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 | ||
48cf06bc HM |
27 | static bool devices_handle_discard_safely = false; |
28 | ||
9d09e663 | 29 | /* |
b12d437b JB |
30 | * The following flags are used by dm-raid.c to set up the array state. |
31 | * They must be cleared before md_run is called. | |
9d09e663 | 32 | */ |
43157840 | 33 | #define FirstUse 10 /* rdev flag */ |
9d09e663 N |
34 | |
35 | struct raid_dev { | |
36 | /* | |
37 | * Two DM devices, one to hold metadata and one to hold the | |
43157840 | 38 | * actual data/parity. The reason for this is to not confuse |
9d09e663 N |
39 | * ti->len and give more flexibility in altering size and |
40 | * characteristics. | |
41 | * | |
42 | * While it is possible for this device to be associated | |
43 | * with a different physical device than the data_dev, it | |
44 | * is intended for it to be the same. | |
45 | * |--------- Physical Device ---------| | |
46 | * |- meta_dev -|------ data_dev ------| | |
47 | */ | |
48 | struct dm_dev *meta_dev; | |
49 | struct dm_dev *data_dev; | |
3cb03002 | 50 | struct md_rdev rdev; |
9d09e663 N |
51 | }; |
52 | ||
53 | /* | |
4286325b | 54 | * Bits for establishing rs->ctr_flags |
702108d1 HM |
55 | * |
56 | * 1 = no flag value | |
57 | * 2 = flag with value | |
9d09e663 | 58 | */ |
4286325b MS |
59 | #define __CTR_FLAG_SYNC 0 /* 1 */ /* Not with raid0! */ |
60 | #define __CTR_FLAG_NOSYNC 1 /* 1 */ /* Not with raid0! */ | |
61 | #define __CTR_FLAG_REBUILD 2 /* 2 */ /* Not with raid0! */ | |
62 | #define __CTR_FLAG_DAEMON_SLEEP 3 /* 2 */ /* Not with raid0! */ | |
63 | #define __CTR_FLAG_MIN_RECOVERY_RATE 4 /* 2 */ /* Not with raid0! */ | |
64 | #define __CTR_FLAG_MAX_RECOVERY_RATE 5 /* 2 */ /* Not with raid0! */ | |
65 | #define __CTR_FLAG_MAX_WRITE_BEHIND 6 /* 2 */ /* Only with raid1! */ | |
66 | #define __CTR_FLAG_WRITE_MOSTLY 7 /* 2 */ /* Only with raid1! */ | |
67 | #define __CTR_FLAG_STRIPE_CACHE 8 /* 2 */ /* Only with raid4/5/6! */ | |
68 | #define __CTR_FLAG_REGION_SIZE 9 /* 2 */ /* Not with raid0! */ | |
69 | #define __CTR_FLAG_RAID10_COPIES 10 /* 2 */ /* Only with raid10 */ | |
70 | #define __CTR_FLAG_RAID10_FORMAT 11 /* 2 */ /* Only with raid10 */ | |
9b6e5423 | 71 | /* New for v1.9.0 */ |
d7ccc2e2 | 72 | #define __CTR_FLAG_DELTA_DISKS 12 /* 2 */ /* Only with reshapable raid1/4/5/6/10! */ |
4286325b MS |
73 | #define __CTR_FLAG_DATA_OFFSET 13 /* 2 */ /* Only with reshapable raid4/5/6/10! */ |
74 | #define __CTR_FLAG_RAID10_USE_NEAR_SETS 14 /* 2 */ /* Only with raid10! */ | |
75 | ||
76 | /* | |
77 | * Flags for rs->ctr_flags field. | |
78 | */ | |
79 | #define CTR_FLAG_SYNC (1 << __CTR_FLAG_SYNC) | |
80 | #define CTR_FLAG_NOSYNC (1 << __CTR_FLAG_NOSYNC) | |
81 | #define CTR_FLAG_REBUILD (1 << __CTR_FLAG_REBUILD) | |
82 | #define CTR_FLAG_DAEMON_SLEEP (1 << __CTR_FLAG_DAEMON_SLEEP) | |
83 | #define CTR_FLAG_MIN_RECOVERY_RATE (1 << __CTR_FLAG_MIN_RECOVERY_RATE) | |
84 | #define CTR_FLAG_MAX_RECOVERY_RATE (1 << __CTR_FLAG_MAX_RECOVERY_RATE) | |
85 | #define CTR_FLAG_MAX_WRITE_BEHIND (1 << __CTR_FLAG_MAX_WRITE_BEHIND) | |
86 | #define CTR_FLAG_WRITE_MOSTLY (1 << __CTR_FLAG_WRITE_MOSTLY) | |
87 | #define CTR_FLAG_STRIPE_CACHE (1 << __CTR_FLAG_STRIPE_CACHE) | |
88 | #define CTR_FLAG_REGION_SIZE (1 << __CTR_FLAG_REGION_SIZE) | |
89 | #define CTR_FLAG_RAID10_COPIES (1 << __CTR_FLAG_RAID10_COPIES) | |
90 | #define CTR_FLAG_RAID10_FORMAT (1 << __CTR_FLAG_RAID10_FORMAT) | |
91 | #define CTR_FLAG_DELTA_DISKS (1 << __CTR_FLAG_DELTA_DISKS) | |
92 | #define CTR_FLAG_DATA_OFFSET (1 << __CTR_FLAG_DATA_OFFSET) | |
93 | #define CTR_FLAG_RAID10_USE_NEAR_SETS (1 << __CTR_FLAG_RAID10_USE_NEAR_SETS) | |
63f33b8d | 94 | |
f090279e HM |
95 | /* |
96 | * Definitions of various constructor flags to | |
97 | * be used in checks of valid / invalid flags | |
98 | * per raid level. | |
99 | */ | |
100 | /* Define all any sync flags */ | |
101 | #define CTR_FLAGS_ANY_SYNC (CTR_FLAG_SYNC | CTR_FLAG_NOSYNC) | |
102 | ||
103 | /* Define flags for options without argument (e.g. 'nosync') */ | |
33e53f06 HM |
104 | #define CTR_FLAG_OPTIONS_NO_ARGS (CTR_FLAGS_ANY_SYNC | \ |
105 | CTR_FLAG_RAID10_USE_NEAR_SETS) | |
f090279e HM |
106 | |
107 | /* Define flags for options with one argument (e.g. 'delta_disks +2') */ | |
108 | #define CTR_FLAG_OPTIONS_ONE_ARG (CTR_FLAG_REBUILD | \ | |
109 | CTR_FLAG_WRITE_MOSTLY | \ | |
110 | CTR_FLAG_DAEMON_SLEEP | \ | |
111 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
112 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
113 | CTR_FLAG_MAX_WRITE_BEHIND | \ | |
114 | CTR_FLAG_STRIPE_CACHE | \ | |
115 | CTR_FLAG_REGION_SIZE | \ | |
116 | CTR_FLAG_RAID10_COPIES | \ | |
33e53f06 HM |
117 | CTR_FLAG_RAID10_FORMAT | \ |
118 | CTR_FLAG_DELTA_DISKS | \ | |
119 | CTR_FLAG_DATA_OFFSET) | |
f090279e | 120 | |
a30cbc0d HM |
121 | /* Valid options definitions per raid level... */ |
122 | ||
123 | /* "raid0" does only accept data offset */ | |
124 | #define RAID0_VALID_FLAGS (CTR_FLAG_DATA_OFFSET) | |
125 | ||
126 | /* "raid1" does not accept stripe cache, data offset, delta_disks or any raid10 options */ | |
127 | #define RAID1_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ | |
128 | CTR_FLAG_REBUILD | \ | |
129 | CTR_FLAG_WRITE_MOSTLY | \ | |
130 | CTR_FLAG_DAEMON_SLEEP | \ | |
131 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
132 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
133 | CTR_FLAG_MAX_WRITE_BEHIND | \ | |
134 | CTR_FLAG_REGION_SIZE | \ | |
7a7c330f | 135 | CTR_FLAG_DELTA_DISKS | \ |
a30cbc0d | 136 | CTR_FLAG_DATA_OFFSET) |
f090279e | 137 | |
a30cbc0d HM |
138 | /* "raid10" does not accept any raid1 or stripe cache options */ |
139 | #define RAID10_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ | |
140 | CTR_FLAG_REBUILD | \ | |
141 | CTR_FLAG_DAEMON_SLEEP | \ | |
142 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
143 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
144 | CTR_FLAG_REGION_SIZE | \ | |
f090279e | 145 | CTR_FLAG_RAID10_COPIES | \ |
33e53f06 HM |
146 | CTR_FLAG_RAID10_FORMAT | \ |
147 | CTR_FLAG_DELTA_DISKS | \ | |
a30cbc0d HM |
148 | CTR_FLAG_DATA_OFFSET | \ |
149 | CTR_FLAG_RAID10_USE_NEAR_SETS) | |
f090279e | 150 | |
f090279e HM |
151 | /* |
152 | * "raid4/5/6" do not accept any raid1 or raid10 specific options | |
153 | * | |
154 | * "raid6" does not accept "nosync", because it is not guaranteed | |
155 | * that both parity and q-syndrome are being written properly with | |
156 | * any writes | |
157 | */ | |
a30cbc0d HM |
158 | #define RAID45_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ |
159 | CTR_FLAG_REBUILD | \ | |
160 | CTR_FLAG_DAEMON_SLEEP | \ | |
161 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
162 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
f090279e | 163 | CTR_FLAG_MAX_WRITE_BEHIND | \ |
a30cbc0d HM |
164 | CTR_FLAG_STRIPE_CACHE | \ |
165 | CTR_FLAG_REGION_SIZE | \ | |
166 | CTR_FLAG_DELTA_DISKS | \ | |
167 | CTR_FLAG_DATA_OFFSET) | |
168 | ||
169 | #define RAID6_VALID_FLAGS (CTR_FLAG_SYNC | \ | |
170 | CTR_FLAG_REBUILD | \ | |
171 | CTR_FLAG_DAEMON_SLEEP | \ | |
172 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
173 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
174 | CTR_FLAG_MAX_WRITE_BEHIND | \ | |
175 | CTR_FLAG_STRIPE_CACHE | \ | |
176 | CTR_FLAG_REGION_SIZE | \ | |
177 | CTR_FLAG_DELTA_DISKS | \ | |
178 | CTR_FLAG_DATA_OFFSET) | |
179 | /* ...valid options definitions per raid level */ | |
f090279e | 180 | |
ecbfb9f1 HM |
181 | /* |
182 | * Flags for rs->runtime_flags field | |
183 | * (RT_FLAG prefix meaning "runtime flag") | |
184 | * | |
185 | * These are all internal and used to define runtime state, | |
186 | * e.g. to prevent another resume from preresume processing | |
187 | * the raid set all over again. | |
188 | */ | |
40ba37e5 HM |
189 | #define RT_FLAG_RS_PRERESUMED 0 |
190 | #define RT_FLAG_RS_RESUMED 1 | |
191 | #define RT_FLAG_RS_BITMAP_LOADED 2 | |
192 | #define RT_FLAG_UPDATE_SBS 3 | |
9dbd1aa3 | 193 | #define RT_FLAG_RESHAPE_RS 4 |
6e20902e | 194 | #define RT_FLAG_KEEP_RS_FROZEN 5 |
ecbfb9f1 | 195 | |
d7ccc2e2 | 196 | /* Array elements of 64 bit needed for rebuild/failed disk bits */ |
33e53f06 HM |
197 | #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8) |
198 | ||
ecbfb9f1 HM |
199 | /* |
200 | * raid set level, layout and chunk sectors backup/restore | |
201 | */ | |
202 | struct rs_layout { | |
203 | int new_level; | |
204 | int new_layout; | |
205 | int new_chunk_sectors; | |
206 | }; | |
207 | ||
9d09e663 N |
208 | struct raid_set { |
209 | struct dm_target *ti; | |
210 | ||
34f8ac6d | 211 | uint32_t bitmap_loaded; |
9dbd1aa3 | 212 | uint32_t stripe_cache_entries; |
4286325b MS |
213 | unsigned long ctr_flags; |
214 | unsigned long runtime_flags; | |
ecbfb9f1 HM |
215 | |
216 | uint64_t rebuild_disks[DISKS_ARRAY_ELEMS]; | |
9d09e663 | 217 | |
33e53f06 HM |
218 | int raid_disks; |
219 | int delta_disks; | |
4763e543 | 220 | int data_offset; |
33e53f06 | 221 | int raid10_copies; |
4257e085 | 222 | int requested_bitmap_chunk_sectors; |
33e53f06 | 223 | |
fd01b88c | 224 | struct mddev md; |
9d09e663 N |
225 | struct raid_type *raid_type; |
226 | struct dm_target_callbacks callbacks; | |
227 | ||
228 | struct raid_dev dev[0]; | |
229 | }; | |
230 | ||
9dbd1aa3 | 231 | static void rs_config_backup(struct raid_set *rs, struct rs_layout *l) |
ecbfb9f1 HM |
232 | { |
233 | struct mddev *mddev = &rs->md; | |
234 | ||
235 | l->new_level = mddev->new_level; | |
236 | l->new_layout = mddev->new_layout; | |
237 | l->new_chunk_sectors = mddev->new_chunk_sectors; | |
238 | } | |
239 | ||
9dbd1aa3 | 240 | static void rs_config_restore(struct raid_set *rs, struct rs_layout *l) |
ecbfb9f1 HM |
241 | { |
242 | struct mddev *mddev = &rs->md; | |
243 | ||
244 | mddev->new_level = l->new_level; | |
245 | mddev->new_layout = l->new_layout; | |
246 | mddev->new_chunk_sectors = l->new_chunk_sectors; | |
247 | } | |
248 | ||
33e53f06 HM |
249 | /* raid10 algorithms (i.e. formats) */ |
250 | #define ALGORITHM_RAID10_DEFAULT 0 | |
251 | #define ALGORITHM_RAID10_NEAR 1 | |
252 | #define ALGORITHM_RAID10_OFFSET 2 | |
253 | #define ALGORITHM_RAID10_FAR 3 | |
254 | ||
9d09e663 N |
255 | /* Supported raid types and properties. */ |
256 | static struct raid_type { | |
257 | const char *name; /* RAID algorithm. */ | |
258 | const char *descr; /* Descriptor text for logging. */ | |
094f394d HM |
259 | const unsigned int parity_devs; /* # of parity devices. */ |
260 | const unsigned int minimal_devs;/* minimal # of devices in set. */ | |
261 | const unsigned int level; /* RAID level. */ | |
262 | const unsigned int algorithm; /* RAID algorithm. */ | |
9d09e663 | 263 | } raid_types[] = { |
43157840 MS |
264 | {"raid0", "raid0 (striping)", 0, 2, 0, 0 /* NONE */}, |
265 | {"raid1", "raid1 (mirroring)", 0, 2, 1, 0 /* NONE */}, | |
266 | {"raid10_far", "raid10 far (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_FAR}, | |
33e53f06 | 267 | {"raid10_offset", "raid10 offset (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_OFFSET}, |
43157840 MS |
268 | {"raid10_near", "raid10 near (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_NEAR}, |
269 | {"raid10", "raid10 (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_DEFAULT}, | |
270 | {"raid4", "raid4 (dedicated last parity disk)", 1, 2, 4, ALGORITHM_PARITY_N}, /* raid4 layout = raid5_n */ | |
271 | {"raid5_n", "raid5 (dedicated last parity disk)", 1, 2, 5, ALGORITHM_PARITY_N}, | |
272 | {"raid5_ls", "raid5 (left symmetric)", 1, 2, 5, ALGORITHM_LEFT_SYMMETRIC}, | |
273 | {"raid5_rs", "raid5 (right symmetric)", 1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC}, | |
274 | {"raid5_la", "raid5 (left asymmetric)", 1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC}, | |
275 | {"raid5_ra", "raid5 (right asymmetric)", 1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC}, | |
276 | {"raid6_zr", "raid6 (zero restart)", 2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART}, | |
277 | {"raid6_nr", "raid6 (N restart)", 2, 4, 6, ALGORITHM_ROTATING_N_RESTART}, | |
278 | {"raid6_nc", "raid6 (N continue)", 2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}, | |
279 | {"raid6_n_6", "raid6 (dedicated parity/Q n/6)", 2, 4, 6, ALGORITHM_PARITY_N_6}, | |
280 | {"raid6_ls_6", "raid6 (left symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_SYMMETRIC_6}, | |
281 | {"raid6_rs_6", "raid6 (right symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_SYMMETRIC_6}, | |
282 | {"raid6_la_6", "raid6 (left asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_ASYMMETRIC_6}, | |
283 | {"raid6_ra_6", "raid6 (right asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_ASYMMETRIC_6} | |
9d09e663 N |
284 | }; |
285 | ||
92c83d79 | 286 | /* True, if @v is in inclusive range [@min, @max] */ |
bb91a63f | 287 | static bool __within_range(long v, long min, long max) |
92c83d79 HM |
288 | { |
289 | return v >= min && v <= max; | |
290 | } | |
291 | ||
702108d1 HM |
292 | /* All table line arguments are defined here */ |
293 | static struct arg_name_flag { | |
4286325b | 294 | const unsigned long flag; |
702108d1 | 295 | const char *name; |
e6ca5e1a | 296 | } __arg_name_flags[] = { |
702108d1 HM |
297 | { CTR_FLAG_SYNC, "sync"}, |
298 | { CTR_FLAG_NOSYNC, "nosync"}, | |
299 | { CTR_FLAG_REBUILD, "rebuild"}, | |
300 | { CTR_FLAG_DAEMON_SLEEP, "daemon_sleep"}, | |
301 | { CTR_FLAG_MIN_RECOVERY_RATE, "min_recovery_rate"}, | |
302 | { CTR_FLAG_MAX_RECOVERY_RATE, "max_recovery_rate"}, | |
303 | { CTR_FLAG_MAX_WRITE_BEHIND, "max_write_behind"}, | |
65359ee6 | 304 | { CTR_FLAG_WRITE_MOSTLY, "write_mostly"}, |
702108d1 HM |
305 | { CTR_FLAG_STRIPE_CACHE, "stripe_cache"}, |
306 | { CTR_FLAG_REGION_SIZE, "region_size"}, | |
307 | { CTR_FLAG_RAID10_COPIES, "raid10_copies"}, | |
308 | { CTR_FLAG_RAID10_FORMAT, "raid10_format"}, | |
4763e543 HM |
309 | { CTR_FLAG_DATA_OFFSET, "data_offset"}, |
310 | { CTR_FLAG_DELTA_DISKS, "delta_disks"}, | |
311 | { CTR_FLAG_RAID10_USE_NEAR_SETS, "raid10_use_near_sets"}, | |
702108d1 HM |
312 | }; |
313 | ||
314 | /* Return argument name string for given @flag */ | |
3fa6cf38 | 315 | static const char *dm_raid_arg_name_by_flag(const uint32_t flag) |
702108d1 HM |
316 | { |
317 | if (hweight32(flag) == 1) { | |
e6ca5e1a | 318 | struct arg_name_flag *anf = __arg_name_flags + ARRAY_SIZE(__arg_name_flags); |
702108d1 | 319 | |
e6ca5e1a | 320 | while (anf-- > __arg_name_flags) |
4286325b | 321 | if (flag & anf->flag) |
702108d1 HM |
322 | return anf->name; |
323 | ||
324 | } else | |
325 | DMERR("%s called with more than one flag!", __func__); | |
326 | ||
327 | return NULL; | |
328 | } | |
329 | ||
33e53f06 | 330 | /* |
d7ccc2e2 HM |
331 | * Bool helpers to test for various raid levels of a raid set. |
332 | * It's level as reported by the superblock rather than | |
33e53f06 HM |
333 | * the requested raid_type passed to the constructor. |
334 | */ | |
335 | /* Return true, if raid set in @rs is raid0 */ | |
336 | static bool rs_is_raid0(struct raid_set *rs) | |
337 | { | |
338 | return !rs->md.level; | |
339 | } | |
340 | ||
9dbd1aa3 HM |
341 | /* Return true, if raid set in @rs is raid1 */ |
342 | static bool rs_is_raid1(struct raid_set *rs) | |
343 | { | |
344 | return rs->md.level == 1; | |
345 | } | |
346 | ||
33e53f06 HM |
347 | /* Return true, if raid set in @rs is raid10 */ |
348 | static bool rs_is_raid10(struct raid_set *rs) | |
349 | { | |
350 | return rs->md.level == 10; | |
351 | } | |
352 | ||
4dff2f1e HM |
353 | /* Return true, if raid set in @rs is level 6 */ |
354 | static bool rs_is_raid6(struct raid_set *rs) | |
355 | { | |
356 | return rs->md.level == 6; | |
357 | } | |
358 | ||
40ba37e5 HM |
359 | /* Return true, if raid set in @rs is level 4, 5 or 6 */ |
360 | static bool rs_is_raid456(struct raid_set *rs) | |
361 | { | |
362 | return __within_range(rs->md.level, 4, 6); | |
363 | } | |
364 | ||
365 | /* Return true, if raid set in @rs is reshapable */ | |
d7ccc2e2 | 366 | static bool __is_raid10_far(int layout); |
40ba37e5 HM |
367 | static bool rs_is_reshapable(struct raid_set *rs) |
368 | { | |
369 | return rs_is_raid456(rs) || | |
370 | (rs_is_raid10(rs) && !__is_raid10_far(rs->md.new_layout)); | |
371 | } | |
372 | ||
9dbd1aa3 HM |
373 | /* Return true, if raid set in @rs is recovering */ |
374 | static bool rs_is_recovering(struct raid_set *rs) | |
375 | { | |
345a6cdc | 376 | return rs->md.recovery_cp < rs->dev[0].rdev.sectors; |
9dbd1aa3 HM |
377 | } |
378 | ||
379 | /* Return true, if raid set in @rs is reshaping */ | |
380 | static bool rs_is_reshaping(struct raid_set *rs) | |
381 | { | |
9dbd1aa3 HM |
382 | return rs->md.reshape_position != MaxSector; |
383 | } | |
384 | ||
f090279e | 385 | /* |
d7ccc2e2 | 386 | * bool helpers to test for various raid levels of a raid type @rt |
f090279e HM |
387 | */ |
388 | ||
389 | /* Return true, if raid type in @rt is raid0 */ | |
390 | static bool rt_is_raid0(struct raid_type *rt) | |
391 | { | |
392 | return !rt->level; | |
393 | } | |
394 | ||
395 | /* Return true, if raid type in @rt is raid1 */ | |
396 | static bool rt_is_raid1(struct raid_type *rt) | |
397 | { | |
398 | return rt->level == 1; | |
399 | } | |
400 | ||
401 | /* Return true, if raid type in @rt is raid10 */ | |
402 | static bool rt_is_raid10(struct raid_type *rt) | |
403 | { | |
404 | return rt->level == 10; | |
405 | } | |
406 | ||
407 | /* Return true, if raid type in @rt is raid4/5 */ | |
408 | static bool rt_is_raid45(struct raid_type *rt) | |
409 | { | |
bb91a63f | 410 | return __within_range(rt->level, 4, 5); |
f090279e HM |
411 | } |
412 | ||
413 | /* Return true, if raid type in @rt is raid6 */ | |
414 | static bool rt_is_raid6(struct raid_type *rt) | |
415 | { | |
416 | return rt->level == 6; | |
417 | } | |
676fa5ad HM |
418 | |
419 | /* Return true, if raid type in @rt is raid4/5/6 */ | |
420 | static bool rt_is_raid456(struct raid_type *rt) | |
421 | { | |
bb91a63f | 422 | return __within_range(rt->level, 4, 6); |
676fa5ad | 423 | } |
f090279e HM |
424 | /* END: raid level bools */ |
425 | ||
a30cbc0d HM |
426 | /* Return valid ctr flags for the raid level of @rs */ |
427 | static unsigned long __valid_flags(struct raid_set *rs) | |
f090279e HM |
428 | { |
429 | if (rt_is_raid0(rs->raid_type)) | |
a30cbc0d | 430 | return RAID0_VALID_FLAGS; |
f090279e | 431 | else if (rt_is_raid1(rs->raid_type)) |
a30cbc0d | 432 | return RAID1_VALID_FLAGS; |
f090279e | 433 | else if (rt_is_raid10(rs->raid_type)) |
a30cbc0d | 434 | return RAID10_VALID_FLAGS; |
f090279e | 435 | else if (rt_is_raid45(rs->raid_type)) |
a30cbc0d | 436 | return RAID45_VALID_FLAGS; |
f090279e | 437 | else if (rt_is_raid6(rs->raid_type)) |
a30cbc0d | 438 | return RAID6_VALID_FLAGS; |
f090279e | 439 | |
d7ccc2e2 | 440 | return 0; |
f090279e HM |
441 | } |
