 | rs->md.new_chunk_sectors = rs->md.chunk_sectors = value; | |
9d09e663 N |
1144 | |
1145 | /* | |
b12d437b JB |
1146 | * We set each individual device as In_sync with a completed |
1147 | * 'recovery_offset'. If there has been a device failure or | |
1148 | * replacement then one of the following cases applies: | |
1149 | * | |
1150 | * 1) User specifies 'rebuild'. | |
43157840 | 1151 | * - Device is reset when param is read. |
b12d437b | 1152 | * 2) A new device is supplied. |
43157840 | 1153 | * - No matching superblock found, resets device. |
b12d437b | 1154 | * 3) Device failure was transient and returns on reload. |
43157840 | 1155 | * - Failure noticed, resets device for bitmap replay. |
b12d437b | 1156 | * 4) Device hadn't completed recovery after previous failure. |
43157840 | 1157 | * - Superblock is read and overrides recovery_offset. |
b12d437b JB |
1158 | * |
1159 | * What is found in the superblocks of the devices is always | |
1160 | * authoritative, unless 'rebuild' or '[no]sync' was specified. | |
9d09e663 | 1161 | */ |
ffeeac75 | 1162 | for (i = 0; i < rs->raid_disks; i++) { |
9d09e663 | 1163 | set_bit(In_sync, &rs->dev[i].rdev.flags); |
b12d437b JB |
1164 | rs->dev[i].rdev.recovery_offset = MaxSector; |
1165 | } | |
9d09e663 | 1166 | |
b12d437b JB |
1167 | /* |
1168 | * Second, parse the unordered optional arguments | |
1169 | */ | |
9d09e663 | 1170 | for (i = 0; i < num_raid_params; i++) { |
4763e543 | 1171 | key = dm_shift_arg(as); |
bd83a4c4 MS |
1172 | if (!key) { |
1173 | rs->ti->error = "Not enough raid parameters given"; | |
1174 | return -EINVAL; | |
1175 | } | |
92c83d79 | 1176 | |
3fa6cf38 | 1177 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC))) { |
4286325b | 1178 | if (test_and_set_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { |
bd83a4c4 MS |
1179 | rs->ti->error = "Only one 'nosync' argument allowed"; |
1180 | return -EINVAL; | |
1181 | } | |
9d09e663 N |
1182 | continue; |
1183 | } | |
3fa6cf38 | 1184 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_SYNC))) { |
4286325b | 1185 | if (test_and_set_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) { |
bd83a4c4 MS |
1186 | rs->ti->error = "Only one 'sync' argument allowed"; |
1187 | return -EINVAL; | |
1188 | } | |
4763e543 HM |
1189 | continue; |
1190 | } | |
3fa6cf38 | 1191 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_USE_NEAR_SETS))) { |
4286325b | 1192 | if (test_and_set_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1193 | rs->ti->error = "Only one 'raid10_use_new_sets' argument allowed"; |
1194 | return -EINVAL; | |
1195 | } | |
9d09e663 N |
1196 | continue; |
1197 | } | |
1198 | ||
92c83d79 HM |
1199 | arg = dm_shift_arg(as); |
1200 | i++; /* Account for the argument pairs */ | |
bd83a4c4 MS |
1201 | if (!arg) { |
1202 | rs->ti->error = "Wrong number of raid parameters given"; | |
1203 | return -EINVAL; | |
1204 | } | |
63f33b8d | 1205 | |
702108d1 HM |
1206 | /* |
1207 | * Parameters that take a string value are checked here. | |
1208 | */ | |
63c32ed4 | 1209 | /* "raid10_format {near|offset|far} */ |
3fa6cf38 | 1210 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT))) { |
4286325b | 1211 | if (test_and_set_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) { |
bd83a4c4 MS |
1212 | rs->ti->error = "Only one 'raid10_format' argument pair allowed"; |
1213 | return -EINVAL; | |
1214 | } | |
1215 | if (!rt_is_raid10(rt)) { | |
1216 | rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type"; | |
1217 | return -EINVAL; | |
1218 | } | |
33e53f06 | 1219 | raid10_format = raid10_name_to_format(arg); |
bd83a4c4 MS |
1220 | if (raid10_format < 0) { |
1221 | rs->ti->error = "Invalid 'raid10_format' value given"; | |
1222 | return raid10_format; | |
1223 | } | |
63f33b8d JB |
1224 | continue; |
1225 | } | |
1226 | ||
6e53636f | 1227 | /* "journal_dev <dev>" */ |
63c32ed4 HM |
1228 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_JOURNAL_DEV))) { |
1229 | int r; | |
1230 | struct md_rdev *jdev; | |
1231 | ||
1232 | if (test_and_set_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { | |
1233 | rs->ti->error = "Only one raid4/5/6 set journaling device allowed"; | |
1234 | return -EINVAL; | |
1235 | } | |
1236 | if (!rt_is_raid456(rt)) { | |
1237 | rs->ti->error = "'journal_dev' is an invalid parameter for this RAID type"; | |
1238 | return -EINVAL; | |
1239 | } | |
1240 | r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), | |
1241 | &rs->journal_dev.dev); | |
1242 | if (r) { | |
1243 | rs->ti->error = "raid4/5/6 journal device lookup failure"; | |
1244 | return r; | |
1245 | } | |
1246 | jdev = &rs->journal_dev.rdev; | |
1247 | md_rdev_init(jdev); | |
1248 | jdev->mddev = &rs->md; | |
1249 | jdev->bdev = rs->journal_dev.dev->bdev; | |
1250 | jdev->sectors = to_sector(i_size_read(jdev->bdev->bd_inode)); | |
1251 | if (jdev->sectors < MIN_RAID456_JOURNAL_SPACE) { | |
1252 | rs->ti->error = "No space for raid4/5/6 journal"; | |
1253 | return -ENOSPC; | |
1254 | } | |
6e53636f | 1255 | rs->journal_dev.mode = R5C_JOURNAL_MODE_WRITE_THROUGH; |
63c32ed4 HM |
1256 | set_bit(Journal, &jdev->flags); |
1257 | continue; | |
1258 | } | |
1259 | ||
6e53636f HM |
1260 | /* "journal_mode <mode>" ("journal_dev" mandatory!) */ |
1261 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_JOURNAL_MODE))) { | |
1262 | int r; | |
1263 | ||
1264 | if (!test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { | |
1265 | rs->ti->error = "raid4/5/6 'journal_mode' is invalid without 'journal_dev'"; | |
1266 | return -EINVAL; | |
1267 | } | |
1268 | if (test_and_set_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) { | |
1269 | rs->ti->error = "Only one raid4/5/6 'journal_mode' argument allowed"; | |
1270 | return -EINVAL; | |
1271 | } | |
1272 | r = dm_raid_journal_mode_to_md(arg); | |
1273 | if (r < 0) { | |
1274 | rs->ti->error = "Invalid 'journal_mode' argument"; | |
1275 | return r; | |
1276 | } | |
1277 | rs->journal_dev.mode = r; | |
1278 | continue; | |
1279 | } | |
1280 | ||
63c32ed4 HM |
1281 | /* |
1282 | * Parameters with number values from here on. | |
1283 | */ | |
9dbd1aa3 | 1284 | if (kstrtoint(arg, 10, &value) < 0) { |
bd83a4c4 MS |
1285 | rs->ti->error = "Bad numerical argument given in raid params"; |
1286 | return -EINVAL; | |
1287 | } | |
702108d1 | 1288 | |
3fa6cf38 | 1289 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD))) { |
702108d1 HM |
1290 | /* |
1291 | * "rebuild" is being passed in by userspace to provide | |
1292 | * indexes of replaced devices and to set up additional | |
1293 | * devices on raid level takeover. | |
43157840 | 1294 | */ |
bb91a63f | 1295 | if (!__within_range(value, 0, rs->raid_disks - 1)) { |
bd83a4c4 MS |
1296 | rs->ti->error = "Invalid rebuild index given"; |
1297 | return -EINVAL; | |
1298 | } | |
702108d1 | 1299 | |
bd83a4c4 MS |
1300 | if (test_and_set_bit(value, (void *) rs->rebuild_disks)) { |
1301 | rs->ti->error = "rebuild for this index already given"; | |
1302 | return -EINVAL; | |
1303 | } | |
ecbfb9f1 | 1304 | |
702108d1 HM |
1305 | rd = rs->dev + value; |
1306 | clear_bit(In_sync, &rd->rdev.flags); | |
1307 | clear_bit(Faulty, &rd->rdev.flags); | |
1308 | rd->rdev.recovery_offset = 0; | |
4286325b | 1309 | set_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags); |
3fa6cf38 | 1310 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY))) { |
bd83a4c4 MS |
1311 | if (!rt_is_raid1(rt)) { |
1312 | rs->ti->error = "write_mostly option is only valid for RAID1"; | |
1313 | return -EINVAL; | |
1314 | } | |
702108d1 | 1315 | |
bb91a63f | 1316 | if (!__within_range(value, 0, rs->md.raid_disks - 1)) { |
bd83a4c4 MS |
1317 | rs->ti->error = "Invalid write_mostly index given"; |
1318 | return -EINVAL; | |
1319 | } | |
9d09e663 | 1320 | |
5fa146b2 | 1321 | write_mostly++; |
46bed2b5 | 1322 | set_bit(WriteMostly, &rs->dev[value].rdev.flags); |
4286325b | 1323 | set_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags); |
3fa6cf38 | 1324 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND))) { |
bd83a4c4 MS |
1325 | if (!rt_is_raid1(rt)) { |
1326 | rs->ti->error = "max_write_behind option is only valid for RAID1"; | |
1327 | return -EINVAL; | |
1328 | } | |
702108d1 | 1329 | |
4286325b | 1330 | if (test_and_set_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) { |
bd83a4c4 MS |
1331 | rs->ti->error = "Only one max_write_behind argument pair allowed"; |
1332 | return -EINVAL; | |
1333 | } | |
9d09e663 N |
1334 | |
1335 | /* | |
1336 | * In device-mapper, we specify things in sectors, but | |
1337 | * MD records this value in kB | |
1338 | */ | |
1339 | value /= 2; | |
bd83a4c4 MS |
1340 | if (value > COUNTER_MAX) { |
1341 | rs->ti->error = "Max write-behind limit out of range"; | |
1342 | return -EINVAL; | |
1343 | } | |
702108d1 | 1344 | |
9d09e663 | 1345 | rs->md.bitmap_info.max_write_behind = value; |
3fa6cf38 | 1346 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP))) { |
4286325b | 1347 | if (test_and_set_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) { |
bd83a4c4 MS |
1348 | rs->ti->error = "Only one daemon_sleep argument pair allowed"; |
1349 | return -EINVAL; | |
1350 | } | |
1351 | if (!value || (value > MAX_SCHEDULE_TIMEOUT)) { | |
1352 | rs->ti->error = "daemon sleep period out of range"; | |
1353 | return -EINVAL; | |
1354 | } | |
9d09e663 | 1355 | rs->md.bitmap_info.daemon_sleep = value; |
3fa6cf38 | 1356 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET))) { |
4763e543 | 1357 | /* Userspace passes new data_offset after having extended the the data image LV */ |
4286325b | 1358 | if (test_and_set_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { |
bd83a4c4 MS |
1359 | rs->ti->error = "Only one data_offset argument pair allowed"; |
1360 | return -EINVAL; | |
1361 | } | |
4763e543 | 1362 | /* Ensure sensible data offset */ |
75dd3b9e HM |
1363 | if (value < 0 || |
1364 | (value && (value < MIN_FREE_RESHAPE_SPACE || value % to_sector(PAGE_SIZE)))) { | |
bd83a4c4 MS |
1365 | rs->ti->error = "Bogus data_offset value"; |
1366 | return -EINVAL; | |
1367 | } | |
4763e543 | 1368 | rs->data_offset = value; |
3fa6cf38 | 1369 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS))) { |
4763e543 | 1370 | /* Define the +/-# of disks to add to/remove from the given raid set */ |
4286325b | 1371 | if (test_and_set_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1372 | rs->ti->error = "Only one delta_disks argument pair allowed"; |
1373 | return -EINVAL; | |
1374 | } | |
4763e543 | 1375 | /* Ensure MAX_RAID_DEVICES and raid type minimal_devs! */ |
bb91a63f | 1376 | if (!__within_range(abs(value), 1, MAX_RAID_DEVICES - rt->minimal_devs)) { |
bd83a4c4 MS |
1377 | rs->ti->error = "Too many delta_disk requested"; |
1378 | return -EINVAL; | |
1379 | } | |
4763e543 HM |
1380 | |
1381 | rs->delta_disks = value; | |
3fa6cf38 | 1382 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE))) { |
4286325b | 1383 | if (test_and_set_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1384 | rs->ti->error = "Only one stripe_cache argument pair allowed"; |
1385 | return -EINVAL; | |
1386 | } | |
1387 | ||
bd83a4c4 MS |
1388 | if (!rt_is_raid456(rt)) { |
1389 | rs->ti->error = "Inappropriate argument: stripe_cache"; | |
1390 | return -EINVAL; | |
1391 | } | |
702108d1 | 1392 | |
9dbd1aa3 | 1393 | rs->stripe_cache_entries = value; |
3fa6cf38 | 1394 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE))) { |
4286325b | 1395 | if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1396 | rs->ti->error = "Only one min_recovery_rate argument pair allowed"; |
1397 | return -EINVAL; | |
1398 | } | |
1399 | if (value > INT_MAX) { | |
1400 | rs->ti->error = "min_recovery_rate out of range"; | |
1401 | return -EINVAL; | |
1402 | } | |
9d09e663 | 1403 | rs->md.sync_speed_min = (int)value; |
3fa6cf38 | 1404 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE))) { |
f15f64d6 | 1405 | if (test_and_set_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1406 | rs->ti->error = "Only one max_recovery_rate argument pair allowed"; |
1407 | return -EINVAL; | |
1408 | } | |
1409 | if (value > INT_MAX) { | |
1410 | rs->ti->error = "max_recovery_rate out of range"; | |
1411 | return -EINVAL; | |
1412 | } | |
9d09e663 | 1413 | rs->md.sync_speed_max = (int)value; |
3fa6cf38 | 1414 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE))) { |
4286325b | 1415 | if (test_and_set_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1416 | rs->ti->error = "Only one region_size argument pair allowed"; |
1417 | return -EINVAL; | |
1418 | } | |
702108d1 | 1419 | |
c1084561 | 1420 | region_size = value; |
4257e085 | 1421 | rs->requested_bitmap_chunk_sectors = value; |
3fa6cf38 | 1422 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES))) { |
4286325b | 1423 | if (test_and_set_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) { |
bd83a4c4 MS |
1424 | rs->ti->error = "Only one raid10_copies argument pair allowed"; |
1425 | return -EINVAL; | |
1426 | } | |
702108d1 | 1427 | |
bb91a63f | 1428 | if (!__within_range(value, 2, rs->md.raid_disks)) { |
bd83a4c4 MS |
1429 | rs->ti->error = "Bad value for 'raid10_copies'"; |
1430 | return -EINVAL; | |
1431 | } | |
702108d1 | 1432 | |
63f33b8d | 1433 | raid10_copies = value; |
9d09e663 N |
1434 | } else { |
1435 | DMERR("Unable to parse RAID parameter: %s", key); | |
bd83a4c4 MS |
1436 | rs->ti->error = "Unable to parse RAID parameter"; |
1437 | return -EINVAL; | |
9d09e663 N |
1438 | } |
1439 | } | |
1440 | ||
0d851d14 HM |
1441 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) && |
1442 | test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { | |
1443 | rs->ti->error = "sync and nosync are mutually exclusive"; | |
1444 | return -EINVAL; | |
1445 | } | |
1446 | ||
37f10be1 HM |
1447 | if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) && |
1448 | (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) || | |
1449 | test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags))) { | |
1450 | rs->ti->error = "sync/nosync and rebuild are mutually exclusive"; | |
1451 | return -EINVAL; | |
1452 | } | |
1453 | ||
5fa146b2 HM |
1454 | if (write_mostly >= rs->md.raid_disks) { |
1455 | rs->ti->error = "Can't set all raid1 devices to write_mostly"; | |
1456 | return -EINVAL; | |
1457 | } | |
1458 | ||
c1084561 JB |
1459 | if (validate_region_size(rs, region_size)) |
1460 | return -EINVAL; | |
1461 | ||
1462 | if (rs->md.chunk_sectors) | |
542f9038 | 1463 | max_io_len = rs->md.chunk_sectors; |
c1084561 | 1464 | else |
