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 */ |
4286325b MS |
72 | #define __CTR_FLAG_DELTA_DISKS 12 /* 2 */ /* Only with reshapable raid4/5/6/10! */ |
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 | \ | |
135 | CTR_FLAG_DATA_OFFSET) | |
f090279e | 136 | |
a30cbc0d HM |
137 | /* "raid10" does not accept any raid1 or stripe cache options */ |
138 | #define RAID10_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ | |
139 | CTR_FLAG_REBUILD | \ | |
140 | CTR_FLAG_DAEMON_SLEEP | \ | |
141 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
142 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
143 | CTR_FLAG_REGION_SIZE | \ | |
f090279e | 144 | CTR_FLAG_RAID10_COPIES | \ |
33e53f06 HM |
145 | CTR_FLAG_RAID10_FORMAT | \ |
146 | CTR_FLAG_DELTA_DISKS | \ | |
a30cbc0d HM |
147 | CTR_FLAG_DATA_OFFSET | \ |
148 | CTR_FLAG_RAID10_USE_NEAR_SETS) | |
f090279e | 149 | |
f090279e HM |
150 | /* |
151 | * "raid4/5/6" do not accept any raid1 or raid10 specific options | |
152 | * | |
153 | * "raid6" does not accept "nosync", because it is not guaranteed | |
154 | * that both parity and q-syndrome are being written properly with | |
155 | * any writes | |
156 | */ | |
a30cbc0d HM |
157 | #define RAID45_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ |
158 | CTR_FLAG_REBUILD | \ | |
159 | CTR_FLAG_DAEMON_SLEEP | \ | |
160 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
161 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
f090279e | 162 | CTR_FLAG_MAX_WRITE_BEHIND | \ |
a30cbc0d HM |
163 | CTR_FLAG_STRIPE_CACHE | \ |
164 | CTR_FLAG_REGION_SIZE | \ | |
165 | CTR_FLAG_DELTA_DISKS | \ | |
166 | CTR_FLAG_DATA_OFFSET) | |
167 | ||
168 | #define RAID6_VALID_FLAGS (CTR_FLAG_SYNC | \ | |
169 | CTR_FLAG_REBUILD | \ | |
170 | CTR_FLAG_DAEMON_SLEEP | \ | |
171 | CTR_FLAG_MIN_RECOVERY_RATE | \ | |
172 | CTR_FLAG_MAX_RECOVERY_RATE | \ | |
173 | CTR_FLAG_MAX_WRITE_BEHIND | \ | |
174 | CTR_FLAG_STRIPE_CACHE | \ | |
175 | CTR_FLAG_REGION_SIZE | \ | |
176 | CTR_FLAG_DELTA_DISKS | \ | |
177 | CTR_FLAG_DATA_OFFSET) | |
178 | /* ...valid options definitions per raid level */ | |
f090279e | 179 | |
ecbfb9f1 HM |
180 | /* |
181 | * Flags for rs->runtime_flags field | |
182 | * (RT_FLAG prefix meaning "runtime flag") | |
183 | * | |
184 | * These are all internal and used to define runtime state, | |
185 | * e.g. to prevent another resume from preresume processing | |
186 | * the raid set all over again. | |
187 | */ | |
40ba37e5 HM |
188 | #define RT_FLAG_RS_PRERESUMED 0 |
189 | #define RT_FLAG_RS_RESUMED 1 | |
190 | #define RT_FLAG_RS_BITMAP_LOADED 2 | |
191 | #define RT_FLAG_UPDATE_SBS 3 | |
9dbd1aa3 | 192 | #define RT_FLAG_RESHAPE_RS 4 |
6e20902e | 193 | #define RT_FLAG_KEEP_RS_FROZEN 5 |
ecbfb9f1 | 194 | |
33e53f06 HM |
195 | /* Array elements of 64 bit needed for rebuild/write_mostly bits */ |
196 | #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8) | |
197 | ||
ecbfb9f1 HM |
198 | /* |
199 | * raid set level, layout and chunk sectors backup/restore | |
200 | */ | |
201 | struct rs_layout { | |
202 | int new_level; | |
203 | int new_layout; | |
204 | int new_chunk_sectors; | |
205 | }; | |
206 | ||
9d09e663 N |
207 | struct raid_set { |
208 | struct dm_target *ti; | |
209 | ||
34f8ac6d | 210 | uint32_t bitmap_loaded; |
9dbd1aa3 | 211 | uint32_t stripe_cache_entries; |
4286325b MS |
212 | unsigned long ctr_flags; |
213 | unsigned long runtime_flags; | |
ecbfb9f1 HM |
214 | |
215 | uint64_t rebuild_disks[DISKS_ARRAY_ELEMS]; | |
9d09e663 | 216 | |
33e53f06 HM |
217 | int raid_disks; |
218 | int delta_disks; | |
4763e543 | 219 | int data_offset; |
33e53f06 | 220 | int raid10_copies; |
4257e085 | 221 | int requested_bitmap_chunk_sectors; |
33e53f06 | 222 | |
fd01b88c | 223 | struct mddev md; |
9d09e663 N |
224 | struct raid_type *raid_type; |
225 | struct dm_target_callbacks callbacks; | |
226 | ||
227 | struct raid_dev dev[0]; | |
228 | }; | |
229 | ||
9dbd1aa3 | 230 | static void rs_config_backup(struct raid_set *rs, struct rs_layout *l) |
ecbfb9f1 HM |
231 | { |
232 | struct mddev *mddev = &rs->md; | |
233 | ||
234 | l->new_level = mddev->new_level; | |
235 | l->new_layout = mddev->new_layout; | |
236 | l->new_chunk_sectors = mddev->new_chunk_sectors; | |
237 | } | |
238 | ||
9dbd1aa3 | 239 | static void rs_config_restore(struct raid_set *rs, struct rs_layout *l) |
ecbfb9f1 HM |
240 | { |
241 | struct mddev *mddev = &rs->md; | |
242 | ||
243 | mddev->new_level = l->new_level; | |
244 | mddev->new_layout = l->new_layout; | |
245 | mddev->new_chunk_sectors = l->new_chunk_sectors; | |
246 | } | |
247 | ||
33e53f06 HM |
248 | /* raid10 algorithms (i.e. formats) */ |
249 | #define ALGORITHM_RAID10_DEFAULT 0 | |
250 | #define ALGORITHM_RAID10_NEAR 1 | |
251 | #define ALGORITHM_RAID10_OFFSET 2 | |
252 | #define ALGORITHM_RAID10_FAR 3 | |
253 | ||
9d09e663 N |
254 | /* Supported raid types and properties. */ |
255 | static struct raid_type { | |
256 | const char *name; /* RAID algorithm. */ | |
257 | const char *descr; /* Descriptor text for logging. */ | |
258 | const unsigned parity_devs; /* # of parity devices. */ | |
259 | const unsigned minimal_devs; /* minimal # of devices in set. */ | |
260 | const unsigned level; /* RAID level. */ | |
261 | const unsigned algorithm; /* RAID algorithm. */ | |
262 | } raid_types[] = { | |
43157840 MS |
263 | {"raid0", "raid0 (striping)", 0, 2, 0, 0 /* NONE */}, |
264 | {"raid1", "raid1 (mirroring)", 0, 2, 1, 0 /* NONE */}, | |
265 | {"raid10_far", "raid10 far (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_FAR}, | |
33e53f06 | 266 | {"raid10_offset", "raid10 offset (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_OFFSET}, |
43157840 MS |
267 | {"raid10_near", "raid10 near (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_NEAR}, |
268 | {"raid10", "raid10 (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_DEFAULT}, | |
269 | {"raid4", "raid4 (dedicated last parity disk)", 1, 2, 4, ALGORITHM_PARITY_N}, /* raid4 layout = raid5_n */ | |
270 | {"raid5_n", "raid5 (dedicated last parity disk)", 1, 2, 5, ALGORITHM_PARITY_N}, | |
271 | {"raid5_ls", "raid5 (left symmetric)", 1, 2, 5, ALGORITHM_LEFT_SYMMETRIC}, | |
272 | {"raid5_rs", "raid5 (right symmetric)", 1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC}, | |
273 | {"raid5_la", "raid5 (left asymmetric)", 1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC}, | |
274 | {"raid5_ra", "raid5 (right asymmetric)", 1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC}, | |
275 | {"raid6_zr", "raid6 (zero restart)", 2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART}, | |
276 | {"raid6_nr", "raid6 (N restart)", 2, 4, 6, ALGORITHM_ROTATING_N_RESTART}, | |
277 | {"raid6_nc", "raid6 (N continue)", 2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}, | |
278 | {"raid6_n_6", "raid6 (dedicated parity/Q n/6)", 2, 4, 6, ALGORITHM_PARITY_N_6}, | |
279 | {"raid6_ls_6", "raid6 (left symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_SYMMETRIC_6}, | |
280 | {"raid6_rs_6", "raid6 (right symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_SYMMETRIC_6}, | |
281 | {"raid6_la_6", "raid6 (left asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_ASYMMETRIC_6}, | |
282 | {"raid6_ra_6", "raid6 (right asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_ASYMMETRIC_6} | |
9d09e663 N |
283 | }; |
284 | ||
92c83d79 | 285 | /* True, if @v is in inclusive range [@min, @max] */ |
bb91a63f | 286 | static bool __within_range(long v, long min, long max) |
92c83d79 HM |
287 | { |
288 | return v >= min && v <= max; | |
289 | } | |
290 | ||
702108d1 HM |
291 | /* All table line arguments are defined here */ |
292 | static struct arg_name_flag { | |
4286325b | 293 | const unsigned long flag; |
702108d1 | 294 | const char *name; |
e6ca5e1a | 295 | } __arg_name_flags[] = { |
702108d1 HM |
296 | { CTR_FLAG_SYNC, "sync"}, |
297 | { CTR_FLAG_NOSYNC, "nosync"}, | |
298 | { CTR_FLAG_REBUILD, "rebuild"}, | |
299 | { CTR_FLAG_DAEMON_SLEEP, "daemon_sleep"}, | |
300 | { CTR_FLAG_MIN_RECOVERY_RATE, "min_recovery_rate"}, | |
301 | { CTR_FLAG_MAX_RECOVERY_RATE, "max_recovery_rate"}, | |
302 | { CTR_FLAG_MAX_WRITE_BEHIND, "max_write_behind"}, | |
303 | { CTR_FLAG_WRITE_MOSTLY, "writemostly"}, | |
304 | { CTR_FLAG_STRIPE_CACHE, "stripe_cache"}, | |
305 | { CTR_FLAG_REGION_SIZE, "region_size"}, | |
306 | { CTR_FLAG_RAID10_COPIES, "raid10_copies"}, | |
307 | { CTR_FLAG_RAID10_FORMAT, "raid10_format"}, | |
4763e543 HM |
308 | { CTR_FLAG_DATA_OFFSET, "data_offset"}, |
309 | { CTR_FLAG_DELTA_DISKS, "delta_disks"}, | |
310 | { CTR_FLAG_RAID10_USE_NEAR_SETS, "raid10_use_near_sets"}, | |
702108d1 HM |
311 | }; |
312 | ||
313 | /* Return argument name string for given @flag */ | |
3fa6cf38 | 314 | static const char *dm_raid_arg_name_by_flag(const uint32_t flag) |
702108d1 HM |
315 | { |
316 | if (hweight32(flag) == 1) { | |
e6ca5e1a | 317 | struct arg_name_flag *anf = __arg_name_flags + ARRAY_SIZE(__arg_name_flags); |
702108d1 | 318 | |
e6ca5e1a | 319 | while (anf-- > __arg_name_flags) |
4286325b | 320 | if (flag & anf->flag) |
702108d1 HM |
321 | return anf->name; |
322 | ||
323 | } else | |
324 | DMERR("%s called with more than one flag!", __func__); | |
325 | ||
326 | return NULL; | |
327 | } | |
328 | ||
33e53f06 HM |
329 | /* |
330 | * bool helpers to test for various raid levels of a raid set, | |
331 | * is. it's level as reported by the superblock rather than | |
332 | * the requested raid_type passed to the constructor. | |
333 | */ | |
334 | /* Return true, if raid set in @rs is raid0 */ | |
335 | static bool rs_is_raid0(struct raid_set *rs) | |
336 | { | |
337 | return !rs->md.level; | |
338 | } | |
339 | ||
9dbd1aa3 HM |
340 | /* Return true, if raid set in @rs is raid1 */ |
341 | static bool rs_is_raid1(struct raid_set *rs) | |
342 | { | |
343 | return rs->md.level == 1; | |
344 | } | |
345 | ||
33e53f06 HM |
346 | /* Return true, if raid set in @rs is raid10 */ |
347 | static bool rs_is_raid10(struct raid_set *rs) | |
348 | { | |
349 | return rs->md.level == 10; | |
350 | } | |
351 | ||
40ba37e5 HM |
352 | /* Return true, if raid set in @rs is level 4, 5 or 6 */ |
353 | static bool rs_is_raid456(struct raid_set *rs) | |
354 | { | |
355 | return __within_range(rs->md.level, 4, 6); | |
356 | } | |
357 | ||
358 | /* Return true, if raid set in @rs is reshapable */ | |
359 | static unsigned int __is_raid10_far(int layout); | |
360 | static bool rs_is_reshapable(struct raid_set *rs) | |
361 | { | |
362 | return rs_is_raid456(rs) || | |
363 | (rs_is_raid10(rs) && !__is_raid10_far(rs->md.new_layout)); | |
364 | } | |
365 | ||
9dbd1aa3 HM |
366 | /* Return true, if raid set in @rs is recovering */ |
367 | static bool rs_is_recovering(struct raid_set *rs) | |
368 | { | |
9dbd1aa3 HM |
369 | return rs->md.recovery_cp != MaxSector; |
370 | } | |
371 | ||
372 | /* Return true, if raid set in @rs is reshaping */ | |
373 | static bool rs_is_reshaping(struct raid_set *rs) | |
374 | { | |
9dbd1aa3 HM |
375 | return rs->md.reshape_position != MaxSector; |
376 | } | |
377 | ||
f090279e HM |
378 | /* |
379 | * bool helpers to test for various raid levels of a raid type | |
380 | */ | |
381 | ||
382 | /* Return true, if raid type in @rt is raid0 */ | |
383 | static bool rt_is_raid0(struct raid_type *rt) | |
384 | { | |
385 | return !rt->level; | |
386 | } | |
387 | ||
388 | /* Return true, if raid type in @rt is raid1 */ | |
389 | static bool rt_is_raid1(struct raid_type *rt) | |
390 | { | |
391 | return rt->level == 1; | |
392 | } | |
393 | ||
394 | /* Return true, if raid type in @rt is raid10 */ | |
395 | static bool rt_is_raid10(struct raid_type *rt) | |
396 | { | |
397 | return rt->level == 10; | |
398 | } | |
399 | ||
400 | /* Return true, if raid type in @rt is raid4/5 */ | |
401 | static bool rt_is_raid45(struct raid_type *rt) | |
402 | { | |
bb91a63f | 403 | return __within_range(rt->level, 4, 5); |
f090279e HM |
404 | } |
405 | ||
406 | /* Return true, if raid type in @rt is raid6 */ | |
407 | static bool rt_is_raid6(struct raid_type *rt) | |
408 | { | |
409 | return rt->level == 6; | |
410 | } | |
676fa5ad HM |
411 | |
412 | /* Return true, if raid type in @rt is raid4/5/6 */ | |
413 | static bool rt_is_raid456(struct raid_type *rt) | |
414 | { | |
bb91a63f | 415 | return __within_range(rt->level, 4, 6); |
676fa5ad | 416 | } |
f090279e HM |
417 | /* END: raid level bools */ |
418 | ||
a30cbc0d HM |
419 | /* Return valid ctr flags for the raid level of @rs */ |
420 | static unsigned long __valid_flags(struct raid_set *rs) | |
f090279e HM |
421 | { |
422 | if (rt_is_raid0(rs->raid_type)) | |
a30cbc0d | 423 | return RAID0_VALID_FLAGS; |
f090279e | 424 | else if (rt_is_raid1(rs->raid_type)) |
a30cbc0d | 425 | return RAID1_VALID_FLAGS; |
f090279e | 426 | else if (rt_is_raid10(rs->raid_type)) |
a30cbc0d | 427 | return RAID10_VALID_FLAGS; |
f090279e | 428 | else if (rt_is_raid45(rs->raid_type)) |
a30cbc0d | 429 | return RAID45_VALID_FLAGS; |
f090279e | 430 | else if (rt_is_raid6(rs->raid_type)) |
a30cbc0d | 431 | return RAID6_VALID_FLAGS; |
f090279e HM |
432 | |
433 | return ~0; | |
434 | } | |
435 | ||
436 | /* | |
a30cbc0d | 437 | * Check for valid flags set on @rs |
f090279e HM |
438 | * |
439 | * Has to be called after parsing of the ctr flags! | |
440 | */ | |
a30cbc0d | 441 | static int rs_check_for_valid_flags(struct raid_set *rs) |
f090279e | 442 | { |
a30cbc0d | 443 | if (rs->ctr_flags & ~__valid_flags(rs)) { |
4286325b | 444 | rs->ti->error = "Invalid flags combination"; |
bd83a4c4 MS |
445 | return -EINVAL; |
446 | } | |
f090279e HM |
447 | |
448 | return 0; | |
449 | } | |
450 | ||
33e53f06 HM |
451 | /* MD raid10 bit definitions and helpers */ |
452 | #define RAID10_OFFSET (1 << 16) /* stripes with data copies area adjacent on devices */ | |
453 | #define RAID10_BROCKEN_USE_FAR_SETS (1 << 17) /* Broken in raid10.c: use sets instead of whole stripe rotation */ | |
454 | #define RAID10_USE_FAR_SETS (1 << 18) /* Use sets instead of whole stripe rotation */ | |
455 | #define RAID10_FAR_COPIES_SHIFT 8 /* raid10 # far copies shift (2nd byte of layout) */ | |
456 | ||
457 | /* Return md raid10 near copies for @layout */ | |
e6ca5e1a | 458 | static unsigned int __raid10_near_copies(int layout) |
33e53f06 HM |
459 | { |
460 | return layout & 0xFF; | |
461 | } | |
462 | ||
463 | /* Return md raid10 far copies for @layout */ | |
e6ca5e1a | 464 | static unsigned int __raid10_far_copies(int layout) |
33e53f06 | 465 | { |
e6ca5e1a | 466 | return __raid10_near_copies(layout >> RAID10_FAR_COPIES_SHIFT); |
33e53f06 HM |
467 | } |
468 | ||
469 | /* Return true if md raid10 offset for @layout */ | |
e6ca5e1a | 470 | static unsigned int __is_raid10_offset(int layout) |
33e53f06 HM |
471 | { |
472 | return layout & RAID10_OFFSET; | |
473 | } | |
474 | ||
475 | /* Return true if md raid10 near for @layout */ | |
e6ca5e1a | 476 | static unsigned int __is_raid10_near(int layout) |
33e53f06 | 477 | { |
e6ca5e1a | 478 | return !__is_raid10_offset(layout) && __raid10_near_copies(layout) > 1; |
33e53f06 HM |
479 | } |
480 | ||
481 | /* Return true if md raid10 far for @layout */ | |
e6ca5e1a | 482 | static unsigned int __is_raid10_far(int layout) |
33e53f06 | 483 | { |
e6ca5e1a | 484 | return !__is_raid10_offset(layout) && __raid10_far_copies(layout) > 1; |
33e53f06 HM |
485 | } |
486 | ||
487 | /* Return md raid10 layout string for @layout */ | |
488 | static const char *raid10_md_layout_to_format(int layout) | |
fe5d2f4a JB |
489 | { |
490 | /* | |
33e53f06 HM |
491 | * Bit 16 stands for "offset" |
492 | * (i.e. adjacent stripes hold copies) | |
493 | * | |
fe5d2f4a JB |
494 | * Refer to MD's raid10.c for details |
495 | */ | |
e6ca5e1a | 496 | if (__is_raid10_offset(layout)) |
fe5d2f4a JB |
497 | return "offset"; |
498 | ||
e6ca5e1a | 499 | if (__raid10_near_copies(layout) > 1) |
fe5d2f4a JB |
500 | return "near"; |
501 | ||
e6ca5e1a | 502 | WARN_ON(__raid10_far_copies(layout) < 2); |
33e53f06 | 503 | |
fe5d2f4a JB |
504 | return "far"; |
505 | } | |
506 | ||
33e53f06 | 507 | /* Return md raid10 algorithm for @name */ |
68c1c4d5 | 508 | static int raid10_name_to_format(const char *name) |
33e53f06 HM |
509 | { |
510 | if (!strcasecmp(name, "near")) | |
511 | return ALGORITHM_RAID10_NEAR; | |
512 | else if (!strcasecmp(name, "offset")) | |
513 | return ALGORITHM_RAID10_OFFSET; | |
514 | else if (!strcasecmp(name, "far")) | |
515 | return ALGORITHM_RAID10_FAR; | |
516 | ||
517 | return -EINVAL; | |
518 | } | |
519 | ||
33e53f06 HM |
520 | /* Return md raid10 copies for @layout */ |
521 | static unsigned int raid10_md_layout_to_copies(int layout) | |
63f33b8d | 522 | { |
e6ca5e1a MS |
523 | return __raid10_near_copies(layout) > 1 ? |
524 | __raid10_near_copies(layout) : __raid10_far_copies(layout); | |
63f33b8d JB |
525 | } |
526 | ||
33e53f06 HM |
527 | /* Return md raid10 format id for @format string */ |
528 | static int raid10_format_to_md_layout(struct raid_set *rs, | |
529 | unsigned int algorithm, | |
530 | unsigned int copies) | |
63f33b8d | 531 | { |
33e53f06 | 532 | unsigned int n = 1, f = 1, r = 0; |
fe5d2f4a | 533 | |
33e53f06 HM |
534 | /* |
535 | * MD resilienece flaw: | |
536 | * | |
537 | * enabling use_far_sets for far/offset formats causes copies | |