442 | ||
443 | /* | |
a30cbc0d | 444 | * Check for valid flags set on @rs |
f090279e HM |
445 | * |
446 | * Has to be called after parsing of the ctr flags! | |
447 | */ | |
a30cbc0d | 448 | static int rs_check_for_valid_flags(struct raid_set *rs) |
f090279e | 449 | { |
a30cbc0d | 450 | if (rs->ctr_flags & ~__valid_flags(rs)) { |
4286325b | 451 | rs->ti->error = "Invalid flags combination"; |
bd83a4c4 MS |
452 | return -EINVAL; |
453 | } | |
f090279e HM |
454 | |
455 | return 0; | |
456 | } | |
457 | ||
33e53f06 HM |
458 | /* MD raid10 bit definitions and helpers */ |
459 | #define RAID10_OFFSET (1 << 16) /* stripes with data copies area adjacent on devices */ | |
460 | #define RAID10_BROCKEN_USE_FAR_SETS (1 << 17) /* Broken in raid10.c: use sets instead of whole stripe rotation */ | |
461 | #define RAID10_USE_FAR_SETS (1 << 18) /* Use sets instead of whole stripe rotation */ | |
462 | #define RAID10_FAR_COPIES_SHIFT 8 /* raid10 # far copies shift (2nd byte of layout) */ | |
463 | ||
464 | /* Return md raid10 near copies for @layout */ | |
e6ca5e1a | 465 | static unsigned int __raid10_near_copies(int layout) |
33e53f06 HM |
466 | { |
467 | return layout & 0xFF; | |
468 | } | |
469 | ||
470 | /* Return md raid10 far copies for @layout */ | |
e6ca5e1a | 471 | static unsigned int __raid10_far_copies(int layout) |
33e53f06 | 472 | { |
e6ca5e1a | 473 | return __raid10_near_copies(layout >> RAID10_FAR_COPIES_SHIFT); |
33e53f06 HM |
474 | } |
475 | ||
476 | /* Return true if md raid10 offset for @layout */ | |
d7ccc2e2 | 477 | static bool __is_raid10_offset(int layout) |
33e53f06 | 478 | { |
d7ccc2e2 | 479 | return !!(layout & RAID10_OFFSET); |
33e53f06 HM |
480 | } |
481 | ||
482 | /* Return true if md raid10 near for @layout */ | |
d7ccc2e2 | 483 | static bool __is_raid10_near(int layout) |
33e53f06 | 484 | { |
e6ca5e1a | 485 | return !__is_raid10_offset(layout) && __raid10_near_copies(layout) > 1; |
33e53f06 HM |
486 | } |
487 | ||
488 | /* Return true if md raid10 far for @layout */ | |
d7ccc2e2 | 489 | static bool __is_raid10_far(int layout) |
33e53f06 | 490 | { |
e6ca5e1a | 491 | return !__is_raid10_offset(layout) && __raid10_far_copies(layout) > 1; |
33e53f06 HM |
492 | } |
493 | ||
494 | /* Return md raid10 layout string for @layout */ | |
495 | static const char *raid10_md_layout_to_format(int layout) | |
fe5d2f4a JB |
496 | { |
497 | /* | |
33e53f06 HM |
498 | * Bit 16 stands for "offset" |
499 | * (i.e. adjacent stripes hold copies) | |
500 | * | |
fe5d2f4a JB |
501 | * Refer to MD's raid10.c for details |
502 | */ | |
e6ca5e1a | 503 | if (__is_raid10_offset(layout)) |
fe5d2f4a JB |
504 | return "offset"; |
505 | ||
e6ca5e1a | 506 | if (__raid10_near_copies(layout) > 1) |
fe5d2f4a JB |
507 | return "near"; |
508 | ||
e6ca5e1a | 509 | WARN_ON(__raid10_far_copies(layout) < 2); |
33e53f06 | 510 | |
fe5d2f4a JB |
511 | return "far"; |
512 | } | |
513 | ||
33e53f06 | 514 | /* Return md raid10 algorithm for @name */ |
68c1c4d5 | 515 | static int raid10_name_to_format(const char *name) |
33e53f06 HM |
516 | { |
517 | if (!strcasecmp(name, "near")) | |
518 | return ALGORITHM_RAID10_NEAR; | |
519 | else if (!strcasecmp(name, "offset")) | |
520 | return ALGORITHM_RAID10_OFFSET; | |
521 | else if (!strcasecmp(name, "far")) | |
522 | return ALGORITHM_RAID10_FAR; | |
523 | ||
524 | return -EINVAL; | |
525 | } | |
526 | ||
33e53f06 HM |
527 | /* Return md raid10 copies for @layout */ |
528 | static unsigned int raid10_md_layout_to_copies(int layout) | |
63f33b8d | 529 | { |
d7ccc2e2 | 530 | return max(__raid10_near_copies(layout), __raid10_far_copies(layout)); |
63f33b8d JB |
531 | } |
532 | ||
33e53f06 HM |
533 | /* Return md raid10 format id for @format string */ |
534 | static int raid10_format_to_md_layout(struct raid_set *rs, | |
535 | unsigned int algorithm, | |
536 | unsigned int copies) | |
63f33b8d | 537 | { |
33e53f06 | 538 | unsigned int n = 1, f = 1, r = 0; |
fe5d2f4a | 539 | |
33e53f06 HM |
540 | /* |
541 | * MD resilienece flaw: | |
542 | * | |
543 | * enabling use_far_sets for far/offset formats causes copies | |
544 | * to be colocated on the same devs together with their origins! | |
545 | * | |
546 | * -> disable it for now in the definition above | |
547 | */ | |
548 | if (algorithm == ALGORITHM_RAID10_DEFAULT || | |
549 | algorithm == ALGORITHM_RAID10_NEAR) | |
fe5d2f4a | 550 | n = copies; |
33e53f06 HM |
551 | |
552 | else if (algorithm == ALGORITHM_RAID10_OFFSET) { | |
553 | f = copies; | |
554 | r = RAID10_OFFSET; | |
4286325b | 555 | if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) |
33e53f06 HM |
556 | r |= RAID10_USE_FAR_SETS; |
557 | ||
558 | } else if (algorithm == ALGORITHM_RAID10_FAR) { | |
fe5d2f4a | 559 | f = copies; |
33e53f06 | 560 | r = !RAID10_OFFSET; |
4286325b | 561 | if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) |
33e53f06 | 562 | r |= RAID10_USE_FAR_SETS; |
fe5d2f4a | 563 | |
33e53f06 HM |
564 | } else |
565 | return -EINVAL; | |
566 | ||
567 | return r | (f << RAID10_FAR_COPIES_SHIFT) | n; | |
568 | } | |
569 | /* END: MD raid10 bit definitions and helpers */ | |
fe5d2f4a | 570 | |
33e53f06 | 571 | /* Check for any of the raid10 algorithms */ |
d7ccc2e2 | 572 | static bool __got_raid10(struct raid_type *rtp, const int layout) |
33e53f06 HM |
573 | { |
574 | if (rtp->level == 10) { | |
575 | switch (rtp->algorithm) { | |
576 | case ALGORITHM_RAID10_DEFAULT: | |
577 | case ALGORITHM_RAID10_NEAR: | |
e6ca5e1a | 578 | return __is_raid10_near(layout); |
33e53f06 | 579 | case ALGORITHM_RAID10_OFFSET: |
e6ca5e1a | 580 | return __is_raid10_offset(layout); |
33e53f06 | 581 | case ALGORITHM_RAID10_FAR: |
e6ca5e1a | 582 | return __is_raid10_far(layout); |
33e53f06 HM |
583 | default: |
584 | break; | |
585 | } | |
586 | } | |
fe5d2f4a | 587 | |
d7ccc2e2 | 588 | return false; |
63f33b8d JB |
589 | } |
590 | ||
33e53f06 | 591 | /* Return raid_type for @name */ |
92c83d79 | 592 | static struct raid_type *get_raid_type(const char *name) |
9d09e663 | 593 | { |
33e53f06 | 594 | struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); |
9d09e663 | 595 | |
33e53f06 HM |
596 | while (rtp-- > raid_types) |
597 | if (!strcasecmp(rtp->name, name)) | |
598 | return rtp; | |
9d09e663 N |
599 | |
600 | return NULL; | |
601 | } | |
602 | ||
33e53f06 HM |
603 | /* Return raid_type for @name based derived from @level and @layout */ |
604 | static struct raid_type *get_raid_type_by_ll(const int level, const int layout) | |
605 | { | |
606 | struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); | |
607 | ||
608 | while (rtp-- > raid_types) { | |
609 | /* RAID10 special checks based on @layout flags/properties */ | |
610 | if (rtp->level == level && | |
e6ca5e1a | 611 | (__got_raid10(rtp, layout) || rtp->algorithm == layout)) |
33e53f06 HM |
612 | return rtp; |
613 | } | |
614 | ||
615 | return NULL; | |
616 | } | |
617 | ||
9dbd1aa3 HM |
618 | /* |
619 | * Conditionally change bdev capacity of @rs | |
620 | * in case of a disk add/remove reshape | |
621 | */ | |
622 | static void rs_set_capacity(struct raid_set *rs) | |
623 | { | |
624 | struct mddev *mddev = &rs->md; | |
fbe6365b | 625 | struct md_rdev *rdev; |
0095dbc9 | 626 | struct gendisk *gendisk = dm_disk(dm_table_get_md(rs->ti->table)); |
9dbd1aa3 | 627 | |
fbe6365b HM |
628 | /* |
629 | * raid10 sets rdev->sector to the device size, which | |
630 | * is unintended in case of out-of-place reshaping | |
631 | */ | |
632 | rdev_for_each(rdev, mddev) | |
633 | rdev->sectors = mddev->dev_sectors; | |
634 | ||
0095dbc9 HM |
635 | set_capacity(gendisk, mddev->array_sectors); |
636 | revalidate_disk(gendisk); | |
9dbd1aa3 HM |
637 | } |
638 | ||
3a1c1ef2 HM |
639 | /* |
640 | * Set the mddev properties in @rs to the current | |
641 | * ones retrieved from the freshest superblock | |
642 | */ | |
643 | static void rs_set_cur(struct raid_set *rs) | |
644 | { | |
645 | struct mddev *mddev = &rs->md; | |
646 | ||
647 | mddev->new_level = mddev->level; | |
648 | mddev->new_layout = mddev->layout; | |
649 | mddev->new_chunk_sectors = mddev->chunk_sectors; | |
650 | } | |
651 | ||
33e53f06 HM |
652 | /* |
653 | * Set the mddev properties in @rs to the new | |
654 | * ones requested by the ctr | |
655 | */ | |
656 | static void rs_set_new(struct raid_set *rs) | |
657 | { | |
658 | struct mddev *mddev = &rs->md; | |
659 | ||
660 | mddev->level = mddev->new_level; | |
661 | mddev->layout = mddev->new_layout; | |
662 | mddev->chunk_sectors = mddev->new_chunk_sectors; | |
3a1c1ef2 | 663 | mddev->raid_disks = rs->raid_disks; |
33e53f06 HM |
664 | mddev->delta_disks = 0; |
665 | } | |
666 | ||
bfcee0e3 | 667 | static struct raid_set *raid_set_alloc(struct dm_target *ti, struct raid_type *raid_type, |
094f394d | 668 | unsigned int raid_devs) |
9d09e663 | 669 | { |
094f394d | 670 | unsigned int i; |
9d09e663 | 671 | struct raid_set *rs; |
9d09e663 | 672 | |
bd83a4c4 MS |
673 | if (raid_devs <= raid_type->parity_devs) { |
674 | ti->error = "Insufficient number of devices"; | |
675 | return ERR_PTR(-EINVAL); | |
676 | } | |
9d09e663 | 677 | |
9d09e663 | 678 | rs = kzalloc(sizeof(*rs) + raid_devs * sizeof(rs->dev[0]), GFP_KERNEL); |
bd83a4c4 MS |
679 | if (!rs) { |
680 | ti->error = "Cannot allocate raid context"; | |
681 | return ERR_PTR(-ENOMEM); | |
682 | } | |
9d09e663 N |
683 | |
684 | mddev_init(&rs->md); | |
685 | ||
33e53f06 HM |
686 | rs->raid_disks = raid_devs; |
687 | rs->delta_disks = 0; | |
688 | ||
9d09e663 N |
689 | rs->ti = ti; |
690 | rs->raid_type = raid_type; | |
9dbd1aa3 | 691 | rs->stripe_cache_entries = 256; |
9d09e663 N |
692 | rs->md.raid_disks = raid_devs; |
693 | rs->md.level = raid_type->level; | |
694 | rs->md.new_level = rs->md.level; | |
9d09e663 N |
695 | rs->md.layout = raid_type->algorithm; |
696 | rs->md.new_layout = rs->md.layout; | |
697 | rs->md.delta_disks = 0; | |
4dff2f1e | 698 | rs->md.recovery_cp = MaxSector; |
9d09e663 N |
699 | |
700 | for (i = 0; i < raid_devs; i++) | |
701 | md_rdev_init(&rs->dev[i].rdev); | |
702 | ||
703 | /* | |
704 | * Remaining items to be initialized by further RAID params: | |
705 | * rs->md.persistent | |
706 | * rs->md.external | |
707 | * rs->md.chunk_sectors | |
708 | * rs->md.new_chunk_sectors | |
c039c332 | 709 | * rs->md.dev_sectors |
9d09e663 N |
710 | */ |
711 | ||
712 | return rs; | |
713 | } | |
714 | ||
bfcee0e3 | 715 | static void raid_set_free(struct raid_set *rs) |
9d09e663 N |
716 | { |
717 | int i; | |
718 | ||
ffeeac75 | 719 | for (i = 0; i < rs->raid_disks; i++) { |
b12d437b JB |
720 | if (rs->dev[i].meta_dev) |
721 | dm_put_device(rs->ti, rs->dev[i].meta_dev); | |
545c8795 | 722 | md_rdev_clear(&rs->dev[i].rdev); |
9d09e663 N |
723 | if (rs->dev[i].data_dev) |
724 | dm_put_device(rs->ti, rs->dev[i].data_dev); | |
b12d437b | 725 | } |
9d09e663 N |
726 | |
727 | kfree(rs); | |
728 | } | |
729 | ||
730 | /* | |
731 | * For every device we have two words | |
732 | * <meta_dev>: meta device name or '-' if missing | |
733 | * <data_dev>: data device name or '-' if missing | |
734 | * | |
b12d437b JB |
735 | * The following are permitted: |
736 | * - - | |
737 | * - <data_dev> | |
738 | * <meta_dev> <data_dev> | |
739 | * | |
740 | * The following is not allowed: | |
741 | * <meta_dev> - | |
742 | * | |
743 | * This code parses those words. If there is a failure, | |
bfcee0e3 | 744 | * the caller must use raid_set_free() to unwind the operations. |
9d09e663 | 745 | */ |
702108d1 | 746 | static int parse_dev_params(struct raid_set *rs, struct dm_arg_set *as) |
9d09e663 N |
747 | { |
748 | int i; | |
749 | int rebuild = 0; | |
750 | int metadata_available = 0; | |
73c6f239 | 751 | int r = 0; |
92c83d79 | 752 | const char *arg; |
9d09e663 | 753 | |
92c83d79 HM |
754 | /* Put off the number of raid devices argument to get to dev pairs */ |
755 | arg = dm_shift_arg(as); | |
756 | if (!arg) | |
757 | return -EINVAL; | |
758 | ||
ffeeac75 | 759 | for (i = 0; i < rs->raid_disks; i++) { |
9d09e663 N |
760 | rs->dev[i].rdev.raid_disk = i; |
761 | ||
762 | rs->dev[i].meta_dev = NULL; | |
763 | rs->dev[i].data_dev = NULL; | |
764 | ||
765 | /* | |
766 | * There are no offsets, since there is a separate device | |
767 | * for data and metadata. | |
768 | */ | |
769 | rs->dev[i].rdev.data_offset = 0; | |
770 | rs->dev[i].rdev.mddev = &rs->md; | |
771 | ||
92c83d79 HM |
772 | arg = dm_shift_arg(as); |
773 | if (!arg) | |
774 | return -EINVAL; | |
775 | ||
776 | if (strcmp(arg, "-")) { | |
bd83a4c4 MS |
777 | r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), |
778 | &rs->dev[i].meta_dev); | |
779 | if (r) { | |
780 | rs->ti->error = "RAID metadata device lookup failure"; | |
781 | return r; | |
782 | } | |
b12d437b JB |
783 | |
784 | rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL); | |
bd83a4c4 MS |
785 | if (!rs->dev[i].rdev.sb_page) { |
786 | rs->ti->error = "Failed to allocate superblock page"; | |
787 | return -ENOMEM; | |
788 | } | |
9d09e663 N |
789 | } |
790 | ||
92c83d79 HM |
791 | arg = dm_shift_arg(as); |
792 | if (!arg) | |
793 | return -EINVAL; | |
794 | ||
795 | if (!strcmp(arg, "-")) { | |
9d09e663 | 796 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags) && |
bd83a4c4 MS |
797 | (!rs->dev[i].rdev.recovery_offset)) { |
798 | rs->ti->error = "Drive designated for rebuild not specified"; | |
799 | return -EINVAL; | |
800 | } | |
9d09e663 | 801 | |
bd83a4c4 MS |
802 | if (rs->dev[i].meta_dev) { |
803 | rs->ti->error = "No data device supplied with metadata device"; | |
804 | return -EINVAL; | |
805 | } | |
b12d437b | 806 | |
9d09e663 N |
807 | continue; |
808 | } | |
809 | ||
bd83a4c4 MS |
810 | r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), |
811 | &rs->dev[i].data_dev); | |
812 | if (r) { | |
813 | rs->ti->error = "RAID device lookup failure"; | |
814 | return r; | |
815 | } | |
9d09e663 | 816 | |
b12d437b JB |
817 | if (rs->dev[i].meta_dev) { |
818 | metadata_available = 1; | |
819 | rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev; | |
820 | } | |
9d09e663 | 821 | rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev; |
3a1c1ef2 | 822 | list_add_tail(&rs->dev[i].rdev.same_set, &rs->md.disks); |
9d09e663 N |
823 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags)) |
824 | rebuild++; | |
825 | } | |
826 | ||
827 | if (metadata_available) { | |
828 | rs->md.external = 0; | |
829 | rs->md.persistent = 1; | |
830 | rs->md.major_version = 2; | |
831 | } else if (rebuild && !rs->md.recovery_cp) { | |
832 | /* | |
833 | * Without metadata, we will not be able to tell if the array | |
834 | * is in-sync or not - we must assume it is not. Therefore, | |
835 | * it is impossible to rebuild a drive. | |
836 | * | |
837 | * Even if there is metadata, the on-disk information may | |
838 | * indicate that the array is not in-sync and it will then | |
839 | * fail at that time. | |
840 | * | |
841 | * User could specify 'nosync' option if desperate. | |
842 | */ | |
bd83a4c4 MS |
843 | rs->ti->error = "Unable to rebuild drive while array is not in-sync"; |
844 | return -EINVAL; | |
9d09e663 N |
845 | } |
846 | ||
847 | return 0; | |
848 | } | |
849 | ||
c1084561 JB |
850 | /* |
851 | * validate_region_size | |
852 | * @rs | |
853 | * @region_size: region size in sectors. If 0, pick a size (4MiB default). | |
854 | * | |
855 | * Set rs->md.bitmap_info.chunksize (which really refers to 'region size'). | |
856 | * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap. | |
857 | * | |
858 | * Returns: 0 on success, -EINVAL on failure. | |
859 | */ | |
860 | static int validate_region_size(struct raid_set *rs, unsigned long region_size) | |
861 | { | |
862 | unsigned long min_region_size = rs->ti->len / (1 << 21); | |
863 | ||
864 | if (!region_size) { | |
865 | /* | |
43157840 | 866 | * Choose a reasonable default. All figures in sectors. |
c1084561 JB |
867 | */ |
868 | if (min_region_size > (1 << 13)) { | |
3a0f9aae | 869 | /* If not a power of 2, make it the next power of 2 */ |
042745ee | 870 | region_size = roundup_pow_of_two(min_region_size); |
c1084561 JB |
871 | DMINFO("Choosing default region size of %lu sectors", |
872 | region_size); | |
c1084561 JB |
873 | } else { |
874 | DMINFO("Choosing default region size of 4MiB"); | |
875 | region_size = 1 << 13; /* sectors */ | |
876 | } | |
877 | } else { | |
878 | /* | |
879 | * Validate user-supplied value. | |
880 | */ | |
bd83a4c4 MS |
881 | if (region_size > rs->ti->len) { |
882 | rs->ti->error = "Supplied region size is too large"; | |
883 | return -EINVAL; | |
884 | } | |
c1084561 JB |
885 | |
886 | if (region_size < min_region_size) { | |
887 | DMERR("Supplied region_size (%lu sectors) below minimum (%lu)", | |
888 | region_size, min_region_size); | |
bd83a4c4 MS |
889 | rs->ti->error = "Supplied region size is too small"; |
890 | return -EINVAL; | |
c1084561 JB |
891 | } |
892 | ||
bd83a4c4 MS |
893 | if (!is_power_of_2(region_size)) { |
894 | rs->ti->error = "Region size is not a power of 2"; | |
895 | return -EINVAL; | |
896 | } | |
c1084561 | 897 | |
bd83a4c4 MS |
898 | if (region_size < rs->md.chunk_sectors) { |
899 | rs->ti->error = "Region size is smaller than the chunk size"; | |
900 | return -EINVAL; | |
901 | } | |
c1084561 JB |
902 | } |
903 | ||
904 | /* | |
905 | * Convert sectors to bytes. | |
906 | */ | |
89d3d9a1 | 907 | rs->md.bitmap_info.chunksize = to_bytes(region_size); |
c1084561 JB |
908 | |
909 | return 0; | |
910 | } | |
911 | ||
eb649123 | 912 | /* |
55ebbb59 | 913 | * validate_raid_redundancy |
eb649123 JB |
914 | * @rs |
915 | * | |
55ebbb59 JB |
916 | * Determine if there are enough devices in the array that haven't |
917 | * failed (or are being rebuilt) to form a usable array. | |
eb649123 JB |
918 | * |
919 | * Returns: 0 on success, -EINVAL on failure. | |
920 | */ | |
55ebbb59 | 921 | static int validate_raid_redundancy(struct raid_set *rs) |
eb649123 | 922 | { |
094f394d HM |
923 | unsigned int i, rebuild_cnt = 0; |
924 | unsigned int rebuilds_per_group = 0, copies; | |
925 | unsigned int group_size, last_group_start; | |
eb649123 | 926 | |
eb649123 | 927 | for (i = 0; i < rs->md.raid_disks; i++) |
55ebbb59 JB |
928 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags) || |
929 | !rs->dev[i].rdev.sb_page) | |
eb649123 JB |
930 | rebuild_cnt++; |
931 | ||
932 | switch (rs->raid_type->level) { | |
933 | case 1: | |
934 | if (rebuild_cnt >= rs->md.raid_disks) | |
935 | goto too_many; | |
936 | break; | |
937 | case 4: | |
938 | case 5: | |
939 | case 6: | |
940 | if (rebuild_cnt > rs->raid_type->parity_devs) | |
941 | goto too_many; | |
942 | break; | |
943 | case 10: | |
9dbd1aa3 | 944 | copies = raid10_md_layout_to_copies(rs->md.new_layout); |
4ec1e369 JB |
945 | if (rebuild_cnt < copies) |
946 | break; | |
947 | ||
948 | /* | |
949 | * It is possible to have a higher rebuild count for RAID10, | |