542f9038 | 1465 | max_io_len = region_size; |
c1084561 | 1466 | |
542f9038 MS |
1467 | if (dm_set_target_max_io_len(rs->ti, max_io_len)) |
1468 | return -EINVAL; | |
32737279 | 1469 | |
33e53f06 | 1470 | if (rt_is_raid10(rt)) { |
bd83a4c4 MS |
1471 | if (raid10_copies > rs->md.raid_disks) { |
1472 | rs->ti->error = "Not enough devices to satisfy specification"; | |
1473 | return -EINVAL; | |
1474 | } | |
63f33b8d | 1475 | |
33e53f06 | 1476 | rs->md.new_layout = raid10_format_to_md_layout(rs, raid10_format, raid10_copies); |
bd83a4c4 MS |
1477 | if (rs->md.new_layout < 0) { |
1478 | rs->ti->error = "Error getting raid10 format"; | |
1479 | return rs->md.new_layout; | |
1480 | } | |
33e53f06 HM |
1481 | |
1482 | rt = get_raid_type_by_ll(10, rs->md.new_layout); | |
bd83a4c4 MS |
1483 | if (!rt) { |
1484 | rs->ti->error = "Failed to recognize new raid10 layout"; | |
1485 | return -EINVAL; | |
1486 | } | |
33e53f06 HM |
1487 | |
1488 | if ((rt->algorithm == ALGORITHM_RAID10_DEFAULT || | |
1489 | rt->algorithm == ALGORITHM_RAID10_NEAR) && | |
4286325b | 1490 | test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1491 | rs->ti->error = "RAID10 format 'near' and 'raid10_use_near_sets' are incompatible"; |
1492 | return -EINVAL; | |
1493 | } | |
bd83a4c4 | 1494 | } |
702108d1 | 1495 | |
33e53f06 | 1496 | rs->raid10_copies = raid10_copies; |
c039c332 | 1497 | |
9d09e663 N |
1498 | /* Assume there are no metadata devices until the drives are parsed */ |
1499 | rs->md.persistent = 0; | |
1500 | rs->md.external = 1; | |
1501 | ||
f090279e | 1502 | /* Check, if any invalid ctr arguments have been passed in for the raid level */ |
a30cbc0d | 1503 | return rs_check_for_valid_flags(rs); |
9d09e663 N |
1504 | } |
1505 | ||
9dbd1aa3 HM |
1506 | /* Set raid4/5/6 cache size */ |
1507 | static int rs_set_raid456_stripe_cache(struct raid_set *rs) | |
1508 | { | |
1509 | int r; | |
1510 | struct r5conf *conf; | |
1511 | struct mddev *mddev = &rs->md; | |
1512 | uint32_t min_stripes = max(mddev->chunk_sectors, mddev->new_chunk_sectors) / 2; | |
1513 | uint32_t nr_stripes = rs->stripe_cache_entries; | |
1514 | ||
1515 | if (!rt_is_raid456(rs->raid_type)) { | |
1516 | rs->ti->error = "Inappropriate raid level; cannot change stripe_cache size"; | |
1517 | return -EINVAL; | |
1518 | } | |
1519 | ||
1520 | if (nr_stripes < min_stripes) { | |
1521 | DMINFO("Adjusting requested %u stripe cache entries to %u to suit stripe size", | |
1522 | nr_stripes, min_stripes); | |
1523 | nr_stripes = min_stripes; | |
1524 | } | |
1525 | ||
1526 | conf = mddev->private; | |
1527 | if (!conf) { | |
1528 | rs->ti->error = "Cannot change stripe_cache size on inactive RAID set"; | |
1529 | return -EINVAL; | |
1530 | } | |
1531 | ||
1532 | /* Try setting number of stripes in raid456 stripe cache */ | |
1533 | if (conf->min_nr_stripes != nr_stripes) { | |
1534 | r = raid5_set_cache_size(mddev, nr_stripes); | |
1535 | if (r) { | |
1536 | rs->ti->error = "Failed to set raid4/5/6 stripe cache size"; | |
1537 | return r; | |
1538 | } | |
1539 | ||
1540 | DMINFO("%u stripe cache entries", nr_stripes); | |
1541 | } | |
1542 | ||
1543 | return 0; | |
1544 | } | |
1545 | ||
3a1c1ef2 HM |
1546 | /* Return # of data stripes as kept in mddev as of @rs (i.e. as of superblock) */ |
1547 | static unsigned int mddev_data_stripes(struct raid_set *rs) | |
1548 | { | |
1549 | return rs->md.raid_disks - rs->raid_type->parity_devs; | |
1550 | } | |
1551 | ||
40ba37e5 HM |
1552 | /* Return # of data stripes of @rs (i.e. as of ctr) */ |
1553 | static unsigned int rs_data_stripes(struct raid_set *rs) | |
1554 | { | |
1555 | return rs->raid_disks - rs->raid_type->parity_devs; | |
1556 | } | |
1557 | ||
50c4feb9 HM |
1558 | /* |
1559 | * Retrieve rdev->sectors from any valid raid device of @rs | |
1560 | * to allow userpace to pass in arbitray "- -" device tupples. | |
1561 | */ | |
1562 | static sector_t __rdev_sectors(struct raid_set *rs) | |
1563 | { | |
1564 | int i; | |
1565 | ||
1566 | for (i = 0; i < rs->md.raid_disks; i++) { | |
1567 | struct md_rdev *rdev = &rs->dev[i].rdev; | |
1568 | ||
63c32ed4 HM |
1569 | if (!test_bit(Journal, &rdev->flags) && |
1570 | rdev->bdev && rdev->sectors) | |
50c4feb9 HM |
1571 | return rdev->sectors; |
1572 | } | |
1573 | ||
4d49f1b4 | 1574 | return 0; |
50c4feb9 HM |
1575 | } |
1576 | ||
40ba37e5 HM |
1577 | /* Calculate the sectors per device and per array used for @rs */ |
1578 | static int rs_set_dev_and_array_sectors(struct raid_set *rs, bool use_mddev) | |
1579 | { | |
1580 | int delta_disks; | |
1581 | unsigned int data_stripes; | |
1582 | struct mddev *mddev = &rs->md; | |
1583 | struct md_rdev *rdev; | |
1584 | sector_t array_sectors = rs->ti->len, dev_sectors = rs->ti->len; | |
1585 | ||
1586 | if (use_mddev) { | |
1587 | delta_disks = mddev->delta_disks; | |
1588 | data_stripes = mddev_data_stripes(rs); | |
1589 | } else { | |
1590 | delta_disks = rs->delta_disks; | |
1591 | data_stripes = rs_data_stripes(rs); | |
1592 | } | |
1593 | ||
1594 | /* Special raid1 case w/o delta_disks support (yet) */ | |
1595 | if (rt_is_raid1(rs->raid_type)) | |
1596 | ; | |
1597 | else if (rt_is_raid10(rs->raid_type)) { | |
1598 | if (rs->raid10_copies < 2 || | |
1599 | delta_disks < 0) { | |
1600 | rs->ti->error = "Bogus raid10 data copies or delta disks"; | |
094f394d | 1601 | return -EINVAL; |
40ba37e5 HM |
1602 | } |
1603 | ||
1604 | dev_sectors *= rs->raid10_copies; | |
1605 | if (sector_div(dev_sectors, data_stripes)) | |
1606 | goto bad; | |
1607 | ||
1608 | array_sectors = (data_stripes + delta_disks) * dev_sectors; | |
1609 | if (sector_div(array_sectors, rs->raid10_copies)) | |
1610 | goto bad; | |
1611 | ||
1612 | } else if (sector_div(dev_sectors, data_stripes)) | |
1613 | goto bad; | |
1614 | ||
1615 | else | |
1616 | /* Striped layouts */ | |
1617 | array_sectors = (data_stripes + delta_disks) * dev_sectors; | |
1618 | ||
1619 | rdev_for_each(rdev, mddev) | |
63c32ed4 HM |
1620 | if (!test_bit(Journal, &rdev->flags)) |
1621 | rdev->sectors = dev_sectors; | |
40ba37e5 HM |
1622 | |
1623 | mddev->array_sectors = array_sectors; | |
1624 | mddev->dev_sectors = dev_sectors; | |
1625 | ||
1626 | return 0; | |
1627 | bad: | |
1628 | rs->ti->error = "Target length not divisible by number of data devices"; | |
094f394d | 1629 | return -EINVAL; |
40ba37e5 HM |
1630 | } |
1631 | ||
4dff2f1e HM |
1632 | /* Setup recovery on @rs */ |
1633 | static void __rs_setup_recovery(struct raid_set *rs, sector_t dev_sectors) | |
1634 | { | |
1635 | /* raid0 does not recover */ | |
1636 | if (rs_is_raid0(rs)) | |
1637 | rs->md.recovery_cp = MaxSector; | |
1638 | /* | |
1639 | * A raid6 set has to be recovered either | |
1640 | * completely or for the grown part to | |
1641 | * ensure proper parity and Q-Syndrome | |
1642 | */ | |
1643 | else if (rs_is_raid6(rs)) | |
1644 | rs->md.recovery_cp = dev_sectors; | |
1645 | /* | |
1646 | * Other raid set types may skip recovery | |
1647 | * depending on the 'nosync' flag. | |
1648 | */ | |
1649 | else | |
1650 | rs->md.recovery_cp = test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags) | |
1651 | ? MaxSector : dev_sectors; | |
1652 | } | |
1653 | ||
1654 | /* Setup recovery on @rs based on raid type, device size and 'nosync' flag */ | |
1655 | static void rs_setup_recovery(struct raid_set *rs, sector_t dev_sectors) | |
1656 | { | |
1657 | if (!dev_sectors) | |
1658 | /* New raid set or 'sync' flag provided */ | |
1659 | __rs_setup_recovery(rs, 0); | |
1660 | else if (dev_sectors == MaxSector) | |
1661 | /* Prevent recovery */ | |
1662 | __rs_setup_recovery(rs, MaxSector); | |
50c4feb9 | 1663 | else if (__rdev_sectors(rs) < dev_sectors) |
4dff2f1e | 1664 | /* Grown raid set */ |
50c4feb9 | 1665 | __rs_setup_recovery(rs, __rdev_sectors(rs)); |
4dff2f1e HM |
1666 | else |
1667 | __rs_setup_recovery(rs, MaxSector); | |
1668 | } | |
1669 | ||
9d09e663 N |
1670 | static void do_table_event(struct work_struct *ws) |
1671 | { | |
1672 | struct raid_set *rs = container_of(ws, struct raid_set, md.event_work); | |
1673 | ||
9d9d939c HM |
1674 | smp_rmb(); /* Make sure we access most actual mddev properties */ |
1675 | if (!rs_is_reshaping(rs)) | |
1676 | rs_set_capacity(rs); | |
9d09e663 N |
1677 | dm_table_event(rs->ti->table); |
1678 | } | |
1679 | ||
1680 | static int raid_is_congested(struct dm_target_callbacks *cb, int bits) | |
1681 | { | |
1682 | struct raid_set *rs = container_of(cb, struct raid_set, callbacks); | |
1683 | ||
5c675f83 | 1684 | return mddev_congested(&rs->md, bits); |
9d09e663 N |
1685 | } |
1686 | ||
ecbfb9f1 HM |
1687 | /* |
1688 | * Make sure a valid takover (level switch) is being requested on @rs | |
1689 | * | |
1690 | * Conversions of raid sets from one MD personality to another | |
1691 | * have to conform to restrictions which are enforced here. | |
ecbfb9f1 HM |
1692 | */ |
1693 | static int rs_check_takeover(struct raid_set *rs) | |
1694 | { | |
1695 | struct mddev *mddev = &rs->md; | |
1696 | unsigned int near_copies; | |
1697 | ||
9dbd1aa3 HM |
1698 | if (rs->md.degraded) { |
1699 | rs->ti->error = "Can't takeover degraded raid set"; | |
1700 | return -EPERM; | |
1701 | } | |
1702 | ||
1703 | if (rs_is_reshaping(rs)) { | |
1704 | rs->ti->error = "Can't takeover reshaping raid set"; | |
1705 | return -EPERM; | |
1706 | } | |
1707 | ||
ecbfb9f1 HM |
1708 | switch (mddev->level) { |
1709 | case 0: | |
1710 | /* raid0 -> raid1/5 with one disk */ | |
1711 | if ((mddev->new_level == 1 || mddev->new_level == 5) && | |
1712 | mddev->raid_disks == 1) | |
1713 | return 0; | |
1714 | ||
1715 | /* raid0 -> raid10 */ | |
1716 | if (mddev->new_level == 10 && | |
9dbd1aa3 | 1717 | !(rs->raid_disks % mddev->raid_disks)) |
ecbfb9f1 HM |
1718 | return 0; |
1719 | ||
1720 | /* raid0 with multiple disks -> raid4/5/6 */ | |
bb91a63f | 1721 | if (__within_range(mddev->new_level, 4, 6) && |
ecbfb9f1 HM |
1722 | mddev->new_layout == ALGORITHM_PARITY_N && |
1723 | mddev->raid_disks > 1) | |
1724 | return 0; | |
1725 | ||
1726 | break; | |
1727 | ||
1728 | case 10: | |
1729 | /* Can't takeover raid10_offset! */ | |
e6ca5e1a | 1730 | if (__is_raid10_offset(mddev->layout)) |
ecbfb9f1 HM |
1731 | break; |
1732 | ||
e6ca5e1a | 1733 | near_copies = __raid10_near_copies(mddev->layout); |
ecbfb9f1 HM |
1734 | |
1735 | /* raid10* -> raid0 */ | |
1736 | if (mddev->new_level == 0) { | |
1737 | /* Can takeover raid10_near with raid disks divisable by data copies! */ | |
1738 | if (near_copies > 1 && | |
1739 | !(mddev->raid_disks % near_copies)) { | |
1740 | mddev->raid_disks /= near_copies; | |
1741 | mddev->delta_disks = mddev->raid_disks; | |
1742 | return 0; | |
1743 | } | |
1744 | ||
1745 | /* Can takeover raid10_far */ | |
1746 | if (near_copies == 1 && | |
e6ca5e1a | 1747 | __raid10_far_copies(mddev->layout) > 1) |
ecbfb9f1 HM |
1748 | return 0; |
1749 | ||
1750 | break; | |
1751 | } | |
1752 | ||
1753 | /* raid10_{near,far} -> raid1 */ | |
1754 | if (mddev->new_level == 1 && | |
e6ca5e1a | 1755 | max(near_copies, __raid10_far_copies(mddev->layout)) == mddev->raid_disks) |
ecbfb9f1 HM |
1756 | return 0; |
1757 | ||
1758 | /* raid10_{near,far} with 2 disks -> raid4/5 */ | |
bb91a63f | 1759 | if (__within_range(mddev->new_level, 4, 5) && |
ecbfb9f1 HM |
1760 | mddev->raid_disks == 2) |
1761 | return 0; | |
1762 | break; | |
1763 | ||
1764 | case 1: | |
1765 | /* raid1 with 2 disks -> raid4/5 */ | |
bb91a63f | 1766 | if (__within_range(mddev->new_level, 4, 5) && |
ecbfb9f1 HM |
1767 | mddev->raid_disks == 2) { |
1768 | mddev->degraded = 1; | |
1769 | return 0; | |
1770 | } | |
1771 | ||
1772 | /* raid1 -> raid0 */ | |
1773 | if (mddev->new_level == 0 && | |
1774 | mddev->raid_disks == 1) | |
1775 | return 0; | |
1776 | ||
1777 | /* raid1 -> raid10 */ | |
1778 | if (mddev->new_level == 10) | |
1779 | return 0; | |
ecbfb9f1 HM |
1780 | break; |
1781 | ||
1782 | case 4: | |
1783 | /* raid4 -> raid0 */ | |
1784 | if (mddev->new_level == 0) | |
1785 | return 0; | |
1786 | ||
1787 | /* raid4 -> raid1/5 with 2 disks */ | |
1788 | if ((mddev->new_level == 1 || mddev->new_level == 5) && | |
1789 | mddev->raid_disks == 2) | |
1790 | return 0; | |
1791 | ||
1792 | /* raid4 -> raid5/6 with parity N */ | |
bb91a63f | 1793 | if (__within_range(mddev->new_level, 5, 6) && |
ecbfb9f1 HM |
1794 | mddev->layout == ALGORITHM_PARITY_N) |
1795 | return 0; | |
1796 | break; | |
1797 | ||
1798 | case 5: | |
1799 | /* raid5 with parity N -> raid0 */ | |
1800 | if (mddev->new_level == 0 && | |
1801 | mddev->layout == ALGORITHM_PARITY_N) | |
1802 | return 0; | |
1803 | ||
1804 | /* raid5 with parity N -> raid4 */ | |
1805 | if (mddev->new_level == 4 && | |
1806 | mddev->layout == ALGORITHM_PARITY_N) | |
1807 | return 0; | |
1808 | ||
1809 | /* raid5 with 2 disks -> raid1/4/10 */ | |
1810 | if ((mddev->new_level == 1 || mddev->new_level == 4 || mddev->new_level == 10) && | |
1811 | mddev->raid_disks == 2) | |
1812 | return 0; | |
1813 | ||
6ee0bae9 | 1814 | /* raid5_* -> raid6_*_6 with Q-Syndrome N (e.g. raid5_ra -> raid6_ra_6 */ |
ecbfb9f1 HM |
1815 | if (mddev->new_level == 6 && |
1816 | ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || | |
bb91a63f | 1817 | __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC_6, ALGORITHM_RIGHT_SYMMETRIC_6))) |
ecbfb9f1 HM |
1818 | return 0; |
1819 | break; | |
1820 | ||
1821 | case 6: | |
1822 | /* raid6 with parity N -> raid0 */ | |
1823 | if (mddev->new_level == 0 && | |
1824 | mddev->layout == ALGORITHM_PARITY_N) | |
1825 | return 0; | |
1826 | ||
1827 | /* raid6 with parity N -> raid4 */ | |
1828 | if (mddev->new_level == 4 && | |
1829 | mddev->layout == ALGORITHM_PARITY_N) | |
1830 | return 0; | |
1831 | ||
6ee0bae9 | 1832 | /* raid6_*_n with Q-Syndrome N -> raid5_* */ |
ecbfb9f1 HM |
1833 | if (mddev->new_level == 5 && |
1834 | ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || | |
bb91a63f | 1835 | __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC, ALGORITHM_RIGHT_SYMMETRIC))) |
ecbfb9f1 HM |
1836 | return 0; |
1837 | ||
1838 | default: | |
1839 | break; | |
1840 | } | |
1841 | ||
bd83a4c4 MS |
1842 | rs->ti->error = "takeover not possible"; |
1843 | return -EINVAL; | |
ecbfb9f1 HM |
1844 | } |
1845 | ||
1846 | /* True if @rs requested to be taken over */ | |
1847 | static bool rs_takeover_requested(struct raid_set *rs) | |
1848 | { | |
1849 | return rs->md.new_level != rs->md.level; | |
1850 | } | |
1851 | ||
40ba37e5 HM |
1852 | /* True if @rs is requested to reshape by ctr */ |