538 | * to be colocated on the same devs together with their origins! | |
539 | * | |
540 | * -> disable it for now in the definition above | |
541 | */ | |
542 | if (algorithm == ALGORITHM_RAID10_DEFAULT || | |
543 | algorithm == ALGORITHM_RAID10_NEAR) | |
fe5d2f4a | 544 | n = copies; |
33e53f06 HM |
545 | |
546 | else if (algorithm == ALGORITHM_RAID10_OFFSET) { | |
547 | f = copies; | |
548 | r = RAID10_OFFSET; | |
4286325b | 549 | if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) |
33e53f06 HM |
550 | r |= RAID10_USE_FAR_SETS; |
551 | ||
552 | } else if (algorithm == ALGORITHM_RAID10_FAR) { | |
fe5d2f4a | 553 | f = copies; |
33e53f06 | 554 | r = !RAID10_OFFSET; |
4286325b | 555 | if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) |
33e53f06 | 556 | r |= RAID10_USE_FAR_SETS; |
fe5d2f4a | 557 | |
33e53f06 HM |
558 | } else |
559 | return -EINVAL; | |
560 | ||
561 | return r | (f << RAID10_FAR_COPIES_SHIFT) | n; | |
562 | } | |
563 | /* END: MD raid10 bit definitions and helpers */ | |
fe5d2f4a | 564 | |
33e53f06 | 565 | /* Check for any of the raid10 algorithms */ |
e6ca5e1a | 566 | static int __got_raid10(struct raid_type *rtp, const int layout) |
33e53f06 HM |
567 | { |
568 | if (rtp->level == 10) { | |
569 | switch (rtp->algorithm) { | |
570 | case ALGORITHM_RAID10_DEFAULT: | |
571 | case ALGORITHM_RAID10_NEAR: | |
e6ca5e1a | 572 | return __is_raid10_near(layout); |
33e53f06 | 573 | case ALGORITHM_RAID10_OFFSET: |
e6ca5e1a | 574 | return __is_raid10_offset(layout); |
33e53f06 | 575 | case ALGORITHM_RAID10_FAR: |
e6ca5e1a | 576 | return __is_raid10_far(layout); |
33e53f06 HM |
577 | default: |
578 | break; | |
579 | } | |
580 | } | |
fe5d2f4a | 581 | |
33e53f06 | 582 | return 0; |
63f33b8d JB |
583 | } |
584 | ||
33e53f06 | 585 | /* Return raid_type for @name */ |
92c83d79 | 586 | static struct raid_type *get_raid_type(const char *name) |
9d09e663 | 587 | { |
33e53f06 | 588 | struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); |
9d09e663 | 589 | |
33e53f06 HM |
590 | while (rtp-- > raid_types) |
591 | if (!strcasecmp(rtp->name, name)) | |
592 | return rtp; | |
9d09e663 N |
593 | |
594 | return NULL; | |
595 | } | |
596 | ||
33e53f06 HM |
597 | /* Return raid_type for @name based derived from @level and @layout */ |
598 | static struct raid_type *get_raid_type_by_ll(const int level, const int layout) | |
599 | { | |
600 | struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); | |
601 | ||
602 | while (rtp-- > raid_types) { | |
603 | /* RAID10 special checks based on @layout flags/properties */ | |
604 | if (rtp->level == level && | |
e6ca5e1a | 605 | (__got_raid10(rtp, layout) || rtp->algorithm == layout)) |
33e53f06 HM |
606 | return rtp; |
607 | } | |
608 | ||
609 | return NULL; | |
610 | } | |
611 | ||
9dbd1aa3 HM |
612 | /* |
613 | * Conditionally change bdev capacity of @rs | |
614 | * in case of a disk add/remove reshape | |
615 | */ | |
616 | static void rs_set_capacity(struct raid_set *rs) | |
617 | { | |
618 | struct mddev *mddev = &rs->md; | |
619 | ||
620 | /* Make sure we access most actual mddev properties */ | |
621 | smp_rmb(); | |
622 | if (rs->ti->len != mddev->array_sectors && !rs_is_reshaping(rs)) { | |
623 | struct gendisk *gendisk = dm_disk(dm_table_get_md(rs->ti->table)); | |
624 | ||
625 | set_capacity(gendisk, mddev->array_sectors); | |
626 | revalidate_disk(gendisk); | |
627 | } | |
628 | } | |
629 | ||
3a1c1ef2 HM |
630 | /* |
631 | * Set the mddev properties in @rs to the current | |
632 | * ones retrieved from the freshest superblock | |
633 | */ | |
634 | static void rs_set_cur(struct raid_set *rs) | |
635 | { | |
636 | struct mddev *mddev = &rs->md; | |
637 | ||
638 | mddev->new_level = mddev->level; | |
639 | mddev->new_layout = mddev->layout; | |
640 | mddev->new_chunk_sectors = mddev->chunk_sectors; | |
641 | } | |
642 | ||
33e53f06 HM |
643 | /* |
644 | * Set the mddev properties in @rs to the new | |
645 | * ones requested by the ctr | |
646 | */ | |
647 | static void rs_set_new(struct raid_set *rs) | |
648 | { | |
649 | struct mddev *mddev = &rs->md; | |
650 | ||
651 | mddev->level = mddev->new_level; | |
652 | mddev->layout = mddev->new_layout; | |
653 | mddev->chunk_sectors = mddev->new_chunk_sectors; | |
3a1c1ef2 | 654 | mddev->raid_disks = rs->raid_disks; |
33e53f06 HM |
655 | mddev->delta_disks = 0; |
656 | } | |
657 | ||
bfcee0e3 MS |
658 | static struct raid_set *raid_set_alloc(struct dm_target *ti, struct raid_type *raid_type, |
659 | unsigned raid_devs) | |
9d09e663 N |
660 | { |
661 | unsigned i; | |
662 | struct raid_set *rs; | |
9d09e663 | 663 | |
bd83a4c4 MS |
664 | if (raid_devs <= raid_type->parity_devs) { |
665 | ti->error = "Insufficient number of devices"; | |
666 | return ERR_PTR(-EINVAL); | |
667 | } | |
9d09e663 | 668 | |
9d09e663 | 669 | rs = kzalloc(sizeof(*rs) + raid_devs * sizeof(rs->dev[0]), GFP_KERNEL); |
bd83a4c4 MS |
670 | if (!rs) { |
671 | ti->error = "Cannot allocate raid context"; | |
672 | return ERR_PTR(-ENOMEM); | |
673 | } | |
9d09e663 N |
674 | |
675 | mddev_init(&rs->md); | |
676 | ||
33e53f06 HM |
677 | rs->raid_disks = raid_devs; |
678 | rs->delta_disks = 0; | |
679 | ||
9d09e663 N |
680 | rs->ti = ti; |
681 | rs->raid_type = raid_type; | |
9dbd1aa3 | 682 | rs->stripe_cache_entries = 256; |
9d09e663 N |
683 | rs->md.raid_disks = raid_devs; |
684 | rs->md.level = raid_type->level; | |
685 | rs->md.new_level = rs->md.level; | |
9d09e663 N |
686 | rs->md.layout = raid_type->algorithm; |
687 | rs->md.new_layout = rs->md.layout; | |
688 | rs->md.delta_disks = 0; | |
ecbfb9f1 | 689 | rs->md.recovery_cp = rs_is_raid0(rs) ? MaxSector : 0; |
9d09e663 N |
690 | |
691 | for (i = 0; i < raid_devs; i++) | |
692 | md_rdev_init(&rs->dev[i].rdev); | |
693 | ||
694 | /* | |
695 | * Remaining items to be initialized by further RAID params: | |
696 | * rs->md.persistent | |
697 | * rs->md.external | |
698 | * rs->md.chunk_sectors | |
699 | * rs->md.new_chunk_sectors | |
c039c332 | 700 | * rs->md.dev_sectors |
9d09e663 N |
701 | */ |
702 | ||
703 | return rs; | |
704 | } | |
705 | ||
bfcee0e3 | 706 | static void raid_set_free(struct raid_set *rs) |
9d09e663 N |
707 | { |
708 | int i; | |
709 | ||
b12d437b JB |
710 | for (i = 0; i < rs->md.raid_disks; i++) { |
711 | if (rs->dev[i].meta_dev) | |
712 | dm_put_device(rs->ti, rs->dev[i].meta_dev); | |
545c8795 | 713 | md_rdev_clear(&rs->dev[i].rdev); |
9d09e663 N |
714 | if (rs->dev[i].data_dev) |
715 | dm_put_device(rs->ti, rs->dev[i].data_dev); | |
b12d437b | 716 | } |
9d09e663 N |
717 | |
718 | kfree(rs); | |
719 | } | |
720 | ||
721 | /* | |
722 | * For every device we have two words | |
723 | * <meta_dev>: meta device name or '-' if missing | |
724 | * <data_dev>: data device name or '-' if missing | |
725 | * | |
b12d437b JB |
726 | * The following are permitted: |
727 | * - - | |
728 | * - <data_dev> | |
729 | * <meta_dev> <data_dev> | |
730 | * | |
731 | * The following is not allowed: | |
732 | * <meta_dev> - | |
733 | * | |
734 | * This code parses those words. If there is a failure, | |
bfcee0e3 | 735 | * the caller must use raid_set_free() to unwind the operations. |
9d09e663 | 736 | */ |
702108d1 | 737 | static int parse_dev_params(struct raid_set *rs, struct dm_arg_set *as) |
9d09e663 N |
738 | { |
739 | int i; | |
740 | int rebuild = 0; | |
741 | int metadata_available = 0; | |
73c6f239 | 742 | int r = 0; |
92c83d79 | 743 | const char *arg; |
9d09e663 | 744 | |
92c83d79 HM |
745 | /* Put off the number of raid devices argument to get to dev pairs */ |
746 | arg = dm_shift_arg(as); | |
747 | if (!arg) | |
748 | return -EINVAL; | |
749 | ||
750 | for (i = 0; i < rs->md.raid_disks; i++) { | |
9d09e663 N |
751 | rs->dev[i].rdev.raid_disk = i; |
752 | ||
753 | rs->dev[i].meta_dev = NULL; | |
754 | rs->dev[i].data_dev = NULL; | |
755 | ||
756 | /* | |
757 | * There are no offsets, since there is a separate device | |
758 | * for data and metadata. | |
759 | */ | |
760 | rs->dev[i].rdev.data_offset = 0; | |
761 | rs->dev[i].rdev.mddev = &rs->md; | |
762 | ||
92c83d79 HM |
763 | arg = dm_shift_arg(as); |
764 | if (!arg) | |
765 | return -EINVAL; | |
766 | ||
767 | if (strcmp(arg, "-")) { | |
bd83a4c4 MS |
768 | r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), |
769 | &rs->dev[i].meta_dev); | |
770 | if (r) { | |
771 | rs->ti->error = "RAID metadata device lookup failure"; | |
772 | return r; | |
773 | } | |
b12d437b JB |
774 | |
775 | rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL); | |
bd83a4c4 MS |
776 | if (!rs->dev[i].rdev.sb_page) { |
777 | rs->ti->error = "Failed to allocate superblock page"; | |
778 | return -ENOMEM; | |
779 | } | |
9d09e663 N |
780 | } |
781 | ||
92c83d79 HM |
782 | arg = dm_shift_arg(as); |
783 | if (!arg) | |
784 | return -EINVAL; | |
785 | ||
786 | if (!strcmp(arg, "-")) { | |
9d09e663 | 787 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags) && |
bd83a4c4 MS |
788 | (!rs->dev[i].rdev.recovery_offset)) { |
789 | rs->ti->error = "Drive designated for rebuild not specified"; | |
790 | return -EINVAL; | |
791 | } | |
9d09e663 | 792 | |
bd83a4c4 MS |
793 | if (rs->dev[i].meta_dev) { |
794 | rs->ti->error = "No data device supplied with metadata device"; | |
795 | return -EINVAL; | |
796 | } | |
b12d437b | 797 | |
9d09e663 N |
798 | continue; |
799 | } | |
800 | ||
bd83a4c4 MS |
801 | r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), |
802 | &rs->dev[i].data_dev); | |
803 | if (r) { | |
804 | rs->ti->error = "RAID device lookup failure"; | |
805 | return r; | |
806 | } | |
9d09e663 | 807 | |
b12d437b JB |
808 | if (rs->dev[i].meta_dev) { |
809 | metadata_available = 1; | |
810 | rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev; | |
811 | } | |
9d09e663 | 812 | rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev; |
3a1c1ef2 | 813 | list_add_tail(&rs->dev[i].rdev.same_set, &rs->md.disks); |
9d09e663 N |
814 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags)) |
815 | rebuild++; | |
816 | } | |
817 | ||
818 | if (metadata_available) { | |
819 | rs->md.external = 0; | |
820 | rs->md.persistent = 1; | |
821 | rs->md.major_version = 2; | |
822 | } else if (rebuild && !rs->md.recovery_cp) { | |
823 | /* | |
824 | * Without metadata, we will not be able to tell if the array | |
825 | * is in-sync or not - we must assume it is not. Therefore, | |
826 | * it is impossible to rebuild a drive. | |
827 | * | |
828 | * Even if there is metadata, the on-disk information may | |
829 | * indicate that the array is not in-sync and it will then | |
830 | * fail at that time. | |
831 | * | |
832 | * User could specify 'nosync' option if desperate. | |
833 | */ | |
bd83a4c4 MS |
834 | rs->ti->error = "Unable to rebuild drive while array is not in-sync"; |
835 | return -EINVAL; | |
9d09e663 N |
836 | } |
837 | ||
838 | return 0; | |
839 | } | |
840 | ||
c1084561 JB |
841 | /* |
842 | * validate_region_size | |
843 | * @rs | |
844 | * @region_size: region size in sectors. If 0, pick a size (4MiB default). | |
845 | * | |
846 | * Set rs->md.bitmap_info.chunksize (which really refers to 'region size'). | |
847 | * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap. | |
848 | * | |
849 | * Returns: 0 on success, -EINVAL on failure. | |
850 | */ | |
851 | static int validate_region_size(struct raid_set *rs, unsigned long region_size) | |
852 | { | |
853 | unsigned long min_region_size = rs->ti->len / (1 << 21); | |
854 | ||
855 | if (!region_size) { | |
856 | /* | |
43157840 | 857 | * Choose a reasonable default. All figures in sectors. |
c1084561 JB |
858 | */ |
859 | if (min_region_size > (1 << 13)) { | |
3a0f9aae | 860 | /* If not a power of 2, make it the next power of 2 */ |
042745ee | 861 | region_size = roundup_pow_of_two(min_region_size); |
c1084561 JB |
862 | DMINFO("Choosing default region size of %lu sectors", |
863 | region_size); | |
c1084561 JB |
864 | } else { |
865 | DMINFO("Choosing default region size of 4MiB"); | |
866 | region_size = 1 << 13; /* sectors */ | |
867 | } | |
868 | } else { | |
869 | /* | |
870 | * Validate user-supplied value. | |
871 | */ | |
bd83a4c4 MS |
872 | if (region_size > rs->ti->len) { |
873 | rs->ti->error = "Supplied region size is too large"; | |
874 | return -EINVAL; | |
875 | } | |
c1084561 JB |
876 | |
877 | if (region_size < min_region_size) { | |
878 | DMERR("Supplied region_size (%lu sectors) below minimum (%lu)", | |
879 | region_size, min_region_size); | |
bd83a4c4 MS |
880 | rs->ti->error = "Supplied region size is too small"; |
881 | return -EINVAL; | |
c1084561 JB |
882 | } |
883 | ||
bd83a4c4 MS |
884 | if (!is_power_of_2(region_size)) { |
885 | rs->ti->error = "Region size is not a power of 2"; | |
886 | return -EINVAL; | |
887 | } | |
c1084561 | 888 | |
bd83a4c4 MS |
889 | if (region_size < rs->md.chunk_sectors) { |
890 | rs->ti->error = "Region size is smaller than the chunk size"; | |
891 | return -EINVAL; | |
892 | } | |
c1084561 JB |
893 | } |
894 | ||
895 | /* | |
896 | * Convert sectors to bytes. | |
897 | */ | |
898 | rs->md.bitmap_info.chunksize = (region_size << 9); | |
899 | ||
900 | return 0; | |
901 | } | |
902 | ||
eb649123 | 903 | /* |
55ebbb59 | 904 | * validate_raid_redundancy |
eb649123 JB |
905 | * @rs |
906 | * | |
55ebbb59 JB |
907 | * Determine if there are enough devices in the array that haven't |
908 | * failed (or are being rebuilt) to form a usable array. | |
eb649123 JB |
909 | * |
910 | * Returns: 0 on success, -EINVAL on failure. | |
911 | */ | |
55ebbb59 | 912 | static int validate_raid_redundancy(struct raid_set *rs) |
eb649123 JB |
913 | { |
914 | unsigned i, rebuild_cnt = 0; | |
9dbd1aa3 | 915 | unsigned rebuilds_per_group = 0, copies; |
fe5d2f4a | 916 | unsigned group_size, last_group_start; |
eb649123 | 917 | |
eb649123 | 918 | for (i = 0; i < rs->md.raid_disks; i++) |
55ebbb59 JB |
919 | if (!test_bit(In_sync, &rs->dev[i].rdev.flags) || |
920 | !rs->dev[i].rdev.sb_page) | |
eb649123 JB |
921 | rebuild_cnt++; |
922 | ||
923 | switch (rs->raid_type->level) { | |
924 | case 1: | |
925 | if (rebuild_cnt >= rs->md.raid_disks) | |
926 | goto too_many; | |
927 | break; | |
928 | case 4: | |
929 | case 5: | |
930 | case 6: | |
931 | if (rebuild_cnt > rs->raid_type->parity_devs) | |
932 | goto too_many; | |
933 | break; | |
934 | case 10: | |
9dbd1aa3 | 935 | copies = raid10_md_layout_to_copies(rs->md.new_layout); |
4ec1e369 JB |
936 | if (rebuild_cnt < copies) |
937 | break; | |
938 | ||
939 | /* | |
940 | * It is possible to have a higher rebuild count for RAID10, | |
941 | * as long as the failed devices occur in different mirror | |
942 | * groups (i.e. different stripes). | |
943 | * | |
4ec1e369 JB |
944 | * When checking "near" format, make sure no adjacent devices |
945 | * have failed beyond what can be handled. In addition to the | |
946 | * simple case where the number of devices is a multiple of the | |
947 | * number of copies, we must also handle cases where the number | |
948 | * of devices is not a multiple of the number of copies. | |
43157840 MS |
949 | * E.g. dev1 dev2 dev3 dev4 dev5 |
950 | * A A B B C | |
951 | * C D D E E | |
4ec1e369 | 952 | */ |
9dbd1aa3 HM |
953 | if (__is_raid10_near(rs->md.new_layout)) { |
954 | for (i = 0; i < rs->raid_disks; i++) { | |
fe5d2f4a JB |
955 | if (!(i % copies)) |
956 | rebuilds_per_group = 0; | |
9dbd1aa3 | 957 | if ((!rs->dev[i].rdev.sb_page || |
40ba37e5 | 958 | !test_bit(In_sync, &rs->dev[i].rdev.flags)) && |
fe5d2f4a JB |
959 | (++rebuilds_per_group >= copies)) |
960 | goto too_many; | |
961 | } | |
962 | break; | |
963 | } | |
964 | ||
965 | /* | |
966 | * When checking "far" and "offset" formats, we need to ensure | |
967 | * that the device that holds its copy is not also dead or | |
968 | * being rebuilt. (Note that "far" and "offset" formats only | |
969 | * support two copies right now. These formats also only ever | |
970 | * use the 'use_far_sets' variant.) | |
971 | * | |
972 | * This check is somewhat complicated by the need to account | |
43157840 | 973 | * for arrays that are not a multiple of (far) copies. This |
fe5d2f4a JB |
974 | * results in the need to treat the last (potentially larger) |
975 | * set differently. | |
976 | */ | |
977 | group_size = (rs->md.raid_disks / copies); | |
978 | last_group_start = (rs->md.raid_disks / group_size) - 1; | |
979 | last_group_start *= group_size; | |
980 | for (i = 0; i < rs->md.raid_disks; i++) { | |
981 | if (!(i % copies) && !(i > last_group_start)) | |
55ebbb59 | 982 | rebuilds_per_group = 0; |
fe5d2f4a JB |
983 | if ((!rs->dev[i].rdev.sb_page || |
984 | !test_bit(In_sync, &rs->dev[i].rdev.flags)) && | |
4ec1e369 | 985 | (++rebuilds_per_group >= copies)) |
fe5d2f4a | 986 | goto too_many; |
4ec1e369 JB |
987 | } |
988 | break; | |
eb649123 | 989 | default: |
55ebbb59 JB |
990 | if (rebuild_cnt) |
991 | return -EINVAL; | |
eb649123 JB |
992 | } |
993 | ||
994 | return 0; | |
995 | ||
996 | too_many: | |
eb649123 JB |
997 | return -EINVAL; |
998 | } | |
999 | ||
9d09e663 N |
1000 | /* |
1001 | * Possible arguments are... | |
9d09e663 N |
1002 | * <chunk_size> [optional_args] |
1003 | * | |
32737279 JB |
1004 | * Argument definitions |
1005 | * <chunk_size> The number of sectors per disk that | |