950 | * as long as the failed devices occur in different mirror | |
951 | * groups (i.e. different stripes). | |
952 | * | |
4ec1e369 JB |
953 | * When checking "near" format, make sure no adjacent devices |
954 | * have failed beyond what can be handled. In addition to the | |
955 | * simple case where the number of devices is a multiple of the | |
956 | * number of copies, we must also handle cases where the number | |
957 | * of devices is not a multiple of the number of copies. | |
43157840 MS |
958 | * E.g. dev1 dev2 dev3 dev4 dev5 |
959 | * A A B B C | |
960 | * C D D E E | |
4ec1e369 | 961 | */ |
9dbd1aa3 | 962 | if (__is_raid10_near(rs->md.new_layout)) { |
ffeeac75 | 963 | for (i = 0; i < rs->md.raid_disks; i++) { |
fe5d2f4a JB |
964 | if (!(i % copies)) |
965 | rebuilds_per_group = 0; | |
9dbd1aa3 | 966 | if ((!rs->dev[i].rdev.sb_page || |
40ba37e5 | 967 | !test_bit(In_sync, &rs->dev[i].rdev.flags)) && |
fe5d2f4a JB |
968 | (++rebuilds_per_group >= copies)) |
969 | goto too_many; | |
970 | } | |
971 | break; | |
972 | } | |
973 | ||
974 | /* | |
975 | * When checking "far" and "offset" formats, we need to ensure | |
976 | * that the device that holds its copy is not also dead or | |
977 | * being rebuilt. (Note that "far" and "offset" formats only | |
978 | * support two copies right now. These formats also only ever | |
979 | * use the 'use_far_sets' variant.) | |
980 | * | |
981 | * This check is somewhat complicated by the need to account | |
43157840 | 982 | * for arrays that are not a multiple of (far) copies. This |
fe5d2f4a JB |
983 | * results in the need to treat the last (potentially larger) |
984 | * set differently. | |
985 | */ | |
986 | group_size = (rs->md.raid_disks / copies); | |
987 | last_group_start = (rs->md.raid_disks / group_size) - 1; | |
988 | last_group_start *= group_size; | |
989 | for (i = 0; i < rs->md.raid_disks; i++) { | |
990 | if (!(i % copies) && !(i > last_group_start)) | |
55ebbb59 | 991 | rebuilds_per_group = 0; |
fe5d2f4a JB |
992 | if ((!rs->dev[i].rdev.sb_page || |
993 | !test_bit(In_sync, &rs->dev[i].rdev.flags)) && | |
4ec1e369 | 994 | (++rebuilds_per_group >= copies)) |
fe5d2f4a | 995 | goto too_many; |
4ec1e369 JB |
996 | } |
997 | break; | |
eb649123 | 998 | default: |
55ebbb59 JB |
999 | if (rebuild_cnt) |
1000 | return -EINVAL; | |
eb649123 JB |
1001 | } |
1002 | ||
1003 | return 0; | |
1004 | ||
1005 | too_many: | |
eb649123 JB |
1006 | return -EINVAL; |
1007 | } | |
1008 | ||
9d09e663 N |
1009 | /* |
1010 | * Possible arguments are... | |
9d09e663 N |
1011 | * <chunk_size> [optional_args] |
1012 | * | |
32737279 JB |
1013 | * Argument definitions |
1014 | * <chunk_size> The number of sectors per disk that | |
43157840 | 1015 | * will form the "stripe" |
32737279 | 1016 | * [[no]sync] Force or prevent recovery of the |
43157840 | 1017 | * entire array |
9d09e663 | 1018 | * [rebuild <idx>] Rebuild the drive indicated by the index |
32737279 | 1019 | * [daemon_sleep <ms>] Time between bitmap daemon work to |
43157840 | 1020 | * clear bits |
9d09e663 N |
1021 | * [min_recovery_rate <kB/sec/disk>] Throttle RAID initialization |
1022 | * [max_recovery_rate <kB/sec/disk>] Throttle RAID initialization | |
46bed2b5 | 1023 | * [write_mostly <idx>] Indicate a write mostly drive via index |
9d09e663 N |
1024 | * [max_write_behind <sectors>] See '-write-behind=' (man mdadm) |
1025 | * [stripe_cache <sectors>] Stripe cache size for higher RAIDs | |
43157840 | 1026 | * [region_size <sectors>] Defines granularity of bitmap |
63f33b8d JB |
1027 | * |
1028 | * RAID10-only options: | |
43157840 | 1029 | * [raid10_copies <# copies>] Number of copies. (Default: 2) |
fe5d2f4a | 1030 | * [raid10_format <near|far|offset>] Layout algorithm. (Default: near) |
9d09e663 | 1031 | */ |
92c83d79 | 1032 | static int parse_raid_params(struct raid_set *rs, struct dm_arg_set *as, |
094f394d | 1033 | unsigned int num_raid_params) |
9d09e663 | 1034 | { |
9dbd1aa3 | 1035 | int value, raid10_format = ALGORITHM_RAID10_DEFAULT; |
094f394d HM |
1036 | unsigned int raid10_copies = 2; |
1037 | unsigned int i, write_mostly = 0; | |
1038 | unsigned int region_size = 0; | |
542f9038 | 1039 | sector_t max_io_len; |
92c83d79 | 1040 | const char *arg, *key; |
702108d1 | 1041 | struct raid_dev *rd; |
33e53f06 | 1042 | struct raid_type *rt = rs->raid_type; |
92c83d79 HM |
1043 | |
1044 | arg = dm_shift_arg(as); | |
1045 | num_raid_params--; /* Account for chunk_size argument */ | |
1046 | ||
9dbd1aa3 | 1047 | if (kstrtoint(arg, 10, &value) < 0) { |
bd83a4c4 MS |
1048 | rs->ti->error = "Bad numerical argument given for chunk_size"; |
1049 | return -EINVAL; | |
1050 | } | |
9d09e663 N |
1051 | |
1052 | /* | |
1053 | * First, parse the in-order required arguments | |
32737279 | 1054 | * "chunk_size" is the only argument of this type. |
9d09e663 | 1055 | */ |
33e53f06 | 1056 | if (rt_is_raid1(rt)) { |
32737279 JB |
1057 | if (value) |
1058 | DMERR("Ignoring chunk size parameter for RAID 1"); | |
1059 | value = 0; | |
bd83a4c4 MS |
1060 | } else if (!is_power_of_2(value)) { |
1061 | rs->ti->error = "Chunk size must be a power of 2"; | |
1062 | return -EINVAL; | |
1063 | } else if (value < 8) { | |
1064 | rs->ti->error = "Chunk size value is too small"; | |
1065 | return -EINVAL; | |
1066 | } | |
9d09e663 N |
1067 | |
1068 | rs->md.new_chunk_sectors = rs->md.chunk_sectors = value; | |
9d09e663 N |
1069 | |
1070 | /* | |
b12d437b JB |
1071 | * We set each individual device as In_sync with a completed |
1072 | * 'recovery_offset'. If there has been a device failure or | |
1073 | * replacement then one of the following cases applies: | |
1074 | * | |
1075 | * 1) User specifies 'rebuild'. | |
43157840 | 1076 | * - Device is reset when param is read. |
b12d437b | 1077 | * 2) A new device is supplied. |
43157840 | 1078 | * - No matching superblock found, resets device. |
b12d437b | 1079 | * 3) Device failure was transient and returns on reload. |
43157840 | 1080 | * - Failure noticed, resets device for bitmap replay. |
b12d437b | 1081 | * 4) Device hadn't completed recovery after previous failure. |
43157840 | 1082 | * - Superblock is read and overrides recovery_offset. |
b12d437b JB |
1083 | * |
1084 | * What is found in the superblocks of the devices is always | |
1085 | * authoritative, unless 'rebuild' or '[no]sync' was specified. | |
9d09e663 | 1086 | */ |
ffeeac75 | 1087 | for (i = 0; i < rs->raid_disks; i++) { |
9d09e663 | 1088 | set_bit(In_sync, &rs->dev[i].rdev.flags); |
b12d437b JB |
1089 | rs->dev[i].rdev.recovery_offset = MaxSector; |
1090 | } | |
9d09e663 | 1091 | |
b12d437b JB |
1092 | /* |
1093 | * Second, parse the unordered optional arguments | |
1094 | */ | |
9d09e663 | 1095 | for (i = 0; i < num_raid_params; i++) { |
4763e543 | 1096 | key = dm_shift_arg(as); |
bd83a4c4 MS |
1097 | if (!key) { |
1098 | rs->ti->error = "Not enough raid parameters given"; | |
1099 | return -EINVAL; | |
1100 | } | |
92c83d79 | 1101 | |
3fa6cf38 | 1102 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC))) { |
4286325b | 1103 | if (test_and_set_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { |
bd83a4c4 MS |
1104 | rs->ti->error = "Only one 'nosync' argument allowed"; |
1105 | return -EINVAL; | |
1106 | } | |
9d09e663 N |
1107 | continue; |
1108 | } | |
3fa6cf38 | 1109 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_SYNC))) { |
4286325b | 1110 | if (test_and_set_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) { |
bd83a4c4 MS |
1111 | rs->ti->error = "Only one 'sync' argument allowed"; |
1112 | return -EINVAL; | |
1113 | } | |
4763e543 HM |
1114 | continue; |
1115 | } | |
3fa6cf38 | 1116 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_USE_NEAR_SETS))) { |
4286325b | 1117 | if (test_and_set_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1118 | rs->ti->error = "Only one 'raid10_use_new_sets' argument allowed"; |
1119 | return -EINVAL; | |
1120 | } | |
9d09e663 N |
1121 | continue; |
1122 | } | |
1123 | ||
92c83d79 HM |
1124 | arg = dm_shift_arg(as); |
1125 | i++; /* Account for the argument pairs */ | |
bd83a4c4 MS |
1126 | if (!arg) { |
1127 | rs->ti->error = "Wrong number of raid parameters given"; | |
1128 | return -EINVAL; | |
1129 | } | |
63f33b8d | 1130 | |
702108d1 HM |
1131 | /* |
1132 | * Parameters that take a string value are checked here. | |
1133 | */ | |
1134 | ||
3fa6cf38 | 1135 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT))) { |
4286325b | 1136 | if (test_and_set_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) { |
bd83a4c4 MS |
1137 | rs->ti->error = "Only one 'raid10_format' argument pair allowed"; |
1138 | return -EINVAL; | |
1139 | } | |
1140 | if (!rt_is_raid10(rt)) { | |
1141 | rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type"; | |
1142 | return -EINVAL; | |
1143 | } | |
33e53f06 | 1144 | raid10_format = raid10_name_to_format(arg); |
bd83a4c4 MS |
1145 | if (raid10_format < 0) { |
1146 | rs->ti->error = "Invalid 'raid10_format' value given"; | |
1147 | return raid10_format; | |
1148 | } | |
63f33b8d JB |
1149 | continue; |
1150 | } | |
1151 | ||
9dbd1aa3 | 1152 | if (kstrtoint(arg, 10, &value) < 0) { |
bd83a4c4 MS |
1153 | rs->ti->error = "Bad numerical argument given in raid params"; |
1154 | return -EINVAL; | |
1155 | } | |
702108d1 | 1156 | |
3fa6cf38 | 1157 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD))) { |
702108d1 HM |
1158 | /* |
1159 | * "rebuild" is being passed in by userspace to provide | |
1160 | * indexes of replaced devices and to set up additional | |
1161 | * devices on raid level takeover. | |
43157840 | 1162 | */ |
bb91a63f | 1163 | if (!__within_range(value, 0, rs->raid_disks - 1)) { |
bd83a4c4 MS |
1164 | rs->ti->error = "Invalid rebuild index given"; |
1165 | return -EINVAL; | |
1166 | } | |
702108d1 | 1167 | |
bd83a4c4 MS |
1168 | if (test_and_set_bit(value, (void *) rs->rebuild_disks)) { |
1169 | rs->ti->error = "rebuild for this index already given"; | |
1170 | return -EINVAL; | |
1171 | } | |
ecbfb9f1 | 1172 | |
702108d1 HM |
1173 | rd = rs->dev + value; |
1174 | clear_bit(In_sync, &rd->rdev.flags); | |
1175 | clear_bit(Faulty, &rd->rdev.flags); | |
1176 | rd->rdev.recovery_offset = 0; | |
4286325b | 1177 | set_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags); |
3fa6cf38 | 1178 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY))) { |
bd83a4c4 MS |
1179 | if (!rt_is_raid1(rt)) { |
1180 | rs->ti->error = "write_mostly option is only valid for RAID1"; | |
1181 | return -EINVAL; | |
1182 | } | |
702108d1 | 1183 | |
bb91a63f | 1184 | if (!__within_range(value, 0, rs->md.raid_disks - 1)) { |
bd83a4c4 MS |
1185 | rs->ti->error = "Invalid write_mostly index given"; |
1186 | return -EINVAL; | |
1187 | } | |
9d09e663 | 1188 | |
5fa146b2 | 1189 | write_mostly++; |
46bed2b5 | 1190 | set_bit(WriteMostly, &rs->dev[value].rdev.flags); |
4286325b | 1191 | set_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags); |
3fa6cf38 | 1192 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND))) { |
bd83a4c4 MS |
1193 | if (!rt_is_raid1(rt)) { |
1194 | rs->ti->error = "max_write_behind option is only valid for RAID1"; | |
1195 | return -EINVAL; | |
1196 | } | |
702108d1 | 1197 | |
4286325b | 1198 | if (test_and_set_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) { |
bd83a4c4 MS |
1199 | rs->ti->error = "Only one max_write_behind argument pair allowed"; |
1200 | return -EINVAL; | |
1201 | } | |
9d09e663 N |
1202 | |
1203 | /* | |
1204 | * In device-mapper, we specify things in sectors, but | |
1205 | * MD records this value in kB | |
1206 | */ | |
1207 | value /= 2; | |
bd83a4c4 MS |
1208 | if (value > COUNTER_MAX) { |
1209 | rs->ti->error = "Max write-behind limit out of range"; | |
1210 | return -EINVAL; | |
1211 | } | |
702108d1 | 1212 | |
9d09e663 | 1213 | rs->md.bitmap_info.max_write_behind = value; |
3fa6cf38 | 1214 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP))) { |
4286325b | 1215 | if (test_and_set_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) { |
bd83a4c4 MS |
1216 | rs->ti->error = "Only one daemon_sleep argument pair allowed"; |
1217 | return -EINVAL; | |
1218 | } | |
1219 | if (!value || (value > MAX_SCHEDULE_TIMEOUT)) { | |
1220 | rs->ti->error = "daemon sleep period out of range"; | |
1221 | return -EINVAL; | |
1222 | } | |
9d09e663 | 1223 | rs->md.bitmap_info.daemon_sleep = value; |
3fa6cf38 | 1224 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET))) { |
4763e543 | 1225 | /* Userspace passes new data_offset after having extended the the data image LV */ |
4286325b | 1226 | if (test_and_set_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { |
bd83a4c4 MS |
1227 | rs->ti->error = "Only one data_offset argument pair allowed"; |
1228 | return -EINVAL; | |
1229 | } | |
4763e543 | 1230 | /* Ensure sensible data offset */ |
75dd3b9e HM |
1231 | if (value < 0 || |
1232 | (value && (value < MIN_FREE_RESHAPE_SPACE || value % to_sector(PAGE_SIZE)))) { | |
bd83a4c4 MS |
1233 | rs->ti->error = "Bogus data_offset value"; |
1234 | return -EINVAL; | |
1235 | } | |
4763e543 | 1236 | rs->data_offset = value; |
3fa6cf38 | 1237 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS))) { |
4763e543 | 1238 | /* Define the +/-# of disks to add to/remove from the given raid set */ |
4286325b | 1239 | if (test_and_set_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1240 | rs->ti->error = "Only one delta_disks argument pair allowed"; |
1241 | return -EINVAL; | |
1242 | } | |
4763e543 | 1243 | /* Ensure MAX_RAID_DEVICES and raid type minimal_devs! */ |
bb91a63f | 1244 | if (!__within_range(abs(value), 1, MAX_RAID_DEVICES - rt->minimal_devs)) { |
bd83a4c4 MS |
1245 | rs->ti->error = "Too many delta_disk requested"; |
1246 | return -EINVAL; | |
1247 | } | |
4763e543 HM |
1248 | |
1249 | rs->delta_disks = value; | |
3fa6cf38 | 1250 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE))) { |
4286325b | 1251 | if (test_and_set_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1252 | rs->ti->error = "Only one stripe_cache argument pair allowed"; |
1253 | return -EINVAL; | |
1254 | } | |
1255 | ||
bd83a4c4 MS |
1256 | if (!rt_is_raid456(rt)) { |
1257 | rs->ti->error = "Inappropriate argument: stripe_cache"; | |
1258 | return -EINVAL; | |
1259 | } | |
702108d1 | 1260 | |
9dbd1aa3 | 1261 | rs->stripe_cache_entries = value; |
3fa6cf38 | 1262 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE))) { |
4286325b | 1263 | if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1264 | rs->ti->error = "Only one min_recovery_rate argument pair allowed"; |
1265 | return -EINVAL; | |
1266 | } | |
1267 | if (value > INT_MAX) { | |
1268 | rs->ti->error = "min_recovery_rate out of range"; | |
1269 | return -EINVAL; | |
1270 | } | |
9d09e663 | 1271 | rs->md.sync_speed_min = (int)value; |
3fa6cf38 | 1272 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE))) { |
f15f64d6 | 1273 | if (test_and_set_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1274 | rs->ti->error = "Only one max_recovery_rate argument pair allowed"; |
1275 | return -EINVAL; | |
1276 | } | |
1277 | if (value > INT_MAX) { | |
1278 | rs->ti->error = "max_recovery_rate out of range"; | |
1279 | return -EINVAL; | |
1280 | } | |
9d09e663 | 1281 | rs->md.sync_speed_max = (int)value; |
3fa6cf38 | 1282 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE))) { |
4286325b | 1283 | if (test_and_set_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1284 | rs->ti->error = "Only one region_size argument pair allowed"; |
1285 | return -EINVAL; | |
1286 | } | |
702108d1 | 1287 | |
c1084561 | 1288 | region_size = value; |
4257e085 | 1289 | rs->requested_bitmap_chunk_sectors = value; |
3fa6cf38 | 1290 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES))) { |
4286325b | 1291 | if (test_and_set_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) { |
bd83a4c4 MS |
1292 | rs->ti->error = "Only one raid10_copies argument pair allowed"; |
1293 | return -EINVAL; | |
1294 | } | |
702108d1 | 1295 | |
bb91a63f | 1296 | if (!__within_range(value, 2, rs->md.raid_disks)) { |
bd83a4c4 MS |
1297 | rs->ti->error = "Bad value for 'raid10_copies'"; |
1298 | return -EINVAL; | |
1299 | } | |
702108d1 | 1300 | |
63f33b8d | 1301 | raid10_copies = value; |
9d09e663 N |
1302 | } else { |
1303 | DMERR("Unable to parse RAID parameter: %s", key); | |
bd83a4c4 MS |
1304 | rs->ti->error = "Unable to parse RAID parameter"; |
1305 | return -EINVAL; | |
9d09e663 N |
1306 | } |
1307 | } | |
1308 | ||
0d851d14 HM |
1309 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) && |
1310 | test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { | |
1311 | rs->ti->error = "sync and nosync are mutually exclusive"; | |
1312 | return -EINVAL; | |
1313 | } | |
1314 | ||
37f10be1 HM |
1315 | if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) && |
1316 | (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) || | |
1317 | test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags))) { | |
1318 | rs->ti->error = "sync/nosync and rebuild are mutually exclusive"; | |
1319 | return -EINVAL; | |
1320 | } | |
1321 | ||
5fa146b2 HM |
1322 | if (write_mostly >= rs->md.raid_disks) { |
1323 | rs->ti->error = "Can't set all raid1 devices to write_mostly"; | |
1324 | return -EINVAL; | |
1325 | } | |
1326 | ||
c1084561 JB |
1327 | if (validate_region_size(rs, region_size)) |
1328 | return -EINVAL; | |
1329 | ||
1330 | if (rs->md.chunk_sectors) | |
542f9038 | 1331 | max_io_len = rs->md.chunk_sectors; |
c1084561 | 1332 | else |
542f9038 | 1333 | max_io_len = region_size; |
c1084561 | 1334 | |
542f9038 MS |
1335 | if (dm_set_target_max_io_len(rs->ti, max_io_len)) |
1336 | return -EINVAL; | |
32737279 | 1337 | |
33e53f06 | 1338 | if (rt_is_raid10(rt)) { |
bd83a4c4 MS |
1339 | if (raid10_copies > rs->md.raid_disks) { |
1340 | rs->ti->error = "Not enough devices to satisfy specification"; | |
1341 | return -EINVAL; | |
1342 | } | |
63f33b8d | 1343 | |
33e53f06 | 1344 | rs->md.new_layout = raid10_format_to_md_layout(rs, raid10_format, raid10_copies); |
bd83a4c4 MS |
1345 | if (rs->md.new_layout < 0) { |
1346 | rs->ti->error = "Error getting raid10 format"; | |
1347 | return rs->md.new_layout; | |
1348 | } | |
33e53f06 HM |
1349 | |
1350 | rt = get_raid_type_by_ll(10, rs->md.new_layout); | |
bd83a4c4 MS |
1351 | if (!rt) { |
1352 | rs->ti->error = "Failed to recognize new raid10 layout"; | |
1353 | return -EINVAL; | |
1354 | } | |
33e53f06 HM |
1355 | |
1356 | if ((rt->algorithm == ALGORITHM_RAID10_DEFAULT || | |
1357 | rt->algorithm == ALGORITHM_RAID10_NEAR) && | |
4286325b | 1358 | test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1359 | rs->ti->error = "RAID10 format 'near' and 'raid10_use_near_sets' are incompatible"; |
1360 | return -EINVAL; | |
1361 | } | |
bd83a4c4 | 1362 | } |
702108d1 | 1363 | |