1853 | static bool rs_reshape_requested(struct raid_set *rs) | |
1854 | { | |
469b304b | 1855 | bool change; |
40ba37e5 HM |
1856 | struct mddev *mddev = &rs->md; |
1857 | ||
469b304b HM |
1858 | if (rs_takeover_requested(rs)) |
1859 | return false; | |
1860 | ||
40ba37e5 HM |
1861 | if (!mddev->level) |
1862 | return false; | |
1863 | ||
469b304b HM |
1864 | change = mddev->new_layout != mddev->layout || |
1865 | mddev->new_chunk_sectors != mddev->chunk_sectors || | |
1866 | rs->delta_disks; | |
1867 | ||
1868 | /* Historical case to support raid1 reshape without delta disks */ | |
7a7c330f HM |
1869 | if (mddev->level == 1) { |
1870 | if (rs->delta_disks) | |
1871 | return !!rs->delta_disks; | |
1872 | ||
469b304b HM |
1873 | return !change && |
1874 | mddev->raid_disks != rs->raid_disks; | |
7a7c330f | 1875 | } |
469b304b HM |
1876 | |
1877 | if (mddev->level == 10) | |
1878 | return change && | |
1879 | !__is_raid10_far(mddev->new_layout) && | |
1880 | rs->delta_disks >= 0; | |
1881 | ||
1882 | return change; | |
40ba37e5 HM |
1883 | } |
1884 | ||
33e53f06 | 1885 | /* Features */ |
9b6e5423 | 1886 | #define FEATURE_FLAG_SUPPORTS_V190 0x1 /* Supports extended superblock */ |
33e53f06 HM |
1887 | |
1888 | /* State flags for sb->flags */ | |
1889 | #define SB_FLAG_RESHAPE_ACTIVE 0x1 | |
1890 | #define SB_FLAG_RESHAPE_BACKWARDS 0x2 | |
1891 | ||
b12d437b JB |
1892 | /* |
1893 | * This structure is never routinely used by userspace, unlike md superblocks. | |
1894 | * Devices with this superblock should only ever be accessed via device-mapper. | |
1895 | */ | |
1896 | #define DM_RAID_MAGIC 0x64526D44 | |
1897 | struct dm_raid_superblock { | |
1898 | __le32 magic; /* "DmRd" */ | |
9b6e5423 | 1899 | __le32 compat_features; /* Used to indicate compatible features (like 1.9.0 ondisk metadata extension) */ |
b12d437b | 1900 | |
33e53f06 HM |
1901 | __le32 num_devices; /* Number of devices in this raid set. (Max 64) */ |
1902 | __le32 array_position; /* The position of this drive in the raid set */ | |
b12d437b JB |
1903 | |
1904 | __le64 events; /* Incremented by md when superblock updated */ | |
9b6e5423 | 1905 | __le64 failed_devices; /* Pre 1.9.0 part of bit field of devices to */ |
33e53f06 | 1906 | /* indicate failures (see extension below) */ |
b12d437b JB |
1907 | |
1908 | /* | |
1909 | * This offset tracks the progress of the repair or replacement of | |
1910 | * an individual drive. | |
1911 | */ | |
1912 | __le64 disk_recovery_offset; | |
1913 | ||
1914 | /* | |
33e53f06 | 1915 | * This offset tracks the progress of the initial raid set |
b12d437b JB |
1916 | * synchronisation/parity calculation. |
1917 | */ | |
1918 | __le64 array_resync_offset; | |
1919 | ||
1920 | /* | |
33e53f06 | 1921 | * raid characteristics |
b12d437b JB |
1922 | */ |
1923 | __le32 level; | |
1924 | __le32 layout; | |
1925 | __le32 stripe_sectors; | |
1926 | ||
33e53f06 | 1927 | /******************************************************************** |
9b6e5423 | 1928 | * BELOW FOLLOW V1.9.0 EXTENSIONS TO THE PRISTINE SUPERBLOCK FORMAT!!! |
33e53f06 | 1929 | * |
c4d097d1 | 1930 | * FEATURE_FLAG_SUPPORTS_V190 in the compat_features member indicates that those exist |
33e53f06 HM |
1931 | */ |
1932 | ||
1933 | __le32 flags; /* Flags defining array states for reshaping */ | |
1934 | ||
1935 | /* | |
1936 | * This offset tracks the progress of a raid | |
1937 | * set reshape in order to be able to restart it | |
1938 | */ | |
1939 | __le64 reshape_position; | |
1940 | ||
1941 | /* | |
1942 | * These define the properties of the array in case of an interrupted reshape | |
1943 | */ | |
1944 | __le32 new_level; | |
1945 | __le32 new_layout; | |
1946 | __le32 new_stripe_sectors; | |
1947 | __le32 delta_disks; | |
1948 | ||
1949 | __le64 array_sectors; /* Array size in sectors */ | |
1950 | ||
1951 | /* | |
1952 | * Sector offsets to data on devices (reshaping). | |
1953 | * Needed to support out of place reshaping, thus | |
1954 | * not writing over any stripes whilst converting | |
1955 | * them from old to new layout | |
1956 | */ | |
1957 | __le64 data_offset; | |
1958 | __le64 new_data_offset; | |
1959 | ||
1960 | __le64 sectors; /* Used device size in sectors */ | |
1961 | ||
1962 | /* | |
1963 | * Additonal Bit field of devices indicating failures to support | |
9b6e5423 | 1964 | * up to 256 devices with the 1.9.0 on-disk metadata format |
33e53f06 HM |
1965 | */ |
1966 | __le64 extended_failed_devices[DISKS_ARRAY_ELEMS - 1]; | |
1967 | ||
1968 | __le32 incompat_features; /* Used to indicate any incompatible features */ | |
1969 | ||
1970 | /* Always set rest up to logical block size to 0 when writing (see get_metadata_device() below). */ | |
b12d437b JB |
1971 | } __packed; |
1972 | ||
9dbd1aa3 HM |
1973 | /* |
1974 | * Check for reshape constraints on raid set @rs: | |
1975 | * | |
1976 | * - reshape function non-existent | |
1977 | * - degraded set | |
1978 | * - ongoing recovery | |
1979 | * - ongoing reshape | |
1980 | * | |
1981 | * Returns 0 if none or -EPERM if given constraint | |
1982 | * and error message reference in @errmsg | |
1983 | */ | |
1984 | static int rs_check_reshape(struct raid_set *rs) | |
1985 | { | |
1986 | struct mddev *mddev = &rs->md; | |
1987 | ||
9dbd1aa3 HM |
1988 | if (!mddev->pers || !mddev->pers->check_reshape) |
1989 | rs->ti->error = "Reshape not supported"; | |
1990 | else if (mddev->degraded) | |
1991 | rs->ti->error = "Can't reshape degraded raid set"; | |
1992 | else if (rs_is_recovering(rs)) | |
1993 | rs->ti->error = "Convert request on recovering raid set prohibited"; | |
469b304b | 1994 | else if (rs_is_reshaping(rs)) |
9dbd1aa3 | 1995 | rs->ti->error = "raid set already reshaping!"; |
7a7c330f HM |
1996 | else if (!(rs_is_raid1(rs) || rs_is_raid10(rs) || rs_is_raid456(rs))) |
1997 | rs->ti->error = "Reshaping only supported for raid1/4/5/6/10"; | |
9dbd1aa3 HM |
1998 | else |
1999 | return 0; | |
2000 | ||
2001 | return -EPERM; | |
2002 | } | |
2003 | ||
e2568465 | 2004 | static int read_disk_sb(struct md_rdev *rdev, int size, bool force_reload) |
b12d437b JB |
2005 | { |
2006 | BUG_ON(!rdev->sb_page); | |
2007 | ||
e2568465 | 2008 | if (rdev->sb_loaded && !force_reload) |
b12d437b JB |
2009 | return 0; |
2010 | ||
e2568465 HM |
2011 | rdev->sb_loaded = 0; |
2012 | ||
0a7b8188 | 2013 | if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, 0, true)) { |
0447568f JB |
2014 | DMERR("Failed to read superblock of device at position %d", |
2015 | rdev->raid_disk); | |
c32fb9e7 | 2016 | md_error(rdev->mddev, rdev); |
e2568465 HM |
2017 | set_bit(Faulty, &rdev->flags); |
2018 | return -EIO; | |
b12d437b JB |
2019 | } |
2020 | ||
2021 | rdev->sb_loaded = 1; | |
2022 | ||
2023 | return 0; | |
2024 | } | |
2025 | ||
33e53f06 HM |
2026 | static void sb_retrieve_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) |
2027 | { | |
2028 | failed_devices[0] = le64_to_cpu(sb->failed_devices); | |
2029 | memset(failed_devices + 1, 0, sizeof(sb->extended_failed_devices)); | |
2030 | ||
4286325b | 2031 | if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { |
33e53f06 HM |
2032 | int i = ARRAY_SIZE(sb->extended_failed_devices); |
2033 | ||
2034 | while (i--) | |
2035 | failed_devices[i+1] = le64_to_cpu(sb->extended_failed_devices[i]); | |
2036 | } | |
2037 | } | |
2038 | ||
7b34df74 HM |
2039 | static void sb_update_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) |
2040 | { | |
2041 | int i = ARRAY_SIZE(sb->extended_failed_devices); | |
2042 | ||
2043 | sb->failed_devices = cpu_to_le64(failed_devices[0]); | |
2044 | while (i--) | |
2045 | sb->extended_failed_devices[i] = cpu_to_le64(failed_devices[i+1]); | |
2046 | } | |
2047 | ||
2048 | /* | |
2049 | * Synchronize the superblock members with the raid set properties | |
2050 | * | |
2051 | * All superblock data is little endian. | |
2052 | */ | |
fd01b88c | 2053 | static void super_sync(struct mddev *mddev, struct md_rdev *rdev) |
b12d437b | 2054 | { |
7b34df74 HM |
2055 | bool update_failed_devices = false; |
2056 | unsigned int i; | |
2057 | uint64_t failed_devices[DISKS_ARRAY_ELEMS]; | |
b12d437b | 2058 | struct dm_raid_superblock *sb; |
81f382f9 | 2059 | struct raid_set *rs = container_of(mddev, struct raid_set, md); |
b12d437b | 2060 | |
7b34df74 HM |
2061 | /* No metadata device, no superblock */ |
2062 | if (!rdev->meta_bdev) | |
2063 | return; | |
2064 | ||
2065 | BUG_ON(!rdev->sb_page); | |
2066 | ||
b12d437b | 2067 | sb = page_address(rdev->sb_page); |
b12d437b | 2068 | |
7b34df74 | 2069 | sb_retrieve_failed_devices(sb, failed_devices); |
b12d437b | 2070 | |
7b34df74 HM |
2071 | for (i = 0; i < rs->raid_disks; i++) |
2072 | if (!rs->dev[i].data_dev || test_bit(Faulty, &rs->dev[i].rdev.flags)) { | |
2073 | update_failed_devices = true; | |
2074 | set_bit(i, (void *) failed_devices); | |
2075 | } | |
2076 | ||
2077 | if (update_failed_devices) | |
2078 | sb_update_failed_devices(sb, failed_devices); | |
b12d437b JB |
2079 | |
2080 | sb->magic = cpu_to_le32(DM_RAID_MAGIC); | |
9b6e5423 | 2081 | sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); |
b12d437b JB |
2082 | |
2083 | sb->num_devices = cpu_to_le32(mddev->raid_disks); | |
2084 | sb->array_position = cpu_to_le32(rdev->raid_disk); | |
2085 | ||
2086 | sb->events = cpu_to_le64(mddev->events); | |
b12d437b JB |
2087 | |
2088 | sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset); | |
2089 | sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp); | |
2090 | ||
2091 | sb->level = cpu_to_le32(mddev->level); | |
2092 | sb->layout = cpu_to_le32(mddev->layout); | |
2093 | sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors); | |
7b34df74 | 2094 | |
c4d097d1 HM |
2095 | /******************************************************************** |
2096 | * BELOW FOLLOW V1.9.0 EXTENSIONS TO THE PRISTINE SUPERBLOCK FORMAT!!! | |
2097 | * | |
2098 | * FEATURE_FLAG_SUPPORTS_V190 in the compat_features member indicates that those exist | |
2099 | */ | |
7b34df74 HM |
2100 | sb->new_level = cpu_to_le32(mddev->new_level); |
2101 | sb->new_layout = cpu_to_le32(mddev->new_layout); | |
2102 | sb->new_stripe_sectors = cpu_to_le32(mddev->new_chunk_sectors); | |
2103 | ||
2104 | sb->delta_disks = cpu_to_le32(mddev->delta_disks); | |
2105 | ||
2106 | smp_rmb(); /* Make sure we access most recent reshape position */ | |
2107 | sb->reshape_position = cpu_to_le64(mddev->reshape_position); | |
2108 | if (le64_to_cpu(sb->reshape_position) != MaxSector) { | |
2109 | /* Flag ongoing reshape */ | |
2110 | sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE); | |
2111 | ||
2112 | if (mddev->delta_disks < 0 || mddev->reshape_backwards) | |
2113 | sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_BACKWARDS); | |
4286325b MS |
2114 | } else { |
2115 | /* Clear reshape flags */ | |
2116 | sb->flags &= ~(cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE|SB_FLAG_RESHAPE_BACKWARDS)); | |
2117 | } | |
7b34df74 HM |
2118 | |
2119 | sb->array_sectors = cpu_to_le64(mddev->array_sectors); | |
2120 | sb->data_offset = cpu_to_le64(rdev->data_offset); | |
2121 | sb->new_data_offset = cpu_to_le64(rdev->new_data_offset); | |
2122 | sb->sectors = cpu_to_le64(rdev->sectors); | |
b2a4872a | 2123 | sb->incompat_features = cpu_to_le32(0); |
7b34df74 HM |
2124 | |
2125 | /* Zero out the rest of the payload after the size of the superblock */ | |
2126 | memset(sb + 1, 0, rdev->sb_size - sizeof(*sb)); | |
b12d437b JB |
2127 | } |
2128 | ||
2129 | /* | |
2130 | * super_load | |
2131 | * | |
2132 | * This function creates a superblock if one is not found on the device | |
2133 | * and will decide which superblock to use if there's a choice. | |
2134 | * | |
2135 | * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise | |
2136 | */ | |
3cb03002 | 2137 | static int super_load(struct md_rdev *rdev, struct md_rdev *refdev) |
b12d437b | 2138 | { |
73c6f239 | 2139 | int r; |
b12d437b JB |
2140 | struct dm_raid_superblock *sb; |
2141 | struct dm_raid_superblock *refsb; | |
2142 | uint64_t events_sb, events_refsb; | |
2143 | ||
2144 | rdev->sb_start = 0; | |
40d43c4b HM |
2145 | rdev->sb_size = bdev_logical_block_size(rdev->meta_bdev); |
2146 | if (rdev->sb_size < sizeof(*sb) || rdev->sb_size > PAGE_SIZE) { | |
2147 | DMERR("superblock size of a logical block is no longer valid"); | |
2148 | return -EINVAL; | |
2149 | } | |
b12d437b | 2150 | |
e2568465 | 2151 | r = read_disk_sb(rdev, rdev->sb_size, false); |
73c6f239 HM |
2152 | if (r) |
2153 | return r; | |
b12d437b JB |
2154 | |
2155 | sb = page_address(rdev->sb_page); | |
3aa3b2b2 JB |
2156 | |
2157 | /* | |
2158 | * Two cases that we want to write new superblocks and rebuild: | |
2159 | * 1) New device (no matching magic number) | |
2160 | * 2) Device specified for rebuild (!In_sync w/ offset == 0) | |
2161 | */ | |
2162 | if ((sb->magic != cpu_to_le32(DM_RAID_MAGIC)) || | |
2163 | (!test_bit(In_sync, &rdev->flags) && !rdev->recovery_offset)) { | |
b12d437b JB |
2164 | super_sync(rdev->mddev, rdev); |
2165 | ||
2166 | set_bit(FirstUse, &rdev->flags); | |
9b6e5423 | 2167 | sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); |
b12d437b JB |
2168 | |
2169 | /* Force writing of superblocks to disk */ | |
2953079c | 2170 | set_bit(MD_SB_CHANGE_DEVS, &rdev->mddev->sb_flags); |
b12d437b JB |
2171 | |
2172 | /* Any superblock is better than none, choose that if given */ | |
2173 | return refdev ? 0 : 1; | |
2174 | } | |
2175 | ||
2176 | if (!refdev) | |
2177 | return 1; | |
2178 | ||
2179 | events_sb = le64_to_cpu(sb->events); | |
2180 | ||
2181 | refsb = page_address(refdev->sb_page); | |
2182 | events_refsb = le64_to_cpu(refsb->events); | |
2183 | ||
2184 | return (events_sb > events_refsb) ? 1 : 0; | |
2185 | } | |
2186 | ||
33e53f06 | 2187 | static int super_init_validation(struct raid_set *rs, struct md_rdev *rdev) |
b12d437b JB |
2188 | { |
2189 | int role; | |
33e53f06 HM |
2190 | unsigned int d; |
2191 | struct mddev *mddev = &rs->md; | |
b12d437b | 2192 | uint64_t events_sb; |
33e53f06 | 2193 | uint64_t failed_devices[DISKS_ARRAY_ELEMS]; |
b12d437b | 2194 | struct dm_raid_superblock *sb; |
33e53f06 | 2195 | uint32_t new_devs = 0, rebuild_and_new = 0, rebuilds = 0; |
dafb20fa | 2196 | struct md_rdev *r; |
b12d437b JB |
2197 | struct dm_raid_superblock *sb2; |
2198 | ||
2199 | sb = page_address(rdev->sb_page); | |
2200 | events_sb = le64_to_cpu(sb->events); | |
b12d437b JB |
2201 | |
2202 | /* | |
2203 | * Initialise to 1 if this is a new superblock. | |
2204 | */ | |
2205 | mddev->events = events_sb ? : 1; | |