43157840 | 1006 | * will form the "stripe" |
32737279 | 1007 | * [[no]sync] Force or prevent recovery of the |
43157840 | 1008 | * entire array |
9d09e663 | 1009 | * [rebuild <idx>] Rebuild the drive indicated by the index |
32737279 | 1010 | * [daemon_sleep <ms>] Time between bitmap daemon work to |
43157840 | 1011 | * clear bits |
9d09e663 N |
1012 | * [min_recovery_rate <kB/sec/disk>] Throttle RAID initialization |
1013 | * [max_recovery_rate <kB/sec/disk>] Throttle RAID initialization | |
46bed2b5 | 1014 | * [write_mostly <idx>] Indicate a write mostly drive via index |
9d09e663 N |
1015 | * [max_write_behind <sectors>] See '-write-behind=' (man mdadm) |
1016 | * [stripe_cache <sectors>] Stripe cache size for higher RAIDs | |
43157840 | 1017 | * [region_size <sectors>] Defines granularity of bitmap |
63f33b8d JB |
1018 | * |
1019 | * RAID10-only options: | |
43157840 | 1020 | * [raid10_copies <# copies>] Number of copies. (Default: 2) |
fe5d2f4a | 1021 | * [raid10_format <near|far|offset>] Layout algorithm. (Default: near) |
9d09e663 | 1022 | */ |
92c83d79 | 1023 | static int parse_raid_params(struct raid_set *rs, struct dm_arg_set *as, |
9d09e663 N |
1024 | unsigned num_raid_params) |
1025 | { | |
9dbd1aa3 | 1026 | int value, raid10_format = ALGORITHM_RAID10_DEFAULT; |
63f33b8d | 1027 | unsigned raid10_copies = 2; |
eb649123 | 1028 | unsigned i; |
9dbd1aa3 | 1029 | unsigned region_size = 0; |
542f9038 | 1030 | sector_t max_io_len; |
92c83d79 | 1031 | const char *arg, *key; |
702108d1 | 1032 | struct raid_dev *rd; |
33e53f06 | 1033 | struct raid_type *rt = rs->raid_type; |
92c83d79 HM |
1034 | |
1035 | arg = dm_shift_arg(as); | |
1036 | num_raid_params--; /* Account for chunk_size argument */ | |
1037 | ||
9dbd1aa3 | 1038 | if (kstrtoint(arg, 10, &value) < 0) { |
bd83a4c4 MS |
1039 | rs->ti->error = "Bad numerical argument given for chunk_size"; |
1040 | return -EINVAL; | |
1041 | } | |
9d09e663 N |
1042 | |
1043 | /* | |
1044 | * First, parse the in-order required arguments | |
32737279 | 1045 | * "chunk_size" is the only argument of this type. |
9d09e663 | 1046 | */ |
33e53f06 | 1047 | if (rt_is_raid1(rt)) { |
32737279 JB |
1048 | if (value) |
1049 | DMERR("Ignoring chunk size parameter for RAID 1"); | |
1050 | value = 0; | |
bd83a4c4 MS |
1051 | } else if (!is_power_of_2(value)) { |
1052 | rs->ti->error = "Chunk size must be a power of 2"; | |
1053 | return -EINVAL; | |
1054 | } else if (value < 8) { | |
1055 | rs->ti->error = "Chunk size value is too small"; | |
1056 | return -EINVAL; | |
1057 | } | |
9d09e663 N |
1058 | |
1059 | rs->md.new_chunk_sectors = rs->md.chunk_sectors = value; | |
9d09e663 N |
1060 | |
1061 | /* | |
b12d437b JB |
1062 | * We set each individual device as In_sync with a completed |
1063 | * 'recovery_offset'. If there has been a device failure or | |
1064 | * replacement then one of the following cases applies: | |
1065 | * | |
1066 | * 1) User specifies 'rebuild'. | |
43157840 | 1067 | * - Device is reset when param is read. |
b12d437b | 1068 | * 2) A new device is supplied. |
43157840 | 1069 | * - No matching superblock found, resets device. |
b12d437b | 1070 | * 3) Device failure was transient and returns on reload. |
43157840 | 1071 | * - Failure noticed, resets device for bitmap replay. |
b12d437b | 1072 | * 4) Device hadn't completed recovery after previous failure. |
43157840 | 1073 | * - Superblock is read and overrides recovery_offset. |
b12d437b JB |
1074 | * |
1075 | * What is found in the superblocks of the devices is always | |
1076 | * authoritative, unless 'rebuild' or '[no]sync' was specified. | |
9d09e663 | 1077 | */ |
b12d437b | 1078 | for (i = 0; i < rs->md.raid_disks; i++) { |
9d09e663 | 1079 | set_bit(In_sync, &rs->dev[i].rdev.flags); |
b12d437b JB |
1080 | rs->dev[i].rdev.recovery_offset = MaxSector; |
1081 | } | |
9d09e663 | 1082 | |
b12d437b JB |
1083 | /* |
1084 | * Second, parse the unordered optional arguments | |
1085 | */ | |
9d09e663 | 1086 | for (i = 0; i < num_raid_params; i++) { |
4763e543 | 1087 | key = dm_shift_arg(as); |
bd83a4c4 MS |
1088 | if (!key) { |
1089 | rs->ti->error = "Not enough raid parameters given"; | |
1090 | return -EINVAL; | |
1091 | } | |
92c83d79 | 1092 | |
3fa6cf38 | 1093 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC))) { |
4286325b | 1094 | if (test_and_set_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { |
bd83a4c4 MS |
1095 | rs->ti->error = "Only one 'nosync' argument allowed"; |
1096 | return -EINVAL; | |
1097 | } | |
9d09e663 | 1098 | rs->md.recovery_cp = MaxSector; |
9d09e663 N |
1099 | continue; |
1100 | } | |
3fa6cf38 | 1101 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_SYNC))) { |
4286325b | 1102 | if (test_and_set_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) { |
bd83a4c4 MS |
1103 | rs->ti->error = "Only one 'sync' argument allowed"; |
1104 | return -EINVAL; | |
1105 | } | |
9d09e663 | 1106 | rs->md.recovery_cp = 0; |
4763e543 HM |
1107 | continue; |
1108 | } | |
3fa6cf38 | 1109 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_USE_NEAR_SETS))) { |
4286325b | 1110 | if (test_and_set_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1111 | rs->ti->error = "Only one 'raid10_use_new_sets' argument allowed"; |
1112 | return -EINVAL; | |
1113 | } | |
9d09e663 N |
1114 | continue; |
1115 | } | |
1116 | ||
92c83d79 HM |
1117 | arg = dm_shift_arg(as); |
1118 | i++; /* Account for the argument pairs */ | |
bd83a4c4 MS |
1119 | if (!arg) { |
1120 | rs->ti->error = "Wrong number of raid parameters given"; | |
1121 | return -EINVAL; | |
1122 | } | |
63f33b8d | 1123 | |
702108d1 HM |
1124 | /* |
1125 | * Parameters that take a string value are checked here. | |
1126 | */ | |
1127 | ||
3fa6cf38 | 1128 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT))) { |
4286325b | 1129 | if (test_and_set_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) { |
bd83a4c4 MS |
1130 | rs->ti->error = "Only one 'raid10_format' argument pair allowed"; |
1131 | return -EINVAL; | |
1132 | } | |
1133 | if (!rt_is_raid10(rt)) { | |
1134 | rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type"; | |
1135 | return -EINVAL; | |
1136 | } | |
33e53f06 | 1137 | raid10_format = raid10_name_to_format(arg); |
bd83a4c4 MS |
1138 | if (raid10_format < 0) { |
1139 | rs->ti->error = "Invalid 'raid10_format' value given"; | |
1140 | return raid10_format; | |
1141 | } | |
63f33b8d JB |
1142 | continue; |
1143 | } | |
1144 | ||
9dbd1aa3 | 1145 | if (kstrtoint(arg, 10, &value) < 0) { |
bd83a4c4 MS |
1146 | rs->ti->error = "Bad numerical argument given in raid params"; |
1147 | return -EINVAL; | |
1148 | } | |
702108d1 | 1149 | |
3fa6cf38 | 1150 | if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD))) { |
702108d1 HM |
1151 | /* |
1152 | * "rebuild" is being passed in by userspace to provide | |
1153 | * indexes of replaced devices and to set up additional | |
1154 | * devices on raid level takeover. | |
43157840 | 1155 | */ |
bb91a63f | 1156 | if (!__within_range(value, 0, rs->raid_disks - 1)) { |
bd83a4c4 MS |
1157 | rs->ti->error = "Invalid rebuild index given"; |
1158 | return -EINVAL; | |
1159 | } | |
702108d1 | 1160 | |
bd83a4c4 MS |
1161 | if (test_and_set_bit(value, (void *) rs->rebuild_disks)) { |
1162 | rs->ti->error = "rebuild for this index already given"; | |
1163 | return -EINVAL; | |
1164 | } | |
ecbfb9f1 | 1165 | |
702108d1 HM |
1166 | rd = rs->dev + value; |
1167 | clear_bit(In_sync, &rd->rdev.flags); | |
1168 | clear_bit(Faulty, &rd->rdev.flags); | |
1169 | rd->rdev.recovery_offset = 0; | |
4286325b | 1170 | set_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags); |
3fa6cf38 | 1171 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY))) { |
bd83a4c4 MS |
1172 | if (!rt_is_raid1(rt)) { |
1173 | rs->ti->error = "write_mostly option is only valid for RAID1"; | |
1174 | return -EINVAL; | |
1175 | } | |
702108d1 | 1176 | |
bb91a63f | 1177 | if (!__within_range(value, 0, rs->md.raid_disks - 1)) { |
bd83a4c4 MS |
1178 | rs->ti->error = "Invalid write_mostly index given"; |
1179 | return -EINVAL; | |
1180 | } | |
9d09e663 | 1181 | |
46bed2b5 | 1182 | set_bit(WriteMostly, &rs->dev[value].rdev.flags); |
4286325b | 1183 | set_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags); |
3fa6cf38 | 1184 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND))) { |
bd83a4c4 MS |
1185 | if (!rt_is_raid1(rt)) { |
1186 | rs->ti->error = "max_write_behind option is only valid for RAID1"; | |
1187 | return -EINVAL; | |
1188 | } | |
702108d1 | 1189 | |
4286325b | 1190 | if (test_and_set_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) { |
bd83a4c4 MS |
1191 | rs->ti->error = "Only one max_write_behind argument pair allowed"; |
1192 | return -EINVAL; | |
1193 | } | |
9d09e663 N |
1194 | |
1195 | /* | |
1196 | * In device-mapper, we specify things in sectors, but | |
1197 | * MD records this value in kB | |
1198 | */ | |
1199 | value /= 2; | |
bd83a4c4 MS |
1200 | if (value > COUNTER_MAX) { |
1201 | rs->ti->error = "Max write-behind limit out of range"; | |
1202 | return -EINVAL; | |
1203 | } | |
702108d1 | 1204 | |
9d09e663 | 1205 | rs->md.bitmap_info.max_write_behind = value; |
3fa6cf38 | 1206 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP))) { |
4286325b | 1207 | if (test_and_set_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) { |
bd83a4c4 MS |
1208 | rs->ti->error = "Only one daemon_sleep argument pair allowed"; |
1209 | return -EINVAL; | |
1210 | } | |
1211 | if (!value || (value > MAX_SCHEDULE_TIMEOUT)) { | |
1212 | rs->ti->error = "daemon sleep period out of range"; | |
1213 | return -EINVAL; | |
1214 | } | |
9d09e663 | 1215 | rs->md.bitmap_info.daemon_sleep = value; |
3fa6cf38 | 1216 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET))) { |
4763e543 | 1217 | /* Userspace passes new data_offset after having extended the the data image LV */ |
4286325b | 1218 | if (test_and_set_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { |
bd83a4c4 MS |
1219 | rs->ti->error = "Only one data_offset argument pair allowed"; |
1220 | return -EINVAL; | |
1221 | } | |
4763e543 | 1222 | /* Ensure sensible data offset */ |
bd83a4c4 MS |
1223 | if (value < 0) { |
1224 | rs->ti->error = "Bogus data_offset value"; | |
1225 | return -EINVAL; | |
1226 | } | |
4763e543 | 1227 | rs->data_offset = value; |
3fa6cf38 | 1228 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS))) { |
4763e543 | 1229 | /* Define the +/-# of disks to add to/remove from the given raid set */ |
4286325b | 1230 | if (test_and_set_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1231 | rs->ti->error = "Only one delta_disks argument pair allowed"; |
1232 | return -EINVAL; | |
1233 | } | |
4763e543 | 1234 | /* Ensure MAX_RAID_DEVICES and raid type minimal_devs! */ |
bb91a63f | 1235 | if (!__within_range(abs(value), 1, MAX_RAID_DEVICES - rt->minimal_devs)) { |
bd83a4c4 MS |
1236 | rs->ti->error = "Too many delta_disk requested"; |
1237 | return -EINVAL; | |
1238 | } | |
4763e543 HM |
1239 | |
1240 | rs->delta_disks = value; | |
3fa6cf38 | 1241 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE))) { |
4286325b | 1242 | if (test_and_set_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1243 | rs->ti->error = "Only one stripe_cache argument pair allowed"; |
1244 | return -EINVAL; | |
1245 | } | |
1246 | ||
bd83a4c4 MS |
1247 | if (!rt_is_raid456(rt)) { |
1248 | rs->ti->error = "Inappropriate argument: stripe_cache"; | |
1249 | return -EINVAL; | |
1250 | } | |
702108d1 | 1251 | |
9dbd1aa3 | 1252 | rs->stripe_cache_entries = value; |
3fa6cf38 | 1253 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE))) { |
4286325b | 1254 | if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1255 | rs->ti->error = "Only one min_recovery_rate argument pair allowed"; |
1256 | return -EINVAL; | |
1257 | } | |
1258 | if (value > INT_MAX) { | |
1259 | rs->ti->error = "min_recovery_rate out of range"; | |
1260 | return -EINVAL; | |
1261 | } | |
9d09e663 | 1262 | rs->md.sync_speed_min = (int)value; |
3fa6cf38 | 1263 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE))) { |
4286325b | 1264 | if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1265 | rs->ti->error = "Only one max_recovery_rate argument pair allowed"; |
1266 | return -EINVAL; | |
1267 | } | |
1268 | if (value > INT_MAX) { | |
1269 | rs->ti->error = "max_recovery_rate out of range"; | |
1270 | return -EINVAL; | |
1271 | } | |
9d09e663 | 1272 | rs->md.sync_speed_max = (int)value; |
3fa6cf38 | 1273 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE))) { |
4286325b | 1274 | if (test_and_set_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) { |
bd83a4c4 MS |
1275 | rs->ti->error = "Only one region_size argument pair allowed"; |
1276 | return -EINVAL; | |
1277 | } | |
702108d1 | 1278 | |
c1084561 | 1279 | region_size = value; |
4257e085 | 1280 | rs->requested_bitmap_chunk_sectors = value; |
3fa6cf38 | 1281 | } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES))) { |
4286325b | 1282 | if (test_and_set_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) { |
bd83a4c4 MS |
1283 | rs->ti->error = "Only one raid10_copies argument pair allowed"; |
1284 | return -EINVAL; | |
1285 | } | |
702108d1 | 1286 | |
bb91a63f | 1287 | if (!__within_range(value, 2, rs->md.raid_disks)) { |
bd83a4c4 MS |
1288 | rs->ti->error = "Bad value for 'raid10_copies'"; |
1289 | return -EINVAL; | |
1290 | } | |
702108d1 | 1291 | |
63f33b8d | 1292 | raid10_copies = value; |
9d09e663 N |
1293 | } else { |
1294 | DMERR("Unable to parse RAID parameter: %s", key); | |
bd83a4c4 MS |
1295 | rs->ti->error = "Unable to parse RAID parameter"; |
1296 | return -EINVAL; | |
9d09e663 N |
1297 | } |
1298 | } | |
1299 | ||
c1084561 JB |
1300 | if (validate_region_size(rs, region_size)) |
1301 | return -EINVAL; | |
1302 | ||
1303 | if (rs->md.chunk_sectors) | |
542f9038 | 1304 | max_io_len = rs->md.chunk_sectors; |
c1084561 | 1305 | else |
542f9038 | 1306 | max_io_len = region_size; |
c1084561 | 1307 | |
542f9038 MS |
1308 | if (dm_set_target_max_io_len(rs->ti, max_io_len)) |
1309 | return -EINVAL; | |
32737279 | 1310 | |
33e53f06 | 1311 | if (rt_is_raid10(rt)) { |
bd83a4c4 MS |
1312 | if (raid10_copies > rs->md.raid_disks) { |
1313 | rs->ti->error = "Not enough devices to satisfy specification"; | |
1314 | return -EINVAL; | |
1315 | } | |
63f33b8d | 1316 | |
33e53f06 | 1317 | rs->md.new_layout = raid10_format_to_md_layout(rs, raid10_format, raid10_copies); |
bd83a4c4 MS |
1318 | if (rs->md.new_layout < 0) { |
1319 | rs->ti->error = "Error getting raid10 format"; | |
1320 | return rs->md.new_layout; | |
1321 | } | |
33e53f06 HM |
1322 | |
1323 | rt = get_raid_type_by_ll(10, rs->md.new_layout); | |
bd83a4c4 MS |
1324 | if (!rt) { |
1325 | rs->ti->error = "Failed to recognize new raid10 layout"; | |
1326 | return -EINVAL; | |
1327 | } | |
33e53f06 HM |
1328 | |
1329 | if ((rt->algorithm == ALGORITHM_RAID10_DEFAULT || | |
1330 | rt->algorithm == ALGORITHM_RAID10_NEAR) && | |
4286325b | 1331 | test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { |
bd83a4c4 MS |
1332 | rs->ti->error = "RAID10 format 'near' and 'raid10_use_near_sets' are incompatible"; |
1333 | return -EINVAL; | |
1334 | } | |
bd83a4c4 | 1335 | } |
702108d1 | 1336 | |
33e53f06 | 1337 | rs->raid10_copies = raid10_copies; |
c039c332 | 1338 | |
9d09e663 N |
1339 | /* Assume there are no metadata devices until the drives are parsed */ |
1340 | rs->md.persistent = 0; | |
1341 | rs->md.external = 1; | |
1342 | ||
f090279e | 1343 | /* Check, if any invalid ctr arguments have been passed in for the raid level */ |
a30cbc0d | 1344 | return rs_check_for_valid_flags(rs); |
9d09e663 N |
1345 | } |
1346 | ||
9dbd1aa3 HM |
1347 | /* Set raid4/5/6 cache size */ |
1348 | static int rs_set_raid456_stripe_cache(struct raid_set *rs) | |
1349 | { | |
1350 | int r; | |
1351 | struct r5conf *conf; | |
1352 | struct mddev *mddev = &rs->md; | |
1353 | uint32_t min_stripes = max(mddev->chunk_sectors, mddev->new_chunk_sectors) / 2; | |
1354 | uint32_t nr_stripes = rs->stripe_cache_entries; | |
1355 | ||
1356 | if (!rt_is_raid456(rs->raid_type)) { | |
1357 | rs->ti->error = "Inappropriate raid level; cannot change stripe_cache size"; | |
1358 | return -EINVAL; | |
1359 | } | |
1360 | ||
1361 | if (nr_stripes < min_stripes) { | |
1362 | DMINFO("Adjusting requested %u stripe cache entries to %u to suit stripe size", | |
1363 | nr_stripes, min_stripes); | |
1364 | nr_stripes = min_stripes; | |
1365 | } | |
1366 | ||
1367 | conf = mddev->private; | |
1368 | if (!conf) { | |
1369 | rs->ti->error = "Cannot change stripe_cache size on inactive RAID set"; | |
1370 | return -EINVAL; | |
1371 | } | |
1372 | ||
1373 | /* Try setting number of stripes in raid456 stripe cache */ | |
1374 | if (conf->min_nr_stripes != nr_stripes) { | |
1375 | r = raid5_set_cache_size(mddev, nr_stripes); | |
1376 | if (r) { | |
1377 | rs->ti->error = "Failed to set raid4/5/6 stripe cache size"; | |
1378 | return r; | |
1379 | } | |
1380 | ||
1381 | DMINFO("%u stripe cache entries", nr_stripes); | |
1382 | } | |
1383 | ||
1384 | return 0; | |
1385 | } | |
1386 | ||
3a1c1ef2 HM |
1387 | /* Return # of data stripes as kept in mddev as of @rs (i.e. as of superblock) */ |