33e53f06 | 1364 | rs->raid10_copies = raid10_copies; |
c039c332 | 1365 | |
9d09e663 N |
1366 | /* Assume there are no metadata devices until the drives are parsed */ |
1367 | rs->md.persistent = 0; | |
1368 | rs->md.external = 1; | |
1369 | ||
f090279e | 1370 | /* Check, if any invalid ctr arguments have been passed in for the raid level */ |
a30cbc0d | 1371 | return rs_check_for_valid_flags(rs); |
9d09e663 N |
1372 | } |
1373 | ||
9dbd1aa3 HM |
1374 | /* Set raid4/5/6 cache size */ |
1375 | static int rs_set_raid456_stripe_cache(struct raid_set *rs) | |
1376 | { | |
1377 | int r; | |
1378 | struct r5conf *conf; | |
1379 | struct mddev *mddev = &rs->md; | |
1380 | uint32_t min_stripes = max(mddev->chunk_sectors, mddev->new_chunk_sectors) / 2; | |
1381 | uint32_t nr_stripes = rs->stripe_cache_entries; | |
1382 | ||
1383 | if (!rt_is_raid456(rs->raid_type)) { | |
1384 | rs->ti->error = "Inappropriate raid level; cannot change stripe_cache size"; | |
1385 | return -EINVAL; | |
1386 | } | |
1387 | ||
1388 | if (nr_stripes < min_stripes) { | |
1389 | DMINFO("Adjusting requested %u stripe cache entries to %u to suit stripe size", | |
1390 | nr_stripes, min_stripes); | |
1391 | nr_stripes = min_stripes; | |
1392 | } | |
1393 | ||
1394 | conf = mddev->private; | |
1395 | if (!conf) { | |
1396 | rs->ti->error = "Cannot change stripe_cache size on inactive RAID set"; | |
1397 | return -EINVAL; | |
1398 | } | |
1399 | ||
1400 | /* Try setting number of stripes in raid456 stripe cache */ | |
1401 | if (conf->min_nr_stripes != nr_stripes) { | |
1402 | r = raid5_set_cache_size(mddev, nr_stripes); | |
1403 | if (r) { | |
1404 | rs->ti->error = "Failed to set raid4/5/6 stripe cache size"; | |
1405 | return r; | |
1406 | } | |
1407 | ||
1408 | DMINFO("%u stripe cache entries", nr_stripes); | |
1409 | } | |
1410 | ||
1411 | return 0; | |
1412 | } | |
1413 | ||
3a1c1ef2 HM |
1414 | /* Return # of data stripes as kept in mddev as of @rs (i.e. as of superblock) */ |
1415 | static unsigned int mddev_data_stripes(struct raid_set *rs) | |
1416 | { | |
1417 | return rs->md.raid_disks - rs->raid_type->parity_devs; | |
1418 | } | |
1419 | ||
40ba37e5 HM |
1420 | /* Return # of data stripes of @rs (i.e. as of ctr) */ |
1421 | static unsigned int rs_data_stripes(struct raid_set *rs) | |
1422 | { | |
1423 | return rs->raid_disks - rs->raid_type->parity_devs; | |
1424 | } | |
1425 | ||
1426 | /* Calculate the sectors per device and per array used for @rs */ | |
1427 | static int rs_set_dev_and_array_sectors(struct raid_set *rs, bool use_mddev) | |
1428 | { | |
1429 | int delta_disks; | |
1430 | unsigned int data_stripes; | |
1431 | struct mddev *mddev = &rs->md; | |
1432 | struct md_rdev *rdev; | |
1433 | sector_t array_sectors = rs->ti->len, dev_sectors = rs->ti->len; | |
1434 | ||
1435 | if (use_mddev) { | |
1436 | delta_disks = mddev->delta_disks; | |
1437 | data_stripes = mddev_data_stripes(rs); | |
1438 | } else { | |
1439 | delta_disks = rs->delta_disks; | |
1440 | data_stripes = rs_data_stripes(rs); | |
1441 | } | |
1442 | ||
1443 | /* Special raid1 case w/o delta_disks support (yet) */ | |
1444 | if (rt_is_raid1(rs->raid_type)) | |
1445 | ; | |
1446 | else if (rt_is_raid10(rs->raid_type)) { | |
1447 | if (rs->raid10_copies < 2 || | |
1448 | delta_disks < 0) { | |
1449 | rs->ti->error = "Bogus raid10 data copies or delta disks"; | |
094f394d | 1450 | return -EINVAL; |
40ba37e5 HM |
1451 | } |
1452 | ||
1453 | dev_sectors *= rs->raid10_copies; | |
1454 | if (sector_div(dev_sectors, data_stripes)) | |
1455 | goto bad; | |
1456 | ||
1457 | array_sectors = (data_stripes + delta_disks) * dev_sectors; | |
1458 | if (sector_div(array_sectors, rs->raid10_copies)) | |
1459 | goto bad; | |
1460 | ||
1461 | } else if (sector_div(dev_sectors, data_stripes)) | |
1462 | goto bad; | |
1463 | ||
1464 | else | |
1465 | /* Striped layouts */ | |
1466 | array_sectors = (data_stripes + delta_disks) * dev_sectors; | |
1467 | ||
1468 | rdev_for_each(rdev, mddev) | |
1469 | rdev->sectors = dev_sectors; | |
1470 | ||
1471 | mddev->array_sectors = array_sectors; | |
1472 | mddev->dev_sectors = dev_sectors; | |
1473 | ||
1474 | return 0; | |
1475 | bad: | |
1476 | rs->ti->error = "Target length not divisible by number of data devices"; | |
094f394d | 1477 | return -EINVAL; |
40ba37e5 HM |
1478 | } |
1479 | ||
4dff2f1e HM |
1480 | /* Setup recovery on @rs */ |
1481 | static void __rs_setup_recovery(struct raid_set *rs, sector_t dev_sectors) | |
1482 | { | |
1483 | /* raid0 does not recover */ | |
1484 | if (rs_is_raid0(rs)) | |
1485 | rs->md.recovery_cp = MaxSector; | |
1486 | /* | |
1487 | * A raid6 set has to be recovered either | |
1488 | * completely or for the grown part to | |
1489 | * ensure proper parity and Q-Syndrome | |
1490 | */ | |
1491 | else if (rs_is_raid6(rs)) | |
1492 | rs->md.recovery_cp = dev_sectors; | |
1493 | /* | |
1494 | * Other raid set types may skip recovery | |
1495 | * depending on the 'nosync' flag. | |
1496 | */ | |
1497 | else | |
1498 | rs->md.recovery_cp = test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags) | |
1499 | ? MaxSector : dev_sectors; | |
1500 | } | |
1501 | ||
1502 | /* Setup recovery on @rs based on raid type, device size and 'nosync' flag */ | |
1503 | static void rs_setup_recovery(struct raid_set *rs, sector_t dev_sectors) | |
1504 | { | |
1505 | if (!dev_sectors) | |
1506 | /* New raid set or 'sync' flag provided */ | |
1507 | __rs_setup_recovery(rs, 0); | |
1508 | else if (dev_sectors == MaxSector) | |
1509 | /* Prevent recovery */ | |
1510 | __rs_setup_recovery(rs, MaxSector); | |
1511 | else if (rs->dev[0].rdev.sectors < dev_sectors) | |
1512 | /* Grown raid set */ | |
1513 | __rs_setup_recovery(rs, rs->dev[0].rdev.sectors); | |
1514 | else | |
1515 | __rs_setup_recovery(rs, MaxSector); | |
1516 | } | |
1517 | ||
9d09e663 N |
1518 | static void do_table_event(struct work_struct *ws) |
1519 | { | |
1520 | struct raid_set *rs = container_of(ws, struct raid_set, md.event_work); | |
1521 | ||
9d9d939c HM |
1522 | smp_rmb(); /* Make sure we access most actual mddev properties */ |
1523 | if (!rs_is_reshaping(rs)) | |
1524 | rs_set_capacity(rs); | |
9d09e663 N |
1525 | dm_table_event(rs->ti->table); |
1526 | } | |
1527 | ||
1528 | static int raid_is_congested(struct dm_target_callbacks *cb, int bits) | |
1529 | { | |
1530 | struct raid_set *rs = container_of(cb, struct raid_set, callbacks); | |
1531 | ||
5c675f83 | 1532 | return mddev_congested(&rs->md, bits); |
9d09e663 N |
1533 | } |
1534 | ||
ecbfb9f1 HM |
1535 | /* |
1536 | * Make sure a valid takover (level switch) is being requested on @rs | |
1537 | * | |
1538 | * Conversions of raid sets from one MD personality to another | |
1539 | * have to conform to restrictions which are enforced here. | |
ecbfb9f1 HM |
1540 | */ |
1541 | static int rs_check_takeover(struct raid_set *rs) | |
1542 | { | |
1543 | struct mddev *mddev = &rs->md; | |
1544 | unsigned int near_copies; | |
1545 | ||
9dbd1aa3 HM |
1546 | if (rs->md.degraded) { |
1547 | rs->ti->error = "Can't takeover degraded raid set"; | |
1548 | return -EPERM; | |
1549 | } | |
1550 | ||
1551 | if (rs_is_reshaping(rs)) { | |
1552 | rs->ti->error = "Can't takeover reshaping raid set"; | |
1553 | return -EPERM; | |
1554 | } | |
1555 | ||
ecbfb9f1 HM |
1556 | switch (mddev->level) { |
1557 | case 0: | |
1558 | /* raid0 -> raid1/5 with one disk */ | |
1559 | if ((mddev->new_level == 1 || mddev->new_level == 5) && | |
1560 | mddev->raid_disks == 1) | |
1561 | return 0; | |
1562 | ||
1563 | /* raid0 -> raid10 */ | |
1564 | if (mddev->new_level == 10 && | |
9dbd1aa3 | 1565 | !(rs->raid_disks % mddev->raid_disks)) |
ecbfb9f1 HM |
1566 | return 0; |
1567 | ||
1568 | /* raid0 with multiple disks -> raid4/5/6 */ | |
bb91a63f | 1569 | if (__within_range(mddev->new_level, 4, 6) && |
ecbfb9f1 HM |
1570 | mddev->new_layout == ALGORITHM_PARITY_N && |
1571 | mddev->raid_disks > 1) | |
1572 | return 0; | |
1573 | ||
1574 | break; | |
1575 | ||
1576 | case 10: | |
1577 | /* Can't takeover raid10_offset! */ | |
e6ca5e1a | 1578 | if (__is_raid10_offset(mddev->layout)) |
ecbfb9f1 HM |
1579 | break; |
1580 | ||
e6ca5e1a | 1581 | near_copies = __raid10_near_copies(mddev->layout); |
ecbfb9f1 HM |
1582 | |
1583 | /* raid10* -> raid0 */ | |
1584 | if (mddev->new_level == 0) { | |
1585 | /* Can takeover raid10_near with raid disks divisable by data copies! */ | |
1586 | if (near_copies > 1 && | |
1587 | !(mddev->raid_disks % near_copies)) { | |
1588 | mddev->raid_disks /= near_copies; | |
1589 | mddev->delta_disks = mddev->raid_disks; | |
1590 | return 0; | |
1591 | } | |
1592 | ||
1593 | /* Can takeover raid10_far */ | |
1594 | if (near_copies == 1 && | |
e6ca5e1a | 1595 | __raid10_far_copies(mddev->layout) > 1) |
ecbfb9f1 HM |
1596 | return 0; |
1597 | ||
1598 | break; | |
1599 | } | |
1600 | ||
1601 | /* raid10_{near,far} -> raid1 */ | |
1602 | if (mddev->new_level == 1 && | |
e6ca5e1a | 1603 | max(near_copies, __raid10_far_copies(mddev->layout)) == mddev->raid_disks) |
ecbfb9f1 HM |
1604 | return 0; |
1605 | ||
1606 | /* raid10_{near,far} with 2 disks -> raid4/5 */ | |
bb91a63f | 1607 | if (__within_range(mddev->new_level, 4, 5) && |
ecbfb9f1 HM |
1608 | mddev->raid_disks == 2) |
1609 | return 0; | |
1610 | break; | |
1611 | ||
1612 | case 1: | |
1613 | /* raid1 with 2 disks -> raid4/5 */ | |
bb91a63f | 1614 | if (__within_range(mddev->new_level, 4, 5) && |
ecbfb9f1 HM |
1615 | mddev->raid_disks == 2) { |
1616 | mddev->degraded = 1; | |
1617 | return 0; | |
1618 | } | |
1619 | ||
1620 | /* raid1 -> raid0 */ | |
1621 | if (mddev->new_level == 0 && | |
1622 | mddev->raid_disks == 1) | |
1623 | return 0; | |
1624 | ||
1625 | /* raid1 -> raid10 */ | |
1626 | if (mddev->new_level == 10) | |
1627 | return 0; | |
ecbfb9f1 HM |
1628 | break; |
1629 | ||
1630 | case 4: | |
1631 | /* raid4 -> raid0 */ | |
1632 | if (mddev->new_level == 0) | |
1633 | return 0; | |
1634 | ||
1635 | /* raid4 -> raid1/5 with 2 disks */ | |
1636 | if ((mddev->new_level == 1 || mddev->new_level == 5) && | |
1637 | mddev->raid_disks == 2) | |
1638 | return 0; | |
1639 | ||
1640 | /* raid4 -> raid5/6 with parity N */ | |
bb91a63f | 1641 | if (__within_range(mddev->new_level, 5, 6) && |
ecbfb9f1 HM |
1642 | mddev->layout == ALGORITHM_PARITY_N) |
1643 | return 0; | |
1644 | break; | |
1645 | ||
1646 | case 5: | |
1647 | /* raid5 with parity N -> raid0 */ | |
1648 | if (mddev->new_level == 0 && | |
1649 | mddev->layout == ALGORITHM_PARITY_N) | |
1650 | return 0; | |
1651 | ||
1652 | /* raid5 with parity N -> raid4 */ | |
1653 | if (mddev->new_level == 4 && | |
1654 | mddev->layout == ALGORITHM_PARITY_N) | |
1655 | return 0; | |
1656 | ||
1657 | /* raid5 with 2 disks -> raid1/4/10 */ | |
1658 | if ((mddev->new_level == 1 || mddev->new_level == 4 || mddev->new_level == 10) && | |
1659 | mddev->raid_disks == 2) | |
1660 | return 0; | |
1661 | ||
6ee0bae9 | 1662 | /* raid5_* -> raid6_*_6 with Q-Syndrome N (e.g. raid5_ra -> raid6_ra_6 */ |
ecbfb9f1 HM |
1663 | if (mddev->new_level == 6 && |
1664 | ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || | |
bb91a63f | 1665 | __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC_6, ALGORITHM_RIGHT_SYMMETRIC_6))) |
ecbfb9f1 HM |
1666 | return 0; |
1667 | break; | |
1668 | ||
1669 | case 6: | |
1670 | /* raid6 with parity N -> raid0 */ | |
1671 | if (mddev->new_level == 0 && | |
1672 | mddev->layout == ALGORITHM_PARITY_N) | |
1673 | return 0; | |
1674 | ||
1675 | /* raid6 with parity N -> raid4 */ | |
1676 | if (mddev->new_level == 4 && | |
1677 | mddev->layout == ALGORITHM_PARITY_N) | |
1678 | return 0; | |
1679 | ||
6ee0bae9 | 1680 | /* raid6_*_n with Q-Syndrome N -> raid5_* */ |
ecbfb9f1 HM |
1681 | if (mddev->new_level == 5 && |
1682 | ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || | |
bb91a63f | 1683 | __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC, ALGORITHM_RIGHT_SYMMETRIC))) |
ecbfb9f1 HM |
1684 | return 0; |
1685 | ||
1686 | default: | |
1687 | break; | |
1688 | } | |
1689 | ||
bd83a4c4 MS |
1690 | rs->ti->error = "takeover not possible"; |
1691 | return -EINVAL; | |
ecbfb9f1 HM |
1692 | } |
1693 | ||
1694 | /* True if @rs requested to be taken over */ | |
1695 | static bool rs_takeover_requested(struct raid_set *rs) | |
1696 | { | |
1697 | return rs->md.new_level != rs->md.level; | |
1698 | } | |
1699 | ||
40ba37e5 HM |
1700 | /* True if @rs is requested to reshape by ctr */ |
1701 | static bool rs_reshape_requested(struct raid_set *rs) | |
1702 | { | |
469b304b | 1703 | bool change; |
40ba37e5 HM |
1704 | struct mddev *mddev = &rs->md; |
1705 | ||
469b304b HM |
1706 | if (rs_takeover_requested(rs)) |
1707 | return false; | |
1708 | ||
40ba37e5 HM |
1709 | if (!mddev->level) |
1710 | return false; | |
1711 | ||
469b304b HM |
1712 | change = mddev->new_layout != mddev->layout || |
1713 | mddev->new_chunk_sectors != mddev->chunk_sectors || | |
1714 | rs->delta_disks; | |
1715 | ||
1716 | /* Historical case to support raid1 reshape without delta disks */ | |
7a7c330f HM |
1717 | if (mddev->level == 1) { |
1718 | if (rs->delta_disks) | |
1719 | return !!rs->delta_disks; | |
1720 | ||
469b304b HM |
1721 | return !change && |
1722 | mddev->raid_disks != rs->raid_disks; | |
7a7c330f | 1723 | } |
469b304b HM |
1724 | |
1725 | if (mddev->level == 10) | |
1726 | return change && | |
1727 | !__is_raid10_far(mddev->new_layout) && | |
1728 | rs->delta_disks >= 0; | |
1729 | ||
1730 | return change; | |
40ba37e5 HM |
1731 | } |
1732 | ||
33e53f06 | 1733 | /* Features */ |
9b6e5423 | 1734 | #define FEATURE_FLAG_SUPPORTS_V190 0x1 /* Supports extended superblock */ |
33e53f06 HM |
1735 | |
1736 | /* State flags for sb->flags */ | |
1737 | #define SB_FLAG_RESHAPE_ACTIVE 0x1 | |
1738 | #define SB_FLAG_RESHAPE_BACKWARDS 0x2 | |
1739 | ||
b12d437b JB |
1740 | /* |
1741 | * This structure is never routinely used by userspace, unlike md superblocks. | |
1742 | * Devices with this superblock should only ever be accessed via device-mapper. | |
1743 | */ | |
1744 | #define DM_RAID_MAGIC 0x64526D44 | |
1745 | struct dm_raid_superblock { | |
1746 | __le32 magic; /* "DmRd" */ | |
9b6e5423 | 1747 | __le32 compat_features; /* Used to indicate compatible features (like 1.9.0 ondisk metadata extension) */ |
b12d437b | 1748 | |
33e53f06 HM |
1749 | __le32 num_devices; /* Number of devices in this raid set. (Max 64) */ |
1750 | __le32 array_position; /* The position of this drive in the raid set */ | |
b12d437b JB |
1751 | |
1752 | __le64 events; /* Incremented by md when superblock updated */ | |
9b6e5423 | 1753 | __le64 failed_devices; /* Pre 1.9.0 part of bit field of devices to */ |
33e53f06 | 1754 | /* indicate failures (see extension below) */ |
b12d437b JB |
1755 | |
1756 | /* | |
1757 | * This offset tracks the progress of the repair or replacement of | |
1758 | * an individual drive. | |
1759 | */ | |
1760 | __le64 disk_recovery_offset; | |
1761 | ||
1762 | /* | |
33e53f06 | 1763 | * This offset tracks the progress of the initial raid set |
b12d437b JB |
1764 | * synchronisation/parity calculation. |
1765 | */ | |
1766 | __le64 array_resync_offset; | |
1767 | ||
1768 | /* | |
33e53f06 | 1769 | * raid characteristics |
b12d437b JB |
1770 | */ |
1771 | __le32 level; | |
1772 | __le32 layout; | |
1773 | __le32 stripe_sectors; | |
1774 | ||
33e53f06 | 1775 | /******************************************************************** |
9b6e5423 | 1776 | * BELOW FOLLOW V1.9.0 EXTENSIONS TO THE PRISTINE SUPERBLOCK FORMAT!!! |
33e53f06 | 1777 | * |
9b6e5423 | 1778 | * FEATURE_FLAG_SUPPORTS_V190 in the features member indicates that those exist |
33e53f06 HM |
1779 | */ |
1780 | ||
1781 | __le32 flags; /* Flags defining array states for reshaping */ | |
1782 | ||
1783 | /* | |
1784 | * This offset tracks the progress of a raid | |
1785 | * set reshape in order to be able to restart it | |
1786 | */ | |
1787 | __le64 reshape_position; | |
1788 | ||
1789 | /* | |
1790 | * These define the properties of the array in case of an interrupted reshape | |
1791 | */ | |
1792 | __le32 new_level; | |
1793 | __le32 new_layout; | |
1794 | __le32 new_stripe_sectors; | |
1795 | __le32 delta_disks; | |
1796 | ||
1797 | __le64 array_sectors; /* Array size in sectors */ | |
1798 | ||
1799 | /* | |
1800 | * Sector offsets to data on devices (reshaping). | |
1801 | * Needed to support out of place reshaping, thus | |
1802 | * not writing over any stripes whilst converting | |
1803 | * them from old to new layout | |
1804 | */ | |
1805 | __le64 data_offset; | |
1806 | __le64 new_data_offset; | |
1807 | ||
1808 | __le64 sectors; /* Used device size in sectors */ | |
1809 | ||
1810 | /* | |
1811 | * Additonal Bit field of devices indicating failures to support | |
9b6e5423 | 1812 | * up to 256 devices with the 1.9.0 on-disk metadata format |
33e53f06 HM |
1813 | */ |
1814 | __le64 extended_failed_devices[DISKS_ARRAY_ELEMS - 1]; | |
1815 | ||
1816 | __le32 incompat_features; /* Used to indicate any incompatible features */ | |
1817 | ||
1818 | /* Always set rest up to logical block size to 0 when writing (see get_metadata_device() below). */ | |
b12d437b JB |
1819 | } __packed; |
1820 | ||
9dbd1aa3 HM |
1821 | /* |
1822 | * Check for reshape constraints on raid set @rs: | |
1823 | * | |
1824 | * - reshape function non-existent | |
1825 | * - degraded set | |
1826 | * - ongoing recovery | |
1827 | * - ongoing reshape | |
1828 | * | |
1829 | * Returns 0 if none or -EPERM if given constraint | |
1830 | * and error message reference in @errmsg | |
1831 | */ | |
1832 | static int rs_check_reshape(struct raid_set *rs) | |
1833 | { | |
1834 | struct mddev *mddev = &rs->md; | |
1835 | ||
9dbd1aa3 HM |
1836 | if (!mddev->pers || !mddev->pers->check_reshape) |
1837 | rs->ti->error = "Reshape not supported"; | |
1838 | else if (mddev->degraded) | |
1839 | rs->ti->error = "Can't reshape degraded raid set"; | |
1840 | else if (rs_is_recovering(rs)) | |
1841 | rs->ti->error = "Convert request on recovering raid set prohibited"; | |
469b304b | 1842 | else if (rs_is_reshaping(rs)) |
9dbd1aa3 | 1843 | rs->ti->error = "raid set already reshaping!"; |
7a7c330f HM |
1844 | else if (!(rs_is_raid1(rs) || rs_is_raid10(rs) || rs_is_raid456(rs))) |
1845 | rs->ti->error = "Reshaping only supported for raid1/4/5/6/10"; | |
9dbd1aa3 HM |
1846 | else |
1847 | return 0; | |
1848 | ||
1849 | return -EPERM; | |
1850 | } | |
1851 | ||
3cb03002 | 1852 | static int read_disk_sb(struct md_rdev *rdev, int size) |
b12d437b JB |
1853 | { |
1854 | BUG_ON(!rdev->sb_page); | |
1855 | ||
1856 | if (rdev->sb_loaded) | |
1857 | return 0; | |
1858 | ||
0a7b8188 | 1859 | if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, 0, true)) { |
0447568f JB |
1860 | DMERR("Failed to read superblock of device at position %d", |
1861 | rdev->raid_disk); | |