2206 | ||
33e53f06 HM |
2207 | mddev->reshape_position = MaxSector; |
2208 | ||
453c2a89 HM |
2209 | mddev->raid_disks = le32_to_cpu(sb->num_devices); |
2210 | mddev->level = le32_to_cpu(sb->level); | |
2211 | mddev->layout = le32_to_cpu(sb->layout); | |
2212 | mddev->chunk_sectors = le32_to_cpu(sb->stripe_sectors); | |
2213 | ||
b12d437b | 2214 | /* |
33e53f06 HM |
2215 | * Reshaping is supported, e.g. reshape_position is valid |
2216 | * in superblock and superblock content is authoritative. | |
b12d437b | 2217 | */ |
4286325b | 2218 | if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { |
33e53f06 | 2219 | /* Superblock is authoritative wrt given raid set layout! */ |
33e53f06 HM |
2220 | mddev->new_level = le32_to_cpu(sb->new_level); |
2221 | mddev->new_layout = le32_to_cpu(sb->new_layout); | |
2222 | mddev->new_chunk_sectors = le32_to_cpu(sb->new_stripe_sectors); | |
2223 | mddev->delta_disks = le32_to_cpu(sb->delta_disks); | |
2224 | mddev->array_sectors = le64_to_cpu(sb->array_sectors); | |
2225 | ||
2226 | /* raid was reshaping and got interrupted */ | |
4286325b MS |
2227 | if (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_ACTIVE) { |
2228 | if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { | |
33e53f06 HM |
2229 | DMERR("Reshape requested but raid set is still reshaping"); |
2230 | return -EINVAL; | |
2231 | } | |
b12d437b | 2232 | |
33e53f06 | 2233 | if (mddev->delta_disks < 0 || |
4286325b | 2234 | (!mddev->delta_disks && (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_BACKWARDS))) |
33e53f06 HM |
2235 | mddev->reshape_backwards = 1; |
2236 | else | |
2237 | mddev->reshape_backwards = 0; | |
2238 | ||
2239 | mddev->reshape_position = le64_to_cpu(sb->reshape_position); | |
2240 | rs->raid_type = get_raid_type_by_ll(mddev->level, mddev->layout); | |
2241 | } | |
2242 | ||
2243 | } else { | |
2244 | /* | |
9b6e5423 | 2245 | * No takeover/reshaping, because we don't have the extended v1.9.0 metadata |
33e53f06 | 2246 | */ |
453c2a89 HM |
2247 | struct raid_type *rt_cur = get_raid_type_by_ll(mddev->level, mddev->layout); |
2248 | struct raid_type *rt_new = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); | |
33e53f06 | 2249 | |
453c2a89 HM |
2250 | if (rs_takeover_requested(rs)) { |
2251 | if (rt_cur && rt_new) | |
2252 | DMERR("Takeover raid sets from %s to %s not yet supported by metadata. (raid level change)", | |
2253 | rt_cur->name, rt_new->name); | |
2254 | else | |
2255 | DMERR("Takeover raid sets not yet supported by metadata. (raid level change)"); | |
2256 | return -EINVAL; | |
2257 | } else if (rs_reshape_requested(rs)) { | |
2258 | DMERR("Reshaping raid sets not yet supported by metadata. (raid layout change keeping level)"); | |
2259 | if (mddev->layout != mddev->new_layout) { | |
2260 | if (rt_cur && rt_new) | |
2261 | DMERR(" current layout %s vs new layout %s", | |
2262 | rt_cur->name, rt_new->name); | |
2263 | else | |
2264 | DMERR(" current layout 0x%X vs new layout 0x%X", | |
2265 | le32_to_cpu(sb->layout), mddev->new_layout); | |
2266 | } | |
2267 | if (mddev->chunk_sectors != mddev->new_chunk_sectors) | |
2268 | DMERR(" current stripe sectors %u vs new stripe sectors %u", | |
2269 | mddev->chunk_sectors, mddev->new_chunk_sectors); | |
2270 | if (rs->delta_disks) | |
2271 | DMERR(" current %u disks vs new %u disks", | |
2272 | mddev->raid_disks, mddev->raid_disks + rs->delta_disks); | |
2273 | if (rs_is_raid10(rs)) { | |
2274 | DMERR(" Old layout: %s w/ %u copies", | |
2275 | raid10_md_layout_to_format(mddev->layout), | |
2276 | raid10_md_layout_to_copies(mddev->layout)); | |
2277 | DMERR(" New layout: %s w/ %u copies", | |
2278 | raid10_md_layout_to_format(mddev->new_layout), | |
2279 | raid10_md_layout_to_copies(mddev->new_layout)); | |
2280 | } | |
33e53f06 HM |
2281 | return -EINVAL; |
2282 | } | |
2283 | ||
b052b07c | 2284 | DMINFO("Discovered old metadata format; upgrading to extended metadata format"); |
b12d437b JB |
2285 | } |
2286 | ||
4286325b | 2287 | if (!test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) |
b12d437b JB |
2288 | mddev->recovery_cp = le64_to_cpu(sb->array_resync_offset); |
2289 | ||
2290 | /* | |
2291 | * During load, we set FirstUse if a new superblock was written. | |
2292 | * There are two reasons we might not have a superblock: | |
33e53f06 | 2293 | * 1) The raid set is brand new - in which case, all of the |
43157840 | 2294 | * devices must have their In_sync bit set. Also, |
b12d437b | 2295 | * recovery_cp must be 0, unless forced. |
33e53f06 | 2296 | * 2) This is a new device being added to an old raid set |
b12d437b JB |
2297 | * and the new device needs to be rebuilt - in which |
2298 | * case the In_sync bit will /not/ be set and | |
2299 | * recovery_cp must be MaxSector. | |
9dbd1aa3 HM |
2300 | * 3) This is/are a new device(s) being added to an old |
2301 | * raid set during takeover to a higher raid level | |
2302 | * to provide capacity for redundancy or during reshape | |
2303 | * to add capacity to grow the raid set. | |
b12d437b | 2304 | */ |
33e53f06 | 2305 | d = 0; |
dafb20fa | 2306 | rdev_for_each(r, mddev) { |
63c32ed4 HM |
2307 | if (test_bit(Journal, &rdev->flags)) |
2308 | continue; | |
2309 | ||
33e53f06 HM |
2310 | if (test_bit(FirstUse, &r->flags)) |
2311 | new_devs++; | |
2312 | ||
b12d437b | 2313 | if (!test_bit(In_sync, &r->flags)) { |
33e53f06 HM |
2314 | DMINFO("Device %d specified for rebuild; clearing superblock", |
2315 | r->raid_disk); | |
b12d437b | 2316 | rebuilds++; |
33e53f06 HM |
2317 | |
2318 | if (test_bit(FirstUse, &r->flags)) | |
2319 | rebuild_and_new++; | |
2320 | } | |
2321 | ||
2322 | d++; | |
b12d437b JB |
2323 | } |
2324 | ||
33e53f06 HM |
2325 | if (new_devs == rs->raid_disks || !rebuilds) { |
2326 | /* Replace a broken device */ | |
2327 | if (new_devs == 1 && !rs->delta_disks) | |
2328 | ; | |
2329 | if (new_devs == rs->raid_disks) { | |
2330 | DMINFO("Superblocks created for new raid set"); | |
b12d437b | 2331 | set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); |
9dbd1aa3 HM |
2332 | } else if (new_devs != rebuilds && |
2333 | new_devs != rs->delta_disks) { | |
33e53f06 HM |
2334 | DMERR("New device injected into existing raid set without " |
2335 | "'delta_disks' or 'rebuild' parameter specified"); | |
b12d437b JB |
2336 | return -EINVAL; |
2337 | } | |
33e53f06 HM |
2338 | } else if (new_devs && new_devs != rebuilds) { |
2339 | DMERR("%u 'rebuild' devices cannot be injected into" | |
2340 | " a raid set with %u other first-time devices", | |
2341 | rebuilds, new_devs); | |
b12d437b | 2342 | return -EINVAL; |
33e53f06 HM |
2343 | } else if (rebuilds) { |
2344 | if (rebuild_and_new && rebuilds != rebuild_and_new) { | |
2345 | DMERR("new device%s provided without 'rebuild'", | |
2346 | new_devs > 1 ? "s" : ""); | |
2347 | return -EINVAL; | |
9dbd1aa3 | 2348 | } else if (rs_is_recovering(rs)) { |
33e53f06 HM |
2349 | DMERR("'rebuild' specified while raid set is not in-sync (recovery_cp=%llu)", |
2350 | (unsigned long long) mddev->recovery_cp); | |
2351 | return -EINVAL; | |
9dbd1aa3 HM |
2352 | } else if (rs_is_reshaping(rs)) { |
2353 | DMERR("'rebuild' specified while raid set is being reshaped (reshape_position=%llu)", | |
2354 | (unsigned long long) mddev->reshape_position); | |
33e53f06 HM |
2355 | return -EINVAL; |
2356 | } | |
b12d437b JB |
2357 | } |
2358 | ||
2359 | /* | |
2360 | * Now we set the Faulty bit for those devices that are | |
2361 | * recorded in the superblock as failed. | |
2362 | */ | |
33e53f06 | 2363 | sb_retrieve_failed_devices(sb, failed_devices); |
dafb20fa | 2364 | rdev_for_each(r, mddev) { |
63c32ed4 HM |
2365 | if (test_bit(Journal, &rdev->flags) || |
2366 | !r->sb_page) | |
b12d437b JB |
2367 | continue; |
2368 | sb2 = page_address(r->sb_page); | |
2369 | sb2->failed_devices = 0; | |
33e53f06 | 2370 | memset(sb2->extended_failed_devices, 0, sizeof(sb2->extended_failed_devices)); |
b12d437b JB |
2371 | |
2372 | /* | |
2373 | * Check for any device re-ordering. | |
2374 | */ | |
2375 | if (!test_bit(FirstUse, &r->flags) && (r->raid_disk >= 0)) { | |
2376 | role = le32_to_cpu(sb2->array_position); | |
33e53f06 HM |
2377 | if (role < 0) |
2378 | continue; | |
2379 | ||
b12d437b | 2380 | if (role != r->raid_disk) { |
453c2a89 | 2381 | if (rs_is_raid10(rs) && __is_raid10_near(mddev->layout)) { |
e6ca5e1a | 2382 | if (mddev->raid_disks % __raid10_near_copies(mddev->layout) || |
bd83a4c4 MS |
2383 | rs->raid_disks % rs->raid10_copies) { |
2384 | rs->ti->error = | |
2385 | "Cannot change raid10 near set to odd # of devices!"; | |
2386 | return -EINVAL; | |
2387 | } | |
33e53f06 HM |
2388 | |
2389 | sb2->array_position = cpu_to_le32(r->raid_disk); | |
2390 | ||
2391 | } else if (!(rs_is_raid10(rs) && rt_is_raid0(rs->raid_type)) && | |
bd83a4c4 MS |
2392 | !(rs_is_raid0(rs) && rt_is_raid10(rs->raid_type)) && |
2393 | !rt_is_raid1(rs->raid_type)) { | |
2394 | rs->ti->error = "Cannot change device positions in raid set"; | |
2395 | return -EINVAL; | |
2396 | } | |
33e53f06 | 2397 | |
bd83a4c4 | 2398 | DMINFO("raid device #%d now at position #%d", role, r->raid_disk); |
b12d437b JB |
2399 | } |
2400 | ||
2401 | /* | |
2402 | * Partial recovery is performed on | |
2403 | * returning failed devices. | |
2404 | */ | |
33e53f06 | 2405 | if (test_bit(role, (void *) failed_devices)) |
b12d437b JB |
2406 | set_bit(Faulty, &r->flags); |
2407 | } | |
2408 | } | |
2409 | ||
2410 | return 0; | |
2411 | } | |
2412 | ||
0cf45031 | 2413 | static int super_validate(struct raid_set *rs, struct md_rdev *rdev) |
b12d437b | 2414 | { |
0cf45031 | 2415 | struct mddev *mddev = &rs->md; |
33e53f06 HM |
2416 | struct dm_raid_superblock *sb; |
2417 | ||
c63ede3b | 2418 | if (rs_is_raid0(rs) || !rdev->sb_page || rdev->raid_disk < 0) |
33e53f06 HM |
2419 | return 0; |
2420 | ||
2421 | sb = page_address(rdev->sb_page); | |
b12d437b JB |
2422 | |
2423 | /* | |
2424 | * If mddev->events is not set, we know we have not yet initialized | |
2425 | * the array. | |
2426 | */ | |
33e53f06 | 2427 | if (!mddev->events && super_init_validation(rs, rdev)) |
b12d437b JB |
2428 | return -EINVAL; |
2429 | ||
5c33677c AW |
2430 | if (le32_to_cpu(sb->compat_features) && |
2431 | le32_to_cpu(sb->compat_features) != FEATURE_FLAG_SUPPORTS_V190) { | |
9b6e5423 MS |
2432 | rs->ti->error = "Unable to assemble array: Unknown flag(s) in compatible feature flags"; |
2433 | return -EINVAL; | |
2434 | } | |
2435 | ||
2436 | if (sb->incompat_features) { | |
ecbfb9f1 | 2437 | rs->ti->error = "Unable to assemble array: No incompatible feature flags supported yet"; |
4c9971ca HM |
2438 | return -EINVAL; |
2439 | } | |
2440 | ||
0cf45031 | 2441 | /* Enable bitmap creation for RAID levels != 0 */ |
676fa5ad | 2442 | mddev->bitmap_info.offset = rt_is_raid0(rs->raid_type) ? 0 : to_sector(4096); |
977f1a0a | 2443 | mddev->bitmap_info.default_offset = mddev->bitmap_info.offset; |
0cf45031 | 2444 | |
33e53f06 | 2445 | if (!test_and_clear_bit(FirstUse, &rdev->flags)) { |
c4d097d1 HM |
2446 | /* |
2447 | * Retrieve rdev size stored in superblock to be prepared for shrink. | |
2448 | * Check extended superblock members are present otherwise the size | |
2449 | * will not be set! | |
2450 | */ | |
2451 | if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) | |
2452 | rdev->sectors = le64_to_cpu(sb->sectors); | |
2453 | ||
b12d437b | 2454 | rdev->recovery_offset = le64_to_cpu(sb->disk_recovery_offset); |
33e53f06 HM |
2455 | if (rdev->recovery_offset == MaxSector) |
2456 | set_bit(In_sync, &rdev->flags); | |
2457 | /* | |
2458 | * If no reshape in progress -> we're recovering single | |
2459 | * disk(s) and have to set the device(s) to out-of-sync | |
2460 | */ | |
9dbd1aa3 | 2461 | else if (!rs_is_reshaping(rs)) |
33e53f06 | 2462 | clear_bit(In_sync, &rdev->flags); /* Mandatory for recovery */ |
b12d437b JB |
2463 | } |
2464 | ||
2465 | /* | |
2466 | * If a device comes back, set it as not In_sync and no longer faulty. | |
2467 | */ | |
33e53f06 HM |
2468 | if (test_and_clear_bit(Faulty, &rdev->flags)) { |
2469 | rdev->recovery_offset = 0; | |
b12d437b JB |
2470 | clear_bit(In_sync, &rdev->flags); |
2471 | rdev->saved_raid_disk = rdev->raid_disk; | |
b12d437b JB |
2472 | } |
2473 | ||
33e53f06 HM |
2474 | /* Reshape support -> restore repective data offsets */ |
2475 | rdev->data_offset = le64_to_cpu(sb->data_offset); | |
2476 | rdev->new_data_offset = le64_to_cpu(sb->new_data_offset); | |
b12d437b JB |
2477 | |
2478 | return 0; | |
2479 | } | |
2480 | ||
2481 | /* | |
2482 | * Analyse superblocks and select the freshest. | |
2483 | */ | |
2484 | static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs) | |
2485 | { | |
73c6f239 | 2486 | int r; |
c63ede3b | 2487 | struct md_rdev *rdev, *freshest; |
fd01b88c | 2488 | struct mddev *mddev = &rs->md; |
b12d437b JB |
2489 | |
2490 | freshest = NULL; | |
c63ede3b | 2491 | rdev_for_each(rdev, mddev) { |
63c32ed4 HM |
2492 | if (test_bit(Journal, &rdev->flags)) |
2493 | continue; | |
2494 | ||
761becff | 2495 | /* |
c76d53f4 | 2496 | * Skipping super_load due to CTR_FLAG_SYNC will cause |
761becff | 2497 | * the array to undergo initialization again as |
43157840 | 2498 | * though it were new. This is the intended effect |
761becff JB |
2499 | * of the "sync" directive. |
2500 | * | |
c63ede3b HM |
2501 | * With reshaping capability added, we must ensure that |
2502 | * that the "sync" directive is disallowed during the reshape. | |
761becff | 2503 | */ |
4286325b | 2504 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) |
761becff JB |
2505 | continue; |
2506 | ||
b12d437b JB |
2507 | if (!rdev->meta_bdev) |
2508 | continue; | |
2509 | ||
73c6f239 | 2510 | r = super_load(rdev, freshest); |
b12d437b | 2511 | |
73c6f239 | 2512 | switch (r) { |
b12d437b JB |
2513 | case 1: |
2514 | freshest = rdev; | |
2515 | break; | |
2516 | case 0: | |
2517 | break; | |
2518 | default: | |
c63ede3b | 2519 | /* This is a failure to read the superblock from the metadata device. */ |
9e7d9367 HM |
2520 | /* |
2521 | * We have to keep any raid0 data/metadata device pairs or | |
2522 | * the MD raid0 personality will fail to start the array. | |
2523 | */ | |
2524 | if (rs_is_raid0(rs)) | |
2525 | continue; | |
2526 | ||
55ebbb59 | 2527 | /* |
c63ede3b HM |
2528 | * We keep the dm_devs to be able to emit the device tuple |
2529 | * properly on the table line in raid_status() (rather than | |
2530 | * mistakenly acting as if '- -' got passed into the constructor). | |
2531 | * | |