1388 | static unsigned int mddev_data_stripes(struct raid_set *rs) | |
1389 | { | |
1390 | return rs->md.raid_disks - rs->raid_type->parity_devs; | |
1391 | } | |
1392 | ||
40ba37e5 HM |
1393 | /* Return # of data stripes of @rs (i.e. as of ctr) */ |
1394 | static unsigned int rs_data_stripes(struct raid_set *rs) | |
1395 | { | |
1396 | return rs->raid_disks - rs->raid_type->parity_devs; | |
1397 | } | |
1398 | ||
1399 | /* Calculate the sectors per device and per array used for @rs */ | |
1400 | static int rs_set_dev_and_array_sectors(struct raid_set *rs, bool use_mddev) | |
1401 | { | |
1402 | int delta_disks; | |
1403 | unsigned int data_stripes; | |
1404 | struct mddev *mddev = &rs->md; | |
1405 | struct md_rdev *rdev; | |
1406 | sector_t array_sectors = rs->ti->len, dev_sectors = rs->ti->len; | |
9dbd1aa3 | 1407 | sector_t cur_dev_sectors = rs->dev[0].rdev.sectors; |
40ba37e5 HM |
1408 | |
1409 | if (use_mddev) { | |
1410 | delta_disks = mddev->delta_disks; | |
1411 | data_stripes = mddev_data_stripes(rs); | |
1412 | } else { | |
1413 | delta_disks = rs->delta_disks; | |
1414 | data_stripes = rs_data_stripes(rs); | |
1415 | } | |
1416 | ||
1417 | /* Special raid1 case w/o delta_disks support (yet) */ | |
1418 | if (rt_is_raid1(rs->raid_type)) | |
1419 | ; | |
1420 | else if (rt_is_raid10(rs->raid_type)) { | |
1421 | if (rs->raid10_copies < 2 || | |
1422 | delta_disks < 0) { | |
1423 | rs->ti->error = "Bogus raid10 data copies or delta disks"; | |
1424 | return EINVAL; | |
1425 | } | |
1426 | ||
1427 | dev_sectors *= rs->raid10_copies; | |
1428 | if (sector_div(dev_sectors, data_stripes)) | |
1429 | goto bad; | |
1430 | ||
1431 | array_sectors = (data_stripes + delta_disks) * dev_sectors; | |
1432 | if (sector_div(array_sectors, rs->raid10_copies)) | |
1433 | goto bad; | |
1434 | ||
1435 | } else if (sector_div(dev_sectors, data_stripes)) | |
1436 | goto bad; | |
1437 | ||
1438 | else | |
1439 | /* Striped layouts */ | |
1440 | array_sectors = (data_stripes + delta_disks) * dev_sectors; | |
1441 | ||
1442 | rdev_for_each(rdev, mddev) | |
1443 | rdev->sectors = dev_sectors; | |
1444 | ||
1445 | mddev->array_sectors = array_sectors; | |
1446 | mddev->dev_sectors = dev_sectors; | |
1447 | ||
9dbd1aa3 HM |
1448 | if (!rs_is_raid0(rs) && dev_sectors > cur_dev_sectors) |
1449 | mddev->recovery_cp = dev_sectors; | |
1450 | ||
40ba37e5 HM |
1451 | return 0; |
1452 | bad: | |
1453 | rs->ti->error = "Target length not divisible by number of data devices"; | |
1454 | return EINVAL; | |
1455 | } | |
1456 | ||
9d09e663 N |
1457 | static void do_table_event(struct work_struct *ws) |
1458 | { | |
1459 | struct raid_set *rs = container_of(ws, struct raid_set, md.event_work); | |
1460 | ||
9dbd1aa3 | 1461 | rs_set_capacity(rs); |
9d09e663 N |
1462 | dm_table_event(rs->ti->table); |
1463 | } | |
1464 | ||
1465 | static int raid_is_congested(struct dm_target_callbacks *cb, int bits) | |
1466 | { | |
1467 | struct raid_set *rs = container_of(cb, struct raid_set, callbacks); | |
1468 | ||
5c675f83 | 1469 | return mddev_congested(&rs->md, bits); |
9d09e663 N |
1470 | } |
1471 | ||
ecbfb9f1 HM |
1472 | /* |
1473 | * Make sure a valid takover (level switch) is being requested on @rs | |
1474 | * | |
1475 | * Conversions of raid sets from one MD personality to another | |
1476 | * have to conform to restrictions which are enforced here. | |
1477 | * | |
1478 | * Degration is already checked for in rs_check_conversion() below. | |
1479 | */ | |
1480 | static int rs_check_takeover(struct raid_set *rs) | |
1481 | { | |
1482 | struct mddev *mddev = &rs->md; | |
1483 | unsigned int near_copies; | |
1484 | ||
9dbd1aa3 HM |
1485 | if (rs->md.degraded) { |
1486 | rs->ti->error = "Can't takeover degraded raid set"; | |
1487 | return -EPERM; | |
1488 | } | |
1489 | ||
1490 | if (rs_is_reshaping(rs)) { | |
1491 | rs->ti->error = "Can't takeover reshaping raid set"; | |
1492 | return -EPERM; | |
1493 | } | |
1494 | ||
ecbfb9f1 HM |
1495 | switch (mddev->level) { |
1496 | case 0: | |
1497 | /* raid0 -> raid1/5 with one disk */ | |
1498 | if ((mddev->new_level == 1 || mddev->new_level == 5) && | |
1499 | mddev->raid_disks == 1) | |
1500 | return 0; | |
1501 | ||
1502 | /* raid0 -> raid10 */ | |
1503 | if (mddev->new_level == 10 && | |
9dbd1aa3 | 1504 | !(rs->raid_disks % mddev->raid_disks)) |
ecbfb9f1 HM |
1505 | return 0; |
1506 | ||
1507 | /* raid0 with multiple disks -> raid4/5/6 */ | |
bb91a63f | 1508 | if (__within_range(mddev->new_level, 4, 6) && |
ecbfb9f1 HM |
1509 | mddev->new_layout == ALGORITHM_PARITY_N && |
1510 | mddev->raid_disks > 1) | |
1511 | return 0; | |
1512 | ||
1513 | break; | |
1514 | ||
1515 | case 10: | |
1516 | /* Can't takeover raid10_offset! */ | |
e6ca5e1a | 1517 | if (__is_raid10_offset(mddev->layout)) |
ecbfb9f1 HM |
1518 | break; |
1519 | ||
e6ca5e1a | 1520 | near_copies = __raid10_near_copies(mddev->layout); |
ecbfb9f1 HM |
1521 | |
1522 | /* raid10* -> raid0 */ | |
1523 | if (mddev->new_level == 0) { | |
1524 | /* Can takeover raid10_near with raid disks divisable by data copies! */ | |
1525 | if (near_copies > 1 && | |
1526 | !(mddev->raid_disks % near_copies)) { | |
1527 | mddev->raid_disks /= near_copies; | |
1528 | mddev->delta_disks = mddev->raid_disks; | |
1529 | return 0; | |
1530 | } | |
1531 | ||
1532 | /* Can takeover raid10_far */ | |
1533 | if (near_copies == 1 && | |
e6ca5e1a | 1534 | __raid10_far_copies(mddev->layout) > 1) |
ecbfb9f1 HM |
1535 | return 0; |
1536 | ||
1537 | break; | |
1538 | } | |
1539 | ||
1540 | /* raid10_{near,far} -> raid1 */ | |
1541 | if (mddev->new_level == 1 && | |
e6ca5e1a | 1542 | max(near_copies, __raid10_far_copies(mddev->layout)) == mddev->raid_disks) |
ecbfb9f1 HM |
1543 | return 0; |
1544 | ||
1545 | /* raid10_{near,far} with 2 disks -> raid4/5 */ | |
bb91a63f | 1546 | if (__within_range(mddev->new_level, 4, 5) && |
ecbfb9f1 HM |
1547 | mddev->raid_disks == 2) |
1548 | return 0; | |
1549 | break; | |
1550 | ||
1551 | case 1: | |
1552 | /* raid1 with 2 disks -> raid4/5 */ | |
bb91a63f | 1553 | if (__within_range(mddev->new_level, 4, 5) && |
ecbfb9f1 HM |
1554 | mddev->raid_disks == 2) { |
1555 | mddev->degraded = 1; | |
1556 | return 0; | |
1557 | } | |
1558 | ||
1559 | /* raid1 -> raid0 */ | |
1560 | if (mddev->new_level == 0 && | |
1561 | mddev->raid_disks == 1) | |
1562 | return 0; | |
1563 | ||
1564 | /* raid1 -> raid10 */ | |
1565 | if (mddev->new_level == 10) | |
1566 | return 0; | |
1567 | ||
1568 | break; | |
1569 | ||
1570 | case 4: | |
1571 | /* raid4 -> raid0 */ | |
1572 | if (mddev->new_level == 0) | |
1573 | return 0; | |
1574 | ||
1575 | /* raid4 -> raid1/5 with 2 disks */ | |
1576 | if ((mddev->new_level == 1 || mddev->new_level == 5) && | |
1577 | mddev->raid_disks == 2) | |
1578 | return 0; | |
1579 | ||
1580 | /* raid4 -> raid5/6 with parity N */ | |
bb91a63f | 1581 | if (__within_range(mddev->new_level, 5, 6) && |
ecbfb9f1 HM |
1582 | mddev->layout == ALGORITHM_PARITY_N) |
1583 | return 0; | |
1584 | break; | |
1585 | ||
1586 | case 5: | |
1587 | /* raid5 with parity N -> raid0 */ | |
1588 | if (mddev->new_level == 0 && | |
1589 | mddev->layout == ALGORITHM_PARITY_N) | |
1590 | return 0; | |
1591 | ||
1592 | /* raid5 with parity N -> raid4 */ | |
1593 | if (mddev->new_level == 4 && | |
1594 | mddev->layout == ALGORITHM_PARITY_N) | |
1595 | return 0; | |
1596 | ||
1597 | /* raid5 with 2 disks -> raid1/4/10 */ | |
1598 | if ((mddev->new_level == 1 || mddev->new_level == 4 || mddev->new_level == 10) && | |
1599 | mddev->raid_disks == 2) | |
1600 | return 0; | |
1601 | ||
1602 | /* raid5 with parity N -> raid6 with parity N */ | |
1603 | if (mddev->new_level == 6 && | |
1604 | ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || | |
bb91a63f | 1605 | __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC_6, ALGORITHM_RIGHT_SYMMETRIC_6))) |
ecbfb9f1 HM |
1606 | return 0; |
1607 | break; | |
1608 | ||
1609 | case 6: | |
1610 | /* raid6 with parity N -> raid0 */ | |
1611 | if (mddev->new_level == 0 && | |
1612 | mddev->layout == ALGORITHM_PARITY_N) | |
1613 | return 0; | |
1614 | ||
1615 | /* raid6 with parity N -> raid4 */ | |
1616 | if (mddev->new_level == 4 && | |
1617 | mddev->layout == ALGORITHM_PARITY_N) | |
1618 | return 0; | |
1619 | ||
1620 | /* raid6_*_n with parity N -> raid5_* */ | |
1621 | if (mddev->new_level == 5 && | |
1622 | ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || | |
bb91a63f | 1623 | __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC, ALGORITHM_RIGHT_SYMMETRIC))) |
ecbfb9f1 HM |
1624 | return 0; |
1625 | ||
1626 | default: | |
1627 | break; | |
1628 | } | |
1629 | ||
bd83a4c4 MS |
1630 | rs->ti->error = "takeover not possible"; |
1631 | return -EINVAL; | |
ecbfb9f1 HM |
1632 | } |
1633 | ||
1634 | /* True if @rs requested to be taken over */ | |
1635 | static bool rs_takeover_requested(struct raid_set *rs) | |
1636 | { | |
1637 | return rs->md.new_level != rs->md.level; | |
1638 | } | |
1639 | ||
40ba37e5 HM |
1640 | /* True if @rs is requested to reshape by ctr */ |
1641 | static bool rs_reshape_requested(struct raid_set *rs) | |
1642 | { | |
1643 | struct mddev *mddev = &rs->md; | |
1644 | ||
1645 | if (!mddev->level) | |
1646 | return false; | |
1647 | ||
1648 | return !__is_raid10_far(mddev->new_layout) && | |
1649 | mddev->new_level == mddev->level && | |
1650 | (mddev->new_layout != mddev->layout || | |
1651 | mddev->new_chunk_sectors != mddev->chunk_sectors || | |
1652 | rs->raid_disks + rs->delta_disks != mddev->raid_disks); | |
1653 | } | |
1654 | ||
33e53f06 | 1655 | /* Features */ |
9b6e5423 | 1656 | #define FEATURE_FLAG_SUPPORTS_V190 0x1 /* Supports extended superblock */ |
33e53f06 HM |
1657 | |
1658 | /* State flags for sb->flags */ | |
1659 | #define SB_FLAG_RESHAPE_ACTIVE 0x1 | |
1660 | #define SB_FLAG_RESHAPE_BACKWARDS 0x2 | |
1661 | ||
b12d437b JB |
1662 | /* |
1663 | * This structure is never routinely used by userspace, unlike md superblocks. | |
1664 | * Devices with this superblock should only ever be accessed via device-mapper. | |
1665 | */ | |
1666 | #define DM_RAID_MAGIC 0x64526D44 | |
1667 | struct dm_raid_superblock { | |
1668 | __le32 magic; /* "DmRd" */ | |
9b6e5423 | 1669 | __le32 compat_features; /* Used to indicate compatible features (like 1.9.0 ondisk metadata extension) */ |
b12d437b | 1670 | |
33e53f06 HM |
1671 | __le32 num_devices; /* Number of devices in this raid set. (Max 64) */ |
1672 | __le32 array_position; /* The position of this drive in the raid set */ | |
b12d437b JB |
1673 | |
1674 | __le64 events; /* Incremented by md when superblock updated */ | |
9b6e5423 | 1675 | __le64 failed_devices; /* Pre 1.9.0 part of bit field of devices to */ |
33e53f06 | 1676 | /* indicate failures (see extension below) */ |
b12d437b JB |
1677 | |
1678 | /* | |
1679 | * This offset tracks the progress of the repair or replacement of | |
1680 | * an individual drive. | |
1681 | */ | |
1682 | __le64 disk_recovery_offset; | |
1683 | ||
1684 | /* | |
33e53f06 | 1685 | * This offset tracks the progress of the initial raid set |
b12d437b JB |
1686 | * synchronisation/parity calculation. |
1687 | */ | |
1688 | __le64 array_resync_offset; | |
1689 | ||
1690 | /* | |
33e53f06 | 1691 | * raid characteristics |
b12d437b JB |
1692 | */ |
1693 | __le32 level; | |
1694 | __le32 layout; | |
1695 | __le32 stripe_sectors; | |
1696 | ||
33e53f06 | 1697 | /******************************************************************** |
9b6e5423 | 1698 | * BELOW FOLLOW V1.9.0 EXTENSIONS TO THE PRISTINE SUPERBLOCK FORMAT!!! |
33e53f06 | 1699 | * |
9b6e5423 | 1700 | * FEATURE_FLAG_SUPPORTS_V190 in the features member indicates that those exist |
33e53f06 HM |
1701 | */ |
1702 | ||
1703 | __le32 flags; /* Flags defining array states for reshaping */ | |
1704 | ||
1705 | /* | |
1706 | * This offset tracks the progress of a raid | |
1707 | * set reshape in order to be able to restart it | |
1708 | */ | |
1709 | __le64 reshape_position; | |
1710 | ||
1711 | /* | |
1712 | * These define the properties of the array in case of an interrupted reshape | |
1713 | */ | |
1714 | __le32 new_level; | |
1715 | __le32 new_layout; | |
1716 | __le32 new_stripe_sectors; | |
1717 | __le32 delta_disks; | |
1718 | ||
1719 | __le64 array_sectors; /* Array size in sectors */ | |
1720 | ||
1721 | /* | |
1722 | * Sector offsets to data on devices (reshaping). | |
1723 | * Needed to support out of place reshaping, thus | |
1724 | * not writing over any stripes whilst converting | |
1725 | * them from old to new layout | |
1726 | */ | |
1727 | __le64 data_offset; | |
1728 | __le64 new_data_offset; | |
1729 | ||
1730 | __le64 sectors; /* Used device size in sectors */ | |
1731 | ||
1732 | /* | |
1733 | * Additonal Bit field of devices indicating failures to support | |
9b6e5423 | 1734 | * up to 256 devices with the 1.9.0 on-disk metadata format |
33e53f06 HM |
1735 | */ |
1736 | __le64 extended_failed_devices[DISKS_ARRAY_ELEMS - 1]; | |
1737 | ||
1738 | __le32 incompat_features; /* Used to indicate any incompatible features */ | |
1739 | ||
1740 | /* Always set rest up to logical block size to 0 when writing (see get_metadata_device() below). */ | |
b12d437b JB |
1741 | } __packed; |
1742 | ||
9dbd1aa3 HM |
1743 | /* |
1744 | * Check for reshape constraints on raid set @rs: | |
1745 | * | |
1746 | * - reshape function non-existent | |
1747 | * - degraded set | |
1748 | * - ongoing recovery | |
1749 | * - ongoing reshape | |
1750 | * | |
1751 | * Returns 0 if none or -EPERM if given constraint | |
1752 | * and error message reference in @errmsg | |
1753 | */ | |
1754 | static int rs_check_reshape(struct raid_set *rs) | |
1755 | { | |
1756 | struct mddev *mddev = &rs->md; | |
1757 | ||
9dbd1aa3 HM |
1758 | if (!mddev->pers || !mddev->pers->check_reshape) |
1759 | rs->ti->error = "Reshape not supported"; | |
1760 | else if (mddev->degraded) | |
1761 | rs->ti->error = "Can't reshape degraded raid set"; | |
1762 | else if (rs_is_recovering(rs)) | |
1763 | rs->ti->error = "Convert request on recovering raid set prohibited"; | |
1764 | else if (mddev->reshape_position && rs_is_reshaping(rs)) | |
1765 | rs->ti->error = "raid set already reshaping!"; | |
1766 | else if (!(rs_is_raid10(rs) || rs_is_raid456(rs))) | |
1767 | rs->ti->error = "Reshaping only supported for raid4/5/6/10"; | |
1768 | else | |
1769 | return 0; | |
1770 | ||
1771 | return -EPERM; | |
1772 | } | |
1773 | ||
3cb03002 | 1774 | static int read_disk_sb(struct md_rdev *rdev, int size) |
b12d437b JB |
1775 | { |
1776 | BUG_ON(!rdev->sb_page); | |
1777 | ||
1778 | if (rdev->sb_loaded) | |
1779 | return 0; | |
1780 | ||
796a5cf0 | 1781 | if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, 0, 1)) { |
0447568f JB |
1782 | DMERR("Failed to read superblock of device at position %d", |
1783 | rdev->raid_disk); | |
c32fb9e7 | 1784 | md_error(rdev->mddev, rdev); |
b12d437b JB |
1785 | return -EINVAL; |
1786 | } | |
1787 | ||
1788 | rdev->sb_loaded = 1; | |
1789 | ||
1790 | return 0; | |
1791 | } | |
1792 | ||
33e53f06 HM |
1793 | static void sb_retrieve_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) |
1794 | { | |
1795 | failed_devices[0] = le64_to_cpu(sb->failed_devices); | |
1796 | memset(failed_devices + 1, 0, sizeof(sb->extended_failed_devices)); | |
1797 | ||
4286325b | 1798 | if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { |
33e53f06 HM |
1799 | int i = ARRAY_SIZE(sb->extended_failed_devices); |
1800 | ||
1801 | while (i--) | |
1802 | failed_devices[i+1] = le64_to_cpu(sb->extended_failed_devices[i]); | |
1803 | } | |
1804 | } | |
1805 | ||
7b34df74 HM |
1806 | static void sb_update_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) |
1807 | { | |
1808 | int i = ARRAY_SIZE(sb->extended_failed_devices); | |
1809 | ||
1810 | sb->failed_devices = cpu_to_le64(failed_devices[0]); | |
1811 | while (i--) | |
1812 | sb->extended_failed_devices[i] = cpu_to_le64(failed_devices[i+1]); | |
1813 | } | |
1814 | ||
1815 | /* | |
1816 | * Synchronize the superblock members with the raid set properties | |
1817 | * | |
1818 | * All superblock data is little endian. | |
1819 | */ | |
fd01b88c | 1820 | static void super_sync(struct mddev *mddev, struct md_rdev *rdev) |
b12d437b | 1821 | { |
7b34df74 HM |
1822 | bool update_failed_devices = false; |
1823 | unsigned int i; | |
1824 | uint64_t failed_devices[DISKS_ARRAY_ELEMS]; | |
b12d437b | 1825 | struct dm_raid_superblock *sb; |
81f382f9 | 1826 | struct raid_set *rs = container_of(mddev, struct raid_set, md); |
b12d437b | 1827 | |
7b34df74 HM |
1828 | /* No metadata device, no superblock */ |
1829 | if (!rdev->meta_bdev) | |
1830 | return; | |
1831 | ||
1832 | BUG_ON(!rdev->sb_page); | |
1833 | ||
b12d437b | 1834 | sb = page_address(rdev->sb_page); |
b12d437b | 1835 | |
7b34df74 | 1836 | sb_retrieve_failed_devices(sb, failed_devices); |
b12d437b | 1837 | |
7b34df74 HM |
1838 | for (i = 0; i < rs->raid_disks; i++) |
1839 | if (!rs->dev[i].data_dev || test_bit(Faulty, &rs->dev[i].rdev.flags)) { | |
1840 | update_failed_devices = true; | |
1841 | set_bit(i, (void *) failed_devices); | |
1842 | } | |
1843 | ||
1844 | if (update_failed_devices) | |
1845 | sb_update_failed_devices(sb, failed_devices); | |
b12d437b JB |
1846 | |