c32fb9e7 | 1862 | md_error(rdev->mddev, rdev); |
b12d437b JB |
1863 | return -EINVAL; |
1864 | } | |
1865 | ||
1866 | rdev->sb_loaded = 1; | |
1867 | ||
1868 | return 0; | |
1869 | } | |
1870 | ||
33e53f06 HM |
1871 | static void sb_retrieve_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) |
1872 | { | |
1873 | failed_devices[0] = le64_to_cpu(sb->failed_devices); | |
1874 | memset(failed_devices + 1, 0, sizeof(sb->extended_failed_devices)); | |
1875 | ||
4286325b | 1876 | if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { |
33e53f06 HM |
1877 | int i = ARRAY_SIZE(sb->extended_failed_devices); |
1878 | ||
1879 | while (i--) | |
1880 | failed_devices[i+1] = le64_to_cpu(sb->extended_failed_devices[i]); | |
1881 | } | |
1882 | } | |
1883 | ||
7b34df74 HM |
1884 | static void sb_update_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) |
1885 | { | |
1886 | int i = ARRAY_SIZE(sb->extended_failed_devices); | |
1887 | ||
1888 | sb->failed_devices = cpu_to_le64(failed_devices[0]); | |
1889 | while (i--) | |
1890 | sb->extended_failed_devices[i] = cpu_to_le64(failed_devices[i+1]); | |
1891 | } | |
1892 | ||
1893 | /* | |
1894 | * Synchronize the superblock members with the raid set properties | |
1895 | * | |
1896 | * All superblock data is little endian. | |
1897 | */ | |
fd01b88c | 1898 | static void super_sync(struct mddev *mddev, struct md_rdev *rdev) |
b12d437b | 1899 | { |
7b34df74 HM |
1900 | bool update_failed_devices = false; |
1901 | unsigned int i; | |
1902 | uint64_t failed_devices[DISKS_ARRAY_ELEMS]; | |
b12d437b | 1903 | struct dm_raid_superblock *sb; |
81f382f9 | 1904 | struct raid_set *rs = container_of(mddev, struct raid_set, md); |
b12d437b | 1905 | |
7b34df74 HM |
1906 | /* No metadata device, no superblock */ |
1907 | if (!rdev->meta_bdev) | |
1908 | return; | |
1909 | ||
1910 | BUG_ON(!rdev->sb_page); | |
1911 | ||
b12d437b | 1912 | sb = page_address(rdev->sb_page); |
b12d437b | 1913 | |
7b34df74 | 1914 | sb_retrieve_failed_devices(sb, failed_devices); |
b12d437b | 1915 | |
7b34df74 HM |
1916 | for (i = 0; i < rs->raid_disks; i++) |
1917 | if (!rs->dev[i].data_dev || test_bit(Faulty, &rs->dev[i].rdev.flags)) { | |
1918 | update_failed_devices = true; | |
1919 | set_bit(i, (void *) failed_devices); | |
1920 | } | |
1921 | ||
1922 | if (update_failed_devices) | |
1923 | sb_update_failed_devices(sb, failed_devices); | |
b12d437b JB |
1924 | |
1925 | sb->magic = cpu_to_le32(DM_RAID_MAGIC); | |
9b6e5423 | 1926 | sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); |
b12d437b JB |
1927 | |
1928 | sb->num_devices = cpu_to_le32(mddev->raid_disks); | |
1929 | sb->array_position = cpu_to_le32(rdev->raid_disk); | |
1930 | ||
1931 | sb->events = cpu_to_le64(mddev->events); | |
b12d437b JB |
1932 | |
1933 | sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset); | |
1934 | sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp); | |
1935 | ||
1936 | sb->level = cpu_to_le32(mddev->level); | |
1937 | sb->layout = cpu_to_le32(mddev->layout); | |
1938 | sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors); | |
7b34df74 HM |
1939 | |
1940 | sb->new_level = cpu_to_le32(mddev->new_level); | |
1941 | sb->new_layout = cpu_to_le32(mddev->new_layout); | |
1942 | sb->new_stripe_sectors = cpu_to_le32(mddev->new_chunk_sectors); | |
1943 | ||
1944 | sb->delta_disks = cpu_to_le32(mddev->delta_disks); | |
1945 | ||
1946 | smp_rmb(); /* Make sure we access most recent reshape position */ | |
1947 | sb->reshape_position = cpu_to_le64(mddev->reshape_position); | |
1948 | if (le64_to_cpu(sb->reshape_position) != MaxSector) { | |
1949 | /* Flag ongoing reshape */ | |
1950 | sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE); | |
1951 | ||
1952 | if (mddev->delta_disks < 0 || mddev->reshape_backwards) | |
1953 | sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_BACKWARDS); | |
4286325b MS |
1954 | } else { |
1955 | /* Clear reshape flags */ | |
1956 | sb->flags &= ~(cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE|SB_FLAG_RESHAPE_BACKWARDS)); | |
1957 | } | |
7b34df74 HM |
1958 | |
1959 | sb->array_sectors = cpu_to_le64(mddev->array_sectors); | |
1960 | sb->data_offset = cpu_to_le64(rdev->data_offset); | |
1961 | sb->new_data_offset = cpu_to_le64(rdev->new_data_offset); | |
1962 | sb->sectors = cpu_to_le64(rdev->sectors); | |
b2a4872a | 1963 | sb->incompat_features = cpu_to_le32(0); |
7b34df74 HM |
1964 | |
1965 | /* Zero out the rest of the payload after the size of the superblock */ | |
1966 | memset(sb + 1, 0, rdev->sb_size - sizeof(*sb)); | |
b12d437b JB |
1967 | } |
1968 | ||
1969 | /* | |
1970 | * super_load | |
1971 | * | |
1972 | * This function creates a superblock if one is not found on the device | |
1973 | * and will decide which superblock to use if there's a choice. | |
1974 | * | |
1975 | * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise | |
1976 | */ | |
3cb03002 | 1977 | static int super_load(struct md_rdev *rdev, struct md_rdev *refdev) |
b12d437b | 1978 | { |
73c6f239 | 1979 | int r; |
b12d437b JB |
1980 | struct dm_raid_superblock *sb; |
1981 | struct dm_raid_superblock *refsb; | |
1982 | uint64_t events_sb, events_refsb; | |
1983 | ||
1984 | rdev->sb_start = 0; | |
40d43c4b HM |
1985 | rdev->sb_size = bdev_logical_block_size(rdev->meta_bdev); |
1986 | if (rdev->sb_size < sizeof(*sb) || rdev->sb_size > PAGE_SIZE) { | |
1987 | DMERR("superblock size of a logical block is no longer valid"); | |
1988 | return -EINVAL; | |
1989 | } | |
b12d437b | 1990 | |
73c6f239 HM |
1991 | r = read_disk_sb(rdev, rdev->sb_size); |
1992 | if (r) | |
1993 | return r; | |
b12d437b JB |
1994 | |
1995 | sb = page_address(rdev->sb_page); | |
3aa3b2b2 JB |
1996 | |
1997 | /* | |
1998 | * Two cases that we want to write new superblocks and rebuild: | |
1999 | * 1) New device (no matching magic number) | |
2000 | * 2) Device specified for rebuild (!In_sync w/ offset == 0) | |
2001 | */ | |
2002 | if ((sb->magic != cpu_to_le32(DM_RAID_MAGIC)) || | |
2003 | (!test_bit(In_sync, &rdev->flags) && !rdev->recovery_offset)) { | |
b12d437b JB |
2004 | super_sync(rdev->mddev, rdev); |
2005 | ||
2006 | set_bit(FirstUse, &rdev->flags); | |
9b6e5423 | 2007 | sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); |
b12d437b JB |
2008 | |
2009 | /* Force writing of superblocks to disk */ | |
2010 | set_bit(MD_CHANGE_DEVS, &rdev->mddev->flags); | |
2011 | ||
2012 | /* Any superblock is better than none, choose that if given */ | |
2013 | return refdev ? 0 : 1; | |
2014 | } | |
2015 | ||
2016 | if (!refdev) | |
2017 | return 1; | |
2018 | ||
2019 | events_sb = le64_to_cpu(sb->events); | |
2020 | ||
2021 | refsb = page_address(refdev->sb_page); | |
2022 | events_refsb = le64_to_cpu(refsb->events); | |
2023 | ||
2024 | return (events_sb > events_refsb) ? 1 : 0; | |
2025 | } | |
2026 | ||
33e53f06 | 2027 | static int super_init_validation(struct raid_set *rs, struct md_rdev *rdev) |
b12d437b JB |
2028 | { |
2029 | int role; | |
33e53f06 HM |
2030 | unsigned int d; |
2031 | struct mddev *mddev = &rs->md; | |
b12d437b | 2032 | uint64_t events_sb; |
33e53f06 | 2033 | uint64_t failed_devices[DISKS_ARRAY_ELEMS]; |
b12d437b | 2034 | struct dm_raid_superblock *sb; |
33e53f06 | 2035 | uint32_t new_devs = 0, rebuild_and_new = 0, rebuilds = 0; |
dafb20fa | 2036 | struct md_rdev *r; |
b12d437b JB |
2037 | struct dm_raid_superblock *sb2; |
2038 | ||
2039 | sb = page_address(rdev->sb_page); | |
2040 | events_sb = le64_to_cpu(sb->events); | |
b12d437b JB |
2041 | |
2042 | /* | |
2043 | * Initialise to 1 if this is a new superblock. | |
2044 | */ | |
2045 | mddev->events = events_sb ? : 1; | |
2046 | ||
33e53f06 HM |
2047 | mddev->reshape_position = MaxSector; |
2048 | ||
b12d437b | 2049 | /* |
33e53f06 HM |
2050 | * Reshaping is supported, e.g. reshape_position is valid |
2051 | * in superblock and superblock content is authoritative. | |
b12d437b | 2052 | */ |
4286325b | 2053 | if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { |
33e53f06 HM |
2054 | /* Superblock is authoritative wrt given raid set layout! */ |
2055 | mddev->raid_disks = le32_to_cpu(sb->num_devices); | |
2056 | mddev->level = le32_to_cpu(sb->level); | |
2057 | mddev->layout = le32_to_cpu(sb->layout); | |
2058 | mddev->chunk_sectors = le32_to_cpu(sb->stripe_sectors); | |
2059 | mddev->new_level = le32_to_cpu(sb->new_level); | |
2060 | mddev->new_layout = le32_to_cpu(sb->new_layout); | |
2061 | mddev->new_chunk_sectors = le32_to_cpu(sb->new_stripe_sectors); | |
2062 | mddev->delta_disks = le32_to_cpu(sb->delta_disks); | |
2063 | mddev->array_sectors = le64_to_cpu(sb->array_sectors); | |
2064 | ||
2065 | /* raid was reshaping and got interrupted */ | |
4286325b MS |
2066 | if (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_ACTIVE) { |
2067 | if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { | |
33e53f06 HM |
2068 | DMERR("Reshape requested but raid set is still reshaping"); |
2069 | return -EINVAL; | |
2070 | } | |
b12d437b | 2071 | |
33e53f06 | 2072 | if (mddev->delta_disks < 0 || |
4286325b | 2073 | (!mddev->delta_disks && (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_BACKWARDS))) |
33e53f06 HM |
2074 | mddev->reshape_backwards = 1; |
2075 | else | |
2076 | mddev->reshape_backwards = 0; | |
2077 | ||
2078 | mddev->reshape_position = le64_to_cpu(sb->reshape_position); | |
2079 | rs->raid_type = get_raid_type_by_ll(mddev->level, mddev->layout); | |
2080 | } | |
2081 | ||
2082 | } else { | |
2083 | /* | |
9b6e5423 | 2084 | * No takeover/reshaping, because we don't have the extended v1.9.0 metadata |
33e53f06 HM |
2085 | */ |
2086 | if (le32_to_cpu(sb->level) != mddev->level) { | |
2087 | DMERR("Reshaping/takeover raid sets not yet supported. (raid level/stripes/size change)"); | |
2088 | return -EINVAL; | |
2089 | } | |
2090 | if (le32_to_cpu(sb->layout) != mddev->layout) { | |
2091 | DMERR("Reshaping raid sets not yet supported. (raid layout change)"); | |
43157840 MS |
2092 | DMERR(" 0x%X vs 0x%X", le32_to_cpu(sb->layout), mddev->layout); |
2093 | DMERR(" Old layout: %s w/ %d copies", | |
33e53f06 HM |
2094 | raid10_md_layout_to_format(le32_to_cpu(sb->layout)), |
2095 | raid10_md_layout_to_copies(le32_to_cpu(sb->layout))); | |
43157840 | 2096 | DMERR(" New layout: %s w/ %d copies", |
33e53f06 HM |
2097 | raid10_md_layout_to_format(mddev->layout), |
2098 | raid10_md_layout_to_copies(mddev->layout)); | |
2099 | return -EINVAL; | |
2100 | } | |
2101 | if (le32_to_cpu(sb->stripe_sectors) != mddev->chunk_sectors) { | |
2102 | DMERR("Reshaping raid sets not yet supported. (stripe sectors change)"); | |
2103 | return -EINVAL; | |
2104 | } | |
2105 | ||
2106 | /* We can only change the number of devices in raid1 with old (i.e. pre 1.0.7) metadata */ | |
2107 | if (!rt_is_raid1(rs->raid_type) && | |
2108 | (le32_to_cpu(sb->num_devices) != mddev->raid_disks)) { | |
2109 | DMERR("Reshaping raid sets not yet supported. (device count change from %u to %u)", | |
2110 | sb->num_devices, mddev->raid_disks); | |
2111 | return -EINVAL; | |
2112 | } | |
2113 | ||
2114 | /* Table line is checked vs. authoritative superblock */ | |
2115 | rs_set_new(rs); | |
b12d437b JB |
2116 | } |
2117 | ||
4286325b | 2118 | if (!test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) |
b12d437b JB |
2119 | mddev->recovery_cp = le64_to_cpu(sb->array_resync_offset); |
2120 | ||
2121 | /* | |
2122 | * During load, we set FirstUse if a new superblock was written. | |
2123 | * There are two reasons we might not have a superblock: | |
33e53f06 | 2124 | * 1) The raid set is brand new - in which case, all of the |
43157840 | 2125 | * devices must have their In_sync bit set. Also, |
b12d437b | 2126 | * recovery_cp must be 0, unless forced. |
33e53f06 | 2127 | * 2) This is a new device being added to an old raid set |
b12d437b JB |
2128 | * and the new device needs to be rebuilt - in which |
2129 | * case the In_sync bit will /not/ be set and | |
2130 | * recovery_cp must be MaxSector. | |
9dbd1aa3 HM |
2131 | * 3) This is/are a new device(s) being added to an old |
2132 | * raid set during takeover to a higher raid level | |
2133 | * to provide capacity for redundancy or during reshape | |
2134 | * to add capacity to grow the raid set. | |
b12d437b | 2135 | */ |
33e53f06 | 2136 | d = 0; |
dafb20fa | 2137 | rdev_for_each(r, mddev) { |
33e53f06 HM |
2138 | if (test_bit(FirstUse, &r->flags)) |
2139 | new_devs++; | |
2140 | ||
b12d437b | 2141 | if (!test_bit(In_sync, &r->flags)) { |
33e53f06 HM |
2142 | DMINFO("Device %d specified for rebuild; clearing superblock", |
2143 | r->raid_disk); | |
b12d437b | 2144 | rebuilds++; |
33e53f06 HM |
2145 | |
2146 | if (test_bit(FirstUse, &r->flags)) | |
2147 | rebuild_and_new++; | |
2148 | } | |
2149 | ||
2150 | d++; | |
b12d437b JB |
2151 | } |
2152 | ||
33e53f06 HM |
2153 | if (new_devs == rs->raid_disks || !rebuilds) { |
2154 | /* Replace a broken device */ | |
2155 | if (new_devs == 1 && !rs->delta_disks) | |
2156 | ; | |
2157 | if (new_devs == rs->raid_disks) { | |
2158 | DMINFO("Superblocks created for new raid set"); | |
b12d437b | 2159 | set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); |
9dbd1aa3 HM |
2160 | } else if (new_devs != rebuilds && |
2161 | new_devs != rs->delta_disks) { | |
33e53f06 HM |
2162 | DMERR("New device injected into existing raid set without " |
2163 | "'delta_disks' or 'rebuild' parameter specified"); | |
b12d437b JB |
2164 | return -EINVAL; |
2165 | } | |
33e53f06 HM |
2166 | } else if (new_devs && new_devs != rebuilds) { |
2167 | DMERR("%u 'rebuild' devices cannot be injected into" | |
2168 | " a raid set with %u other first-time devices", | |
2169 | rebuilds, new_devs); | |
b12d437b | 2170 | return -EINVAL; |
33e53f06 HM |
2171 | } else if (rebuilds) { |
2172 | if (rebuild_and_new && rebuilds != rebuild_and_new) { | |
2173 | DMERR("new device%s provided without 'rebuild'", | |
2174 | new_devs > 1 ? "s" : ""); | |
2175 | return -EINVAL; | |
9dbd1aa3 | 2176 | } else if (rs_is_recovering(rs)) { |
33e53f06 HM |
2177 | DMERR("'rebuild' specified while raid set is not in-sync (recovery_cp=%llu)", |
2178 | (unsigned long long) mddev->recovery_cp); | |
2179 | return -EINVAL; | |
9dbd1aa3 HM |
2180 | } else if (rs_is_reshaping(rs)) { |
2181 | DMERR("'rebuild' specified while raid set is being reshaped (reshape_position=%llu)", | |
2182 | (unsigned long long) mddev->reshape_position); | |
33e53f06 HM |
2183 | return -EINVAL; |
2184 | } | |
b12d437b JB |
2185 | } |
2186 | ||
2187 | /* | |
2188 | * Now we set the Faulty bit for those devices that are | |
2189 | * recorded in the superblock as failed. | |
2190 | */ | |
33e53f06 | 2191 | sb_retrieve_failed_devices(sb, failed_devices); |
dafb20fa | 2192 | rdev_for_each(r, mddev) { |
b12d437b JB |
2193 | if (!r->sb_page) |
2194 | continue; | |
2195 | sb2 = page_address(r->sb_page); | |
2196 | sb2->failed_devices = 0; | |
33e53f06 | 2197 | memset(sb2->extended_failed_devices, 0, sizeof(sb2->extended_failed_devices)); |
b12d437b JB |
2198 | |
2199 | /* | |
2200 | * Check for any device re-ordering. | |
2201 | */ | |
2202 | if (!test_bit(FirstUse, &r->flags) && (r->raid_disk >= 0)) { | |
2203 | role = le32_to_cpu(sb2->array_position); | |
33e53f06 HM |
2204 | if (role < 0) |
2205 | continue; | |
2206 | ||
b12d437b | 2207 | if (role != r->raid_disk) { |
e6ca5e1a MS |
2208 | if (__is_raid10_near(mddev->layout)) { |
2209 | if (mddev->raid_disks % __raid10_near_copies(mddev->layout) || | |
bd83a4c4 MS |
2210 | rs->raid_disks % rs->raid10_copies) { |
2211 | rs->ti->error = | |
2212 | "Cannot change raid10 near set to odd # of devices!"; | |
2213 | return -EINVAL; | |
2214 | } | |
33e53f06 HM |
2215 | |
2216 | sb2->array_position = cpu_to_le32(r->raid_disk); | |
2217 | ||
2218 | } else if (!(rs_is_raid10(rs) && rt_is_raid0(rs->raid_type)) && | |
bd83a4c4 MS |
2219 | !(rs_is_raid0(rs) && rt_is_raid10(rs->raid_type)) && |
2220 | !rt_is_raid1(rs->raid_type)) { | |
2221 | rs->ti->error = "Cannot change device positions in raid set"; | |
2222 | return -EINVAL; | |
2223 | } | |
33e53f06 | 2224 | |
bd83a4c4 | 2225 | DMINFO("raid device #%d now at position #%d", role, r->raid_disk); |
b12d437b JB |
2226 | } |
2227 | ||
2228 | /* | |
2229 | * Partial recovery is performed on | |
2230 | * returning failed devices. | |
2231 | */ | |
33e53f06 | 2232 | if (test_bit(role, (void *) failed_devices)) |
b12d437b JB |
2233 | set_bit(Faulty, &r->flags); |
2234 | } | |
2235 | } | |
2236 | ||
2237 | return 0; | |
2238 | } | |
2239 | ||
0cf45031 | 2240 | static int super_validate(struct raid_set *rs, struct md_rdev *rdev) |
b12d437b | 2241 | { |
0cf45031 | 2242 | struct mddev *mddev = &rs->md; |
33e53f06 HM |
2243 | struct dm_raid_superblock *sb; |
2244 | ||
3a1c1ef2 | 2245 | if (rs_is_raid0(rs) || !rdev->sb_page) |
33e53f06 HM |
2246 | return 0; |
2247 | ||
2248 | sb = page_address(rdev->sb_page); | |
b12d437b JB |
2249 | |
2250 | /* | |
2251 | * If mddev->events is not set, we know we have not yet initialized | |
2252 | * the array. | |
2253 | */ | |
33e53f06 | 2254 | if (!mddev->events && super_init_validation(rs, rdev)) |
b12d437b JB |
2255 | return -EINVAL; |
2256 | ||
9b6e5423 MS |
2257 | if (le32_to_cpu(sb->compat_features) != FEATURE_FLAG_SUPPORTS_V190) { |
2258 | rs->ti->error = "Unable to assemble array: Unknown flag(s) in compatible feature flags"; | |
2259 | return -EINVAL; | |
2260 | } | |
2261 | ||
2262 | if (sb->incompat_features) { | |
ecbfb9f1 | 2263 | rs->ti->error = "Unable to assemble array: No incompatible feature flags supported yet"; |
4c9971ca HM |
2264 | return -EINVAL; |
2265 | } | |
2266 | ||
0cf45031 | 2267 | /* Enable bitmap creation for RAID levels != 0 */ |
676fa5ad | 2268 | mddev->bitmap_info.offset = rt_is_raid0(rs->raid_type) ? 0 : to_sector(4096); |
0cf45031 HM |
2269 | rdev->mddev->bitmap_info.default_offset = mddev->bitmap_info.offset; |
2270 | ||
33e53f06 HM |
2271 | if (!test_and_clear_bit(FirstUse, &rdev->flags)) { |
2272 | /* Retrieve device size stored in superblock to be prepared for shrink */ | |
2273 | rdev->sectors = le64_to_cpu(sb->sectors); | |
b12d437b | 2274 | rdev->recovery_offset = le64_to_cpu(sb->disk_recovery_offset); |
33e53f06 HM |
2275 | if (rdev->recovery_offset == MaxSector) |
2276 | set_bit(In_sync, &rdev->flags); | |
2277 | /* | |
2278 | * If no reshape in progress -> we're recovering single | |
2279 | * disk(s) and have to set the device(s) to out-of-sync | |
2280 | */ | |
9dbd1aa3 | 2281 | else if (!rs_is_reshaping(rs)) |
33e53f06 | 2282 | clear_bit(In_sync, &rdev->flags); /* Mandatory for recovery */ |
b12d437b JB |
2283 | } |
2284 | ||
2285 | /* | |
2286 | * If a device comes back, set it as not In_sync and no longer faulty. | |
2287 | */ | |
33e53f06 HM |
2288 | if (test_and_clear_bit(Faulty, &rdev->flags)) { |
2289 | rdev->recovery_offset = 0; | |
b12d437b JB |