2532 | * The rdev has to stay on the same_set list to allow for | |
2533 | * the attempt to restore faulty devices on second resume. | |
55ebbb59 | 2534 | */ |
c63ede3b HM |
2535 | rdev->raid_disk = rdev->saved_raid_disk = -1; |
2536 | break; | |
b12d437b JB |
2537 | } |
2538 | } | |
2539 | ||
2540 | if (!freshest) | |
2541 | return 0; | |
2542 | ||
bd83a4c4 MS |
2543 | if (validate_raid_redundancy(rs)) { |
2544 | rs->ti->error = "Insufficient redundancy to activate array"; | |
2545 | return -EINVAL; | |
2546 | } | |
55ebbb59 | 2547 | |
b12d437b JB |
2548 | /* |
2549 | * Validation of the freshest device provides the source of | |
2550 | * validation for the remaining devices. | |
2551 | */ | |
9dbd1aa3 HM |
2552 | rs->ti->error = "Unable to assemble array: Invalid superblocks"; |
2553 | if (super_validate(rs, freshest)) | |
bd83a4c4 | 2554 | return -EINVAL; |
b12d437b | 2555 | |
dafb20fa | 2556 | rdev_for_each(rdev, mddev) |
63c32ed4 HM |
2557 | if (!test_bit(Journal, &rdev->flags) && |
2558 | rdev != freshest && | |
2559 | super_validate(rs, rdev)) | |
b12d437b | 2560 | return -EINVAL; |
b12d437b JB |
2561 | return 0; |
2562 | } | |
2563 | ||
40ba37e5 HM |
2564 | /* |
2565 | * Adjust data_offset and new_data_offset on all disk members of @rs | |
2566 | * for out of place reshaping if requested by contructor | |
2567 | * | |
2568 | * We need free space at the beginning of each raid disk for forward | |
2569 | * and at the end for backward reshapes which userspace has to provide | |
2570 | * via remapping/reordering of space. | |
2571 | */ | |
2572 | static int rs_adjust_data_offsets(struct raid_set *rs) | |
2573 | { | |
2574 | sector_t data_offset = 0, new_data_offset = 0; | |
2575 | struct md_rdev *rdev; | |
2576 | ||
2577 | /* Constructor did not request data offset change */ | |
2578 | if (!test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { | |
2579 | if (!rs_is_reshapable(rs)) | |
2580 | goto out; | |
2581 | ||
2582 | return 0; | |
2583 | } | |
2584 | ||
2585 | /* HM FIXME: get InSync raid_dev? */ | |
2586 | rdev = &rs->dev[0].rdev; | |
2587 | ||
2588 | if (rs->delta_disks < 0) { | |
2589 | /* | |
2590 | * Removing disks (reshaping backwards): | |
2591 | * | |
2592 | * - before reshape: data is at offset 0 and free space | |
2593 | * is at end of each component LV | |
2594 | * | |
2595 | * - after reshape: data is at offset rs->data_offset != 0 on each component LV | |
2596 | */ | |
2597 | data_offset = 0; | |
2598 | new_data_offset = rs->data_offset; | |
2599 | ||
2600 | } else if (rs->delta_disks > 0) { | |
2601 | /* | |
2602 | * Adding disks (reshaping forwards): | |
2603 | * | |
2604 | * - before reshape: data is at offset rs->data_offset != 0 and | |
2605 | * free space is at begin of each component LV | |
2606 | * | |
2607 | * - after reshape: data is at offset 0 on each component LV | |
2608 | */ | |
2609 | data_offset = rs->data_offset; | |
2610 | new_data_offset = 0; | |
2611 | ||
2612 | } else { | |
2613 | /* | |
2614 | * User space passes in 0 for data offset after having removed reshape space | |
2615 | * | |
2616 | * - or - (data offset != 0) | |
2617 | * | |
2618 | * Changing RAID layout or chunk size -> toggle offsets | |
2619 | * | |
2620 | * - before reshape: data is at offset rs->data_offset 0 and | |
2621 | * free space is at end of each component LV | |
2622 | * -or- | |
2623 | * data is at offset rs->data_offset != 0 and | |
2624 | * free space is at begin of each component LV | |
2625 | * | |
2527b56e HM |
2626 | * - after reshape: data is at offset 0 if it was at offset != 0 |
2627 | * or at offset != 0 if it was at offset 0 | |
40ba37e5 HM |
2628 | * on each component LV |
2629 | * | |
2630 | */ | |
2631 | data_offset = rs->data_offset ? rdev->data_offset : 0; | |
2632 | new_data_offset = data_offset ? 0 : rs->data_offset; | |
2633 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
2634 | } | |
2635 | ||
2636 | /* | |
2637 | * Make sure we got a minimum amount of free sectors per device | |
2638 | */ | |
2639 | if (rs->data_offset && | |
2640 | to_sector(i_size_read(rdev->bdev->bd_inode)) - rdev->sectors < MIN_FREE_RESHAPE_SPACE) { | |
2641 | rs->ti->error = data_offset ? "No space for forward reshape" : | |
2642 | "No space for backward reshape"; | |
2643 | return -ENOSPC; | |
2644 | } | |
2645 | out: | |
63c32ed4 | 2646 | /* Adjust data offsets on all rdevs but on any raid4/5/6 journal device */ |
40ba37e5 | 2647 | rdev_for_each(rdev, &rs->md) { |
63c32ed4 HM |
2648 | if (!test_bit(Journal, &rdev->flags)) { |
2649 | rdev->data_offset = data_offset; | |
2650 | rdev->new_data_offset = new_data_offset; | |
2651 | } | |
40ba37e5 HM |
2652 | } |
2653 | ||
2654 | return 0; | |
2655 | } | |
2656 | ||
ecbfb9f1 | 2657 | /* Userpace reordered disks -> adjust raid_disk indexes in @rs */ |
e6ca5e1a | 2658 | static void __reorder_raid_disk_indexes(struct raid_set *rs) |
ecbfb9f1 HM |
2659 | { |
2660 | int i = 0; | |
2661 | struct md_rdev *rdev; | |
2662 | ||
2663 | rdev_for_each(rdev, &rs->md) { | |
63c32ed4 HM |
2664 | if (!test_bit(Journal, &rdev->flags)) { |
2665 | rdev->raid_disk = i++; | |
2666 | rdev->saved_raid_disk = rdev->new_raid_disk = -1; | |
2667 | } | |
ecbfb9f1 HM |
2668 | } |
2669 | } | |
2670 | ||
2671 | /* | |
2672 | * Setup @rs for takeover by a different raid level | |
2673 | */ | |
2674 | static int rs_setup_takeover(struct raid_set *rs) | |
2675 | { | |
2676 | struct mddev *mddev = &rs->md; | |
2677 | struct md_rdev *rdev; | |
2678 | unsigned int d = mddev->raid_disks = rs->raid_disks; | |
2679 | sector_t new_data_offset = rs->dev[0].rdev.data_offset ? 0 : rs->data_offset; | |
2680 | ||
2681 | if (rt_is_raid10(rs->raid_type)) { | |
2682 | if (mddev->level == 0) { | |
2683 | /* Userpace reordered disks -> adjust raid_disk indexes */ | |
e6ca5e1a | 2684 | __reorder_raid_disk_indexes(rs); |
ecbfb9f1 HM |
2685 | |
2686 | /* raid0 -> raid10_far layout */ | |
2687 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_FAR, | |
2688 | rs->raid10_copies); | |
2689 | } else if (mddev->level == 1) | |
2690 | /* raid1 -> raid10_near layout */ | |
2691 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, | |
2692 | rs->raid_disks); | |
094f394d | 2693 | else |
ecbfb9f1 HM |
2694 | return -EINVAL; |
2695 | ||
2696 | } | |
2697 | ||
2698 | clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags); | |
2699 | mddev->recovery_cp = MaxSector; | |
2700 | ||
2701 | while (d--) { | |
2702 | rdev = &rs->dev[d].rdev; | |
2703 | ||
2704 | if (test_bit(d, (void *) rs->rebuild_disks)) { | |
2705 | clear_bit(In_sync, &rdev->flags); | |
2706 | clear_bit(Faulty, &rdev->flags); | |
2707 | mddev->recovery_cp = rdev->recovery_offset = 0; | |
2708 | /* Bitmap has to be created when we do an "up" takeover */ | |
2709 | set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); | |
2710 | } | |
2711 | ||
2712 | rdev->new_data_offset = new_data_offset; | |
2713 | } | |
2714 | ||
ecbfb9f1 HM |
2715 | return 0; |
2716 | } | |
2717 | ||
469b304b HM |
2718 | /* Prepare @rs for reshape */ |
2719 | static int rs_prepare_reshape(struct raid_set *rs) | |
2720 | { | |
2721 | bool reshape; | |
2722 | struct mddev *mddev = &rs->md; | |
2723 | ||
2724 | if (rs_is_raid10(rs)) { | |
2725 | if (rs->raid_disks != mddev->raid_disks && | |
2726 | __is_raid10_near(mddev->layout) && | |
2727 | rs->raid10_copies && | |
2728 | rs->raid10_copies != __raid10_near_copies(mddev->layout)) { | |
2729 | /* | |
2730 | * raid disk have to be multiple of data copies to allow this conversion, | |
2731 | * | |
2732 | * This is actually not a reshape it is a | |
2733 | * rebuild of any additional mirrors per group | |
2734 | */ | |
2735 | if (rs->raid_disks % rs->raid10_copies) { | |
2736 | rs->ti->error = "Can't reshape raid10 mirror groups"; | |
2737 | return -EINVAL; | |
2738 | } | |
2739 | ||
2740 | /* Userpace reordered disks to add/remove mirrors -> adjust raid_disk indexes */ | |
2741 | __reorder_raid_disk_indexes(rs); | |
2742 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, | |
2743 | rs->raid10_copies); | |
2744 | mddev->new_layout = mddev->layout; | |
2745 | reshape = false; | |
2746 | } else | |
2747 | reshape = true; | |
2748 | ||
2749 | } else if (rs_is_raid456(rs)) | |
2750 | reshape = true; | |
2751 | ||
469b304b | 2752 | else if (rs_is_raid1(rs)) { |
7a7c330f HM |
2753 | if (rs->delta_disks) { |
2754 | /* Process raid1 via delta_disks */ | |
2755 | mddev->degraded = rs->delta_disks < 0 ? -rs->delta_disks : rs->delta_disks; | |
2756 | reshape = true; | |
2757 | } else { | |
2758 | /* Process raid1 without delta_disks */ | |
2759 | mddev->raid_disks = rs->raid_disks; | |
7a7c330f HM |
2760 | reshape = false; |
2761 | } | |
469b304b HM |
2762 | } else { |
2763 | rs->ti->error = "Called with bogus raid type"; | |
2764 | return -EINVAL; | |
2765 | } | |
2766 | ||
2767 | if (reshape) { | |
2768 | set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags); | |
2769 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
7a7c330f HM |
2770 | } else if (mddev->raid_disks < rs->raid_disks) |
2771 | /* Create new superblocks and bitmaps, if any new disks */ | |
469b304b | 2772 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); |
469b304b HM |
2773 | |
2774 | return 0; | |
2775 | } | |
2776 | ||
9dbd1aa3 HM |
2777 | /* |
2778 | * | |
2779 | * - change raid layout | |
2780 | * - change chunk size | |
2781 | * - add disks | |
2782 | * - remove disks | |
2783 | */ | |
2784 | static int rs_setup_reshape(struct raid_set *rs) | |
2785 | { | |
2786 | int r = 0; | |
2787 | unsigned int cur_raid_devs, d; | |
2788 | struct mddev *mddev = &rs->md; | |
2789 | struct md_rdev *rdev; | |
2790 | ||
2791 | mddev->delta_disks = rs->delta_disks; | |
2792 | cur_raid_devs = mddev->raid_disks; | |
2793 | ||
2794 | /* Ignore impossible layout change whilst adding/removing disks */ | |
2795 | if (mddev->delta_disks && | |
2796 | mddev->layout != mddev->new_layout) { | |
2797 | DMINFO("Ignoring invalid layout change with delta_disks=%d", rs->delta_disks); | |
2798 | mddev->new_layout = mddev->layout; | |
2799 | } | |
2800 | ||
2801 | /* | |
2802 | * Adjust array size: | |
2803 | * | |
2804 | * - in case of adding disks, array size has | |
2805 | * to grow after the disk adding reshape, | |
2806 | * which'll hapen in the event handler; | |
2807 | * reshape will happen forward, so space has to | |
2808 | * be available at the beginning of each disk | |
2809 | * | |
2810 | * - in case of removing disks, array size | |
2811 | * has to shrink before starting the reshape, | |
2812 | * which'll happen here; | |
2813 | * reshape will happen backward, so space has to | |
2814 | * be available at the end of each disk | |
2815 | * | |
2816 | * - data_offset and new_data_offset are | |
ae3c6cff | 2817 | * adjusted for aforementioned out of place |
9dbd1aa3 HM |
2818 | * reshaping based on userspace passing in |
2819 | * the "data_offset <sectors>" key/value | |
ae3c6cff | 2820 | * pair via the constructor |
9dbd1aa3 HM |
2821 | */ |
2822 | ||
2823 | /* Add disk(s) */ | |
2824 | if (rs->delta_disks > 0) { | |
2825 | /* Prepare disks for check in raid4/5/6/10 {check|start}_reshape */ | |
2826 | for (d = cur_raid_devs; d < rs->raid_disks; d++) { | |
2827 | rdev = &rs->dev[d].rdev; | |
2828 | clear_bit(In_sync, &rdev->flags); | |
2829 | ||
2830 | /* | |
2831 | * save_raid_disk needs to be -1, or recovery_offset will be set to 0 | |
2832 | * by md, which'll store that erroneously in the superblock on reshape | |
2833 | */ | |
2834 | rdev->saved_raid_disk = -1; | |
2835 | rdev->raid_disk = d; | |
2836 | ||
2837 | rdev->sectors = mddev->dev_sectors; | |
7a7c330f | 2838 | rdev->recovery_offset = rs_is_raid1(rs) ? 0 : MaxSector; |
9dbd1aa3 HM |
2839 | } |
2840 | ||
2841 | mddev->reshape_backwards = 0; /* adding disks -> forward reshape */ | |
2842 | ||
2843 | /* Remove disk(s) */ | |
2844 | } else if (rs->delta_disks < 0) { | |
2845 | r = rs_set_dev_and_array_sectors(rs, true); | |
2846 | mddev->reshape_backwards = 1; /* removing disk(s) -> backward reshape */ | |
2847 | ||
2848 | /* Change layout and/or chunk size */ | |
2849 | } else { | |
2850 | /* | |
2851 | * Reshape layout (e.g. raid5_ls -> raid5_n) and/or chunk size: | |
2852 | * | |
2853 | * keeping number of disks and do layout change -> | |
2854 | * | |
2855 | * toggle reshape_backward depending on data_offset: | |
2856 | * | |
2857 | * - free space upfront -> reshape forward | |
2858 | * | |
2859 | * - free space at the end -> reshape backward | |
2860 | * | |
2861 | * | |
2862 | * This utilizes free reshape space avoiding the need | |
2863 | * for userspace to move (parts of) LV segments in | |
2864 | * case of layout/chunksize change (for disk | |
2865 | * adding/removing reshape space has to be at | |
2866 | * the proper address (see above with delta_disks): | |
2867 | * | |
2868 | * add disk(s) -> begin | |
2869 | * remove disk(s)-> end | |
2870 | */ | |
2871 | mddev->reshape_backwards = rs->dev[0].rdev.data_offset ? 0 : 1; | |
2872 | } | |
2873 | ||
2874 | return r; | |
2875 | } | |
2876 | ||
75b8e04b | 2877 | /* |
48cf06bc HM |
2878 | * Enable/disable discard support on RAID set depending on |
2879 | * RAID level and discard properties of underlying RAID members. | |
75b8e04b | 2880 | */ |
ecbfb9f1 | 2881 | static void configure_discard_support(struct raid_set *rs) |
75b8e04b | 2882 | { |
48cf06bc HM |
2883 | int i; |
2884 | bool raid456; | |
ecbfb9f1 | 2885 | struct dm_target *ti = rs->ti; |
48cf06bc | 2886 | |
75b8e04b HM |
2887 | /* Assume discards not supported until after checks below. */ |
2888 | ti->discards_supported = false; | |
2889 | ||
48920ff2 CH |
2890 | /* |
2891 | * XXX: RAID level 4,5,6 require zeroing for safety. | |
2892 | */ | |
48cf06bc | 2893 | raid456 = (rs->md.level == 4 || rs->md.level == 5 || rs->md.level == 6); |
75b8e04b | 2894 | |
ffeeac75 | 2895 | for (i = 0; i < rs->raid_disks; i++) { |
d20c4b08 | 2896 | struct request_queue *q; |
48cf06bc | 2897 | |
d20c4b08 HM |
2898 | if (!rs->dev[i].rdev.bdev) |
2899 | continue; | |
2900 | ||
2901 | q = bdev_get_queue(rs->dev[i].rdev.bdev); | |
48cf06bc HM |
2902 | if (!q || !blk_queue_discard(q)) |
2903 | return; | |
2904 | ||
2905 | if (raid456) { | |
48cf06bc HM |
2906 | if (!devices_handle_discard_safely) { |
2907 | DMERR("raid456 discard support disabled due to discard_zeroes_data uncertainty."); | |
2908 | DMERR("Set dm-raid.devices_handle_discard_safely=Y to override."); | |
2909 | return; | |
2910 | } | |
2911 | } | |
2912 | } | |
2913 | ||
