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