2290 | clear_bit(In_sync, &rdev->flags); |
2291 | rdev->saved_raid_disk = rdev->raid_disk; | |
b12d437b JB |
2292 | } |
2293 | ||
33e53f06 HM |
2294 | /* Reshape support -> restore repective data offsets */ |
2295 | rdev->data_offset = le64_to_cpu(sb->data_offset); | |
2296 | rdev->new_data_offset = le64_to_cpu(sb->new_data_offset); | |
b12d437b JB |
2297 | |
2298 | return 0; | |
2299 | } | |
2300 | ||
2301 | /* | |
2302 | * Analyse superblocks and select the freshest. | |
2303 | */ | |
2304 | static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs) | |
2305 | { | |
73c6f239 | 2306 | int r; |
0447568f | 2307 | struct raid_dev *dev; |
a9ad8526 | 2308 | struct md_rdev *rdev, *tmp, *freshest; |
fd01b88c | 2309 | struct mddev *mddev = &rs->md; |
b12d437b JB |
2310 | |
2311 | freshest = NULL; | |
a9ad8526 | 2312 | rdev_for_each_safe(rdev, tmp, mddev) { |
761becff | 2313 | /* |
c76d53f4 | 2314 | * Skipping super_load due to CTR_FLAG_SYNC will cause |
761becff | 2315 | * the array to undergo initialization again as |
43157840 | 2316 | * though it were new. This is the intended effect |
761becff JB |
2317 | * of the "sync" directive. |
2318 | * | |
2319 | * When reshaping capability is added, we must ensure | |
2320 | * that the "sync" directive is disallowed during the | |
2321 | * reshape. | |
2322 | */ | |
4286325b | 2323 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) |
761becff JB |
2324 | continue; |
2325 | ||
b12d437b JB |
2326 | if (!rdev->meta_bdev) |
2327 | continue; | |
2328 | ||
73c6f239 | 2329 | r = super_load(rdev, freshest); |
b12d437b | 2330 | |
73c6f239 | 2331 | switch (r) { |
b12d437b JB |
2332 | case 1: |
2333 | freshest = rdev; | |
2334 | break; | |
2335 | case 0: | |
2336 | break; | |
2337 | default: | |
0447568f | 2338 | dev = container_of(rdev, struct raid_dev, rdev); |
55ebbb59 JB |
2339 | if (dev->meta_dev) |
2340 | dm_put_device(ti, dev->meta_dev); | |
0447568f | 2341 | |
55ebbb59 JB |
2342 | dev->meta_dev = NULL; |
2343 | rdev->meta_bdev = NULL; | |
0447568f | 2344 | |
55ebbb59 JB |
2345 | if (rdev->sb_page) |
2346 | put_page(rdev->sb_page); | |
0447568f | 2347 | |
55ebbb59 | 2348 | rdev->sb_page = NULL; |
0447568f | 2349 | |
55ebbb59 | 2350 | rdev->sb_loaded = 0; |
0447568f | 2351 | |
55ebbb59 JB |
2352 | /* |
2353 | * We might be able to salvage the data device | |
2354 | * even though the meta device has failed. For | |
2355 | * now, we behave as though '- -' had been | |
2356 | * set for this device in the table. | |
2357 | */ | |
2358 | if (dev->data_dev) | |
2359 | dm_put_device(ti, dev->data_dev); | |
0447568f | 2360 | |
55ebbb59 JB |
2361 | dev->data_dev = NULL; |
2362 | rdev->bdev = NULL; | |
0447568f | 2363 | |
55ebbb59 | 2364 | list_del(&rdev->same_set); |
b12d437b JB |
2365 | } |
2366 | } | |
2367 | ||
2368 | if (!freshest) | |
2369 | return 0; | |
2370 | ||
bd83a4c4 MS |
2371 | if (validate_raid_redundancy(rs)) { |
2372 | rs->ti->error = "Insufficient redundancy to activate array"; | |
2373 | return -EINVAL; | |
2374 | } | |
55ebbb59 | 2375 | |
b12d437b JB |
2376 | /* |
2377 | * Validation of the freshest device provides the source of | |
2378 | * validation for the remaining devices. | |
2379 | */ | |
9dbd1aa3 HM |
2380 | rs->ti->error = "Unable to assemble array: Invalid superblocks"; |
2381 | if (super_validate(rs, freshest)) | |
bd83a4c4 | 2382 | return -EINVAL; |
b12d437b | 2383 | |
dafb20fa | 2384 | rdev_for_each(rdev, mddev) |
0cf45031 | 2385 | if ((rdev != freshest) && super_validate(rs, rdev)) |
b12d437b | 2386 | return -EINVAL; |
b12d437b JB |
2387 | return 0; |
2388 | } | |
2389 | ||
40ba37e5 HM |
2390 | /* |
2391 | * Adjust data_offset and new_data_offset on all disk members of @rs | |
2392 | * for out of place reshaping if requested by contructor | |
2393 | * | |
2394 | * We need free space at the beginning of each raid disk for forward | |
2395 | * and at the end for backward reshapes which userspace has to provide | |
2396 | * via remapping/reordering of space. | |
2397 | */ | |
2398 | static int rs_adjust_data_offsets(struct raid_set *rs) | |
2399 | { | |
2400 | sector_t data_offset = 0, new_data_offset = 0; | |
2401 | struct md_rdev *rdev; | |
2402 | ||
2403 | /* Constructor did not request data offset change */ | |
2404 | if (!test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { | |
2405 | if (!rs_is_reshapable(rs)) | |
2406 | goto out; | |
2407 | ||
2408 | return 0; | |
2409 | } | |
2410 | ||
2411 | /* HM FIXME: get InSync raid_dev? */ | |
2412 | rdev = &rs->dev[0].rdev; | |
2413 | ||
2414 | if (rs->delta_disks < 0) { | |
2415 | /* | |
2416 | * Removing disks (reshaping backwards): | |
2417 | * | |
2418 | * - before reshape: data is at offset 0 and free space | |
2419 | * is at end of each component LV | |
2420 | * | |
2421 | * - after reshape: data is at offset rs->data_offset != 0 on each component LV | |
2422 | */ | |
2423 | data_offset = 0; | |
2424 | new_data_offset = rs->data_offset; | |
2425 | ||
2426 | } else if (rs->delta_disks > 0) { | |
2427 | /* | |
2428 | * Adding disks (reshaping forwards): | |
2429 | * | |
2430 | * - before reshape: data is at offset rs->data_offset != 0 and | |
2431 | * free space is at begin of each component LV | |
2432 | * | |
2433 | * - after reshape: data is at offset 0 on each component LV | |
2434 | */ | |
2435 | data_offset = rs->data_offset; | |
2436 | new_data_offset = 0; | |
2437 | ||
2438 | } else { | |
2439 | /* | |
2440 | * User space passes in 0 for data offset after having removed reshape space | |
2441 | * | |
2442 | * - or - (data offset != 0) | |
2443 | * | |
2444 | * Changing RAID layout or chunk size -> toggle offsets | |
2445 | * | |
2446 | * - before reshape: data is at offset rs->data_offset 0 and | |
2447 | * free space is at end of each component LV | |
2448 | * -or- | |
2449 | * data is at offset rs->data_offset != 0 and | |
2450 | * free space is at begin of each component LV | |
2451 | * | |
2527b56e HM |
2452 | * - after reshape: data is at offset 0 if it was at offset != 0 |
2453 | * or at offset != 0 if it was at offset 0 | |
40ba37e5 HM |
2454 | * on each component LV |
2455 | * | |
2456 | */ | |
2457 | data_offset = rs->data_offset ? rdev->data_offset : 0; | |
2458 | new_data_offset = data_offset ? 0 : rs->data_offset; | |
2459 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
2460 | } | |
2461 | ||
2462 | /* | |
2463 | * Make sure we got a minimum amount of free sectors per device | |
2464 | */ | |
2465 | if (rs->data_offset && | |
2466 | to_sector(i_size_read(rdev->bdev->bd_inode)) - rdev->sectors < MIN_FREE_RESHAPE_SPACE) { | |
2467 | rs->ti->error = data_offset ? "No space for forward reshape" : | |
2468 | "No space for backward reshape"; | |
2469 | return -ENOSPC; | |
2470 | } | |
2471 | out: | |
2472 | /* Adjust data offsets on all rdevs */ | |
2473 | rdev_for_each(rdev, &rs->md) { | |
2474 | rdev->data_offset = data_offset; | |
2475 | rdev->new_data_offset = new_data_offset; | |
2476 | } | |
2477 | ||
2478 | return 0; | |
2479 | } | |
2480 | ||
ecbfb9f1 | 2481 | /* Userpace reordered disks -> adjust raid_disk indexes in @rs */ |
e6ca5e1a | 2482 | static void __reorder_raid_disk_indexes(struct raid_set *rs) |
ecbfb9f1 HM |
2483 | { |
2484 | int i = 0; | |
2485 | struct md_rdev *rdev; | |
2486 | ||
2487 | rdev_for_each(rdev, &rs->md) { | |
2488 | rdev->raid_disk = i++; | |
2489 | rdev->saved_raid_disk = rdev->new_raid_disk = -1; | |
2490 | } | |
2491 | } | |
2492 | ||
2493 | /* | |
2494 | * Setup @rs for takeover by a different raid level | |
2495 | */ | |
2496 | static int rs_setup_takeover(struct raid_set *rs) | |
2497 | { | |
2498 | struct mddev *mddev = &rs->md; | |
2499 | struct md_rdev *rdev; | |
2500 | unsigned int d = mddev->raid_disks = rs->raid_disks; | |
2501 | sector_t new_data_offset = rs->dev[0].rdev.data_offset ? 0 : rs->data_offset; | |
2502 | ||
2503 | if (rt_is_raid10(rs->raid_type)) { | |
2504 | if (mddev->level == 0) { | |
2505 | /* Userpace reordered disks -> adjust raid_disk indexes */ | |
e6ca5e1a | 2506 | __reorder_raid_disk_indexes(rs); |
ecbfb9f1 HM |
2507 | |
2508 | /* raid0 -> raid10_far layout */ | |
2509 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_FAR, | |
2510 | rs->raid10_copies); | |
2511 | } else if (mddev->level == 1) | |
2512 | /* raid1 -> raid10_near layout */ | |
2513 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, | |
2514 | rs->raid_disks); | |
094f394d | 2515 | else |
ecbfb9f1 HM |
2516 | return -EINVAL; |
2517 | ||
2518 | } | |
2519 | ||
2520 | clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags); | |
2521 | mddev->recovery_cp = MaxSector; | |
2522 | ||
2523 | while (d--) { | |
2524 | rdev = &rs->dev[d].rdev; | |
2525 | ||
2526 | if (test_bit(d, (void *) rs->rebuild_disks)) { | |
2527 | clear_bit(In_sync, &rdev->flags); | |
2528 | clear_bit(Faulty, &rdev->flags); | |
2529 | mddev->recovery_cp = rdev->recovery_offset = 0; | |
2530 | /* Bitmap has to be created when we do an "up" takeover */ | |
2531 | set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); | |
2532 | } | |
2533 | ||
2534 | rdev->new_data_offset = new_data_offset; | |
2535 | } | |
2536 | ||
ecbfb9f1 HM |
2537 | return 0; |
2538 | } | |
2539 | ||
469b304b HM |
2540 | /* Prepare @rs for reshape */ |
2541 | static int rs_prepare_reshape(struct raid_set *rs) | |
2542 | { | |
2543 | bool reshape; | |
2544 | struct mddev *mddev = &rs->md; | |
2545 | ||
2546 | if (rs_is_raid10(rs)) { | |
2547 | if (rs->raid_disks != mddev->raid_disks && | |
2548 | __is_raid10_near(mddev->layout) && | |
2549 | rs->raid10_copies && | |
2550 | rs->raid10_copies != __raid10_near_copies(mddev->layout)) { | |
2551 | /* | |
2552 | * raid disk have to be multiple of data copies to allow this conversion, | |
2553 | * | |
2554 | * This is actually not a reshape it is a | |
2555 | * rebuild of any additional mirrors per group | |
2556 | */ | |
2557 | if (rs->raid_disks % rs->raid10_copies) { | |
2558 | rs->ti->error = "Can't reshape raid10 mirror groups"; | |
2559 | return -EINVAL; | |
2560 | } | |
2561 | ||
2562 | /* Userpace reordered disks to add/remove mirrors -> adjust raid_disk indexes */ | |
2563 | __reorder_raid_disk_indexes(rs); | |
2564 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, | |
2565 | rs->raid10_copies); | |
2566 | mddev->new_layout = mddev->layout; | |
2567 | reshape = false; | |
2568 | } else | |
2569 | reshape = true; | |
2570 | ||
2571 | } else if (rs_is_raid456(rs)) | |
2572 | reshape = true; | |
2573 | ||
469b304b | 2574 | else if (rs_is_raid1(rs)) { |
7a7c330f HM |
2575 | if (rs->delta_disks) { |
2576 | /* Process raid1 via delta_disks */ | |
2577 | mddev->degraded = rs->delta_disks < 0 ? -rs->delta_disks : rs->delta_disks; | |
2578 | reshape = true; | |
2579 | } else { | |
2580 | /* Process raid1 without delta_disks */ | |
2581 | mddev->raid_disks = rs->raid_disks; | |
2582 | set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); | |
2583 | reshape = false; | |
2584 | } | |
469b304b HM |
2585 | } else { |
2586 | rs->ti->error = "Called with bogus raid type"; | |
2587 | return -EINVAL; | |
2588 | } | |
2589 | ||
2590 | if (reshape) { | |
2591 | set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags); | |
2592 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
2593 | set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); | |
7a7c330f HM |
2594 | } else if (mddev->raid_disks < rs->raid_disks) |
2595 | /* Create new superblocks and bitmaps, if any new disks */ | |
469b304b | 2596 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); |
469b304b HM |
2597 | |
2598 | return 0; | |
2599 | } | |
2600 | ||
9dbd1aa3 HM |
2601 | /* |
2602 | * | |
2603 | * - change raid layout | |
2604 | * - change chunk size | |
2605 | * - add disks | |
2606 | * - remove disks | |
2607 | */ | |
2608 | static int rs_setup_reshape(struct raid_set *rs) | |
2609 | { | |
2610 | int r = 0; | |
2611 | unsigned int cur_raid_devs, d; | |
2612 | struct mddev *mddev = &rs->md; | |
2613 | struct md_rdev *rdev; | |
2614 | ||
2615 | mddev->delta_disks = rs->delta_disks; | |
2616 | cur_raid_devs = mddev->raid_disks; | |
2617 | ||
2618 | /* Ignore impossible layout change whilst adding/removing disks */ | |
2619 | if (mddev->delta_disks && | |
2620 | mddev->layout != mddev->new_layout) { | |
2621 | DMINFO("Ignoring invalid layout change with delta_disks=%d", rs->delta_disks); | |
2622 | mddev->new_layout = mddev->layout; | |
2623 | } | |
2624 | ||
2625 | /* | |
2626 | * Adjust array size: | |
2627 | * | |
2628 | * - in case of adding disks, array size has | |
2629 | * to grow after the disk adding reshape, | |
2630 | * which'll hapen in the event handler; | |
2631 | * reshape will happen forward, so space has to | |
2632 | * be available at the beginning of each disk | |
2633 | * | |
2634 | * - in case of removing disks, array size | |
2635 | * has to shrink before starting the reshape, | |
2636 | * which'll happen here; | |
2637 | * reshape will happen backward, so space has to | |
2638 | * be available at the end of each disk | |
2639 | * | |
2640 | * - data_offset and new_data_offset are | |
ae3c6cff | 2641 | * adjusted for aforementioned out of place |
9dbd1aa3 HM |
2642 | * reshaping based on userspace passing in |
2643 | * the "data_offset <sectors>" key/value | |
ae3c6cff | 2644 | * pair via the constructor |
9dbd1aa3 HM |
2645 | */ |
2646 | ||
2647 | /* Add disk(s) */ | |
2648 | if (rs->delta_disks > 0) { | |
2649 | /* Prepare disks for check in raid4/5/6/10 {check|start}_reshape */ | |
2650 | for (d = cur_raid_devs; d < rs->raid_disks; d++) { | |
2651 | rdev = &rs->dev[d].rdev; | |
2652 | clear_bit(In_sync, &rdev->flags); | |
2653 | ||
2654 | /* | |
2655 | * save_raid_disk needs to be -1, or recovery_offset will be set to 0 | |
2656 | * by md, which'll store that erroneously in the superblock on reshape | |
2657 | */ | |
2658 | rdev->saved_raid_disk = -1; | |
2659 | rdev->raid_disk = d; | |
2660 | ||
2661 | rdev->sectors = mddev->dev_sectors; | |
7a7c330f | 2662 | rdev->recovery_offset = rs_is_raid1(rs) ? 0 : MaxSector; |
9dbd1aa3 HM |
2663 | } |
2664 | ||
2665 | mddev->reshape_backwards = 0; /* adding disks -> forward reshape */ | |
2666 | ||
2667 | /* Remove disk(s) */ | |
2668 | } else if (rs->delta_disks < 0) { | |
2669 | r = rs_set_dev_and_array_sectors(rs, true); | |
2670 | mddev->reshape_backwards = 1; /* removing disk(s) -> backward reshape */ | |
2671 | ||
2672 | /* Change layout and/or chunk size */ | |
2673 | } else { | |
2674 | /* | |
2675 | * Reshape layout (e.g. raid5_ls -> raid5_n) and/or chunk size: | |
2676 | * | |
2677 | * keeping number of disks and do layout change -> | |
2678 | * | |
2679 | * toggle reshape_backward depending on data_offset: | |
2680 | * | |
2681 | * - free space upfront -> reshape forward | |
2682 | * | |
2683 | * - free space at the end -> reshape backward | |
2684 | * | |
2685 | * | |
2686 | * This utilizes free reshape space avoiding the need | |
2687 | * for userspace to move (parts of) LV segments in | |
2688 | * case of layout/chunksize change (for disk | |
2689 | * adding/removing reshape space has to be at | |
2690 | * the proper address (see above with delta_disks): | |
2691 | * | |
2692 | * add disk(s) -> begin | |
2693 | * remove disk(s)-> end | |
2694 | */ | |
2695 | mddev->reshape_backwards = rs->dev[0].rdev.data_offset ? 0 : 1; | |
2696 | } | |
2697 | ||
2698 | return r; | |
2699 | } | |
2700 | ||
75b8e04b | 2701 | /* |
48cf06bc HM |
2702 | * Enable/disable discard support on RAID set depending on |
2703 | * RAID level and discard properties of underlying RAID members. | |
75b8e04b | 2704 | */ |
ecbfb9f1 | 2705 | static void configure_discard_support(struct raid_set *rs) |
75b8e04b | 2706 | { |
48cf06bc HM |
2707 | int i; |
2708 | bool raid456; | |
ecbfb9f1 | 2709 | struct dm_target *ti = rs->ti; |
48cf06bc | 2710 | |
75b8e04b HM |
2711 | /* Assume discards not supported until after checks below. */ |
2712 | ti->discards_supported = false; | |
2713 | ||
2714 | /* RAID level 4,5,6 require discard_zeroes_data for data integrity! */ | |
48cf06bc | 2715 | raid456 = (rs->md.level == 4 || rs->md.level == 5 || rs->md.level == 6); |
75b8e04b | 2716 | |
ffeeac75 | 2717 | for (i = 0; i < rs->raid_disks; i++) { |
d20c4b08 | 2718 | struct request_queue *q; |
48cf06bc | 2719 | |
d20c4b08 HM |
2720 | if (!rs->dev[i].rdev.bdev) |
2721 | continue; | |
2722 | ||
2723 | q = bdev_get_queue(rs->dev[i].rdev.bdev); | |
48cf06bc HM |
2724 | if (!q || !blk_queue_discard(q)) |
2725 | return; | |
2726 | ||
2727 | if (raid456) { | |
2728 | if (!q->limits.discard_zeroes_data) | |
2729 | return; | |
2730 | if (!devices_handle_discard_safely) { | |
2731 | DMERR("raid456 discard support disabled due to discard_zeroes_data uncertainty."); | |
2732 | DMERR("Set dm-raid.devices_handle_discard_safely=Y to override."); | |
2733 | return; | |
2734 | } | |
2735 | } | |
2736 | } | |
2737 | ||
2738 | /* All RAID members properly support discards */ | |
75b8e04b HM |
2739 | ti->discards_supported = true; |
2740 | ||
2741 | /* | |
2742 | * RAID1 and RAID10 personalities require bio splitting, | |
48cf06bc | 2743 | * RAID0/4/5/6 don't and process large discard bios properly. |
75b8e04b | 2744 | */ |
48cf06bc | 2745 | ti->split_discard_bios = !!(rs->md.level == 1 || rs->md.level == 10); |
75b8e04b HM |
2746 | ti->num_discard_bios = 1; |
2747 | } | |
2748 | ||
9d09e663 | 2749 | /* |
73c6f239 | 2750 | * Construct a RAID0/1/10/4/5/6 mapping: |
9d09e663 | 2751 | * Args: |
43157840 MS |
2752 | * <raid_type> <#raid_params> <raid_params>{0,} \ |
2753 | * <#raid_devs> [<meta_dev1> <dev1>]{1,} | |
9d09e663 | 2754 | * |
43157840 | 2755 | * <raid_params> varies by <raid_type>. See 'parse_raid_params' for |
9d09e663 | 2756 | * details on possible <raid_params>. |
73c6f239 HM |
2757 | * |
2758 | * Userspace is free to initialize the metadata devices, hence the superblocks to | |
2759 | * enforce recreation based on the passed in table parameters. | |
2760 | * | |
9d09e663 | 2761 | */ |
094f394d | 2762 | static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) |
9d09e663 | 2763 | { |
73c6f239 | 2764 | int r; |
469b304b | 2765 | bool resize; |
9d09e663 | 2766 | struct raid_type *rt; |
094f394d | 2767 | unsigned int num_raid_params, num_raid_devs; |
4dff2f1e | 2768 | sector_t calculated_dev_sectors; |
9d09e663 | 2769 | struct raid_set *rs = NULL; |
92c83d79 | 2770 | const char *arg; |
9dbd1aa3 | 2771 | struct rs_layout rs_layout; |
92c83d79 HM |
2772 | struct dm_arg_set as = { argc, argv }, as_nrd; |
2773 | struct dm_arg _args[] = { | |
2774 | { 0, as.argc, "Cannot understand number of raid parameters" }, | |
2775 | { 1, 254, "Cannot understand number of raid devices parameters" } | |
2776 | }; | |
2777 | ||
2778 | /* Must have <raid_type> */ | |
2779 | arg = dm_shift_arg(&as); | |
bd83a4c4 MS |
2780 | if (!arg) { |