2914 | /* All RAID members properly support discards */ | |
75b8e04b HM |
2915 | ti->discards_supported = true; |
2916 | ||
2917 | /* | |
2918 | * RAID1 and RAID10 personalities require bio splitting, | |
48cf06bc | 2919 | * RAID0/4/5/6 don't and process large discard bios properly. |
75b8e04b | 2920 | */ |
48cf06bc | 2921 | ti->split_discard_bios = !!(rs->md.level == 1 || rs->md.level == 10); |
75b8e04b HM |
2922 | ti->num_discard_bios = 1; |
2923 | } | |
2924 | ||
9d09e663 | 2925 | /* |
73c6f239 | 2926 | * Construct a RAID0/1/10/4/5/6 mapping: |
9d09e663 | 2927 | * Args: |
43157840 MS |
2928 | * <raid_type> <#raid_params> <raid_params>{0,} \ |
2929 | * <#raid_devs> [<meta_dev1> <dev1>]{1,} | |
9d09e663 | 2930 | * |
43157840 | 2931 | * <raid_params> varies by <raid_type>. See 'parse_raid_params' for |
9d09e663 | 2932 | * details on possible <raid_params>. |
73c6f239 HM |
2933 | * |
2934 | * Userspace is free to initialize the metadata devices, hence the superblocks to | |
2935 | * enforce recreation based on the passed in table parameters. | |
2936 | * | |
9d09e663 | 2937 | */ |
094f394d | 2938 | static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) |
9d09e663 | 2939 | { |
73c6f239 | 2940 | int r; |
469b304b | 2941 | bool resize; |
9d09e663 | 2942 | struct raid_type *rt; |
094f394d | 2943 | unsigned int num_raid_params, num_raid_devs; |
4d49f1b4 | 2944 | sector_t calculated_dev_sectors, rdev_sectors; |
9d09e663 | 2945 | struct raid_set *rs = NULL; |
92c83d79 | 2946 | const char *arg; |
9dbd1aa3 | 2947 | struct rs_layout rs_layout; |
92c83d79 HM |
2948 | struct dm_arg_set as = { argc, argv }, as_nrd; |
2949 | struct dm_arg _args[] = { | |
2950 | { 0, as.argc, "Cannot understand number of raid parameters" }, | |
2951 | { 1, 254, "Cannot understand number of raid devices parameters" } | |
2952 | }; | |
2953 | ||
2954 | /* Must have <raid_type> */ | |
2955 | arg = dm_shift_arg(&as); | |
bd83a4c4 MS |
2956 | if (!arg) { |
2957 | ti->error = "No arguments"; | |
2958 | return -EINVAL; | |
2959 | } | |
9d09e663 | 2960 | |
92c83d79 | 2961 | rt = get_raid_type(arg); |
bd83a4c4 MS |
2962 | if (!rt) { |
2963 | ti->error = "Unrecognised raid_type"; | |
2964 | return -EINVAL; | |
2965 | } | |
9d09e663 | 2966 | |
92c83d79 HM |
2967 | /* Must have <#raid_params> */ |
2968 | if (dm_read_arg_group(_args, &as, &num_raid_params, &ti->error)) | |
43157840 | 2969 | return -EINVAL; |
9d09e663 | 2970 | |
92c83d79 HM |
2971 | /* number of raid device tupples <meta_dev data_dev> */ |
2972 | as_nrd = as; | |
2973 | dm_consume_args(&as_nrd, num_raid_params); | |
2974 | _args[1].max = (as_nrd.argc - 1) / 2; | |
2975 | if (dm_read_arg(_args + 1, &as_nrd, &num_raid_devs, &ti->error)) | |
43157840 | 2976 | return -EINVAL; |
9d09e663 | 2977 | |
bb91a63f | 2978 | if (!__within_range(num_raid_devs, 1, MAX_RAID_DEVICES)) { |
bd83a4c4 MS |
2979 | ti->error = "Invalid number of supplied raid devices"; |
2980 | return -EINVAL; | |
2981 | } | |
3ca5a21a | 2982 | |
bfcee0e3 | 2983 | rs = raid_set_alloc(ti, rt, num_raid_devs); |
9d09e663 N |
2984 | if (IS_ERR(rs)) |
2985 | return PTR_ERR(rs); | |
2986 | ||
92c83d79 | 2987 | r = parse_raid_params(rs, &as, num_raid_params); |
73c6f239 | 2988 | if (r) |
9d09e663 N |
2989 | goto bad; |
2990 | ||
702108d1 | 2991 | r = parse_dev_params(rs, &as); |
73c6f239 | 2992 | if (r) |
9d09e663 N |
2993 | goto bad; |
2994 | ||
b12d437b | 2995 | rs->md.sync_super = super_sync; |
ecbfb9f1 | 2996 | |
2527b56e HM |
2997 | /* |
2998 | * Calculate ctr requested array and device sizes to allow | |
2999 | * for superblock analysis needing device sizes defined. | |
3000 | * | |
3001 | * Any existing superblock will overwrite the array and device sizes | |
3002 | */ | |
40ba37e5 HM |
3003 | r = rs_set_dev_and_array_sectors(rs, false); |
3004 | if (r) | |
b1956dc4 | 3005 | goto bad; |
40ba37e5 | 3006 | |
50c4feb9 | 3007 | calculated_dev_sectors = rs->md.dev_sectors; |
4dff2f1e | 3008 | |
ecbfb9f1 HM |
3009 | /* |
3010 | * Backup any new raid set level, layout, ... | |
3011 | * requested to be able to compare to superblock | |
3012 | * members for conversion decisions. | |
3013 | */ | |
9dbd1aa3 | 3014 | rs_config_backup(rs, &rs_layout); |
ecbfb9f1 | 3015 | |
73c6f239 HM |
3016 | r = analyse_superblocks(ti, rs); |
3017 | if (r) | |
b12d437b JB |
3018 | goto bad; |
3019 | ||
4d49f1b4 HM |
3020 | rdev_sectors = __rdev_sectors(rs); |
3021 | if (!rdev_sectors) { | |
3022 | ti->error = "Invalid rdev size"; | |
3023 | r = -EINVAL; | |
3024 | goto bad; | |
3025 | } | |
3026 | ||
3027 | resize = calculated_dev_sectors != rdev_sectors; | |
4dff2f1e | 3028 | |
9d09e663 | 3029 | INIT_WORK(&rs->md.event_work, do_table_event); |
9d09e663 | 3030 | ti->private = rs; |
55a62eef | 3031 | ti->num_flush_bios = 1; |
9d09e663 | 3032 | |
ecbfb9f1 | 3033 | /* Restore any requested new layout for conversion decision */ |
9dbd1aa3 | 3034 | rs_config_restore(rs, &rs_layout); |
ecbfb9f1 | 3035 | |
469b304b HM |
3036 | /* |
3037 | * Now that we have any superblock metadata available, | |
3038 | * check for new, recovering, reshaping, to be taken over, | |
3039 | * to be reshaped or an existing, unchanged raid set to | |
3040 | * run in sequence. | |
3041 | */ | |
9dbd1aa3 | 3042 | if (test_bit(MD_ARRAY_FIRST_USE, &rs->md.flags)) { |
2d92a3c2 HM |
3043 | /* A new raid6 set has to be recovered to ensure proper parity and Q-Syndrome */ |
3044 | if (rs_is_raid6(rs) && | |
3045 | test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { | |
3046 | ti->error = "'nosync' not allowed for new raid6 set"; | |
b1956dc4 HM |
3047 | r = -EINVAL; |
3048 | goto bad; | |
2d92a3c2 HM |
3049 | } |
3050 | rs_setup_recovery(rs, 0); | |
2a5556c2 HM |
3051 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); |
3052 | rs_set_new(rs); | |
3053 | } else if (rs_is_recovering(rs)) { | |
469b304b | 3054 | /* A recovering raid set may be resized */ |
2a5556c2 HM |
3055 | ; /* skip setup rs */ |
3056 | } else if (rs_is_reshaping(rs)) { | |
3057 | /* Have to reject size change request during reshape */ | |
3058 | if (resize) { | |
3059 | ti->error = "Can't resize a reshaping raid set"; | |
b1956dc4 HM |
3060 | r = -EPERM; |
3061 | goto bad; | |
2d92a3c2 | 3062 | } |
469b304b | 3063 | /* skip setup rs */ |
2d92a3c2 | 3064 | } else if (rs_takeover_requested(rs)) { |
9dbd1aa3 HM |
3065 | if (rs_is_reshaping(rs)) { |
3066 | ti->error = "Can't takeover a reshaping raid set"; | |
b1956dc4 HM |
3067 | r = -EPERM; |
3068 | goto bad; | |
9dbd1aa3 HM |
3069 | } |
3070 | ||
63c32ed4 HM |
3071 | /* We can't takeover a journaled raid4/5/6 */ |
3072 | if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { | |
3073 | ti->error = "Can't takeover a journaled raid4/5/6 set"; | |
3074 | r = -EPERM; | |
3075 | goto bad; | |
3076 | } | |
3077 | ||
9dbd1aa3 | 3078 | /* |
2527b56e | 3079 | * If a takeover is needed, userspace sets any additional |
469b304b HM |
3080 | * devices to rebuild and we can check for a valid request here. |
3081 | * | |
3082 | * If acceptible, set the level to the new requested | |
2a5556c2 HM |
3083 | * one, prohibit requesting recovery, allow the raid |
3084 | * set to run and store superblocks during resume. | |
9dbd1aa3 | 3085 | */ |
ecbfb9f1 HM |
3086 | r = rs_check_takeover(rs); |
3087 | if (r) | |
b1956dc4 | 3088 | goto bad; |
ecbfb9f1 HM |
3089 | |
3090 | r = rs_setup_takeover(rs); | |
3091 | if (r) | |
b1956dc4 | 3092 | goto bad; |
ecbfb9f1 | 3093 | |
4286325b | 3094 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); |
469b304b | 3095 | /* Takeover ain't recovery, so disable recovery */ |
2a5556c2 | 3096 | rs_setup_recovery(rs, MaxSector); |
3a1c1ef2 | 3097 | rs_set_new(rs); |
40ba37e5 | 3098 | } else if (rs_reshape_requested(rs)) { |
63c32ed4 HM |
3099 | /* |
3100 | * No need to check for 'ongoing' takeover here, because takeover | |
3101 | * is an instant operation as oposed to an ongoing reshape. | |
3102 | */ | |
3103 | ||
3104 | /* We can't reshape a journaled raid4/5/6 */ | |
3105 | if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) { | |
3106 | ti->error = "Can't reshape a journaled raid4/5/6 set"; | |
3107 | r = -EPERM; | |
3108 | goto bad; | |
3109 | } | |
3110 | ||
9dbd1aa3 | 3111 | /* |
469b304b HM |
3112 | * We can only prepare for a reshape here, because the |
3113 | * raid set needs to run to provide the repective reshape | |
3114 | * check functions via its MD personality instance. | |
3115 | * | |
3116 | * So do the reshape check after md_run() succeeded. | |
3117 | */ | |
3118 | r = rs_prepare_reshape(rs); | |
3119 | if (r) | |
3120 | return r; | |
9dbd1aa3 | 3121 | |
469b304b | 3122 | /* Reshaping ain't recovery, so disable recovery */ |
4dff2f1e | 3123 | rs_setup_recovery(rs, MaxSector); |
3a1c1ef2 | 3124 | rs_set_cur(rs); |
2a5556c2 HM |
3125 | } else { |
3126 | /* May not set recovery when a device rebuild is requested */ | |
37f10be1 HM |
3127 | if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) { |
3128 | rs_setup_recovery(rs, MaxSector); | |
3129 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
3130 | } else | |
3131 | rs_setup_recovery(rs, test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) ? | |
2a5556c2 HM |
3132 | 0 : (resize ? calculated_dev_sectors : MaxSector)); |
3133 | rs_set_cur(rs); | |
4dff2f1e | 3134 | } |
ecbfb9f1 | 3135 | |
40ba37e5 HM |
3136 | /* If constructor requested it, change data and new_data offsets */ |
3137 | r = rs_adjust_data_offsets(rs); | |
3138 | if (r) | |
b1956dc4 | 3139 | goto bad; |
40ba37e5 | 3140 | |
ecbfb9f1 HM |
3141 | /* Start raid set read-only and assumed clean to change in raid_resume() */ |
3142 | rs->md.ro = 1; | |
3143 | rs->md.in_sync = 1; | |
3144 | set_bit(MD_RECOVERY_FROZEN, &rs->md.recovery); | |
75b8e04b | 3145 | |
0cf45031 HM |
3146 | /* Has to be held on running the array */ |
3147 | mddev_lock_nointr(&rs->md); | |
73c6f239 | 3148 | r = md_run(&rs->md); |
9d09e663 | 3149 | rs->md.in_sync = 0; /* Assume already marked dirty */ |
9d09e663 | 3150 | |
73c6f239 | 3151 | if (r) { |
9dbd1aa3 HM |
3152 | ti->error = "Failed to run raid array"; |
3153 | mddev_unlock(&rs->md); | |
9d09e663 N |
3154 | goto bad; |
3155 | } | |
3156 | ||
3157 | rs->callbacks.congested_fn = raid_is_congested; | |
9d09e663 N |
3158 | dm_table_add_target_callbacks(ti->table, &rs->callbacks); |
3159 | ||
6e53636f HM |
3160 | /* If raid4/5/6 journal mode explictely requested (only possible with journal dev) -> set it */ |
3161 | if (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) { | |
3162 | r = r5c_journal_mode_set(&rs->md, rs->journal_dev.mode); | |
3163 | if (r) { | |
3164 | ti->error = "Failed to set raid4/5/6 journal mode"; | |
3165 | mddev_unlock(&rs->md); | |
3166 | goto bad_journal_mode_set; | |
3167 | } | |
3168 | } | |
3169 | ||
32737279 | 3170 | mddev_suspend(&rs->md); |
9dbd1aa3 HM |
3171 | |
3172 | /* Try to adjust the raid4/5/6 stripe cache size to the stripe size */ | |
3173 | if (rs_is_raid456(rs)) { | |
3174 | r = rs_set_raid456_stripe_cache(rs); | |
3175 | if (r) | |
3176 | goto bad_stripe_cache; | |
3177 | } | |
3178 | ||
3179 | /* Now do an early reshape check */ | |
3180 | if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { | |
3181 | r = rs_check_reshape(rs); | |
3182 | if (r) | |
b1956dc4 | 3183 | goto bad_check_reshape; |
9dbd1aa3 HM |
3184 | |
3185 | /* Restore new, ctr requested layout to perform check */ | |
3186 | rs_config_restore(rs, &rs_layout); | |
3187 | ||
7a7c330f HM |
3188 | if (rs->md.pers->start_reshape) { |
3189 | r = rs->md.pers->check_reshape(&rs->md); | |
3190 | if (r) { | |
3191 | ti->error = "Reshape check failed"; | |
3192 | goto bad_check_reshape; | |
3193 | } | |
9dbd1aa3 HM |
3194 | } |
3195 | } | |
3196 | ||
11e29684 HM |
3197 | /* Disable/enable discard support on raid set. */ |
3198 | configure_discard_support(rs); | |
3199 | ||
9dbd1aa3 | 3200 | mddev_unlock(&rs->md); |
9d09e663 N |
3201 | return 0; |
3202 | ||
6e53636f | 3203 | bad_journal_mode_set: |
9dbd1aa3 HM |
3204 | bad_stripe_cache: |
3205 | bad_check_reshape: | |
63f33b8d | 3206 | md_stop(&rs->md); |
9d09e663 | 3207 | bad: |
bfcee0e3 | 3208 | raid_set_free(rs); |
9d09e663 | 3209 | |
73c6f239 | 3210 | return r; |
9d09e663 N |
3211 | } |
3212 | ||
3213 | static void raid_dtr(struct dm_target *ti) | |
3214 | { | |
3215 | struct raid_set *rs = ti->private; | |
3216 | ||
3217 | list_del_init(&rs->callbacks.list); | |
3218 | md_stop(&rs->md); | |
bfcee0e3 | 3219 | raid_set_free(rs); |
9d09e663 N |
3220 | } |
3221 | ||
7de3ee57 | 3222 | static int raid_map(struct dm_target *ti, struct bio *bio) |
9d09e663 N |
3223 | { |
3224 | struct raid_set *rs = ti->private; | |
fd01b88c | 3225 | struct mddev *mddev = &rs->md; |
9d09e663 | 3226 | |
9dbd1aa3 HM |
3227 | /* |
3228 | * If we're reshaping to add disk(s)), ti->len and | |
3229 | * mddev->array_sectors will differ during the process | |
3230 | * (ti->len > mddev->array_sectors), so we have to requeue | |
3231 | * bios with addresses > mddev->array_sectors here or | |
2527b56e | 3232 | * there will occur accesses past EOD of the component |
9dbd1aa3 HM |
3233 | * data images thus erroring the raid set. |
3234 | */ | |
3235 | if (unlikely(bio_end_sector(bio) > mddev->array_sectors)) | |
3236 | return DM_MAPIO_REQUEUE; | |
3237 | ||
9d09e663 N |
3238 | mddev->pers->make_request(mddev, bio); |
3239 | ||
3240 | return DM_MAPIO_SUBMITTED; | |
3241 | } | |
3242 | ||
3a1c1ef2 | 3243 | /* Return string describing the current sync action of @mddev */ |
be83651f JB |
3244 | static const char *decipher_sync_action(struct mddev *mddev) |
3245 | { | |
3246 | if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) | |
3247 | return "frozen"; | |
3248 | ||
3249 | if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || | |
3250 | (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))) { | |
3251 | if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) | |
3252 | return "reshape"; | |
3253 | ||
3254 | if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { | |
3255 | if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) | |
3256 | return "resync"; | |
3257 | else if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) | |
3258 | return "check"; | |
3259 | return "repair"; | |
3260 | } | |
3261 | ||
3262 | if (test_bit(MD_RECOVERY_RECOVER, &mddev->recovery)) | |
3263 | return "recover"; | |
3264 | } | |
3265 | ||
3266 | return "idle"; | |
3267 | } | |
3268 | ||
3a1c1ef2 | 3269 | /* |
63c32ed4 | 3270 | * Return status string for @rdev |
3a1c1ef2 HM |
3271 | * |
3272 | * Status characters: | |