1847 | sb->magic = cpu_to_le32(DM_RAID_MAGIC); | |
9b6e5423 | 1848 | sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); |
b12d437b JB |
1849 | |
1850 | sb->num_devices = cpu_to_le32(mddev->raid_disks); | |
1851 | sb->array_position = cpu_to_le32(rdev->raid_disk); | |
1852 | ||
1853 | sb->events = cpu_to_le64(mddev->events); | |
b12d437b JB |
1854 | |
1855 | sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset); | |
1856 | sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp); | |
1857 | ||
1858 | sb->level = cpu_to_le32(mddev->level); | |
1859 | sb->layout = cpu_to_le32(mddev->layout); | |
1860 | sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors); | |
7b34df74 HM |
1861 | |
1862 | sb->new_level = cpu_to_le32(mddev->new_level); | |
1863 | sb->new_layout = cpu_to_le32(mddev->new_layout); | |
1864 | sb->new_stripe_sectors = cpu_to_le32(mddev->new_chunk_sectors); | |
1865 | ||
1866 | sb->delta_disks = cpu_to_le32(mddev->delta_disks); | |
1867 | ||
1868 | smp_rmb(); /* Make sure we access most recent reshape position */ | |
1869 | sb->reshape_position = cpu_to_le64(mddev->reshape_position); | |
1870 | if (le64_to_cpu(sb->reshape_position) != MaxSector) { | |
1871 | /* Flag ongoing reshape */ | |
1872 | sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE); | |
1873 | ||
1874 | if (mddev->delta_disks < 0 || mddev->reshape_backwards) | |
1875 | sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_BACKWARDS); | |
4286325b MS |
1876 | } else { |
1877 | /* Clear reshape flags */ | |
1878 | sb->flags &= ~(cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE|SB_FLAG_RESHAPE_BACKWARDS)); | |
1879 | } | |
7b34df74 HM |
1880 | |
1881 | sb->array_sectors = cpu_to_le64(mddev->array_sectors); | |
1882 | sb->data_offset = cpu_to_le64(rdev->data_offset); | |
1883 | sb->new_data_offset = cpu_to_le64(rdev->new_data_offset); | |
1884 | sb->sectors = cpu_to_le64(rdev->sectors); | |
1885 | ||
1886 | /* Zero out the rest of the payload after the size of the superblock */ | |
1887 | memset(sb + 1, 0, rdev->sb_size - sizeof(*sb)); | |
b12d437b JB |
1888 | } |
1889 | ||
1890 | /* | |
1891 | * super_load | |
1892 | * | |
1893 | * This function creates a superblock if one is not found on the device | |
1894 | * and will decide which superblock to use if there's a choice. | |
1895 | * | |
1896 | * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise | |
1897 | */ | |
3cb03002 | 1898 | static int super_load(struct md_rdev *rdev, struct md_rdev *refdev) |
b12d437b | 1899 | { |
73c6f239 | 1900 | int r; |
b12d437b JB |
1901 | struct dm_raid_superblock *sb; |
1902 | struct dm_raid_superblock *refsb; | |
1903 | uint64_t events_sb, events_refsb; | |
1904 | ||
1905 | rdev->sb_start = 0; | |
40d43c4b HM |
1906 | rdev->sb_size = bdev_logical_block_size(rdev->meta_bdev); |
1907 | if (rdev->sb_size < sizeof(*sb) || rdev->sb_size > PAGE_SIZE) { | |
1908 | DMERR("superblock size of a logical block is no longer valid"); | |
1909 | return -EINVAL; | |
1910 | } | |
b12d437b | 1911 | |
73c6f239 HM |
1912 | r = read_disk_sb(rdev, rdev->sb_size); |
1913 | if (r) | |
1914 | return r; | |
b12d437b JB |
1915 | |
1916 | sb = page_address(rdev->sb_page); | |
3aa3b2b2 JB |
1917 | |
1918 | /* | |
1919 | * Two cases that we want to write new superblocks and rebuild: | |
1920 | * 1) New device (no matching magic number) | |
1921 | * 2) Device specified for rebuild (!In_sync w/ offset == 0) | |
1922 | */ | |
1923 | if ((sb->magic != cpu_to_le32(DM_RAID_MAGIC)) || | |
1924 | (!test_bit(In_sync, &rdev->flags) && !rdev->recovery_offset)) { | |
b12d437b JB |
1925 | super_sync(rdev->mddev, rdev); |
1926 | ||
1927 | set_bit(FirstUse, &rdev->flags); | |
9b6e5423 | 1928 | sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); |
b12d437b JB |
1929 | |
1930 | /* Force writing of superblocks to disk */ | |
1931 | set_bit(MD_CHANGE_DEVS, &rdev->mddev->flags); | |
1932 | ||
1933 | /* Any superblock is better than none, choose that if given */ | |
1934 | return refdev ? 0 : 1; | |
1935 | } | |
1936 | ||
1937 | if (!refdev) | |
1938 | return 1; | |
1939 | ||
1940 | events_sb = le64_to_cpu(sb->events); | |
1941 | ||
1942 | refsb = page_address(refdev->sb_page); | |
1943 | events_refsb = le64_to_cpu(refsb->events); | |
1944 | ||
1945 | return (events_sb > events_refsb) ? 1 : 0; | |
1946 | } | |
1947 | ||
33e53f06 | 1948 | static int super_init_validation(struct raid_set *rs, struct md_rdev *rdev) |
b12d437b JB |
1949 | { |
1950 | int role; | |
33e53f06 HM |
1951 | unsigned int d; |
1952 | struct mddev *mddev = &rs->md; | |
b12d437b | 1953 | uint64_t events_sb; |
33e53f06 | 1954 | uint64_t failed_devices[DISKS_ARRAY_ELEMS]; |
b12d437b | 1955 | struct dm_raid_superblock *sb; |
33e53f06 | 1956 | uint32_t new_devs = 0, rebuild_and_new = 0, rebuilds = 0; |
dafb20fa | 1957 | struct md_rdev *r; |
b12d437b JB |
1958 | struct dm_raid_superblock *sb2; |
1959 | ||
1960 | sb = page_address(rdev->sb_page); | |
1961 | events_sb = le64_to_cpu(sb->events); | |
b12d437b JB |
1962 | |
1963 | /* | |
1964 | * Initialise to 1 if this is a new superblock. | |
1965 | */ | |
1966 | mddev->events = events_sb ? : 1; | |
1967 | ||
33e53f06 HM |
1968 | mddev->reshape_position = MaxSector; |
1969 | ||
b12d437b | 1970 | /* |
33e53f06 HM |
1971 | * Reshaping is supported, e.g. reshape_position is valid |
1972 | * in superblock and superblock content is authoritative. | |
b12d437b | 1973 | */ |
4286325b | 1974 | if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { |
33e53f06 HM |
1975 | /* Superblock is authoritative wrt given raid set layout! */ |
1976 | mddev->raid_disks = le32_to_cpu(sb->num_devices); | |
1977 | mddev->level = le32_to_cpu(sb->level); | |
1978 | mddev->layout = le32_to_cpu(sb->layout); | |
1979 | mddev->chunk_sectors = le32_to_cpu(sb->stripe_sectors); | |
1980 | mddev->new_level = le32_to_cpu(sb->new_level); | |
1981 | mddev->new_layout = le32_to_cpu(sb->new_layout); | |
1982 | mddev->new_chunk_sectors = le32_to_cpu(sb->new_stripe_sectors); | |
1983 | mddev->delta_disks = le32_to_cpu(sb->delta_disks); | |
1984 | mddev->array_sectors = le64_to_cpu(sb->array_sectors); | |
1985 | ||
1986 | /* raid was reshaping and got interrupted */ | |
4286325b MS |
1987 | if (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_ACTIVE) { |
1988 | if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { | |
33e53f06 HM |
1989 | DMERR("Reshape requested but raid set is still reshaping"); |
1990 | return -EINVAL; | |
1991 | } | |
b12d437b | 1992 | |
33e53f06 | 1993 | if (mddev->delta_disks < 0 || |
4286325b | 1994 | (!mddev->delta_disks && (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_BACKWARDS))) |
33e53f06 HM |
1995 | mddev->reshape_backwards = 1; |
1996 | else | |
1997 | mddev->reshape_backwards = 0; | |
1998 | ||
1999 | mddev->reshape_position = le64_to_cpu(sb->reshape_position); | |
2000 | rs->raid_type = get_raid_type_by_ll(mddev->level, mddev->layout); | |
2001 | } | |
2002 | ||
2003 | } else { | |
2004 | /* | |
9b6e5423 | 2005 | * No takeover/reshaping, because we don't have the extended v1.9.0 metadata |
33e53f06 HM |
2006 | */ |
2007 | if (le32_to_cpu(sb->level) != mddev->level) { | |
2008 | DMERR("Reshaping/takeover raid sets not yet supported. (raid level/stripes/size change)"); | |
2009 | return -EINVAL; | |
2010 | } | |
2011 | if (le32_to_cpu(sb->layout) != mddev->layout) { | |
2012 | DMERR("Reshaping raid sets not yet supported. (raid layout change)"); | |
43157840 MS |
2013 | DMERR(" 0x%X vs 0x%X", le32_to_cpu(sb->layout), mddev->layout); |
2014 | DMERR(" Old layout: %s w/ %d copies", | |
33e53f06 HM |
2015 | raid10_md_layout_to_format(le32_to_cpu(sb->layout)), |
2016 | raid10_md_layout_to_copies(le32_to_cpu(sb->layout))); | |
43157840 | 2017 | DMERR(" New layout: %s w/ %d copies", |
33e53f06 HM |
2018 | raid10_md_layout_to_format(mddev->layout), |
2019 | raid10_md_layout_to_copies(mddev->layout)); | |
2020 | return -EINVAL; | |
2021 | } | |
2022 | if (le32_to_cpu(sb->stripe_sectors) != mddev->chunk_sectors) { | |
2023 | DMERR("Reshaping raid sets not yet supported. (stripe sectors change)"); | |
2024 | return -EINVAL; | |
2025 | } | |
2026 | ||
2027 | /* We can only change the number of devices in raid1 with old (i.e. pre 1.0.7) metadata */ | |
2028 | if (!rt_is_raid1(rs->raid_type) && | |
2029 | (le32_to_cpu(sb->num_devices) != mddev->raid_disks)) { | |
2030 | DMERR("Reshaping raid sets not yet supported. (device count change from %u to %u)", | |
2031 | sb->num_devices, mddev->raid_disks); | |
2032 | return -EINVAL; | |
2033 | } | |
2034 | ||
2035 | /* Table line is checked vs. authoritative superblock */ | |
2036 | rs_set_new(rs); | |
b12d437b JB |
2037 | } |
2038 | ||
4286325b | 2039 | if (!test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) |
b12d437b JB |
2040 | mddev->recovery_cp = le64_to_cpu(sb->array_resync_offset); |
2041 | ||
2042 | /* | |
2043 | * During load, we set FirstUse if a new superblock was written. | |
2044 | * There are two reasons we might not have a superblock: | |
33e53f06 | 2045 | * 1) The raid set is brand new - in which case, all of the |
43157840 | 2046 | * devices must have their In_sync bit set. Also, |
b12d437b | 2047 | * recovery_cp must be 0, unless forced. |
33e53f06 | 2048 | * 2) This is a new device being added to an old raid set |
b12d437b JB |
2049 | * and the new device needs to be rebuilt - in which |
2050 | * case the In_sync bit will /not/ be set and | |
2051 | * recovery_cp must be MaxSector. | |
9dbd1aa3 HM |
2052 | * 3) This is/are a new device(s) being added to an old |
2053 | * raid set during takeover to a higher raid level | |
2054 | * to provide capacity for redundancy or during reshape | |
2055 | * to add capacity to grow the raid set. | |
b12d437b | 2056 | */ |
33e53f06 | 2057 | d = 0; |
dafb20fa | 2058 | rdev_for_each(r, mddev) { |
33e53f06 HM |
2059 | if (test_bit(FirstUse, &r->flags)) |
2060 | new_devs++; | |
2061 | ||
b12d437b | 2062 | if (!test_bit(In_sync, &r->flags)) { |
33e53f06 HM |
2063 | DMINFO("Device %d specified for rebuild; clearing superblock", |
2064 | r->raid_disk); | |
b12d437b | 2065 | rebuilds++; |
33e53f06 HM |
2066 | |
2067 | if (test_bit(FirstUse, &r->flags)) | |
2068 | rebuild_and_new++; | |
2069 | } | |
2070 | ||
2071 | d++; | |
b12d437b JB |
2072 | } |
2073 | ||
33e53f06 HM |
2074 | if (new_devs == rs->raid_disks || !rebuilds) { |
2075 | /* Replace a broken device */ | |
2076 | if (new_devs == 1 && !rs->delta_disks) | |
2077 | ; | |
2078 | if (new_devs == rs->raid_disks) { | |
2079 | DMINFO("Superblocks created for new raid set"); | |
b12d437b | 2080 | set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); |
33e53f06 | 2081 | mddev->recovery_cp = 0; |
9dbd1aa3 HM |
2082 | } else if (new_devs != rebuilds && |
2083 | new_devs != rs->delta_disks) { | |
33e53f06 HM |
2084 | DMERR("New device injected into existing raid set without " |
2085 | "'delta_disks' or 'rebuild' parameter specified"); | |
b12d437b JB |
2086 | return -EINVAL; |
2087 | } | |
33e53f06 HM |
2088 | } else if (new_devs && new_devs != rebuilds) { |
2089 | DMERR("%u 'rebuild' devices cannot be injected into" | |
2090 | " a raid set with %u other first-time devices", | |
2091 | rebuilds, new_devs); | |
b12d437b | 2092 | return -EINVAL; |
33e53f06 HM |
2093 | } else if (rebuilds) { |
2094 | if (rebuild_and_new && rebuilds != rebuild_and_new) { | |
2095 | DMERR("new device%s provided without 'rebuild'", | |
2096 | new_devs > 1 ? "s" : ""); | |
2097 | return -EINVAL; | |
9dbd1aa3 | 2098 | } else if (rs_is_recovering(rs)) { |
33e53f06 HM |
2099 | DMERR("'rebuild' specified while raid set is not in-sync (recovery_cp=%llu)", |
2100 | (unsigned long long) mddev->recovery_cp); | |
2101 | return -EINVAL; | |
9dbd1aa3 HM |
2102 | } else if (rs_is_reshaping(rs)) { |
2103 | DMERR("'rebuild' specified while raid set is being reshaped (reshape_position=%llu)", | |
2104 | (unsigned long long) mddev->reshape_position); | |
33e53f06 HM |
2105 | return -EINVAL; |
2106 | } | |
b12d437b JB |
2107 | } |
2108 | ||
2109 | /* | |
2110 | * Now we set the Faulty bit for those devices that are | |
2111 | * recorded in the superblock as failed. | |
2112 | */ | |
33e53f06 | 2113 | sb_retrieve_failed_devices(sb, failed_devices); |
dafb20fa | 2114 | rdev_for_each(r, mddev) { |
b12d437b JB |
2115 | if (!r->sb_page) |
2116 | continue; | |
2117 | sb2 = page_address(r->sb_page); | |
2118 | sb2->failed_devices = 0; | |
33e53f06 | 2119 | memset(sb2->extended_failed_devices, 0, sizeof(sb2->extended_failed_devices)); |
b12d437b JB |
2120 | |
2121 | /* | |
2122 | * Check for any device re-ordering. | |
2123 | */ | |
2124 | if (!test_bit(FirstUse, &r->flags) && (r->raid_disk >= 0)) { | |
2125 | role = le32_to_cpu(sb2->array_position); | |
33e53f06 HM |
2126 | if (role < 0) |
2127 | continue; | |
2128 | ||
b12d437b | 2129 | if (role != r->raid_disk) { |
e6ca5e1a MS |
2130 | if (__is_raid10_near(mddev->layout)) { |
2131 | if (mddev->raid_disks % __raid10_near_copies(mddev->layout) || | |
bd83a4c4 MS |
2132 | rs->raid_disks % rs->raid10_copies) { |
2133 | rs->ti->error = | |
2134 | "Cannot change raid10 near set to odd # of devices!"; | |
2135 | return -EINVAL; | |
2136 | } | |
33e53f06 HM |
2137 | |
2138 | sb2->array_position = cpu_to_le32(r->raid_disk); | |
2139 | ||
2140 | } else if (!(rs_is_raid10(rs) && rt_is_raid0(rs->raid_type)) && | |
bd83a4c4 MS |
2141 | !(rs_is_raid0(rs) && rt_is_raid10(rs->raid_type)) && |
2142 | !rt_is_raid1(rs->raid_type)) { | |
2143 | rs->ti->error = "Cannot change device positions in raid set"; | |
2144 | return -EINVAL; | |
2145 | } | |
33e53f06 | 2146 | |
bd83a4c4 | 2147 | DMINFO("raid device #%d now at position #%d", role, r->raid_disk); |
b12d437b JB |
2148 | } |
2149 | ||
2150 | /* | |
2151 | * Partial recovery is performed on | |
2152 | * returning failed devices. | |
2153 | */ | |
33e53f06 | 2154 | if (test_bit(role, (void *) failed_devices)) |
b12d437b JB |
2155 | set_bit(Faulty, &r->flags); |
2156 | } | |
2157 | } | |
2158 | ||
2159 | return 0; | |
2160 | } | |
2161 | ||
0cf45031 | 2162 | static int super_validate(struct raid_set *rs, struct md_rdev *rdev) |
b12d437b | 2163 | { |
0cf45031 | 2164 | struct mddev *mddev = &rs->md; |
33e53f06 HM |
2165 | struct dm_raid_superblock *sb; |
2166 | ||
3a1c1ef2 | 2167 | if (rs_is_raid0(rs) || !rdev->sb_page) |
33e53f06 HM |
2168 | return 0; |
2169 | ||
2170 | sb = page_address(rdev->sb_page); | |
b12d437b JB |
2171 | |
2172 | /* | |
2173 | * If mddev->events is not set, we know we have not yet initialized | |
2174 | * the array. | |
2175 | */ | |
33e53f06 | 2176 | if (!mddev->events && super_init_validation(rs, rdev)) |
b12d437b JB |
2177 | return -EINVAL; |
2178 | ||
9b6e5423 MS |
2179 | if (le32_to_cpu(sb->compat_features) != FEATURE_FLAG_SUPPORTS_V190) { |
2180 | rs->ti->error = "Unable to assemble array: Unknown flag(s) in compatible feature flags"; | |
2181 | return -EINVAL; | |
2182 | } | |
2183 | ||
2184 | if (sb->incompat_features) { | |
ecbfb9f1 | 2185 | rs->ti->error = "Unable to assemble array: No incompatible feature flags supported yet"; |
4c9971ca HM |
2186 | return -EINVAL; |
2187 | } | |
2188 | ||
0cf45031 | 2189 | /* Enable bitmap creation for RAID levels != 0 */ |
676fa5ad | 2190 | mddev->bitmap_info.offset = rt_is_raid0(rs->raid_type) ? 0 : to_sector(4096); |
0cf45031 HM |
2191 | rdev->mddev->bitmap_info.default_offset = mddev->bitmap_info.offset; |
2192 | ||
33e53f06 HM |
2193 | if (!test_and_clear_bit(FirstUse, &rdev->flags)) { |
2194 | /* Retrieve device size stored in superblock to be prepared for shrink */ | |
2195 | rdev->sectors = le64_to_cpu(sb->sectors); | |
b12d437b | 2196 | rdev->recovery_offset = le64_to_cpu(sb->disk_recovery_offset); |
33e53f06 HM |
2197 | if (rdev->recovery_offset == MaxSector) |
2198 | set_bit(In_sync, &rdev->flags); | |
2199 | /* | |
2200 | * If no reshape in progress -> we're recovering single | |
2201 | * disk(s) and have to set the device(s) to out-of-sync | |
2202 | */ | |
9dbd1aa3 | 2203 | else if (!rs_is_reshaping(rs)) |
33e53f06 | 2204 | clear_bit(In_sync, &rdev->flags); /* Mandatory for recovery */ |
b12d437b JB |
2205 | } |
2206 | ||
2207 | /* | |
2208 | * If a device comes back, set it as not In_sync and no longer faulty. | |
2209 | */ | |
33e53f06 HM |
2210 | if (test_and_clear_bit(Faulty, &rdev->flags)) { |
2211 | rdev->recovery_offset = 0; | |
b12d437b JB |
2212 | clear_bit(In_sync, &rdev->flags); |
2213 | rdev->saved_raid_disk = rdev->raid_disk; | |
b12d437b JB |
2214 | } |
2215 | ||
33e53f06 HM |
2216 | /* Reshape support -> restore repective data offsets */ |
2217 | rdev->data_offset = le64_to_cpu(sb->data_offset); | |
2218 | rdev->new_data_offset = le64_to_cpu(sb->new_data_offset); | |
b12d437b JB |
2219 | |
2220 | return 0; | |
2221 | } | |
2222 | ||
2223 | /* | |
2224 | * Analyse superblocks and select the freshest. | |
2225 | */ | |
2226 | static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs) | |
2227 | { | |
73c6f239 | 2228 | int r; |
0447568f | 2229 | struct raid_dev *dev; |
a9ad8526 | 2230 | struct md_rdev *rdev, *tmp, *freshest; |
fd01b88c | 2231 | struct mddev *mddev = &rs->md; |
b12d437b JB |
2232 | |
2233 | freshest = NULL; | |
a9ad8526 | 2234 | rdev_for_each_safe(rdev, tmp, mddev) { |
761becff | 2235 | /* |