2781 | ti->error = "No arguments"; | |
2782 | return -EINVAL; | |
2783 | } | |
9d09e663 | 2784 | |
92c83d79 | 2785 | rt = get_raid_type(arg); |
bd83a4c4 MS |
2786 | if (!rt) { |
2787 | ti->error = "Unrecognised raid_type"; | |
2788 | return -EINVAL; | |
2789 | } | |
9d09e663 | 2790 | |
92c83d79 HM |
2791 | /* Must have <#raid_params> */ |
2792 | if (dm_read_arg_group(_args, &as, &num_raid_params, &ti->error)) | |
43157840 | 2793 | return -EINVAL; |
9d09e663 | 2794 | |
92c83d79 HM |
2795 | /* number of raid device tupples <meta_dev data_dev> */ |
2796 | as_nrd = as; | |
2797 | dm_consume_args(&as_nrd, num_raid_params); | |
2798 | _args[1].max = (as_nrd.argc - 1) / 2; | |
2799 | if (dm_read_arg(_args + 1, &as_nrd, &num_raid_devs, &ti->error)) | |
43157840 | 2800 | return -EINVAL; |
9d09e663 | 2801 | |
bb91a63f | 2802 | if (!__within_range(num_raid_devs, 1, MAX_RAID_DEVICES)) { |
bd83a4c4 MS |
2803 | ti->error = "Invalid number of supplied raid devices"; |
2804 | return -EINVAL; | |
2805 | } | |
3ca5a21a | 2806 | |
bfcee0e3 | 2807 | rs = raid_set_alloc(ti, rt, num_raid_devs); |
9d09e663 N |
2808 | if (IS_ERR(rs)) |
2809 | return PTR_ERR(rs); | |
2810 | ||
92c83d79 | 2811 | r = parse_raid_params(rs, &as, num_raid_params); |
73c6f239 | 2812 | if (r) |
9d09e663 N |
2813 | goto bad; |
2814 | ||
702108d1 | 2815 | r = parse_dev_params(rs, &as); |
73c6f239 | 2816 | if (r) |
9d09e663 N |
2817 | goto bad; |
2818 | ||
b12d437b | 2819 | rs->md.sync_super = super_sync; |
ecbfb9f1 | 2820 | |
2527b56e HM |
2821 | /* |
2822 | * Calculate ctr requested array and device sizes to allow | |
2823 | * for superblock analysis needing device sizes defined. | |
2824 | * | |
2825 | * Any existing superblock will overwrite the array and device sizes | |
2826 | */ | |
40ba37e5 HM |
2827 | r = rs_set_dev_and_array_sectors(rs, false); |
2828 | if (r) | |
b1956dc4 | 2829 | goto bad; |
40ba37e5 | 2830 | |
4dff2f1e HM |
2831 | calculated_dev_sectors = rs->dev[0].rdev.sectors; |
2832 | ||
ecbfb9f1 HM |
2833 | /* |
2834 | * Backup any new raid set level, layout, ... | |
2835 | * requested to be able to compare to superblock | |
2836 | * members for conversion decisions. | |
2837 | */ | |
9dbd1aa3 | 2838 | rs_config_backup(rs, &rs_layout); |
ecbfb9f1 | 2839 | |
73c6f239 HM |
2840 | r = analyse_superblocks(ti, rs); |
2841 | if (r) | |
b12d437b JB |
2842 | goto bad; |
2843 | ||
2a5556c2 | 2844 | resize = calculated_dev_sectors != rs->dev[0].rdev.sectors; |
4dff2f1e | 2845 | |
9d09e663 | 2846 | INIT_WORK(&rs->md.event_work, do_table_event); |
9d09e663 | 2847 | ti->private = rs; |
55a62eef | 2848 | ti->num_flush_bios = 1; |
9d09e663 | 2849 | |
ecbfb9f1 | 2850 | /* Restore any requested new layout for conversion decision */ |
9dbd1aa3 | 2851 | rs_config_restore(rs, &rs_layout); |
ecbfb9f1 | 2852 | |
469b304b HM |
2853 | /* |
2854 | * Now that we have any superblock metadata available, | |
2855 | * check for new, recovering, reshaping, to be taken over, | |
2856 | * to be reshaped or an existing, unchanged raid set to | |
2857 | * run in sequence. | |
2858 | */ | |
9dbd1aa3 | 2859 | if (test_bit(MD_ARRAY_FIRST_USE, &rs->md.flags)) { |
2d92a3c2 HM |
2860 | /* A new raid6 set has to be recovered to ensure proper parity and Q-Syndrome */ |
2861 | if (rs_is_raid6(rs) && | |
2862 | test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { | |
2863 | ti->error = "'nosync' not allowed for new raid6 set"; | |
b1956dc4 HM |
2864 | r = -EINVAL; |
2865 | goto bad; | |
2d92a3c2 HM |
2866 | } |
2867 | rs_setup_recovery(rs, 0); | |
2a5556c2 HM |
2868 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); |
2869 | rs_set_new(rs); | |
2870 | } else if (rs_is_recovering(rs)) { | |
469b304b | 2871 | /* A recovering raid set may be resized */ |
2a5556c2 HM |
2872 | ; /* skip setup rs */ |
2873 | } else if (rs_is_reshaping(rs)) { | |
2874 | /* Have to reject size change request during reshape */ | |
2875 | if (resize) { | |
2876 | ti->error = "Can't resize a reshaping raid set"; | |
b1956dc4 HM |
2877 | r = -EPERM; |
2878 | goto bad; | |
2d92a3c2 | 2879 | } |
469b304b | 2880 | /* skip setup rs */ |
2d92a3c2 | 2881 | } else if (rs_takeover_requested(rs)) { |
9dbd1aa3 HM |
2882 | if (rs_is_reshaping(rs)) { |
2883 | ti->error = "Can't takeover a reshaping raid set"; | |
b1956dc4 HM |
2884 | r = -EPERM; |
2885 | goto bad; | |
9dbd1aa3 HM |
2886 | } |
2887 | ||
2888 | /* | |
2527b56e | 2889 | * If a takeover is needed, userspace sets any additional |
469b304b HM |
2890 | * devices to rebuild and we can check for a valid request here. |
2891 | * | |
2892 | * If acceptible, set the level to the new requested | |
2a5556c2 HM |
2893 | * one, prohibit requesting recovery, allow the raid |
2894 | * set to run and store superblocks during resume. | |
9dbd1aa3 | 2895 | */ |
ecbfb9f1 HM |
2896 | r = rs_check_takeover(rs); |
2897 | if (r) | |
b1956dc4 | 2898 | goto bad; |
ecbfb9f1 HM |
2899 | |
2900 | r = rs_setup_takeover(rs); | |
2901 | if (r) | |
b1956dc4 | 2902 | goto bad; |
ecbfb9f1 | 2903 | |
4286325b | 2904 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); |
6e20902e | 2905 | set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); |
469b304b | 2906 | /* Takeover ain't recovery, so disable recovery */ |
2a5556c2 | 2907 | rs_setup_recovery(rs, MaxSector); |
3a1c1ef2 | 2908 | rs_set_new(rs); |
40ba37e5 | 2909 | } else if (rs_reshape_requested(rs)) { |
9dbd1aa3 | 2910 | /* |
469b304b HM |
2911 | * We can only prepare for a reshape here, because the |
2912 | * raid set needs to run to provide the repective reshape | |
2913 | * check functions via its MD personality instance. | |
2914 | * | |
2915 | * So do the reshape check after md_run() succeeded. | |
2916 | */ | |
2917 | r = rs_prepare_reshape(rs); | |
2918 | if (r) | |
2919 | return r; | |
9dbd1aa3 | 2920 | |
469b304b | 2921 | /* Reshaping ain't recovery, so disable recovery */ |
4dff2f1e | 2922 | rs_setup_recovery(rs, MaxSector); |
3a1c1ef2 | 2923 | rs_set_cur(rs); |
2a5556c2 HM |
2924 | } else { |
2925 | /* May not set recovery when a device rebuild is requested */ | |
37f10be1 HM |
2926 | if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) { |
2927 | rs_setup_recovery(rs, MaxSector); | |
2928 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
2929 | } else | |
2930 | rs_setup_recovery(rs, test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) ? | |
2a5556c2 HM |
2931 | 0 : (resize ? calculated_dev_sectors : MaxSector)); |
2932 | rs_set_cur(rs); | |
4dff2f1e | 2933 | } |
ecbfb9f1 | 2934 | |
40ba37e5 HM |
2935 | /* If constructor requested it, change data and new_data offsets */ |
2936 | r = rs_adjust_data_offsets(rs); | |
2937 | if (r) | |
b1956dc4 | 2938 | goto bad; |
40ba37e5 | 2939 | |
ecbfb9f1 HM |
2940 | /* Start raid set read-only and assumed clean to change in raid_resume() */ |
2941 | rs->md.ro = 1; | |
2942 | rs->md.in_sync = 1; | |
2943 | set_bit(MD_RECOVERY_FROZEN, &rs->md.recovery); | |
75b8e04b | 2944 | |
0cf45031 HM |
2945 | /* Has to be held on running the array */ |
2946 | mddev_lock_nointr(&rs->md); | |
73c6f239 | 2947 | r = md_run(&rs->md); |
9d09e663 | 2948 | rs->md.in_sync = 0; /* Assume already marked dirty */ |
9d09e663 | 2949 | |
73c6f239 | 2950 | if (r) { |
9dbd1aa3 HM |
2951 | ti->error = "Failed to run raid array"; |
2952 | mddev_unlock(&rs->md); | |
9d09e663 N |
2953 | goto bad; |
2954 | } | |
2955 | ||
2956 | rs->callbacks.congested_fn = raid_is_congested; | |
9d09e663 N |
2957 | dm_table_add_target_callbacks(ti->table, &rs->callbacks); |
2958 | ||
32737279 | 2959 | mddev_suspend(&rs->md); |
9dbd1aa3 HM |
2960 | |
2961 | /* Try to adjust the raid4/5/6 stripe cache size to the stripe size */ | |
2962 | if (rs_is_raid456(rs)) { | |
2963 | r = rs_set_raid456_stripe_cache(rs); | |
2964 | if (r) | |
2965 | goto bad_stripe_cache; | |
2966 | } | |
2967 | ||
2968 | /* Now do an early reshape check */ | |
2969 | if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { | |
2970 | r = rs_check_reshape(rs); | |
2971 | if (r) | |
b1956dc4 | 2972 | goto bad_check_reshape; |
9dbd1aa3 HM |
2973 | |
2974 | /* Restore new, ctr requested layout to perform check */ | |
2975 | rs_config_restore(rs, &rs_layout); | |
2976 | ||
7a7c330f HM |
2977 | if (rs->md.pers->start_reshape) { |
2978 | r = rs->md.pers->check_reshape(&rs->md); | |
2979 | if (r) { | |
2980 | ti->error = "Reshape check failed"; | |
2981 | goto bad_check_reshape; | |
2982 | } | |
9dbd1aa3 HM |
2983 | } |
2984 | } | |
2985 | ||
2986 | mddev_unlock(&rs->md); | |
9d09e663 N |
2987 | return 0; |
2988 | ||
9dbd1aa3 HM |
2989 | bad_stripe_cache: |
2990 | bad_check_reshape: | |
63f33b8d | 2991 | md_stop(&rs->md); |
9d09e663 | 2992 | bad: |
bfcee0e3 | 2993 | raid_set_free(rs); |
9d09e663 | 2994 | |
73c6f239 | 2995 | return r; |
9d09e663 N |
2996 | } |
2997 | ||
2998 | static void raid_dtr(struct dm_target *ti) | |
2999 | { | |
3000 | struct raid_set *rs = ti->private; | |
3001 | ||
3002 | list_del_init(&rs->callbacks.list); | |
3003 | md_stop(&rs->md); | |
bfcee0e3 | 3004 | raid_set_free(rs); |
9d09e663 N |
3005 | } |
3006 | ||
7de3ee57 | 3007 | static int raid_map(struct dm_target *ti, struct bio *bio) |
9d09e663 N |
3008 | { |
3009 | struct raid_set *rs = ti->private; | |
fd01b88c | 3010 | struct mddev *mddev = &rs->md; |
9d09e663 | 3011 | |
9dbd1aa3 HM |
3012 | /* |
3013 | * If we're reshaping to add disk(s)), ti->len and | |
3014 | * mddev->array_sectors will differ during the process | |
3015 | * (ti->len > mddev->array_sectors), so we have to requeue | |
3016 | * bios with addresses > mddev->array_sectors here or | |
2527b56e | 3017 | * there will occur accesses past EOD of the component |
9dbd1aa3 HM |
3018 | * data images thus erroring the raid set. |
3019 | */ | |
3020 | if (unlikely(bio_end_sector(bio) > mddev->array_sectors)) | |
3021 | return DM_MAPIO_REQUEUE; | |
3022 | ||
9d09e663 N |
3023 | mddev->pers->make_request(mddev, bio); |
3024 | ||
3025 | return DM_MAPIO_SUBMITTED; | |
3026 | } | |
3027 | ||
3a1c1ef2 | 3028 | /* Return string describing the current sync action of @mddev */ |
be83651f JB |
3029 | static const char *decipher_sync_action(struct mddev *mddev) |
3030 | { | |
3031 | if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) | |
3032 | return "frozen"; | |
3033 | ||
3034 | if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || | |
3035 | (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))) { | |
3036 | if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) | |
3037 | return "reshape"; | |
3038 | ||
3039 | if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { | |
3040 | if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) | |
3041 | return "resync"; | |
3042 | else if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) | |
3043 | return "check"; | |
3044 | return "repair"; | |
3045 | } | |
3046 | ||
3047 | if (test_bit(MD_RECOVERY_RECOVER, &mddev->recovery)) | |
3048 | return "recover"; | |
3049 | } | |
3050 | ||
3051 | return "idle"; | |
3052 | } | |
3053 | ||
3a1c1ef2 HM |
3054 | /* |
3055 | * Return status string @rdev | |
3056 | * | |
3057 | * Status characters: | |
3058 | * | |
3059 | * 'D' = Dead/Failed device | |
3060 | * 'a' = Alive but not in-sync | |
3061 | * 'A' = Alive and in-sync | |
3062 | */ | |
e6ca5e1a | 3063 | static const char *__raid_dev_status(struct md_rdev *rdev, bool array_in_sync) |
9d09e663 | 3064 | { |
3a1c1ef2 HM |
3065 | if (test_bit(Faulty, &rdev->flags)) |
3066 | return "D"; | |
3067 | else if (!array_in_sync || !test_bit(In_sync, &rdev->flags)) | |
3068 | return "a"; | |
3069 | else | |
3070 | return "A"; | |
3071 | } | |
9d09e663 | 3072 | |
3a1c1ef2 HM |
3073 | /* Helper to return resync/reshape progress for @rs and @array_in_sync */ |
3074 | static sector_t rs_get_progress(struct raid_set *rs, | |
3075 | sector_t resync_max_sectors, bool *array_in_sync) | |
3076 | { | |
3077 | sector_t r, recovery_cp, curr_resync_completed; | |
3078 | struct mddev *mddev = &rs->md; | |
9d09e663 | 3079 | |
3a1c1ef2 HM |
3080 | curr_resync_completed = mddev->curr_resync_completed ?: mddev->recovery_cp; |
3081 | recovery_cp = mddev->recovery_cp; | |
3082 | *array_in_sync = false; | |
3083 | ||
3084 | if (rs_is_raid0(rs)) { | |
3085 | r = resync_max_sectors; | |
3086 | *array_in_sync = true; | |
3087 | ||
3088 | } else { | |
3089 | r = mddev->reshape_position; | |
3090 | ||
3091 | /* Reshape is relative to the array size */ | |
3092 | if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) || | |
3093 | r != MaxSector) { | |
3094 | if (r == MaxSector) { | |
3095 | *array_in_sync = true; | |
3096 | r = resync_max_sectors; | |
0cf45031 | 3097 | } else { |
3a1c1ef2 HM |
3098 | /* Got to reverse on backward reshape */ |
3099 | if (mddev->reshape_backwards) | |
3100 | r = mddev->array_sectors - r; | |
3101 | ||
3102 | /* Devide by # of data stripes */ | |
3103 | sector_div(r, mddev_data_stripes(rs)); | |
0cf45031 | 3104 | } |
3a1c1ef2 HM |
3105 | |
3106 | /* Sync is relative to the component device size */ | |
3107 | } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) | |
3108 | r = curr_resync_completed; | |
3109 | else | |
3110 | r = recovery_cp; | |
3111 | ||
3112 | if (r == MaxSector) { | |
3113 | /* | |
3114 | * Sync complete. | |
3115 | */ | |
3116 | *array_in_sync = true; | |
3117 | r = resync_max_sectors; | |
3118 | } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) { | |
3119 | /* | |
3120 | * If "check" or "repair" is occurring, the raid set has | |
3121 | * undergone an initial sync and the health characters | |
3122 | * should not be 'a' anymore. | |
3123 | */ | |
3124 | *array_in_sync = true; | |
0cf45031 | 3125 | } else { |
3a1c1ef2 | 3126 | struct md_rdev *rdev; |
be83651f | 3127 | |
3a1c1ef2 HM |
3128 | /* |
3129 | * The raid set may be doing an initial sync, or it may | |
43157840 | 3130 | * be rebuilding individual components. If all the |
3a1c1ef2 HM |
3131 | * devices are In_sync, then it is the raid set that is |
3132 | * being initialized. | |
3133 | */ | |
3134 | rdev_for_each(rdev, mddev) | |
3135 | if (!test_bit(In_sync, &rdev->flags)) | |
3136 | *array_in_sync = true; | |
3137 | #if 0 | |
3138 | r = 0; /* HM FIXME: TESTME: https://bugzilla.redhat.com/show_bug.cgi?id=1210637 ? */ | |
3139 | #endif | |
2e727c3c | 3140 | } |
3a1c1ef2 HM |
3141 | } |
3142 | ||
3143 | return r; | |
3144 | } | |
3145 | ||
3146 | /* Helper to return @dev name or "-" if !@dev */ | |
e6ca5e1a | 3147 | static const char *__get_dev_name(struct dm_dev *dev) |
3a1c1ef2 HM |
3148 | { |
3149 | return dev ? dev->name : "-"; | |
3150 | } | |
3151 | ||
3152 | static void raid_status(struct dm_target *ti, status_type_t type, | |
3153 | unsigned int status_flags, char *result, unsigned int maxlen) | |
3154 | { | |
3155 | struct raid_set *rs = ti->private; | |
3156 | struct mddev *mddev = &rs->md; | |
3157 | struct r5conf *conf = mddev->private; | |
7a7c330f | 3158 | int i, max_nr_stripes = conf ? conf->max_nr_stripes : 0; |
3a1c1ef2 HM |
3159 | bool array_in_sync; |
3160 | unsigned int raid_param_cnt = 1; /* at least 1 for chunksize */ | |
3161 | unsigned int sz = 0; | |
7a7c330f | 3162 | unsigned int rebuild_disks; |
3a1c1ef2 HM |
3163 | unsigned int write_mostly_params = 0; |
3164 | sector_t progress, resync_max_sectors, resync_mismatches; | |
3165 | const char *sync_action; | |
3166 | struct raid_type *rt; | |
3167 | struct md_rdev *rdev; | |
3168 | ||
3169 | switch (type) { | |
3170 | case STATUSTYPE_INFO: | |
3171 | /* *Should* always succeed */ | |
3172 | rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); | |
3173 | if (!rt) | |
3174 | return; | |
3175 | ||
9dbd1aa3 | 3176 | DMEMIT("%s %d ", rt->name, mddev->raid_disks); |
3a1c1ef2 HM |
3177 | |
3178 | /* Access most recent mddev properties for status output */ | |
3179 | smp_rmb(); | |
3180 | /* Get sensible max sectors even if raid set not yet started */ | |
4286325b | 3181 | resync_max_sectors = test_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags) ? |
3a1c1ef2 HM |
3182 | mddev->resync_max_sectors : mddev->dev_sectors; |
3183 | progress = rs_get_progress(rs, resync_max_sectors, &array_in_sync); | |
3184 | resync_mismatches = (mddev->last_sync_action && !strcasecmp(mddev->last_sync_action, "check")) ? | |
9dbd1aa3 | 3185 | atomic64_read(&mddev->resync_mismatches) : 0; |
3a1c1ef2 HM |
3186 | sync_action = decipher_sync_action(&rs->md); |
3187 | ||
3188 | /* HM FIXME: do we want another state char for raid0? It shows 'D' or 'A' now */ | |
32682409 HM |
3189 | rdev_for_each(rdev, mddev) |
3190 | DMEMIT(__raid_dev_status(rdev, array_in_sync)); | |
9d09e663 | 3191 | |
2e727c3c | 3192 | /* |
3a1c1ef2 | 3193 | * In-sync/Reshape ratio: |
2e727c3c | 3194 | * The in-sync ratio shows the progress of: |
3a1c1ef2 HM |
3195 | * - Initializing the raid set |
3196 | * - Rebuilding a subset of devices of the raid set | |
2e727c3c JB |
3197 | * The user can distinguish between the two by referring |
3198 | * to the status characters. | |
3a1c1ef2 HM |
3199 | * |
3200 | * The reshape ratio shows the progress of | |
3201 | * changing the raid layout or the number of | |
3202 | * disks of a raid set | |
2e727c3c | 3203 | */ |
3a1c1ef2 HM |
3204 | DMEMIT(" %llu/%llu", (unsigned long long) progress, |
3205 | (unsigned long long) resync_max_sectors); | |
9d09e663 | 3206 | |
be83651f | 3207 | /* |
3a1c1ef2 HM |
3208 | * v1.5.0+: |
3209 | * | |
be83651f | 3210 | * Sync action: |
3a1c1ef2 | 3211 | * See Documentation/device-mapper/dm-raid.txt for |
be83651f JB |
3212 | * information on each of these states. |
3213 | */ | |
3a1c1ef2 | 3214 | DMEMIT(" %s", sync_action); |
be83651f JB |
3215 | |
3216 | /* | |
3a1c1ef2 HM |
3217 | * v1.5.0+: |
3218 | * | |
be83651f JB |
3219 | * resync_mismatches/mismatch_cnt |
3220 | * This field shows the number of discrepancies found when | |
3a1c1ef2 | 3221 | * performing a "check" of the raid set. |
be83651f | 3222 | */ |
3a1c1ef2 | 3223 | DMEMIT(" %llu", (unsigned long long) resync_mismatches); |
9d09e663 | 3224 | |
3a1c1ef2 | 3225 | /* |
9b6e5423 | 3226 | * v1.9.0+: |
3a1c1ef2 HM |
3227 | * |
3228 | * data_offset (needed for out of space reshaping) | |
3229 | * This field shows the data offset into the data | |
3230 | * image LV where the first stripes data starts. | |
3231 | * | |
3232 | * We keep data_offset equal on all raid disks of the set, | |
3233 | * so retrieving it from the first raid disk is sufficient. | |
3234 | */ | |
3235 | DMEMIT(" %llu", (unsigned long long) rs->dev[0].rdev.data_offset); | |
3236 | break; | |
9d09e663 | 3237 | |
3a1c1ef2 HM |
3238 | case STATUSTYPE_TABLE: |
3239 | /* Report the table line string you would use to construct this raid set */ | |
3240 | ||
3241 | /* Calculate raid parameter count */ | |
7a7c330f HM |
3242 | for (i = 0; i < rs->raid_disks; i++) |
3243 | if (test_bit(WriteMostly, &rs->dev[i].rdev.flags)) | |