3273 | * | |
63c32ed4 | 3274 | * 'D' = Dead/Failed raid set component or raid4/5/6 journal device |
6e53636f HM |
3275 | * 'a' = Alive but not in-sync raid set component _or_ alive raid4/5/6 'write_back' journal device |
3276 | * 'A' = Alive and in-sync raid set component _or_ alive raid4/5/6 'write_through' journal device | |
c63ede3b | 3277 | * '-' = Non-existing device (i.e. uspace passed '- -' into the ctr) |
3a1c1ef2 | 3278 | */ |
6e53636f | 3279 | static const char *__raid_dev_status(struct raid_set *rs, struct md_rdev *rdev, bool array_in_sync) |
9d09e663 | 3280 | { |
c63ede3b HM |
3281 | if (!rdev->bdev) |
3282 | return "-"; | |
3283 | else if (test_bit(Faulty, &rdev->flags)) | |
3a1c1ef2 | 3284 | return "D"; |
63c32ed4 | 3285 | else if (test_bit(Journal, &rdev->flags)) |
6e53636f | 3286 | return (rs->journal_dev.mode == R5C_JOURNAL_MODE_WRITE_THROUGH) ? "A" : "a"; |
3a1c1ef2 HM |
3287 | else if (!array_in_sync || !test_bit(In_sync, &rdev->flags)) |
3288 | return "a"; | |
3289 | else | |
3290 | return "A"; | |
3291 | } | |
9d09e663 | 3292 | |
3a1c1ef2 HM |
3293 | /* Helper to return resync/reshape progress for @rs and @array_in_sync */ |
3294 | static sector_t rs_get_progress(struct raid_set *rs, | |
3295 | sector_t resync_max_sectors, bool *array_in_sync) | |
3296 | { | |
3297 | sector_t r, recovery_cp, curr_resync_completed; | |
3298 | struct mddev *mddev = &rs->md; | |
9d09e663 | 3299 | |
3a1c1ef2 HM |
3300 | curr_resync_completed = mddev->curr_resync_completed ?: mddev->recovery_cp; |
3301 | recovery_cp = mddev->recovery_cp; | |
3302 | *array_in_sync = false; | |
3303 | ||
3304 | if (rs_is_raid0(rs)) { | |
3305 | r = resync_max_sectors; | |
3306 | *array_in_sync = true; | |
3307 | ||
3308 | } else { | |
3309 | r = mddev->reshape_position; | |
3310 | ||
3311 | /* Reshape is relative to the array size */ | |
3312 | if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) || | |
3313 | r != MaxSector) { | |
3314 | if (r == MaxSector) { | |
3315 | *array_in_sync = true; | |
3316 | r = resync_max_sectors; | |
0cf45031 | 3317 | } else { |
3a1c1ef2 HM |
3318 | /* Got to reverse on backward reshape */ |
3319 | if (mddev->reshape_backwards) | |
3320 | r = mddev->array_sectors - r; | |
3321 | ||
3322 | /* Devide by # of data stripes */ | |
3323 | sector_div(r, mddev_data_stripes(rs)); | |
0cf45031 | 3324 | } |
3a1c1ef2 HM |
3325 | |
3326 | /* Sync is relative to the component device size */ | |
3327 | } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) | |
3328 | r = curr_resync_completed; | |
3329 | else | |
3330 | r = recovery_cp; | |
3331 | ||
3332 | if (r == MaxSector) { | |
3333 | /* | |
3334 | * Sync complete. | |
3335 | */ | |
3336 | *array_in_sync = true; | |
3337 | r = resync_max_sectors; | |
3338 | } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) { | |
3339 | /* | |
3340 | * If "check" or "repair" is occurring, the raid set has | |
3341 | * undergone an initial sync and the health characters | |
3342 | * should not be 'a' anymore. | |
3343 | */ | |
3344 | *array_in_sync = true; | |
0cf45031 | 3345 | } else { |
3a1c1ef2 | 3346 | struct md_rdev *rdev; |
be83651f | 3347 | |
3a1c1ef2 HM |
3348 | /* |
3349 | * The raid set may be doing an initial sync, or it may | |
43157840 | 3350 | * be rebuilding individual components. If all the |
3a1c1ef2 HM |
3351 | * devices are In_sync, then it is the raid set that is |
3352 | * being initialized. | |
3353 | */ | |
3354 | rdev_for_each(rdev, mddev) | |
63c32ed4 HM |
3355 | if (!test_bit(Journal, &rdev->flags) && |
3356 | !test_bit(In_sync, &rdev->flags)) | |
3a1c1ef2 HM |
3357 | *array_in_sync = true; |
3358 | #if 0 | |
3359 | r = 0; /* HM FIXME: TESTME: https://bugzilla.redhat.com/show_bug.cgi?id=1210637 ? */ | |
3360 | #endif | |
2e727c3c | 3361 | } |
3a1c1ef2 HM |
3362 | } |
3363 | ||
3364 | return r; | |
3365 | } | |
3366 | ||
3367 | /* Helper to return @dev name or "-" if !@dev */ | |
e6ca5e1a | 3368 | static const char *__get_dev_name(struct dm_dev *dev) |
3a1c1ef2 HM |
3369 | { |
3370 | return dev ? dev->name : "-"; | |
3371 | } | |
3372 | ||
3373 | static void raid_status(struct dm_target *ti, status_type_t type, | |
3374 | unsigned int status_flags, char *result, unsigned int maxlen) | |
3375 | { | |
3376 | struct raid_set *rs = ti->private; | |
3377 | struct mddev *mddev = &rs->md; | |
3378 | struct r5conf *conf = mddev->private; | |
7a7c330f | 3379 | int i, max_nr_stripes = conf ? conf->max_nr_stripes : 0; |
3a1c1ef2 HM |
3380 | bool array_in_sync; |
3381 | unsigned int raid_param_cnt = 1; /* at least 1 for chunksize */ | |
3382 | unsigned int sz = 0; | |
7a7c330f | 3383 | unsigned int rebuild_disks; |
3a1c1ef2 HM |
3384 | unsigned int write_mostly_params = 0; |
3385 | sector_t progress, resync_max_sectors, resync_mismatches; | |
3386 | const char *sync_action; | |
3387 | struct raid_type *rt; | |
3a1c1ef2 HM |
3388 | |
3389 | switch (type) { | |
3390 | case STATUSTYPE_INFO: | |
3391 | /* *Should* always succeed */ | |
3392 | rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); | |
3393 | if (!rt) | |
3394 | return; | |
3395 | ||
9dbd1aa3 | 3396 | DMEMIT("%s %d ", rt->name, mddev->raid_disks); |
3a1c1ef2 HM |
3397 | |
3398 | /* Access most recent mddev properties for status output */ | |
3399 | smp_rmb(); | |
3400 | /* Get sensible max sectors even if raid set not yet started */ | |
4286325b | 3401 | resync_max_sectors = test_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags) ? |
3a1c1ef2 HM |
3402 | mddev->resync_max_sectors : mddev->dev_sectors; |
3403 | progress = rs_get_progress(rs, resync_max_sectors, &array_in_sync); | |
3404 | resync_mismatches = (mddev->last_sync_action && !strcasecmp(mddev->last_sync_action, "check")) ? | |
9dbd1aa3 | 3405 | atomic64_read(&mddev->resync_mismatches) : 0; |
3a1c1ef2 HM |
3406 | sync_action = decipher_sync_action(&rs->md); |
3407 | ||
c63ede3b HM |
3408 | /* HM FIXME: do we want another state char for raid0? It shows 'D'/'A'/'-' now */ |
3409 | for (i = 0; i < rs->raid_disks; i++) | |
6e53636f | 3410 | DMEMIT(__raid_dev_status(rs, &rs->dev[i].rdev, array_in_sync)); |
9d09e663 | 3411 | |
2e727c3c | 3412 | /* |
3a1c1ef2 | 3413 | * In-sync/Reshape ratio: |
2e727c3c | 3414 | * The in-sync ratio shows the progress of: |
3a1c1ef2 HM |
3415 | * - Initializing the raid set |
3416 | * - Rebuilding a subset of devices of the raid set | |
2e727c3c JB |
3417 | * The user can distinguish between the two by referring |
3418 | * to the status characters. | |
3a1c1ef2 HM |
3419 | * |
3420 | * The reshape ratio shows the progress of | |
3421 | * changing the raid layout or the number of | |
3422 | * disks of a raid set | |
2e727c3c | 3423 | */ |
3a1c1ef2 HM |
3424 | DMEMIT(" %llu/%llu", (unsigned long long) progress, |
3425 | (unsigned long long) resync_max_sectors); | |
9d09e663 | 3426 | |
be83651f | 3427 | /* |
3a1c1ef2 HM |
3428 | * v1.5.0+: |
3429 | * | |
be83651f | 3430 | * Sync action: |
3a1c1ef2 | 3431 | * See Documentation/device-mapper/dm-raid.txt for |
be83651f JB |
3432 | * information on each of these states. |
3433 | */ | |
3a1c1ef2 | 3434 | DMEMIT(" %s", sync_action); |
be83651f JB |
3435 | |
3436 | /* | |
3a1c1ef2 HM |
3437 | * v1.5.0+: |
3438 | * | |
be83651f JB |
3439 | * resync_mismatches/mismatch_cnt |
3440 | * This field shows the number of discrepancies found when | |
3a1c1ef2 | 3441 | * performing a "check" of the raid set. |
be83651f | 3442 | */ |
3a1c1ef2 | 3443 | DMEMIT(" %llu", (unsigned long long) resync_mismatches); |
9d09e663 | 3444 | |
3a1c1ef2 | 3445 | /* |
9b6e5423 | 3446 | * v1.9.0+: |
3a1c1ef2 HM |
3447 | * |
3448 | * data_offset (needed for out of space reshaping) | |
3449 | * This field shows the data offset into the data | |
3450 | * image LV where the first stripes data starts. | |
3451 | * | |
3452 | * We keep data_offset equal on all raid disks of the set, | |
3453 | * so retrieving it from the first raid disk is sufficient. | |
3454 | */ | |
3455 | DMEMIT(" %llu", (unsigned long long) rs->dev[0].rdev.data_offset); | |
63c32ed4 HM |
3456 | |
3457 | /* | |
3458 | * v1.10.0+: | |
3459 | */ | |
3460 | DMEMIT(" %s", test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags) ? | |
6e53636f | 3461 | __raid_dev_status(rs, &rs->journal_dev.rdev, 0) : "-"); |
3a1c1ef2 | 3462 | break; |
9d09e663 | 3463 | |
3a1c1ef2 HM |
3464 | case STATUSTYPE_TABLE: |
3465 | /* Report the table line string you would use to construct this raid set */ | |
3466 | ||
3467 | /* Calculate raid parameter count */ | |
7a7c330f HM |
3468 | for (i = 0; i < rs->raid_disks; i++) |
3469 | if (test_bit(WriteMostly, &rs->dev[i].rdev.flags)) | |
3a1c1ef2 | 3470 | write_mostly_params += 2; |
7a7c330f HM |
3471 | rebuild_disks = memweight(rs->rebuild_disks, DISKS_ARRAY_ELEMS * sizeof(*rs->rebuild_disks)); |
3472 | raid_param_cnt += rebuild_disks * 2 + | |
3a1c1ef2 HM |
3473 | write_mostly_params + |
3474 | hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_NO_ARGS) + | |
63c32ed4 | 3475 | hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2 + |
6e53636f HM |
3476 | (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags) ? 2 : 0) + |
3477 | (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags) ? 2 : 0); | |
4464e36e | 3478 | |
3a1c1ef2 | 3479 | /* Emit table line */ |
4464e36e | 3480 | /* This has to be in the documented order for userspace! */ |
3a1c1ef2 | 3481 | DMEMIT("%s %u %u", rs->raid_type->name, raid_param_cnt, mddev->new_chunk_sectors); |
4286325b | 3482 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) |
3fa6cf38 | 3483 | DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_SYNC)); |
4464e36e HM |
3484 | if (test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) |
3485 | DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC)); | |
7a7c330f HM |
3486 | if (rebuild_disks) |
3487 | for (i = 0; i < rs->raid_disks; i++) | |
3488 | if (test_bit(rs->dev[i].rdev.raid_disk, (void *) rs->rebuild_disks)) | |
3489 | DMEMIT(" %s %u", dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD), | |
3490 | rs->dev[i].rdev.raid_disk); | |
4464e36e HM |
3491 | if (test_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) |
3492 | DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP), | |
3493 | mddev->bitmap_info.daemon_sleep); | |
3494 | if (test_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) | |
3495 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE), | |
3496 | mddev->sync_speed_min); | |
3497 | if (test_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) | |
3498 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE), | |
3499 | mddev->sync_speed_max); | |
7a7c330f HM |
3500 | if (write_mostly_params) |
3501 | for (i = 0; i < rs->raid_disks; i++) | |
3502 | if (test_bit(WriteMostly, &rs->dev[i].rdev.flags)) | |
3503 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY), | |
3504 | rs->dev[i].rdev.raid_disk); | |
4286325b | 3505 | if (test_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) |
3fa6cf38 | 3506 | DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND), |
3a1c1ef2 | 3507 | mddev->bitmap_info.max_write_behind); |
4464e36e HM |
3508 | if (test_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) |
3509 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE), | |
3510 | max_nr_stripes); | |
3511 | if (test_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) | |
3512 | DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE), | |
3513 | (unsigned long long) to_sector(mddev->bitmap_info.chunksize)); | |
3514 | if (test_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) | |
3515 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES), | |
3516 | raid10_md_layout_to_copies(mddev->layout)); | |
3517 | if (test_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) | |
3518 | DMEMIT(" %s %s", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT), | |
3519 | raid10_md_layout_to_format(mddev->layout)); | |
3520 | if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) | |
3521 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS), | |
3522 | max(rs->delta_disks, mddev->delta_disks)); | |
3523 | if (test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) | |
3524 | DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET), | |
3525 | (unsigned long long) rs->data_offset); | |
63c32ed4 HM |
3526 | if (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags)) |
3527 | DMEMIT(" %s %s", dm_raid_arg_name_by_flag(CTR_FLAG_JOURNAL_DEV), | |
3528 | __get_dev_name(rs->journal_dev.dev)); | |
6e53636f HM |
3529 | if (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags)) |
3530 | DMEMIT(" %s %s", dm_raid_arg_name_by_flag(CTR_FLAG_JOURNAL_MODE), | |
3531 | md_journal_mode_to_dm_raid(rs->journal_dev.mode)); | |
3a1c1ef2 | 3532 | DMEMIT(" %d", rs->raid_disks); |
7a7c330f HM |
3533 | for (i = 0; i < rs->raid_disks; i++) |
3534 | DMEMIT(" %s %s", __get_dev_name(rs->dev[i].meta_dev), | |
3535 | __get_dev_name(rs->dev[i].data_dev)); | |
9d09e663 | 3536 | } |
9d09e663 N |
3537 | } |
3538 | ||
094f394d | 3539 | static int raid_message(struct dm_target *ti, unsigned int argc, char **argv) |
be83651f JB |
3540 | { |
3541 | struct raid_set *rs = ti->private; | |
3542 | struct mddev *mddev = &rs->md; | |
3543 | ||
be83651f JB |
3544 | if (!mddev->pers || !mddev->pers->sync_request) |
3545 | return -EINVAL; | |
3546 | ||
3547 | if (!strcasecmp(argv[0], "frozen")) | |
3548 | set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3549 | else | |
3550 | clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3551 | ||
3552 | if (!strcasecmp(argv[0], "idle") || !strcasecmp(argv[0], "frozen")) { | |
3553 | if (mddev->sync_thread) { | |
3554 | set_bit(MD_RECOVERY_INTR, &mddev->recovery); | |
3555 | md_reap_sync_thread(mddev); | |
3556 | } | |
3557 | } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || | |
3558 | test_bit(MD_RECOVERY_NEEDED, &mddev->recovery)) | |
3559 | return -EBUSY; | |
3560 | else if (!strcasecmp(argv[0], "resync")) | |
3a1c1ef2 HM |
3561 | ; /* MD_RECOVERY_NEEDED set below */ |
3562 | else if (!strcasecmp(argv[0], "recover")) | |
be83651f | 3563 | set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); |
3a1c1ef2 | 3564 | else { |
ad470472 | 3565 | if (!strcasecmp(argv[0], "check")) { |
be83651f | 3566 | set_bit(MD_RECOVERY_CHECK, &mddev->recovery); |
ad470472 MS |
3567 | set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); |
3568 | set_bit(MD_RECOVERY_SYNC, &mddev->recovery); | |
3569 | } else if (!strcasecmp(argv[0], "repair")) { | |
105db599 MS |