c76d53f4 | 2236 | * Skipping super_load due to CTR_FLAG_SYNC will cause |
761becff | 2237 | * the array to undergo initialization again as |
43157840 | 2238 | * though it were new. This is the intended effect |
761becff JB |
2239 | * of the "sync" directive. |
2240 | * | |
2241 | * When reshaping capability is added, we must ensure | |
2242 | * that the "sync" directive is disallowed during the | |
2243 | * reshape. | |
2244 | */ | |
4286325b | 2245 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) |
761becff JB |
2246 | continue; |
2247 | ||
b12d437b JB |
2248 | if (!rdev->meta_bdev) |
2249 | continue; | |
2250 | ||
73c6f239 | 2251 | r = super_load(rdev, freshest); |
b12d437b | 2252 | |
73c6f239 | 2253 | switch (r) { |
b12d437b JB |
2254 | case 1: |
2255 | freshest = rdev; | |
2256 | break; | |
2257 | case 0: | |
2258 | break; | |
2259 | default: | |
0447568f | 2260 | dev = container_of(rdev, struct raid_dev, rdev); |
55ebbb59 JB |
2261 | if (dev->meta_dev) |
2262 | dm_put_device(ti, dev->meta_dev); | |
0447568f | 2263 | |
55ebbb59 JB |
2264 | dev->meta_dev = NULL; |
2265 | rdev->meta_bdev = NULL; | |
0447568f | 2266 | |
55ebbb59 JB |
2267 | if (rdev->sb_page) |
2268 | put_page(rdev->sb_page); | |
0447568f | 2269 | |
55ebbb59 | 2270 | rdev->sb_page = NULL; |
0447568f | 2271 | |
55ebbb59 | 2272 | rdev->sb_loaded = 0; |
0447568f | 2273 | |
55ebbb59 JB |
2274 | /* |
2275 | * We might be able to salvage the data device | |
2276 | * even though the meta device has failed. For | |
2277 | * now, we behave as though '- -' had been | |
2278 | * set for this device in the table. | |
2279 | */ | |
2280 | if (dev->data_dev) | |
2281 | dm_put_device(ti, dev->data_dev); | |
0447568f | 2282 | |
55ebbb59 JB |
2283 | dev->data_dev = NULL; |
2284 | rdev->bdev = NULL; | |
0447568f | 2285 | |
55ebbb59 | 2286 | list_del(&rdev->same_set); |
b12d437b JB |
2287 | } |
2288 | } | |
2289 | ||
2290 | if (!freshest) | |
2291 | return 0; | |
2292 | ||
bd83a4c4 MS |
2293 | if (validate_raid_redundancy(rs)) { |
2294 | rs->ti->error = "Insufficient redundancy to activate array"; | |
2295 | return -EINVAL; | |
2296 | } | |
55ebbb59 | 2297 | |
b12d437b JB |
2298 | /* |
2299 | * Validation of the freshest device provides the source of | |
2300 | * validation for the remaining devices. | |
2301 | */ | |
9dbd1aa3 HM |
2302 | rs->ti->error = "Unable to assemble array: Invalid superblocks"; |
2303 | if (super_validate(rs, freshest)) | |
bd83a4c4 | 2304 | return -EINVAL; |
b12d437b | 2305 | |
dafb20fa | 2306 | rdev_for_each(rdev, mddev) |
0cf45031 | 2307 | if ((rdev != freshest) && super_validate(rs, rdev)) |
b12d437b | 2308 | return -EINVAL; |
b12d437b JB |
2309 | return 0; |
2310 | } | |
2311 | ||
40ba37e5 HM |
2312 | /* |
2313 | * Adjust data_offset and new_data_offset on all disk members of @rs | |
2314 | * for out of place reshaping if requested by contructor | |
2315 | * | |
2316 | * We need free space at the beginning of each raid disk for forward | |
2317 | * and at the end for backward reshapes which userspace has to provide | |
2318 | * via remapping/reordering of space. | |
2319 | */ | |
2320 | static int rs_adjust_data_offsets(struct raid_set *rs) | |
2321 | { | |
2322 | sector_t data_offset = 0, new_data_offset = 0; | |
2323 | struct md_rdev *rdev; | |
2324 | ||
2325 | /* Constructor did not request data offset change */ | |
2326 | if (!test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { | |
2327 | if (!rs_is_reshapable(rs)) | |
2328 | goto out; | |
2329 | ||
2330 | return 0; | |
2331 | } | |
2332 | ||
2333 | /* HM FIXME: get InSync raid_dev? */ | |
2334 | rdev = &rs->dev[0].rdev; | |
2335 | ||
2336 | if (rs->delta_disks < 0) { | |
2337 | /* | |
2338 | * Removing disks (reshaping backwards): | |
2339 | * | |
2340 | * - before reshape: data is at offset 0 and free space | |
2341 | * is at end of each component LV | |
2342 | * | |
2343 | * - after reshape: data is at offset rs->data_offset != 0 on each component LV | |
2344 | */ | |
2345 | data_offset = 0; | |
2346 | new_data_offset = rs->data_offset; | |
2347 | ||
2348 | } else if (rs->delta_disks > 0) { | |
2349 | /* | |
2350 | * Adding disks (reshaping forwards): | |
2351 | * | |
2352 | * - before reshape: data is at offset rs->data_offset != 0 and | |
2353 | * free space is at begin of each component LV | |
2354 | * | |
2355 | * - after reshape: data is at offset 0 on each component LV | |
2356 | */ | |
2357 | data_offset = rs->data_offset; | |
2358 | new_data_offset = 0; | |
2359 | ||
2360 | } else { | |
2361 | /* | |
2362 | * User space passes in 0 for data offset after having removed reshape space | |
2363 | * | |
2364 | * - or - (data offset != 0) | |
2365 | * | |
2366 | * Changing RAID layout or chunk size -> toggle offsets | |
2367 | * | |
2368 | * - before reshape: data is at offset rs->data_offset 0 and | |
2369 | * free space is at end of each component LV | |
2370 | * -or- | |
2371 | * data is at offset rs->data_offset != 0 and | |
2372 | * free space is at begin of each component LV | |
2373 | * | |
2374 | * - after reshape: data is at offset 0 if i was at offset != 0 | |
2375 | * of at offset != 0 if it was at offset 0 | |
2376 | * on each component LV | |
2377 | * | |
2378 | */ | |
2379 | data_offset = rs->data_offset ? rdev->data_offset : 0; | |
2380 | new_data_offset = data_offset ? 0 : rs->data_offset; | |
2381 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
2382 | } | |
2383 | ||
2384 | /* | |
2385 | * Make sure we got a minimum amount of free sectors per device | |
2386 | */ | |
2387 | if (rs->data_offset && | |
2388 | to_sector(i_size_read(rdev->bdev->bd_inode)) - rdev->sectors < MIN_FREE_RESHAPE_SPACE) { | |
2389 | rs->ti->error = data_offset ? "No space for forward reshape" : | |
2390 | "No space for backward reshape"; | |
2391 | return -ENOSPC; | |
2392 | } | |
2393 | out: | |
2394 | /* Adjust data offsets on all rdevs */ | |
2395 | rdev_for_each(rdev, &rs->md) { | |
2396 | rdev->data_offset = data_offset; | |
2397 | rdev->new_data_offset = new_data_offset; | |
2398 | } | |
2399 | ||
2400 | return 0; | |
2401 | } | |
2402 | ||
ecbfb9f1 | 2403 | /* Userpace reordered disks -> adjust raid_disk indexes in @rs */ |
e6ca5e1a | 2404 | static void __reorder_raid_disk_indexes(struct raid_set *rs) |
ecbfb9f1 HM |
2405 | { |
2406 | int i = 0; | |
2407 | struct md_rdev *rdev; | |
2408 | ||
2409 | rdev_for_each(rdev, &rs->md) { | |
2410 | rdev->raid_disk = i++; | |
2411 | rdev->saved_raid_disk = rdev->new_raid_disk = -1; | |
2412 | } | |
2413 | } | |
2414 | ||
2415 | /* | |
2416 | * Setup @rs for takeover by a different raid level | |
2417 | */ | |
2418 | static int rs_setup_takeover(struct raid_set *rs) | |
2419 | { | |
2420 | struct mddev *mddev = &rs->md; | |
2421 | struct md_rdev *rdev; | |
2422 | unsigned int d = mddev->raid_disks = rs->raid_disks; | |
2423 | sector_t new_data_offset = rs->dev[0].rdev.data_offset ? 0 : rs->data_offset; | |
2424 | ||
2425 | if (rt_is_raid10(rs->raid_type)) { | |
2426 | if (mddev->level == 0) { | |
2427 | /* Userpace reordered disks -> adjust raid_disk indexes */ | |
e6ca5e1a | 2428 | __reorder_raid_disk_indexes(rs); |
ecbfb9f1 HM |
2429 | |
2430 | /* raid0 -> raid10_far layout */ | |
2431 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_FAR, | |
2432 | rs->raid10_copies); | |
2433 | } else if (mddev->level == 1) | |
2434 | /* raid1 -> raid10_near layout */ | |
2435 | mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, | |
2436 | rs->raid_disks); | |
2437 | else | |
2438 | return -EINVAL; | |
2439 | ||
2440 | } | |
2441 | ||
2442 | clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags); | |
2443 | mddev->recovery_cp = MaxSector; | |
2444 | ||
2445 | while (d--) { | |
2446 | rdev = &rs->dev[d].rdev; | |
2447 | ||
2448 | if (test_bit(d, (void *) rs->rebuild_disks)) { | |
2449 | clear_bit(In_sync, &rdev->flags); | |
2450 | clear_bit(Faulty, &rdev->flags); | |
2451 | mddev->recovery_cp = rdev->recovery_offset = 0; | |
2452 | /* Bitmap has to be created when we do an "up" takeover */ | |
2453 | set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); | |
2454 | } | |
2455 | ||
2456 | rdev->new_data_offset = new_data_offset; | |
2457 | } | |
2458 | ||
ecbfb9f1 HM |
2459 | return 0; |
2460 | } | |
2461 | ||
9dbd1aa3 HM |
2462 | /* |
2463 | * | |
2464 | * - change raid layout | |
2465 | * - change chunk size | |
2466 | * - add disks | |
2467 | * - remove disks | |
2468 | */ | |
2469 | static int rs_setup_reshape(struct raid_set *rs) | |
2470 | { | |
2471 | int r = 0; | |
2472 | unsigned int cur_raid_devs, d; | |
2473 | struct mddev *mddev = &rs->md; | |
2474 | struct md_rdev *rdev; | |
2475 | ||
2476 | mddev->delta_disks = rs->delta_disks; | |
2477 | cur_raid_devs = mddev->raid_disks; | |
2478 | ||
2479 | /* Ignore impossible layout change whilst adding/removing disks */ | |
2480 | if (mddev->delta_disks && | |
2481 | mddev->layout != mddev->new_layout) { | |
2482 | DMINFO("Ignoring invalid layout change with delta_disks=%d", rs->delta_disks); | |
2483 | mddev->new_layout = mddev->layout; | |
2484 | } | |
2485 | ||
2486 | /* | |
2487 | * Adjust array size: | |
2488 | * | |
2489 | * - in case of adding disks, array size has | |
2490 | * to grow after the disk adding reshape, | |
2491 | * which'll hapen in the event handler; | |
2492 | * reshape will happen forward, so space has to | |
2493 | * be available at the beginning of each disk | |
2494 | * | |
2495 | * - in case of removing disks, array size | |
2496 | * has to shrink before starting the reshape, | |
2497 | * which'll happen here; | |
2498 | * reshape will happen backward, so space has to | |
2499 | * be available at the end of each disk | |
2500 | * | |
2501 | * - data_offset and new_data_offset are | |
2502 | * adjusted for afreentioned out of place | |
2503 | * reshaping based on userspace passing in | |
2504 | * the "data_offset <sectors>" key/value | |
2505 | * pair via te constructor | |
2506 | */ | |
2507 | ||
2508 | /* Add disk(s) */ | |
2509 | if (rs->delta_disks > 0) { | |
2510 | /* Prepare disks for check in raid4/5/6/10 {check|start}_reshape */ | |
2511 | for (d = cur_raid_devs; d < rs->raid_disks; d++) { | |
2512 | rdev = &rs->dev[d].rdev; | |
2513 | clear_bit(In_sync, &rdev->flags); | |
2514 | ||
2515 | /* | |
2516 | * save_raid_disk needs to be -1, or recovery_offset will be set to 0 | |
2517 | * by md, which'll store that erroneously in the superblock on reshape | |
2518 | */ | |
2519 | rdev->saved_raid_disk = -1; | |
2520 | rdev->raid_disk = d; | |
2521 | ||
2522 | rdev->sectors = mddev->dev_sectors; | |
2523 | rdev->recovery_offset = MaxSector; | |
2524 | } | |
2525 | ||
2526 | mddev->reshape_backwards = 0; /* adding disks -> forward reshape */ | |
2527 | ||
2528 | /* Remove disk(s) */ | |
2529 | } else if (rs->delta_disks < 0) { | |
2530 | r = rs_set_dev_and_array_sectors(rs, true); | |
2531 | mddev->reshape_backwards = 1; /* removing disk(s) -> backward reshape */ | |
2532 | ||
2533 | /* Change layout and/or chunk size */ | |
2534 | } else { | |
2535 | /* | |
2536 | * Reshape layout (e.g. raid5_ls -> raid5_n) and/or chunk size: | |
2537 | * | |
2538 | * keeping number of disks and do layout change -> | |
2539 | * | |
2540 | * toggle reshape_backward depending on data_offset: | |
2541 | * | |
2542 | * - free space upfront -> reshape forward | |
2543 | * | |
2544 | * - free space at the end -> reshape backward | |
2545 | * | |
2546 | * | |
2547 | * This utilizes free reshape space avoiding the need | |
2548 | * for userspace to move (parts of) LV segments in | |
2549 | * case of layout/chunksize change (for disk | |
2550 | * adding/removing reshape space has to be at | |
2551 | * the proper address (see above with delta_disks): | |
2552 | * | |
2553 | * add disk(s) -> begin | |
2554 | * remove disk(s)-> end | |
2555 | */ | |
2556 | mddev->reshape_backwards = rs->dev[0].rdev.data_offset ? 0 : 1; | |
2557 | } | |
2558 | ||
2559 | return r; | |
2560 | } | |
2561 | ||
75b8e04b | 2562 | /* |
48cf06bc HM |
2563 | * Enable/disable discard support on RAID set depending on |
2564 | * RAID level and discard properties of underlying RAID members. | |
75b8e04b | 2565 | */ |
ecbfb9f1 | 2566 | static void configure_discard_support(struct raid_set *rs) |
75b8e04b | 2567 | { |
48cf06bc HM |
2568 | int i; |
2569 | bool raid456; | |
ecbfb9f1 | 2570 | struct dm_target *ti = rs->ti; |
48cf06bc | 2571 | |
75b8e04b HM |
2572 | /* Assume discards not supported until after checks below. */ |
2573 | ti->discards_supported = false; | |
2574 | ||
2575 | /* RAID level 4,5,6 require discard_zeroes_data for data integrity! */ | |
48cf06bc | 2576 | raid456 = (rs->md.level == 4 || rs->md.level == 5 || rs->md.level == 6); |
75b8e04b | 2577 | |
48cf06bc | 2578 | for (i = 0; i < rs->md.raid_disks; i++) { |
d20c4b08 | 2579 | struct request_queue *q; |
48cf06bc | 2580 | |
d20c4b08 HM |
2581 | if (!rs->dev[i].rdev.bdev) |
2582 | continue; | |
2583 | ||
2584 | q = bdev_get_queue(rs->dev[i].rdev.bdev); | |
48cf06bc HM |
2585 | if (!q || !blk_queue_discard(q)) |
2586 | return; | |
2587 | ||
2588 | if (raid456) { | |
2589 | if (!q->limits.discard_zeroes_data) | |
2590 | return; | |
2591 | if (!devices_handle_discard_safely) { | |
2592 | DMERR("raid456 discard support disabled due to discard_zeroes_data uncertainty."); | |
2593 | DMERR("Set dm-raid.devices_handle_discard_safely=Y to override."); | |
2594 | return; | |
2595 | } | |
2596 | } | |
2597 | } | |
2598 | ||
2599 | /* All RAID members properly support discards */ | |
75b8e04b HM |
2600 | ti->discards_supported = true; |
2601 | ||
2602 | /* | |
2603 | * RAID1 and RAID10 personalities require bio splitting, | |
48cf06bc | 2604 | * RAID0/4/5/6 don't and process large discard bios properly. |
75b8e04b | 2605 | */ |
48cf06bc | 2606 | ti->split_discard_bios = !!(rs->md.level == 1 || rs->md.level == 10); |
75b8e04b HM |
2607 | ti->num_discard_bios = 1; |
2608 | } | |
2609 | ||
9d09e663 | 2610 | /* |
73c6f239 | 2611 | * Construct a RAID0/1/10/4/5/6 mapping: |
9d09e663 | 2612 | * Args: |
43157840 MS |
2613 | * <raid_type> <#raid_params> <raid_params>{0,} \ |
2614 | * <#raid_devs> [<meta_dev1> <dev1>]{1,} | |
9d09e663 | 2615 | * |
43157840 | 2616 | * <raid_params> varies by <raid_type>. See 'parse_raid_params' for |
9d09e663 | 2617 | * details on possible <raid_params>. |
73c6f239 HM |
2618 | * |
2619 | * Userspace is free to initialize the metadata devices, hence the superblocks to | |
2620 | * enforce recreation based on the passed in table parameters. | |
2621 | * | |
9d09e663 N |
2622 | */ |
2623 | static int raid_ctr(struct dm_target *ti, unsigned argc, char **argv) | |
2624 | { | |
73c6f239 | 2625 | int r; |
9d09e663 | 2626 | struct raid_type *rt; |
92c83d79 | 2627 | unsigned num_raid_params, num_raid_devs; |
9d09e663 | 2628 | struct raid_set *rs = NULL; |
92c83d79 | 2629 | const char *arg; |
9dbd1aa3 | 2630 | struct rs_layout rs_layout; |
92c83d79 HM |
2631 | struct dm_arg_set as = { argc, argv }, as_nrd; |
2632 | struct dm_arg _args[] = { | |
2633 | { 0, as.argc, "Cannot understand number of raid parameters" }, | |
2634 | { 1, 254, "Cannot understand number of raid devices parameters" } | |
2635 | }; | |
2636 | ||
2637 | /* Must have <raid_type> */ | |
2638 | arg = dm_shift_arg(&as); | |
bd83a4c4 MS |
2639 | if (!arg) { |
2640 | ti->error = "No arguments"; | |
2641 | return -EINVAL; | |
2642 | } | |
9d09e663 | 2643 | |
92c83d79 | 2644 | rt = get_raid_type(arg); |
bd83a4c4 MS |
2645 | if (!rt) { |
2646 | ti->error = "Unrecognised raid_type"; | |
2647 | return -EINVAL; | |
2648 | } | |
9d09e663 | 2649 | |
92c83d79 HM |
2650 | /* Must have <#raid_params> */ |
2651 | if (dm_read_arg_group(_args, &as, &num_raid_params, &ti->error)) | |
43157840 | 2652 | return -EINVAL; |
9d09e663 | 2653 | |
92c83d79 HM |
2654 | /* number of raid device tupples <meta_dev data_dev> */ |
2655 | as_nrd = as; | |
2656 | dm_consume_args(&as_nrd, num_raid_params); | |
2657 | _args[1].max = (as_nrd.argc - 1) / 2; | |
2658 | if (dm_read_arg(_args + 1, &as_nrd, &num_raid_devs, &ti->error)) | |
43157840 | 2659 | return -EINVAL; |
9d09e663 | 2660 | |
bb91a63f | 2661 | if (!__within_range(num_raid_devs, 1, MAX_RAID_DEVICES)) { |
bd83a4c4 MS |
2662 | ti->error = "Invalid number of supplied raid devices"; |
2663 | return -EINVAL; | |
2664 | } | |
3ca5a21a | 2665 | |
bfcee0e3 | 2666 | rs = raid_set_alloc(ti, rt, num_raid_devs); |
9d09e663 N |
2667 | if (IS_ERR(rs)) |
2668 | return PTR_ERR(rs); | |
2669 | ||
92c83d79 | 2670 | r = parse_raid_params(rs, &as, num_raid_params); |
73c6f239 | 2671 | if (r) |
9d09e663 N |
2672 | goto bad; |
2673 | ||
702108d1 | 2674 | r = parse_dev_params(rs, &as); |
73c6f239 | 2675 | if (r) |
9d09e663 N |
2676 | goto bad; |
2677 | ||
b12d437b | 2678 | rs->md.sync_super = super_sync; |
ecbfb9f1 | 2679 | |
40ba37e5 HM |
2680 | r = rs_set_dev_and_array_sectors(rs, false); |
2681 | if (r) | |
2682 | return r; | |
2683 | ||
ecbfb9f1 HM |
2684 | /* |
2685 | * Backup any new raid set level, layout, ... | |
2686 | * requested to be able to compare to superblock | |
2687 | * members for conversion decisions. | |
2688 | */ | |
9dbd1aa3 | 2689 | rs_config_backup(rs, &rs_layout); |
ecbfb9f1 | 2690 | |
73c6f239 HM |