3a1c1ef2 | 3244 | write_mostly_params += 2; |
7a7c330f HM |
3245 | rebuild_disks = memweight(rs->rebuild_disks, DISKS_ARRAY_ELEMS * sizeof(*rs->rebuild_disks)); |
3246 | raid_param_cnt += rebuild_disks * 2 + | |
3a1c1ef2 HM |
3247 | write_mostly_params + |
3248 | hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_NO_ARGS) + | |
3249 | hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2; | |
3250 | /* Emit table line */ | |
3251 | DMEMIT("%s %u %u", rs->raid_type->name, raid_param_cnt, mddev->new_chunk_sectors); | |
4286325b | 3252 | if (test_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) |
3fa6cf38 | 3253 | DMEMIT(" %s %s", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT), |
3a1c1ef2 | 3254 | raid10_md_layout_to_format(mddev->layout)); |
4286325b | 3255 | if (test_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) |
3fa6cf38 | 3256 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES), |
3a1c1ef2 | 3257 | raid10_md_layout_to_copies(mddev->layout)); |
4286325b | 3258 | if (test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) |
3fa6cf38 | 3259 | DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC)); |
4286325b | 3260 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) |
3fa6cf38 | 3261 | DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_SYNC)); |
4286325b | 3262 | if (test_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) |
3fa6cf38 | 3263 | DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE), |
3a1c1ef2 | 3264 | (unsigned long long) to_sector(mddev->bitmap_info.chunksize)); |
4286325b | 3265 | if (test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) |
3fa6cf38 | 3266 | DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET), |
3a1c1ef2 | 3267 | (unsigned long long) rs->data_offset); |
4286325b | 3268 | if (test_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) |
3fa6cf38 | 3269 | DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP), |
3a1c1ef2 | 3270 | mddev->bitmap_info.daemon_sleep); |
4286325b | 3271 | if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) |
3fa6cf38 | 3272 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS), |
7a7c330f | 3273 | max(rs->delta_disks, mddev->delta_disks)); |
4286325b | 3274 | if (test_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) |
3fa6cf38 | 3275 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE), |
3a1c1ef2 | 3276 | max_nr_stripes); |
7a7c330f HM |
3277 | if (rebuild_disks) |
3278 | for (i = 0; i < rs->raid_disks; i++) | |
3279 | if (test_bit(rs->dev[i].rdev.raid_disk, (void *) rs->rebuild_disks)) | |
3280 | DMEMIT(" %s %u", dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD), | |
3281 | rs->dev[i].rdev.raid_disk); | |
3282 | if (write_mostly_params) | |
3283 | for (i = 0; i < rs->raid_disks; i++) | |
3284 | if (test_bit(WriteMostly, &rs->dev[i].rdev.flags)) | |
3285 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY), | |
3286 | rs->dev[i].rdev.raid_disk); | |
4286325b | 3287 | if (test_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) |
3fa6cf38 | 3288 | DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND), |
3a1c1ef2 | 3289 | mddev->bitmap_info.max_write_behind); |
4286325b | 3290 | if (test_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) |
3fa6cf38 | 3291 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE), |
3a1c1ef2 | 3292 | mddev->sync_speed_max); |
4286325b | 3293 | if (test_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) |
3fa6cf38 | 3294 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE), |
3a1c1ef2 HM |
3295 | mddev->sync_speed_min); |
3296 | DMEMIT(" %d", rs->raid_disks); | |
7a7c330f HM |
3297 | for (i = 0; i < rs->raid_disks; i++) |
3298 | DMEMIT(" %s %s", __get_dev_name(rs->dev[i].meta_dev), | |
3299 | __get_dev_name(rs->dev[i].data_dev)); | |
9d09e663 | 3300 | } |
9d09e663 N |
3301 | } |
3302 | ||
094f394d | 3303 | static int raid_message(struct dm_target *ti, unsigned int argc, char **argv) |
be83651f JB |
3304 | { |
3305 | struct raid_set *rs = ti->private; | |
3306 | struct mddev *mddev = &rs->md; | |
3307 | ||
be83651f JB |
3308 | if (!mddev->pers || !mddev->pers->sync_request) |
3309 | return -EINVAL; | |
3310 | ||
3311 | if (!strcasecmp(argv[0], "frozen")) | |
3312 | set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3313 | else | |
3314 | clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3315 | ||
3316 | if (!strcasecmp(argv[0], "idle") || !strcasecmp(argv[0], "frozen")) { | |
3317 | if (mddev->sync_thread) { | |
3318 | set_bit(MD_RECOVERY_INTR, &mddev->recovery); | |
3319 | md_reap_sync_thread(mddev); | |
3320 | } | |
3321 | } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || | |
3322 | test_bit(MD_RECOVERY_NEEDED, &mddev->recovery)) | |
3323 | return -EBUSY; | |
3324 | else if (!strcasecmp(argv[0], "resync")) | |
3a1c1ef2 HM |
3325 | ; /* MD_RECOVERY_NEEDED set below */ |
3326 | else if (!strcasecmp(argv[0], "recover")) | |
be83651f | 3327 | set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); |
3a1c1ef2 | 3328 | else { |
be83651f JB |
3329 | if (!strcasecmp(argv[0], "check")) |
3330 | set_bit(MD_RECOVERY_CHECK, &mddev->recovery); | |
3331 | else if (!!strcasecmp(argv[0], "repair")) | |
3332 | return -EINVAL; | |
3333 | set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); | |
3334 | set_bit(MD_RECOVERY_SYNC, &mddev->recovery); | |
3335 | } | |
3336 | if (mddev->ro == 2) { | |
3337 | /* A write to sync_action is enough to justify | |
3338 | * canceling read-auto mode | |
3339 | */ | |
3340 | mddev->ro = 0; | |
3a1c1ef2 | 3341 | if (!mddev->suspended && mddev->sync_thread) |
be83651f JB |
3342 | md_wakeup_thread(mddev->sync_thread); |
3343 | } | |
3344 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
3a1c1ef2 | 3345 | if (!mddev->suspended && mddev->thread) |
be83651f JB |
3346 | md_wakeup_thread(mddev->thread); |
3347 | ||
3348 | return 0; | |
3349 | } | |
3350 | ||
3351 | static int raid_iterate_devices(struct dm_target *ti, | |
3352 | iterate_devices_callout_fn fn, void *data) | |
9d09e663 N |
3353 | { |
3354 | struct raid_set *rs = ti->private; | |
094f394d | 3355 | unsigned int i; |
73c6f239 | 3356 | int r = 0; |
9d09e663 | 3357 | |
73c6f239 | 3358 | for (i = 0; !r && i < rs->md.raid_disks; i++) |
9d09e663 | 3359 | if (rs->dev[i].data_dev) |
73c6f239 | 3360 | r = fn(ti, |
9d09e663 N |
3361 | rs->dev[i].data_dev, |
3362 | 0, /* No offset on data devs */ | |
3363 | rs->md.dev_sectors, | |
3364 | data); | |
3365 | ||
73c6f239 | 3366 | return r; |
9d09e663 N |
3367 | } |
3368 | ||
3369 | static void raid_io_hints(struct dm_target *ti, struct queue_limits *limits) | |
3370 | { | |
3371 | struct raid_set *rs = ti->private; | |
89d3d9a1 | 3372 | unsigned int chunk_size = to_bytes(rs->md.chunk_sectors); |
9d09e663 N |
3373 | |
3374 | blk_limits_io_min(limits, chunk_size); | |
89d3d9a1 | 3375 | blk_limits_io_opt(limits, chunk_size * mddev_data_stripes(rs)); |
9d09e663 N |
3376 | } |
3377 | ||
3378 | static void raid_presuspend(struct dm_target *ti) | |
3379 | { | |
3380 | struct raid_set *rs = ti->private; | |
3381 | ||
3382 | md_stop_writes(&rs->md); | |
3383 | } | |
3384 | ||
3385 | static void raid_postsuspend(struct dm_target *ti) | |
3386 | { | |
3387 | struct raid_set *rs = ti->private; | |
3388 | ||
6e20902e HM |
3389 | if (test_and_clear_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { |
3390 | if (!rs->md.suspended) | |
3391 | mddev_suspend(&rs->md); | |
3392 | rs->md.ro = 1; | |
3393 | } | |
9d09e663 N |
3394 | } |
3395 | ||
f381e71b | 3396 | static void attempt_restore_of_faulty_devices(struct raid_set *rs) |
9d09e663 | 3397 | { |
9092c02d JB |
3398 | int i; |
3399 | uint64_t failed_devices, cleared_failed_devices = 0; | |
3400 | unsigned long flags; | |
3401 | struct dm_raid_superblock *sb; | |
9092c02d | 3402 | struct md_rdev *r; |
9d09e663 | 3403 | |
f381e71b JB |
3404 | for (i = 0; i < rs->md.raid_disks; i++) { |
3405 | r = &rs->dev[i].rdev; | |
3406 | if (test_bit(Faulty, &r->flags) && r->sb_page && | |
0a7b8188 HM |
3407 | sync_page_io(r, 0, r->sb_size, r->sb_page, |
3408 | REQ_OP_READ, 0, true)) { | |
f381e71b JB |
3409 | DMINFO("Faulty %s device #%d has readable super block." |
3410 | " Attempting to revive it.", | |
3411 | rs->raid_type->name, i); | |
a4dc163a JB |
3412 | |
3413 | /* | |
3414 | * Faulty bit may be set, but sometimes the array can | |
3415 | * be suspended before the personalities can respond | |
3416 | * by removing the device from the array (i.e. calling | |
43157840 | 3417 | * 'hot_remove_disk'). If they haven't yet removed |
a4dc163a JB |
3418 | * the failed device, its 'raid_disk' number will be |
3419 | * '>= 0' - meaning we must call this function | |
3420 | * ourselves. | |
3421 | */ | |
3422 | if ((r->raid_disk >= 0) && | |
3423 | (r->mddev->pers->hot_remove_disk(r->mddev, r) != 0)) | |
3424 | /* Failed to revive this device, try next */ | |
3425 | continue; | |
3426 | ||
f381e71b JB |
3427 | r->raid_disk = i; |
3428 | r->saved_raid_disk = i; | |
3429 | flags = r->flags; | |
3430 | clear_bit(Faulty, &r->flags); | |
3431 | clear_bit(WriteErrorSeen, &r->flags); | |
3432 | clear_bit(In_sync, &r->flags); | |
3433 | if (r->mddev->pers->hot_add_disk(r->mddev, r)) { | |
3434 | r->raid_disk = -1; | |
3435 | r->saved_raid_disk = -1; | |
3436 | r->flags = flags; | |
3437 | } else { | |
3438 | r->recovery_offset = 0; | |
3439 | cleared_failed_devices |= 1 << i; | |
3440 | } | |
3441 | } | |
3442 | } | |
3443 | if (cleared_failed_devices) { | |
3444 | rdev_for_each(r, &rs->md) { | |
3445 | sb = page_address(r->sb_page); | |
3446 | failed_devices = le64_to_cpu(sb->failed_devices); | |
3447 | failed_devices &= ~cleared_failed_devices; | |
3448 | sb->failed_devices = cpu_to_le64(failed_devices); | |
3449 | } | |
3450 | } | |
3451 | } | |
3452 | ||
e6ca5e1a | 3453 | static int __load_dirty_region_bitmap(struct raid_set *rs) |
ecbfb9f1 HM |
3454 | { |
3455 | int r = 0; | |
3456 | ||
3457 | /* Try loading the bitmap unless "raid0", which does not have one */ | |
3458 | if (!rs_is_raid0(rs) && | |
4286325b | 3459 | !test_and_set_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags)) { |
ecbfb9f1 HM |
3460 | r = bitmap_load(&rs->md); |
3461 | if (r) | |
3462 | DMERR("Failed to load bitmap"); | |
3463 | } | |
3464 | ||
3465 | return r; | |
3466 | } | |
3467 | ||
6e20902e HM |
3468 | /* Enforce updating all superblocks */ |
3469 | static void rs_update_sbs(struct raid_set *rs) | |
3470 | { | |
3471 | struct mddev *mddev = &rs->md; | |
3472 | int ro = mddev->ro; | |
3473 | ||
3474 | set_bit(MD_CHANGE_DEVS, &mddev->flags); | |
3475 | mddev->ro = 0; | |
3476 | md_update_sb(mddev, 1); | |
3477 | mddev->ro = ro; | |
3478 | } | |
3479 | ||
9dbd1aa3 HM |
3480 | /* |
3481 | * Reshape changes raid algorithm of @rs to new one within personality | |
3482 | * (e.g. raid6_zr -> raid6_nc), changes stripe size, adds/removes | |
3483 | * disks from a raid set thus growing/shrinking it or resizes the set | |
3484 | * | |
3485 | * Call mddev_lock_nointr() before! | |
3486 | */ | |
3487 | static int rs_start_reshape(struct raid_set *rs) | |
3488 | { | |
3489 | int r; | |
3490 | struct mddev *mddev = &rs->md; | |
3491 | struct md_personality *pers = mddev->pers; | |
3492 | ||
3493 | r = rs_setup_reshape(rs); | |
3494 | if (r) | |
3495 | return r; | |
3496 | ||
3497 | /* Need to be resumed to be able to start reshape, recovery is frozen until raid_resume() though */ | |
3498 | if (mddev->suspended) | |
3499 | mddev_resume(mddev); | |
3500 | ||
3501 | /* | |
3502 | * Check any reshape constraints enforced by the personalility | |
3503 | * | |
3504 | * May as well already kick the reshape off so that * pers->start_reshape() becomes optional. | |
3505 | */ | |
3506 | r = pers->check_reshape(mddev); | |
3507 | if (r) { | |
3508 | rs->ti->error = "pers->check_reshape() failed"; | |
3509 | return r; | |
3510 | } | |
3511 | ||
3512 | /* | |
3513 | * Personality may not provide start reshape method in which | |
3514 | * case check_reshape above has already covered everything | |
3515 | */ | |
3516 | if (pers->start_reshape) { | |
3517 | r = pers->start_reshape(mddev); | |
3518 | if (r) { | |
3519 | rs->ti->error = "pers->start_reshape() failed"; | |
3520 | return r; | |
3521 | } | |
3522 | } | |
3523 | ||
3524 | /* Suspend because a resume will happen in raid_resume() */ | |
3525 | if (!mddev->suspended) | |
3526 | mddev_suspend(mddev); | |
3527 | ||
6e20902e HM |
3528 | /* |
3529 | * Now reshape got set up, update superblocks to | |
3530 | * reflect the fact so that a table reload will | |
3531 | * access proper superblock content in the ctr. | |
3532 | */ | |
3533 | rs_update_sbs(rs); | |
9dbd1aa3 HM |
3534 | |
3535 | return 0; | |
3536 | } | |
3537 | ||
ecbfb9f1 HM |
3538 | static int raid_preresume(struct dm_target *ti) |
3539 | { | |
9dbd1aa3 | 3540 | int r; |
ecbfb9f1 HM |
3541 | struct raid_set *rs = ti->private; |
3542 | struct mddev *mddev = &rs->md; | |
3543 | ||
3544 | /* This is a resume after a suspend of the set -> it's already started */ | |
4286325b | 3545 | if (test_and_set_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags)) |
ecbfb9f1 HM |
3546 | return 0; |
3547 | ||
3548 | /* | |
3549 | * The superblocks need to be updated on disk if the | |
6e20902e HM |
3550 | * array is new or new devices got added (thus zeroed |
3551 | * out by userspace) or __load_dirty_region_bitmap | |
3552 | * will overwrite them in core with old data or fail. | |
ecbfb9f1 | 3553 | */ |
6e20902e HM |
3554 | if (test_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags)) |
3555 | rs_update_sbs(rs); | |
ecbfb9f1 HM |
3556 | |
3557 | /* | |
3558 | * Disable/enable discard support on raid set after any | |
3559 | * conversion, because devices can have been added | |
3560 | */ | |
3561 | configure_discard_support(rs); | |
3562 | ||
3563 | /* Load the bitmap from disk unless raid0 */ | |
9dbd1aa3 HM |
3564 | r = __load_dirty_region_bitmap(rs); |
3565 | if (r) | |
3566 | return r; | |
3567 | ||
4257e085 HM |
3568 | /* Resize bitmap to adjust to changed region size (aka MD bitmap chunksize) */ |
3569 | if (test_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags) && | |
3570 | mddev->bitmap_info.chunksize != to_bytes(rs->requested_bitmap_chunk_sectors)) { | |
3571 | r = bitmap_resize(mddev->bitmap, mddev->dev_sectors, | |
3572 | to_bytes(rs->requested_bitmap_chunk_sectors), 0); | |
3573 | if (r) | |
3574 | DMERR("Failed to resize bitmap"); | |
3575 | } | |
3576 | ||
9dbd1aa3 HM |
3577 | /* Check for any resize/reshape on @rs and adjust/initiate */ |
3578 | /* Be prepared for mddev_resume() in raid_resume() */ | |
3579 | set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3580 | if (mddev->recovery_cp && mddev->recovery_cp < MaxSector) { | |
9dbd1aa3 HM |
3581 | set_bit(MD_RECOVERY_SYNC, &mddev->recovery); |
3582 | mddev->resync_min = mddev->recovery_cp; | |
3583 | } | |
3584 | ||
3585 | rs_set_capacity(rs); | |
3586 | ||
345a6cdc | 3587 | /* Check for any reshape request unless new raid set */ |
9dbd1aa3 HM |
3588 | if (test_and_clear_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { |
3589 | /* Initiate a reshape. */ | |
3590 | mddev_lock_nointr(mddev); | |
3591 | r = rs_start_reshape(rs); | |
3592 | mddev_unlock(mddev); | |
3593 | if (r) | |
3594 | DMWARN("Failed to check/start reshape, continuing without change"); | |
3595 | r = 0; | |
3596 | } | |
3597 | ||
3598 | return r; | |
ecbfb9f1 HM |
3599 | } |
3600 | ||
f381e71b JB |
3601 | static void raid_resume(struct dm_target *ti) |
3602 | { | |
3603 | struct raid_set *rs = ti->private; | |
ecbfb9f1 | 3604 | struct mddev *mddev = &rs->md; |
f381e71b | 3605 | |
4286325b | 3606 | if (test_and_set_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { |
ecbfb9f1 HM |
3607 | /* |
3608 | * A secondary resume while the device is active. | |
3609 | * Take this opportunity to check whether any failed | |
3610 | * devices are reachable again. | |
3611 | */ | |
3612 | attempt_restore_of_faulty_devices(rs); | |
6e20902e HM |
3613 | } else { |
3614 | mddev->ro = 0; | |
3615 | mddev->in_sync = 0; | |
34f8ac6d | 3616 | |
6e20902e HM |
3617 | /* |
3618 | * When passing in flags to the ctr, we expect userspace | |
3619 | * to reset them because they made it to the superblocks | |
3620 | * and reload the mapping anyway. | |
3621 | * | |
3622 | * -> only unfreeze recovery in case of a table reload or | |
3623 | * we'll have a bogus recovery/reshape position | |
3624 | * retrieved from the superblock by the ctr because | |
3625 | * the ongoing recovery/reshape will change it after read. | |
3626 | */ | |
3627 | if (!test_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags)) | |
3628 | clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3a1c1ef2 | 3629 | |
6e20902e HM |
3630 | if (mddev->suspended) |
3631 | mddev_resume(mddev); | |
3632 | } | |
9d09e663 N |
3633 | } |
3634 | ||
3635 | static struct target_type raid_target = { | |
3636 | .name = "raid", | |
9b6e5423 | 3637 | .version = {1, 9, 0}, |
9d09e663 N |
3638 | .module = THIS_MODULE, |
3639 | .ctr = raid_ctr, | |
3640 | .dtr = raid_dtr, | |
3641 | .map = raid_map, | |
3642 | .status = raid_status, | |
be83651f | 3643 | .message = raid_message, |
9d09e663 N |
3644 | .iterate_devices = raid_iterate_devices, |
3645 | .io_hints = raid_io_hints, | |
3646 | .presuspend = raid_presuspend, | |
3647 | .postsuspend = raid_postsuspend, | |
ecbfb9f1 | 3648 | .preresume = raid_preresume, |
9d09e663 N |
3649 | .resume = raid_resume, |
3650 | }; | |
3651 | ||
3652 | static int __init dm_raid_init(void) | |
3653 | { | |
fe5d2f4a JB |
3654 | DMINFO("Loading target version %u.%u.%u", |
3655 | raid_target.version[0], | |
3656 | raid_target.version[1], | |
3657 | raid_target.version[2]); | |
9d09e663 N |
3658 | return dm_register_target(&raid_target); |
3659 | } | |
3660 | ||
3661 | static void __exit dm_raid_exit(void) | |
3662 | { | |
3663 | dm_unregister_target(&raid_target); | |
3664 | } | |
3665 | ||
3666 | module_init(dm_raid_init); | |
3667 | module_exit(dm_raid_exit); | |
3668 | ||
48cf06bc HM |
3669 | module_param(devices_handle_discard_safely, bool, 0644); |
3670 | MODULE_PARM_DESC(devices_handle_discard_safely, | |
3671 | "Set to Y if all devices in each array reliably return zeroes on reads from discarded regions"); | |
3672 | ||
ef9b85a6 MS |
3673 | MODULE_DESCRIPTION(DM_NAME " raid0/1/10/4/5/6 target"); |
3674 | MODULE_ALIAS("dm-raid0"); | |
63f33b8d JB |
3675 | MODULE_ALIAS("dm-raid1"); |
3676 | MODULE_ALIAS("dm-raid10"); | |
9d09e663 N |
3677 | MODULE_ALIAS("dm-raid4"); |
3678 | MODULE_ALIAS("dm-raid5"); | |
3679 | MODULE_ALIAS("dm-raid6"); | |
3680 | MODULE_AUTHOR("Neil Brown <dm-devel@redhat.com>"); | |
3a1c1ef2 | 3681 | MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@redhat.com>"); |
9d09e663 | 3682 | MODULE_LICENSE("GPL"); |