3570 | set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); |
3571 | set_bit(MD_RECOVERY_SYNC, &mddev->recovery); | |
3572 | } else | |
be83651f | 3573 | return -EINVAL; |
be83651f JB |
3574 | } |
3575 | if (mddev->ro == 2) { | |
3576 | /* A write to sync_action is enough to justify | |
3577 | * canceling read-auto mode | |
3578 | */ | |
3579 | mddev->ro = 0; | |
3a1c1ef2 | 3580 | if (!mddev->suspended && mddev->sync_thread) |
be83651f JB |
3581 | md_wakeup_thread(mddev->sync_thread); |
3582 | } | |
3583 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
3a1c1ef2 | 3584 | if (!mddev->suspended && mddev->thread) |
be83651f JB |
3585 | md_wakeup_thread(mddev->thread); |
3586 | ||
3587 | return 0; | |
3588 | } | |
3589 | ||
3590 | static int raid_iterate_devices(struct dm_target *ti, | |
3591 | iterate_devices_callout_fn fn, void *data) | |
9d09e663 N |
3592 | { |
3593 | struct raid_set *rs = ti->private; | |
094f394d | 3594 | unsigned int i; |
73c6f239 | 3595 | int r = 0; |
9d09e663 | 3596 | |
73c6f239 | 3597 | for (i = 0; !r && i < rs->md.raid_disks; i++) |
9d09e663 | 3598 | if (rs->dev[i].data_dev) |
73c6f239 | 3599 | r = fn(ti, |
9d09e663 N |
3600 | rs->dev[i].data_dev, |
3601 | 0, /* No offset on data devs */ | |
3602 | rs->md.dev_sectors, | |
3603 | data); | |
3604 | ||
73c6f239 | 3605 | return r; |
9d09e663 N |
3606 | } |
3607 | ||
3608 | static void raid_io_hints(struct dm_target *ti, struct queue_limits *limits) | |
3609 | { | |
3610 | struct raid_set *rs = ti->private; | |
89d3d9a1 | 3611 | unsigned int chunk_size = to_bytes(rs->md.chunk_sectors); |
9d09e663 N |
3612 | |
3613 | blk_limits_io_min(limits, chunk_size); | |
89d3d9a1 | 3614 | blk_limits_io_opt(limits, chunk_size * mddev_data_stripes(rs)); |
9d09e663 N |
3615 | } |
3616 | ||
3617 | static void raid_presuspend(struct dm_target *ti) | |
3618 | { | |
3619 | struct raid_set *rs = ti->private; | |
3620 | ||
3621 | md_stop_writes(&rs->md); | |
3622 | } | |
3623 | ||
3624 | static void raid_postsuspend(struct dm_target *ti) | |
3625 | { | |
3626 | struct raid_set *rs = ti->private; | |
3627 | ||
31e10a41 HM |
3628 | if (!rs->md.suspended) |
3629 | mddev_suspend(&rs->md); | |
3630 | ||
3631 | rs->md.ro = 1; | |
9d09e663 N |
3632 | } |
3633 | ||
f381e71b | 3634 | static void attempt_restore_of_faulty_devices(struct raid_set *rs) |
9d09e663 | 3635 | { |
9092c02d | 3636 | int i; |
a3c06a38 | 3637 | uint64_t cleared_failed_devices[DISKS_ARRAY_ELEMS]; |
9092c02d | 3638 | unsigned long flags; |
a3c06a38 | 3639 | bool cleared = false; |
9092c02d | 3640 | struct dm_raid_superblock *sb; |
a3c06a38 | 3641 | struct mddev *mddev = &rs->md; |
9092c02d | 3642 | struct md_rdev *r; |
9d09e663 | 3643 | |
a3c06a38 HM |
3644 | /* RAID personalities have to provide hot add/remove methods or we need to bail out. */ |
3645 | if (!mddev->pers || !mddev->pers->hot_add_disk || !mddev->pers->hot_remove_disk) | |
3646 | return; | |
3647 | ||
3648 | memset(cleared_failed_devices, 0, sizeof(cleared_failed_devices)); | |
3649 | ||
c63ede3b | 3650 | for (i = 0; i < mddev->raid_disks; i++) { |
f381e71b | 3651 | r = &rs->dev[i].rdev; |
63c32ed4 HM |
3652 | /* HM FIXME: enhance journal device recovery processing */ |
3653 | if (test_bit(Journal, &r->flags)) | |
3654 | continue; | |
3655 | ||
e2568465 HM |
3656 | if (test_bit(Faulty, &r->flags) && |
3657 | r->meta_bdev && !read_disk_sb(r, r->sb_size, true)) { | |
f381e71b JB |
3658 | DMINFO("Faulty %s device #%d has readable super block." |
3659 | " Attempting to revive it.", | |
3660 | rs->raid_type->name, i); | |
a4dc163a JB |
3661 | |
3662 | /* | |
3663 | * Faulty bit may be set, but sometimes the array can | |
3664 | * be suspended before the personalities can respond | |
3665 | * by removing the device from the array (i.e. calling | |
43157840 | 3666 | * 'hot_remove_disk'). If they haven't yet removed |
a4dc163a JB |
3667 | * the failed device, its 'raid_disk' number will be |
3668 | * '>= 0' - meaning we must call this function | |
3669 | * ourselves. | |
3670 | */ | |
f381e71b | 3671 | flags = r->flags; |
c63ede3b HM |
3672 | clear_bit(In_sync, &r->flags); /* Mandatory for hot remove. */ |
3673 | if (r->raid_disk >= 0) { | |
3674 | if (mddev->pers->hot_remove_disk(mddev, r)) { | |
3675 | /* Failed to revive this device, try next */ | |
3676 | r->flags = flags; | |
3677 | continue; | |
3678 | } | |
3679 | } else | |
3680 | r->raid_disk = r->saved_raid_disk = i; | |
3681 | ||
f381e71b JB |
3682 | clear_bit(Faulty, &r->flags); |
3683 | clear_bit(WriteErrorSeen, &r->flags); | |
c63ede3b | 3684 | |
a3c06a38 | 3685 | if (mddev->pers->hot_add_disk(mddev, r)) { |
c63ede3b HM |
3686 | /* Failed to revive this device, try next */ |
3687 | r->raid_disk = r->saved_raid_disk = -1; | |
f381e71b JB |
3688 | r->flags = flags; |
3689 | } else { | |
c63ede3b | 3690 | clear_bit(In_sync, &r->flags); |
f381e71b | 3691 | r->recovery_offset = 0; |
a3c06a38 HM |
3692 | set_bit(i, (void *) cleared_failed_devices); |
3693 | cleared = true; | |
f381e71b JB |
3694 | } |
3695 | } | |
3696 | } | |
a3c06a38 HM |
3697 | |
3698 | /* If any failed devices could be cleared, update all sbs failed_devices bits */ | |
3699 | if (cleared) { | |
3700 | uint64_t failed_devices[DISKS_ARRAY_ELEMS]; | |
3701 | ||
f381e71b | 3702 | rdev_for_each(r, &rs->md) { |
63c32ed4 HM |
3703 | if (test_bit(Journal, &r->flags)) |
3704 | continue; | |
3705 | ||
f381e71b | 3706 | sb = page_address(r->sb_page); |
a3c06a38 HM |
3707 | sb_retrieve_failed_devices(sb, failed_devices); |
3708 | ||
3709 | for (i = 0; i < DISKS_ARRAY_ELEMS; i++) | |
3710 | failed_devices[i] &= ~cleared_failed_devices[i]; | |
3711 | ||
3712 | sb_update_failed_devices(sb, failed_devices); | |
f381e71b JB |
3713 | } |
3714 | } | |
3715 | } | |
3716 | ||
e6ca5e1a | 3717 | static int __load_dirty_region_bitmap(struct raid_set *rs) |
ecbfb9f1 HM |
3718 | { |
3719 | int r = 0; | |
3720 | ||
3721 | /* Try loading the bitmap unless "raid0", which does not have one */ | |
3722 | if (!rs_is_raid0(rs) && | |
4286325b | 3723 | !test_and_set_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags)) { |
ecbfb9f1 HM |
3724 | r = bitmap_load(&rs->md); |
3725 | if (r) | |
3726 | DMERR("Failed to load bitmap"); | |
3727 | } | |
3728 | ||
3729 | return r; | |
3730 | } | |
3731 | ||
6e20902e HM |
3732 | /* Enforce updating all superblocks */ |
3733 | static void rs_update_sbs(struct raid_set *rs) | |
3734 | { | |
3735 | struct mddev *mddev = &rs->md; | |
3736 | int ro = mddev->ro; | |
3737 | ||
2953079c | 3738 | set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags); |
6e20902e HM |
3739 | mddev->ro = 0; |
3740 | md_update_sb(mddev, 1); | |
3741 | mddev->ro = ro; | |
3742 | } | |
3743 | ||
9dbd1aa3 HM |
3744 | /* |
3745 | * Reshape changes raid algorithm of @rs to new one within personality | |
3746 | * (e.g. raid6_zr -> raid6_nc), changes stripe size, adds/removes | |
3747 | * disks from a raid set thus growing/shrinking it or resizes the set | |
3748 | * | |
3749 | * Call mddev_lock_nointr() before! | |
3750 | */ | |
3751 | static int rs_start_reshape(struct raid_set *rs) | |
3752 | { | |
3753 | int r; | |
3754 | struct mddev *mddev = &rs->md; | |
3755 | struct md_personality *pers = mddev->pers; | |
3756 | ||
3757 | r = rs_setup_reshape(rs); | |
3758 | if (r) | |
3759 | return r; | |
3760 | ||
3761 | /* Need to be resumed to be able to start reshape, recovery is frozen until raid_resume() though */ | |
3762 | if (mddev->suspended) | |
3763 | mddev_resume(mddev); | |
3764 | ||
3765 | /* | |
3766 | * Check any reshape constraints enforced by the personalility | |
3767 | * | |
3768 | * May as well already kick the reshape off so that * pers->start_reshape() becomes optional. | |
3769 | */ | |
3770 | r = pers->check_reshape(mddev); | |
3771 | if (r) { | |
3772 | rs->ti->error = "pers->check_reshape() failed"; | |
3773 | return r; | |
3774 | } | |
3775 | ||
3776 | /* | |
3777 | * Personality may not provide start reshape method in which | |
3778 | * case check_reshape above has already covered everything | |
3779 | */ | |
3780 | if (pers->start_reshape) { | |
3781 | r = pers->start_reshape(mddev); | |
3782 | if (r) { | |
3783 | rs->ti->error = "pers->start_reshape() failed"; | |
3784 | return r; | |
3785 | } | |
3786 | } | |
3787 | ||
3788 | /* Suspend because a resume will happen in raid_resume() */ | |
3789 | if (!mddev->suspended) | |
3790 | mddev_suspend(mddev); | |
3791 | ||
6e20902e HM |
3792 | /* |
3793 | * Now reshape got set up, update superblocks to | |
3794 | * reflect the fact so that a table reload will | |
3795 | * access proper superblock content in the ctr. | |
3796 | */ | |
3797 | rs_update_sbs(rs); | |
9dbd1aa3 HM |
3798 | |
3799 | return 0; | |
3800 | } | |
3801 | ||
ecbfb9f1 HM |
3802 | static int raid_preresume(struct dm_target *ti) |
3803 | { | |
9dbd1aa3 | 3804 | int r; |
ecbfb9f1 HM |
3805 | struct raid_set *rs = ti->private; |
3806 | struct mddev *mddev = &rs->md; | |
3807 | ||
3808 | /* This is a resume after a suspend of the set -> it's already started */ | |
4286325b | 3809 | if (test_and_set_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags)) |
ecbfb9f1 HM |
3810 | return 0; |
3811 | ||
3812 | /* | |
3813 | * The superblocks need to be updated on disk if the | |
6e20902e HM |
3814 | * array is new or new devices got added (thus zeroed |
3815 | * out by userspace) or __load_dirty_region_bitmap | |
3816 | * will overwrite them in core with old data or fail. | |
ecbfb9f1 | 3817 | */ |
6e20902e HM |
3818 | if (test_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags)) |
3819 | rs_update_sbs(rs); | |
ecbfb9f1 | 3820 | |
ecbfb9f1 | 3821 | /* Load the bitmap from disk unless raid0 */ |
9dbd1aa3 HM |
3822 | r = __load_dirty_region_bitmap(rs); |
3823 | if (r) | |
3824 | return r; | |
3825 | ||
4257e085 | 3826 | /* Resize bitmap to adjust to changed region size (aka MD bitmap chunksize) */ |
7a0c5c5b | 3827 | if (test_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags) && mddev->bitmap && |
4257e085 HM |
3828 | mddev->bitmap_info.chunksize != to_bytes(rs->requested_bitmap_chunk_sectors)) { |
3829 | r = bitmap_resize(mddev->bitmap, mddev->dev_sectors, | |
3830 | to_bytes(rs->requested_bitmap_chunk_sectors), 0); | |
3831 | if (r) | |
3832 | DMERR("Failed to resize bitmap"); | |
3833 | } | |
3834 | ||
9dbd1aa3 HM |
3835 | /* Check for any resize/reshape on @rs and adjust/initiate */ |
3836 | /* Be prepared for mddev_resume() in raid_resume() */ | |
3837 | set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3838 | if (mddev->recovery_cp && mddev->recovery_cp < MaxSector) { | |
9dbd1aa3 HM |
3839 | set_bit(MD_RECOVERY_SYNC, &mddev->recovery); |
3840 | mddev->resync_min = mddev->recovery_cp; | |
3841 | } | |
3842 | ||
3843 | rs_set_capacity(rs); | |
3844 | ||
345a6cdc | 3845 | /* Check for any reshape request unless new raid set */ |
9dbd1aa3 HM |
3846 | if (test_and_clear_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { |
3847 | /* Initiate a reshape. */ | |
3848 | mddev_lock_nointr(mddev); | |
3849 | r = rs_start_reshape(rs); | |
3850 | mddev_unlock(mddev); | |
3851 | if (r) | |
3852 | DMWARN("Failed to check/start reshape, continuing without change"); | |
3853 | r = 0; | |
3854 | } | |
3855 | ||
3856 | return r; | |
ecbfb9f1 HM |
3857 | } |
3858 | ||
f381e71b JB |
3859 | static void raid_resume(struct dm_target *ti) |
3860 | { | |
3861 | struct raid_set *rs = ti->private; | |
ecbfb9f1 | 3862 | struct mddev *mddev = &rs->md; |
f381e71b | 3863 | |
4286325b | 3864 | if (test_and_set_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { |
ecbfb9f1 HM |
3865 | /* |
3866 | * A secondary resume while the device is active. | |
3867 | * Take this opportunity to check whether any failed | |
3868 | * devices are reachable again. | |
3869 | */ | |
3870 | attempt_restore_of_faulty_devices(rs); | |
31e10a41 | 3871 | } |
34f8ac6d | 3872 | |
31e10a41 HM |
3873 | mddev->ro = 0; |
3874 | mddev->in_sync = 0; | |
3a1c1ef2 | 3875 | |
d36a1954 HM |
3876 | /* |
3877 | * Keep the RAID set frozen if reshape/rebuild flags are set. | |
3878 | * The RAID set is unfrozen once the next table load/resume, | |
3879 | * which clears the reshape/rebuild flags, occurs. | |
3880 | * This ensures that the constructor for the inactive table | |
3881 | * retrieves an up-to-date reshape_position. | |
3882 | */ | |
3883 | if (!(rs->ctr_flags & RESUME_STAY_FROZEN_FLAGS)) | |
3884 | clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
31e10a41 HM |
3885 | |
3886 | if (mddev->suspended) | |
3887 | mddev_resume(mddev); | |
9d09e663 N |
3888 | } |
3889 | ||
3890 | static struct target_type raid_target = { | |
3891 | .name = "raid", | |
6e53636f | 3892 | .version = {1, 11, 1}, |
9d09e663 N |
3893 | .module = THIS_MODULE, |
3894 | .ctr = raid_ctr, | |
3895 | .dtr = raid_dtr, | |
3896 | .map = raid_map, | |
3897 | .status = raid_status, | |
be83651f | 3898 | .message = raid_message, |
9d09e663 N |
3899 | .iterate_devices = raid_iterate_devices, |
3900 | .io_hints = raid_io_hints, | |
3901 | .presuspend = raid_presuspend, | |
3902 | .postsuspend = raid_postsuspend, | |
ecbfb9f1 | 3903 | .preresume = raid_preresume, |
9d09e663 N |
3904 | .resume = raid_resume, |
3905 | }; | |
3906 | ||
3907 | static int __init dm_raid_init(void) | |
3908 | { | |
fe5d2f4a JB |
3909 | DMINFO("Loading target version %u.%u.%u", |
3910 | raid_target.version[0], | |
3911 | raid_target.version[1], | |
3912 | raid_target.version[2]); | |
9d09e663 N |
3913 | return dm_register_target(&raid_target); |
3914 | } | |
3915 | ||
3916 | static void __exit dm_raid_exit(void) | |
3917 | { | |
3918 | dm_unregister_target(&raid_target); | |
3919 | } | |
3920 | ||
3921 | module_init(dm_raid_init); | |
3922 | module_exit(dm_raid_exit); | |
3923 | ||
48cf06bc HM |
3924 | module_param(devices_handle_discard_safely, bool, 0644); |
3925 | MODULE_PARM_DESC(devices_handle_discard_safely, | |
3926 | "Set to Y if all devices in each array reliably return zeroes on reads from discarded regions"); | |
3927 | ||
ef9b85a6 MS |
3928 | MODULE_DESCRIPTION(DM_NAME " raid0/1/10/4/5/6 target"); |
3929 | MODULE_ALIAS("dm-raid0"); | |
63f33b8d JB |
3930 | MODULE_ALIAS("dm-raid1"); |
3931 | MODULE_ALIAS("dm-raid10"); | |
9d09e663 N |
3932 | MODULE_ALIAS("dm-raid4"); |
3933 | MODULE_ALIAS("dm-raid5"); | |
3934 | MODULE_ALIAS("dm-raid6"); | |
3935 | MODULE_AUTHOR("Neil Brown <dm-devel@redhat.com>"); | |
3a1c1ef2 | 3936 | MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@redhat.com>"); |
9d09e663 | 3937 | MODULE_LICENSE("GPL"); |