2691 | r = analyse_superblocks(ti, rs); |
2692 | if (r) | |
b12d437b JB |
2693 | goto bad; |
2694 | ||
9d09e663 | 2695 | INIT_WORK(&rs->md.event_work, do_table_event); |
9d09e663 | 2696 | ti->private = rs; |
55a62eef | 2697 | ti->num_flush_bios = 1; |
9d09e663 | 2698 | |
ecbfb9f1 | 2699 | /* Restore any requested new layout for conversion decision */ |
9dbd1aa3 | 2700 | rs_config_restore(rs, &rs_layout); |
ecbfb9f1 | 2701 | |
9dbd1aa3 HM |
2702 | if (test_bit(MD_ARRAY_FIRST_USE, &rs->md.flags)) { |
2703 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
2704 | rs_set_new(rs); | |
2705 | } else if (rs_is_reshaping(rs)) | |
2706 | ; /* skip rs setup */ | |
2707 | else if (rs_takeover_requested(rs)) { | |
2708 | if (rs_is_reshaping(rs)) { | |
2709 | ti->error = "Can't takeover a reshaping raid set"; | |
2710 | return -EPERM; | |
2711 | } | |
2712 | ||
2713 | /* | |
2714 | * If a takeover is needed, just set the level to | |
2715 | * the new requested one and allow the raid set to run. | |
2716 | */ | |
ecbfb9f1 HM |
2717 | r = rs_check_takeover(rs); |
2718 | if (r) | |
2719 | return r; | |
2720 | ||
2721 | r = rs_setup_takeover(rs); | |
2722 | if (r) | |
2723 | return r; | |
2724 | ||
4286325b | 2725 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); |
6e20902e | 2726 | set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); |
3a1c1ef2 | 2727 | rs_set_new(rs); |
40ba37e5 | 2728 | } else if (rs_reshape_requested(rs)) { |
9dbd1aa3 HM |
2729 | if (rs_is_reshaping(rs)) { |
2730 | ti->error = "raid set already reshaping!"; | |
2731 | return -EPERM; | |
2732 | } | |
2733 | ||
2734 | if (rs_is_raid10(rs)) { | |
2735 | if (rs->raid_disks != rs->md.raid_disks && | |
2736 | __is_raid10_near(rs->md.layout) && | |
2737 | rs->raid10_copies && | |
2738 | rs->raid10_copies != __raid10_near_copies(rs->md.layout)) { | |
2739 | /* | |
2740 | * raid disk have to be multiple of data copies to allow this conversion, | |
2741 | * | |
2742 | * This is actually not a reshape it is a | |
2743 | * rebuild of any additional mirrors per group | |
2744 | */ | |
2745 | if (rs->raid_disks % rs->raid10_copies) { | |
2746 | ti->error = "Can't reshape raid10 mirror groups"; | |
2747 | return -EINVAL; | |
2748 | } | |
2749 | ||
2750 | /* Userpace reordered disks to add/remove mirrors -> adjust raid_disk indexes */ | |
2751 | __reorder_raid_disk_indexes(rs); | |
2752 | rs->md.layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, | |
2753 | rs->raid10_copies); | |
2754 | rs->md.new_layout = rs->md.layout; | |
2755 | ||
2756 | } else | |
2757 | set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags); | |
2758 | ||
2759 | } else if (rs_is_raid456(rs)) | |
2760 | set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags); | |
2761 | ||
2762 | /* | |
2763 | * HM FIXME: process raid1 via delta_disks as well? | |
2764 | * Would cause allocations in raid1->check_reshape | |
2765 | * though, thus more issues with potential failures | |
2766 | */ | |
6e20902e HM |
2767 | else if (rs_is_raid1(rs)) { |
2768 | set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); | |
9dbd1aa3 | 2769 | rs->md.raid_disks = rs->raid_disks; |
6e20902e HM |
2770 | } |
2771 | ||
2772 | if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { | |
2773 | set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); | |
2774 | set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); | |
2775 | } | |
9dbd1aa3 HM |
2776 | |
2777 | if (rs->md.raid_disks < rs->raid_disks) | |
2778 | set_bit(MD_ARRAY_FIRST_USE, &rs->md.flags); | |
2779 | ||
9dbd1aa3 | 2780 | rs_set_cur(rs); |
3a1c1ef2 HM |
2781 | } else |
2782 | rs_set_cur(rs); | |
ecbfb9f1 | 2783 | |
40ba37e5 HM |
2784 | /* If constructor requested it, change data and new_data offsets */ |
2785 | r = rs_adjust_data_offsets(rs); | |
2786 | if (r) | |
2787 | return r; | |
2788 | ||
ecbfb9f1 HM |
2789 | /* Start raid set read-only and assumed clean to change in raid_resume() */ |
2790 | rs->md.ro = 1; | |
2791 | rs->md.in_sync = 1; | |
2792 | set_bit(MD_RECOVERY_FROZEN, &rs->md.recovery); | |
75b8e04b | 2793 | |
0cf45031 HM |
2794 | /* Has to be held on running the array */ |
2795 | mddev_lock_nointr(&rs->md); | |
73c6f239 | 2796 | r = md_run(&rs->md); |
9d09e663 | 2797 | rs->md.in_sync = 0; /* Assume already marked dirty */ |
9d09e663 | 2798 | |
73c6f239 | 2799 | if (r) { |
9dbd1aa3 HM |
2800 | ti->error = "Failed to run raid array"; |
2801 | mddev_unlock(&rs->md); | |
9d09e663 N |
2802 | goto bad; |
2803 | } | |
2804 | ||
2805 | rs->callbacks.congested_fn = raid_is_congested; | |
9d09e663 N |
2806 | dm_table_add_target_callbacks(ti->table, &rs->callbacks); |
2807 | ||
32737279 | 2808 | mddev_suspend(&rs->md); |
9dbd1aa3 HM |
2809 | |
2810 | /* Try to adjust the raid4/5/6 stripe cache size to the stripe size */ | |
2811 | if (rs_is_raid456(rs)) { | |
2812 | r = rs_set_raid456_stripe_cache(rs); | |
2813 | if (r) | |
2814 | goto bad_stripe_cache; | |
2815 | } | |
2816 | ||
2817 | /* Now do an early reshape check */ | |
2818 | if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { | |
2819 | r = rs_check_reshape(rs); | |
2820 | if (r) | |
2821 | return r; | |
2822 | ||
2823 | /* Restore new, ctr requested layout to perform check */ | |
2824 | rs_config_restore(rs, &rs_layout); | |
2825 | ||
2826 | r = rs->md.pers->check_reshape(&rs->md); | |
2827 | if (r) { | |
2828 | ti->error = "Reshape check failed"; | |
2829 | goto bad_check_reshape; | |
2830 | } | |
2831 | } | |
2832 | ||
2833 | mddev_unlock(&rs->md); | |
9d09e663 N |
2834 | return 0; |
2835 | ||
9dbd1aa3 HM |
2836 | bad_stripe_cache: |
2837 | bad_check_reshape: | |
63f33b8d | 2838 | md_stop(&rs->md); |
9d09e663 | 2839 | bad: |
bfcee0e3 | 2840 | raid_set_free(rs); |
9d09e663 | 2841 | |
73c6f239 | 2842 | return r; |
9d09e663 N |
2843 | } |
2844 | ||
2845 | static void raid_dtr(struct dm_target *ti) | |
2846 | { | |
2847 | struct raid_set *rs = ti->private; | |
2848 | ||
2849 | list_del_init(&rs->callbacks.list); | |
2850 | md_stop(&rs->md); | |
bfcee0e3 | 2851 | raid_set_free(rs); |
9d09e663 N |
2852 | } |
2853 | ||
7de3ee57 | 2854 | static int raid_map(struct dm_target *ti, struct bio *bio) |
9d09e663 N |
2855 | { |
2856 | struct raid_set *rs = ti->private; | |
fd01b88c | 2857 | struct mddev *mddev = &rs->md; |
9d09e663 | 2858 | |
9dbd1aa3 HM |
2859 | /* |
2860 | * If we're reshaping to add disk(s)), ti->len and | |
2861 | * mddev->array_sectors will differ during the process | |
2862 | * (ti->len > mddev->array_sectors), so we have to requeue | |
2863 | * bios with addresses > mddev->array_sectors here or | |
2864 | * or there will occur accesses past EOD of the component | |
2865 | * data images thus erroring the raid set. | |
2866 | */ | |
2867 | if (unlikely(bio_end_sector(bio) > mddev->array_sectors)) | |
2868 | return DM_MAPIO_REQUEUE; | |
2869 | ||
9d09e663 N |
2870 | mddev->pers->make_request(mddev, bio); |
2871 | ||
2872 | return DM_MAPIO_SUBMITTED; | |
2873 | } | |
2874 | ||
3a1c1ef2 | 2875 | /* Return string describing the current sync action of @mddev */ |
be83651f JB |
2876 | static const char *decipher_sync_action(struct mddev *mddev) |
2877 | { | |
2878 | if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) | |
2879 | return "frozen"; | |
2880 | ||
2881 | if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || | |
2882 | (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))) { | |
2883 | if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) | |
2884 | return "reshape"; | |
2885 | ||
2886 | if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { | |
2887 | if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) | |
2888 | return "resync"; | |
2889 | else if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) | |
2890 | return "check"; | |
2891 | return "repair"; | |
2892 | } | |
2893 | ||
2894 | if (test_bit(MD_RECOVERY_RECOVER, &mddev->recovery)) | |
2895 | return "recover"; | |
2896 | } | |
2897 | ||
2898 | return "idle"; | |
2899 | } | |
2900 | ||
3a1c1ef2 HM |
2901 | /* |
2902 | * Return status string @rdev | |
2903 | * | |
2904 | * Status characters: | |
2905 | * | |
2906 | * 'D' = Dead/Failed device | |
2907 | * 'a' = Alive but not in-sync | |
2908 | * 'A' = Alive and in-sync | |
2909 | */ | |
e6ca5e1a | 2910 | static const char *__raid_dev_status(struct md_rdev *rdev, bool array_in_sync) |
9d09e663 | 2911 | { |
3a1c1ef2 HM |
2912 | if (test_bit(Faulty, &rdev->flags)) |
2913 | return "D"; | |
2914 | else if (!array_in_sync || !test_bit(In_sync, &rdev->flags)) | |
2915 | return "a"; | |
2916 | else | |
2917 | return "A"; | |
2918 | } | |
9d09e663 | 2919 | |
3a1c1ef2 HM |
2920 | /* Helper to return resync/reshape progress for @rs and @array_in_sync */ |
2921 | static sector_t rs_get_progress(struct raid_set *rs, | |
2922 | sector_t resync_max_sectors, bool *array_in_sync) | |
2923 | { | |
2924 | sector_t r, recovery_cp, curr_resync_completed; | |
2925 | struct mddev *mddev = &rs->md; | |
9d09e663 | 2926 | |
3a1c1ef2 HM |
2927 | curr_resync_completed = mddev->curr_resync_completed ?: mddev->recovery_cp; |
2928 | recovery_cp = mddev->recovery_cp; | |
2929 | *array_in_sync = false; | |
2930 | ||
2931 | if (rs_is_raid0(rs)) { | |
2932 | r = resync_max_sectors; | |
2933 | *array_in_sync = true; | |
2934 | ||
2935 | } else { | |
2936 | r = mddev->reshape_position; | |
2937 | ||
2938 | /* Reshape is relative to the array size */ | |
2939 | if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) || | |
2940 | r != MaxSector) { | |
2941 | if (r == MaxSector) { | |
2942 | *array_in_sync = true; | |
2943 | r = resync_max_sectors; | |
0cf45031 | 2944 | } else { |
3a1c1ef2 HM |
2945 | /* Got to reverse on backward reshape */ |
2946 | if (mddev->reshape_backwards) | |
2947 | r = mddev->array_sectors - r; | |
2948 | ||
2949 | /* Devide by # of data stripes */ | |
2950 | sector_div(r, mddev_data_stripes(rs)); | |
0cf45031 | 2951 | } |
3a1c1ef2 HM |
2952 | |
2953 | /* Sync is relative to the component device size */ | |
2954 | } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) | |
2955 | r = curr_resync_completed; | |
2956 | else | |
2957 | r = recovery_cp; | |
2958 | ||
2959 | if (r == MaxSector) { | |
2960 | /* | |
2961 | * Sync complete. | |
2962 | */ | |
2963 | *array_in_sync = true; | |
2964 | r = resync_max_sectors; | |
2965 | } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) { | |
2966 | /* | |
2967 | * If "check" or "repair" is occurring, the raid set has | |
2968 | * undergone an initial sync and the health characters | |
2969 | * should not be 'a' anymore. | |
2970 | */ | |
2971 | *array_in_sync = true; | |
0cf45031 | 2972 | } else { |
3a1c1ef2 | 2973 | struct md_rdev *rdev; |
be83651f | 2974 | |
3a1c1ef2 HM |
2975 | /* |
2976 | * The raid set may be doing an initial sync, or it may | |
43157840 | 2977 | * be rebuilding individual components. If all the |
3a1c1ef2 HM |
2978 | * devices are In_sync, then it is the raid set that is |
2979 | * being initialized. | |
2980 | */ | |
2981 | rdev_for_each(rdev, mddev) | |
2982 | if (!test_bit(In_sync, &rdev->flags)) | |
2983 | *array_in_sync = true; | |
2984 | #if 0 | |
2985 | r = 0; /* HM FIXME: TESTME: https://bugzilla.redhat.com/show_bug.cgi?id=1210637 ? */ | |
2986 | #endif | |
2e727c3c | 2987 | } |
3a1c1ef2 HM |
2988 | } |
2989 | ||
2990 | return r; | |
2991 | } | |
2992 | ||
2993 | /* Helper to return @dev name or "-" if !@dev */ | |
e6ca5e1a | 2994 | static const char *__get_dev_name(struct dm_dev *dev) |
3a1c1ef2 HM |
2995 | { |
2996 | return dev ? dev->name : "-"; | |
2997 | } | |
2998 | ||
2999 | static void raid_status(struct dm_target *ti, status_type_t type, | |
3000 | unsigned int status_flags, char *result, unsigned int maxlen) | |
3001 | { | |
3002 | struct raid_set *rs = ti->private; | |
3003 | struct mddev *mddev = &rs->md; | |
3004 | struct r5conf *conf = mddev->private; | |
3005 | int max_nr_stripes = conf ? conf->max_nr_stripes : 0; | |
3006 | bool array_in_sync; | |
3007 | unsigned int raid_param_cnt = 1; /* at least 1 for chunksize */ | |
3008 | unsigned int sz = 0; | |
3009 | unsigned int write_mostly_params = 0; | |
3010 | sector_t progress, resync_max_sectors, resync_mismatches; | |
3011 | const char *sync_action; | |
3012 | struct raid_type *rt; | |
3013 | struct md_rdev *rdev; | |
3014 | ||
3015 | switch (type) { | |
3016 | case STATUSTYPE_INFO: | |
3017 | /* *Should* always succeed */ | |
3018 | rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); | |
3019 | if (!rt) | |
3020 | return; | |
3021 | ||
9dbd1aa3 | 3022 | DMEMIT("%s %d ", rt->name, mddev->raid_disks); |
3a1c1ef2 HM |
3023 | |
3024 | /* Access most recent mddev properties for status output */ | |
3025 | smp_rmb(); | |
3026 | /* Get sensible max sectors even if raid set not yet started */ | |
4286325b | 3027 | resync_max_sectors = test_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags) ? |
3a1c1ef2 HM |
3028 | mddev->resync_max_sectors : mddev->dev_sectors; |
3029 | progress = rs_get_progress(rs, resync_max_sectors, &array_in_sync); | |
3030 | resync_mismatches = (mddev->last_sync_action && !strcasecmp(mddev->last_sync_action, "check")) ? | |
9dbd1aa3 | 3031 | atomic64_read(&mddev->resync_mismatches) : 0; |
3a1c1ef2 HM |
3032 | sync_action = decipher_sync_action(&rs->md); |
3033 | ||
3034 | /* HM FIXME: do we want another state char for raid0? It shows 'D' or 'A' now */ | |
3035 | rdev_for_each(rdev, mddev) | |
e6ca5e1a | 3036 | DMEMIT(__raid_dev_status(rdev, array_in_sync)); |
9d09e663 | 3037 | |
2e727c3c | 3038 | /* |
3a1c1ef2 | 3039 | * In-sync/Reshape ratio: |
2e727c3c | 3040 | * The in-sync ratio shows the progress of: |
3a1c1ef2 HM |
3041 | * - Initializing the raid set |
3042 | * - Rebuilding a subset of devices of the raid set | |
2e727c3c JB |
3043 | * The user can distinguish between the two by referring |
3044 | * to the status characters. | |
3a1c1ef2 HM |
3045 | * |
3046 | * The reshape ratio shows the progress of | |
3047 | * changing the raid layout or the number of | |
3048 | * disks of a raid set | |
2e727c3c | 3049 | */ |
3a1c1ef2 HM |
3050 | DMEMIT(" %llu/%llu", (unsigned long long) progress, |
3051 | (unsigned long long) resync_max_sectors); | |
9d09e663 | 3052 | |
be83651f | 3053 | /* |
3a1c1ef2 HM |
3054 | * v1.5.0+: |
3055 | * | |
be83651f | 3056 | * Sync action: |
3a1c1ef2 | 3057 | * See Documentation/device-mapper/dm-raid.txt for |
be83651f JB |
3058 | * information on each of these states. |
3059 | */ | |
3a1c1ef2 | 3060 | DMEMIT(" %s", sync_action); |
be83651f JB |
3061 | |
3062 | /* | |
3a1c1ef2 HM |
3063 | * v1.5.0+: |
3064 | * | |
be83651f JB |
3065 | * resync_mismatches/mismatch_cnt |
3066 | * This field shows the number of discrepancies found when | |
3a1c1ef2 | 3067 | * performing a "check" of the raid set. |
be83651f | 3068 | */ |
3a1c1ef2 | 3069 | DMEMIT(" %llu", (unsigned long long) resync_mismatches); |
9d09e663 | 3070 | |
3a1c1ef2 | 3071 | /* |
9b6e5423 | 3072 | * v1.9.0+: |
3a1c1ef2 HM |
3073 | * |
3074 | * data_offset (needed for out of space reshaping) | |
3075 | * This field shows the data offset into the data | |
3076 | * image LV where the first stripes data starts. | |
3077 | * | |
3078 | * We keep data_offset equal on all raid disks of the set, | |
3079 | * so retrieving it from the first raid disk is sufficient. | |
3080 | */ | |
3081 | DMEMIT(" %llu", (unsigned long long) rs->dev[0].rdev.data_offset); | |
3082 | break; | |
9d09e663 | 3083 | |
3a1c1ef2 HM |
3084 | case STATUSTYPE_TABLE: |
3085 | /* Report the table line string you would use to construct this raid set */ | |
3086 | ||
3087 | /* Calculate raid parameter count */ | |
3088 | rdev_for_each(rdev, mddev) | |
3089 | if (test_bit(WriteMostly, &rdev->flags)) | |
3090 | write_mostly_params += 2; | |
3091 | raid_param_cnt += memweight(rs->rebuild_disks, | |
3092 | DISKS_ARRAY_ELEMS * sizeof(*rs->rebuild_disks)) * 2 + | |
3093 | write_mostly_params + | |
3094 | hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_NO_ARGS) + | |
3095 | hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2; | |
3096 | /* Emit table line */ | |
3097 | DMEMIT("%s %u %u", rs->raid_type->name, raid_param_cnt, mddev->new_chunk_sectors); | |
4286325b | 3098 | if (test_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) |
3fa6cf38 | 3099 | DMEMIT(" %s %s", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT), |
3a1c1ef2 | 3100 | raid10_md_layout_to_format(mddev->layout)); |
4286325b | 3101 | if (test_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) |
3fa6cf38 | 3102 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES), |
3a1c1ef2 | 3103 | raid10_md_layout_to_copies(mddev->layout)); |
4286325b | 3104 | if (test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) |
3fa6cf38 | 3105 | DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC)); |
4286325b | 3106 | if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) |
3fa6cf38 | 3107 | DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_SYNC)); |
4286325b | 3108 | if (test_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) |
3fa6cf38 | 3109 | DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE), |
3a1c1ef2 | 3110 | (unsigned long long) to_sector(mddev->bitmap_info.chunksize)); |
4286325b | 3111 | if (test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) |
3fa6cf38 | 3112 | DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET), |
3a1c1ef2 | 3113 | (unsigned long long) rs->data_offset); |
4286325b | 3114 | if (test_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) |
3fa6cf38 | 3115 | DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP), |
3a1c1ef2 | 3116 | mddev->bitmap_info.daemon_sleep); |
4286325b | 3117 | if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) |
3fa6cf38 | 3118 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS), |
3a1c1ef2 | 3119 | mddev->delta_disks); |
4286325b | 3120 | if (test_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) |
3fa6cf38 | 3121 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE), |
3a1c1ef2 HM |
3122 | max_nr_stripes); |
3123 | rdev_for_each(rdev, mddev) | |
3124 | if (test_bit(rdev->raid_disk, (void *) rs->rebuild_disks)) | |
3fa6cf38 | 3125 | DMEMIT(" %s %u", dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD), |
3a1c1ef2 HM |
3126 | rdev->raid_disk); |
3127 | rdev_for_each(rdev, mddev) | |
3128 | if (test_bit(WriteMostly, &rdev->flags)) | |
3fa6cf38 | 3129 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY), |
3a1c1ef2 | 3130 | rdev->raid_disk); |
4286325b | 3131 | if (test_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) |
3fa6cf38 | 3132 | DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND), |
3a1c1ef2 | 3133 | mddev->bitmap_info.max_write_behind); |
4286325b | 3134 | if (test_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) |
3fa6cf38 | 3135 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE), |
3a1c1ef2 | 3136 | mddev->sync_speed_max); |
4286325b | 3137 | if (test_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) |
3fa6cf38 | 3138 | DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE), |
3a1c1ef2 HM |
3139 | mddev->sync_speed_min); |
3140 | DMEMIT(" %d", rs->raid_disks); | |
3141 | rdev_for_each(rdev, mddev) { | |
3142 | struct raid_dev *rd = container_of(rdev, struct raid_dev, rdev); | |
3143 | ||
e6ca5e1a MS |
3144 | DMEMIT(" %s %s", __get_dev_name(rd->meta_dev), |
3145 | __get_dev_name(rd->data_dev)); | |
9d09e663 N |
3146 | } |
3147 | } | |
9d09e663 N |
3148 | } |
3149 | ||
be83651f JB |
3150 | static int raid_message(struct dm_target *ti, unsigned argc, char **argv) |
3151 | { | |
3152 | struct raid_set *rs = ti->private; | |
3153 | struct mddev *mddev = &rs->md; | |
3154 | ||
be83651f JB |
3155 | if (!mddev->pers || !mddev->pers->sync_request) |
3156 | return -EINVAL; | |
3157 | ||
3158 | if (!strcasecmp(argv[0], "frozen")) | |
3159 | set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3160 | else | |
3161 | clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3162 | ||
3163 | if (!strcasecmp(argv[0], "idle") || !strcasecmp(argv[0], "frozen")) { | |
3164 | if (mddev->sync_thread) { | |
3165 | set_bit(MD_RECOVERY_INTR, &mddev->recovery); | |
3166 | md_reap_sync_thread(mddev); | |
3167 | } | |
3168 | } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || | |
3169 | test_bit(MD_RECOVERY_NEEDED, &mddev->recovery)) | |
3170 | return -EBUSY; | |
3171 | else if (!strcasecmp(argv[0], "resync")) | |
3a1c1ef2 HM |
3172 | ; /* MD_RECOVERY_NEEDED set below */ |
3173 | else if (!strcasecmp(argv[0], "recover")) | |
be83651f | 3174 | set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); |
3a1c1ef2 | 3175 | else { |
be83651f JB |
3176 | if (!strcasecmp(argv[0], "check")) |
3177 | set_bit(MD_RECOVERY_CHECK, &mddev->recovery); | |
3178 | else if (!!strcasecmp(argv[0], "repair")) | |
3179 | return -EINVAL; | |
3180 | set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); | |
3181 | set_bit(MD_RECOVERY_SYNC, &mddev->recovery); | |
3182 | } | |
3183 | if (mddev->ro == 2) { | |
3184 | /* A write to sync_action is enough to justify | |
3185 | * canceling read-auto mode | |
3186 | */ | |
3187 | mddev->ro = 0; | |
3a1c1ef2 | 3188 | if (!mddev->suspended && mddev->sync_thread) |
be83651f JB |
3189 | md_wakeup_thread(mddev->sync_thread); |
3190 | } | |
3191 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
3a1c1ef2 | 3192 | if (!mddev->suspended && mddev->thread) |
be83651f JB |
3193 | md_wakeup_thread(mddev->thread); |
3194 | ||
3195 | return 0; | |
3196 | } | |
3197 | ||
3198 | static int raid_iterate_devices(struct dm_target *ti, | |
3199 | iterate_devices_callout_fn fn, void *data) | |
9d09e663 N |
3200 | { |
3201 | struct raid_set *rs = ti->private; | |
3202 | unsigned i; | |
73c6f239 | 3203 | int r = 0; |
9d09e663 | 3204 | |
73c6f239 | 3205 | for (i = 0; !r && i < rs->md.raid_disks; i++) |
9d09e663 | 3206 | if (rs->dev[i].data_dev) |
73c6f239 | 3207 | r = fn(ti, |
9d09e663 N |
3208 | rs->dev[i].data_dev, |
3209 | 0, /* No offset on data devs */ | |
3210 | rs->md.dev_sectors, | |
3211 | data); | |
3212 | ||
73c6f239 | 3213 | return r; |
9d09e663 N |
3214 | } |
3215 | ||
3216 | static void raid_io_hints(struct dm_target *ti, struct queue_limits *limits) | |
3217 | { | |
3218 | struct raid_set *rs = ti->private; | |
3219 | unsigned chunk_size = rs->md.chunk_sectors << 9; | |
d1688a6d | 3220 | struct r5conf *conf = rs->md.private; |
9d09e663 N |
3221 | |
3222 | blk_limits_io_min(limits, chunk_size); | |
3223 | blk_limits_io_opt(limits, chunk_size * (conf->raid_disks - conf->max_degraded)); | |
3224 | } | |
3225 | ||
3226 | static void raid_presuspend(struct dm_target *ti) | |
3227 | { | |
3228 | struct raid_set *rs = ti->private; | |
3229 | ||
3230 | md_stop_writes(&rs->md); | |
3231 | } | |
3232 | ||
3233 | static void raid_postsuspend(struct dm_target *ti) | |
3234 | { | |
3235 | struct raid_set *rs = ti->private; | |
3236 | ||
6e20902e HM |
3237 | if (test_and_clear_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { |
3238 | if (!rs->md.suspended) | |
3239 | mddev_suspend(&rs->md); | |
3240 | rs->md.ro = 1; | |
3241 | } | |
9d09e663 N |
3242 | } |
3243 | ||
f381e71b | 3244 | static void attempt_restore_of_faulty_devices(struct raid_set *rs) |
9d09e663 | 3245 | { |
9092c02d JB |
3246 | int i; |
3247 | uint64_t failed_devices, cleared_failed_devices = 0; | |
3248 | unsigned long flags; | |
3249 | struct dm_raid_superblock *sb; | |
9092c02d | 3250 | struct md_rdev *r; |
9d09e663 | 3251 | |
f381e71b JB |
3252 | for (i = 0; i < rs->md.raid_disks; i++) { |
3253 | r = &rs->dev[i].rdev; | |
3254 | if (test_bit(Faulty, &r->flags) && r->sb_page && | |
796a5cf0 MC |
3255 | sync_page_io(r, 0, r->sb_size, r->sb_page, REQ_OP_READ, 0, |
3256 | 1)) { | |
f381e71b JB |
3257 | DMINFO("Faulty %s device #%d has readable super block." |
3258 | " Attempting to revive it.", | |
3259 | rs->raid_type->name, i); | |
a4dc163a JB |
3260 | |
3261 | /* | |
3262 | * Faulty bit may be set, but sometimes the array can | |
3263 | * be suspended before the personalities can respond | |
3264 | * by removing the device from the array (i.e. calling | |
43157840 | 3265 | * 'hot_remove_disk'). If they haven't yet removed |
a4dc163a JB |
3266 | * the failed device, its 'raid_disk' number will be |
3267 | * '>= 0' - meaning we must call this function | |
3268 | * ourselves. | |
3269 | */ | |
3270 | if ((r->raid_disk >= 0) && | |
3271 | (r->mddev->pers->hot_remove_disk(r->mddev, r) != 0)) | |
3272 | /* Failed to revive this device, try next */ | |
3273 | continue; | |
3274 | ||
f381e71b JB |
3275 | r->raid_disk = i; |
3276 | r->saved_raid_disk = i; | |
3277 | flags = r->flags; | |
3278 | clear_bit(Faulty, &r->flags); | |
3279 | clear_bit(WriteErrorSeen, &r->flags); | |
3280 | clear_bit(In_sync, &r->flags); | |
3281 | if (r->mddev->pers->hot_add_disk(r->mddev, r)) { | |
3282 | r->raid_disk = -1; | |
3283 | r->saved_raid_disk = -1; | |
3284 | r->flags = flags; | |
3285 | } else { | |
3286 | r->recovery_offset = 0; | |
3287 | cleared_failed_devices |= 1 << i; | |
3288 | } | |
3289 | } | |
3290 | } | |
3291 | if (cleared_failed_devices) { | |
3292 | rdev_for_each(r, &rs->md) { | |
3293 | sb = page_address(r->sb_page); | |
3294 | failed_devices = le64_to_cpu(sb->failed_devices); | |
3295 | failed_devices &= ~cleared_failed_devices; | |
3296 | sb->failed_devices = cpu_to_le64(failed_devices); | |
3297 | } | |
3298 | } | |
3299 | } | |
3300 | ||
e6ca5e1a | 3301 | static int __load_dirty_region_bitmap(struct raid_set *rs) |
ecbfb9f1 HM |
3302 | { |
3303 | int r = 0; | |
3304 | ||
3305 | /* Try loading the bitmap unless "raid0", which does not have one */ | |
3306 | if (!rs_is_raid0(rs) && | |
4286325b | 3307 | !test_and_set_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags)) { |
ecbfb9f1 HM |
3308 | r = bitmap_load(&rs->md); |
3309 | if (r) | |
3310 | DMERR("Failed to load bitmap"); | |
3311 | } | |
3312 | ||
3313 | return r; | |
3314 | } | |
3315 | ||
6e20902e HM |
3316 | /* Enforce updating all superblocks */ |
3317 | static void rs_update_sbs(struct raid_set *rs) | |
3318 | { | |
3319 | struct mddev *mddev = &rs->md; | |
3320 | int ro = mddev->ro; | |
3321 | ||
3322 | set_bit(MD_CHANGE_DEVS, &mddev->flags); | |
3323 | mddev->ro = 0; | |
3324 | md_update_sb(mddev, 1); | |
3325 | mddev->ro = ro; | |
3326 | } | |
3327 | ||
9dbd1aa3 HM |
3328 | /* |
3329 | * Reshape changes raid algorithm of @rs to new one within personality | |
3330 | * (e.g. raid6_zr -> raid6_nc), changes stripe size, adds/removes | |
3331 | * disks from a raid set thus growing/shrinking it or resizes the set | |
3332 | * | |
3333 | * Call mddev_lock_nointr() before! | |
3334 | */ | |
3335 | static int rs_start_reshape(struct raid_set *rs) | |
3336 | { | |
3337 | int r; | |
3338 | struct mddev *mddev = &rs->md; | |
3339 | struct md_personality *pers = mddev->pers; | |
3340 | ||
3341 | r = rs_setup_reshape(rs); | |
3342 | if (r) | |
3343 | return r; | |
3344 | ||
3345 | /* Need to be resumed to be able to start reshape, recovery is frozen until raid_resume() though */ | |
3346 | if (mddev->suspended) | |
3347 | mddev_resume(mddev); | |
3348 | ||
3349 | /* | |
3350 | * Check any reshape constraints enforced by the personalility | |
3351 | * | |
3352 | * May as well already kick the reshape off so that * pers->start_reshape() becomes optional. | |
3353 | */ | |
3354 | r = pers->check_reshape(mddev); | |
3355 | if (r) { | |
3356 | rs->ti->error = "pers->check_reshape() failed"; | |
3357 | return r; | |
3358 | } | |
3359 | ||
3360 | /* | |
3361 | * Personality may not provide start reshape method in which | |
3362 | * case check_reshape above has already covered everything | |
3363 | */ | |
3364 | if (pers->start_reshape) { | |
3365 | r = pers->start_reshape(mddev); | |
3366 | if (r) { | |
3367 | rs->ti->error = "pers->start_reshape() failed"; | |
3368 | return r; | |
3369 | } | |
3370 | } | |
3371 | ||
3372 | /* Suspend because a resume will happen in raid_resume() */ | |
3373 | if (!mddev->suspended) | |
3374 | mddev_suspend(mddev); | |
3375 | ||
6e20902e HM |
3376 | /* |
3377 | * Now reshape got set up, update superblocks to | |
3378 | * reflect the fact so that a table reload will | |
3379 | * access proper superblock content in the ctr. | |
3380 | */ | |
3381 | rs_update_sbs(rs); | |
9dbd1aa3 HM |
3382 | |
3383 | return 0; | |
3384 | } | |
3385 | ||
ecbfb9f1 HM |
3386 | static int raid_preresume(struct dm_target *ti) |
3387 | { | |
9dbd1aa3 | 3388 | int r; |
ecbfb9f1 HM |
3389 | struct raid_set *rs = ti->private; |
3390 | struct mddev *mddev = &rs->md; | |
3391 | ||
3392 | /* This is a resume after a suspend of the set -> it's already started */ | |
4286325b | 3393 | if (test_and_set_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags)) |
ecbfb9f1 HM |
3394 | return 0; |
3395 | ||
3396 | /* | |
3397 | * The superblocks need to be updated on disk if the | |
6e20902e HM |
3398 | * array is new or new devices got added (thus zeroed |
3399 | * out by userspace) or __load_dirty_region_bitmap | |
3400 | * will overwrite them in core with old data or fail. | |
ecbfb9f1 | 3401 | */ |
6e20902e HM |
3402 | if (test_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags)) |
3403 | rs_update_sbs(rs); | |
ecbfb9f1 HM |
3404 | |
3405 | /* | |
3406 | * Disable/enable discard support on raid set after any | |
3407 | * conversion, because devices can have been added | |
3408 | */ | |
3409 | configure_discard_support(rs); | |
3410 | ||
3411 | /* Load the bitmap from disk unless raid0 */ | |
9dbd1aa3 HM |
3412 | r = __load_dirty_region_bitmap(rs); |
3413 | if (r) | |
3414 | return r; | |
3415 | ||
4257e085 HM |
3416 | /* Resize bitmap to adjust to changed region size (aka MD bitmap chunksize) */ |
3417 | if (test_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags) && | |
3418 | mddev->bitmap_info.chunksize != to_bytes(rs->requested_bitmap_chunk_sectors)) { | |
3419 | r = bitmap_resize(mddev->bitmap, mddev->dev_sectors, | |
3420 | to_bytes(rs->requested_bitmap_chunk_sectors), 0); | |
3421 | if (r) | |
3422 | DMERR("Failed to resize bitmap"); | |
3423 | } | |
3424 | ||
9dbd1aa3 HM |
3425 | /* Check for any resize/reshape on @rs and adjust/initiate */ |
3426 | /* Be prepared for mddev_resume() in raid_resume() */ | |
3427 | set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3428 | if (mddev->recovery_cp && mddev->recovery_cp < MaxSector) { | |
3429 | set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); | |
3430 | set_bit(MD_RECOVERY_SYNC, &mddev->recovery); | |
3431 | mddev->resync_min = mddev->recovery_cp; | |
3432 | } | |
3433 | ||
3434 | rs_set_capacity(rs); | |
3435 | ||
3436 | /* Check for any reshape request and region size change unless new raid set */ | |
3437 | if (test_and_clear_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { | |
3438 | /* Initiate a reshape. */ | |
3439 | mddev_lock_nointr(mddev); | |
3440 | r = rs_start_reshape(rs); | |
3441 | mddev_unlock(mddev); | |
3442 | if (r) | |
3443 | DMWARN("Failed to check/start reshape, continuing without change"); | |
3444 | r = 0; | |
3445 | } | |
3446 | ||
3447 | return r; | |
ecbfb9f1 HM |
3448 | } |
3449 | ||
f381e71b JB |
3450 | static void raid_resume(struct dm_target *ti) |
3451 | { | |
3452 | struct raid_set *rs = ti->private; | |
ecbfb9f1 | 3453 | struct mddev *mddev = &rs->md; |
f381e71b | 3454 | |
4286325b | 3455 | if (test_and_set_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { |
ecbfb9f1 HM |
3456 | /* |
3457 | * A secondary resume while the device is active. | |
3458 | * Take this opportunity to check whether any failed | |
3459 | * devices are reachable again. | |
3460 | */ | |
3461 | attempt_restore_of_faulty_devices(rs); | |
6e20902e HM |
3462 | } else { |
3463 | mddev->ro = 0; | |
3464 | mddev->in_sync = 0; | |
34f8ac6d | 3465 | |
6e20902e HM |
3466 | /* |
3467 | * When passing in flags to the ctr, we expect userspace | |
3468 | * to reset them because they made it to the superblocks | |
3469 | * and reload the mapping anyway. | |
3470 | * | |
3471 | * -> only unfreeze recovery in case of a table reload or | |
3472 | * we'll have a bogus recovery/reshape position | |
3473 | * retrieved from the superblock by the ctr because | |
3474 | * the ongoing recovery/reshape will change it after read. | |
3475 | */ | |
3476 | if (!test_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags)) | |
3477 | clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); | |
3a1c1ef2 | 3478 | |
6e20902e HM |
3479 | if (mddev->suspended) |
3480 | mddev_resume(mddev); | |
3481 | } | |
9d09e663 N |
3482 | } |
3483 | ||
3484 | static struct target_type raid_target = { | |
3485 | .name = "raid", | |
9b6e5423 | 3486 | .version = {1, 9, 0}, |
9d09e663 N |
3487 | .module = THIS_MODULE, |
3488 | .ctr = raid_ctr, | |
3489 | .dtr = raid_dtr, | |
3490 | .map = raid_map, | |
3491 | .status = raid_status, | |
be83651f | 3492 | .message = raid_message, |
9d09e663 N |
3493 | .iterate_devices = raid_iterate_devices, |
3494 | .io_hints = raid_io_hints, | |
3495 | .presuspend = raid_presuspend, | |
3496 | .postsuspend = raid_postsuspend, | |
ecbfb9f1 | 3497 | .preresume = raid_preresume, |
9d09e663 N |
3498 | .resume = raid_resume, |
3499 | }; | |
3500 | ||
3501 | static int __init dm_raid_init(void) | |
3502 | { | |
fe5d2f4a JB |
3503 | DMINFO("Loading target version %u.%u.%u", |
3504 | raid_target.version[0], | |
3505 | raid_target.version[1], | |
3506 | raid_target.version[2]); | |
9d09e663 N |
3507 | return dm_register_target(&raid_target); |
3508 | } | |
3509 | ||
3510 | static void __exit dm_raid_exit(void) | |
3511 | { | |
3512 | dm_unregister_target(&raid_target); | |
3513 | } | |
3514 | ||
3515 | module_init(dm_raid_init); | |
3516 | module_exit(dm_raid_exit); | |
3517 | ||
48cf06bc HM |
3518 | module_param(devices_handle_discard_safely, bool, 0644); |
3519 | MODULE_PARM_DESC(devices_handle_discard_safely, | |
3520 | "Set to Y if all devices in each array reliably return zeroes on reads from discarded regions"); | |
3521 | ||
ef9b85a6 MS |
3522 | MODULE_DESCRIPTION(DM_NAME " raid0/1/10/4/5/6 target"); |
3523 | MODULE_ALIAS("dm-raid0"); | |
63f33b8d JB |
3524 | MODULE_ALIAS("dm-raid1"); |
3525 | MODULE_ALIAS("dm-raid10"); | |
9d09e663 N |
3526 | MODULE_ALIAS("dm-raid4"); |
3527 | MODULE_ALIAS("dm-raid5"); | |
3528 | MODULE_ALIAS("dm-raid6"); | |
3529 | MODULE_AUTHOR("Neil Brown <dm-devel@redhat.com>"); | |
3a1c1ef2 | 3530 | MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@redhat.com>"); |
9d09e663 | 3531 | MODULE_LICENSE("GPL"); |