]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb DW |
43 | |
44 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
45 | #define MPB_ATTRIB_PM __cpu_to_le32(0x40000000) | |
46 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
47 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
48 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
51 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
52 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
53 | ||
c2c087e6 DW |
54 | #define MPB_SECTOR_CNT 418 |
55 | #define IMSM_RESERVED_SECTORS 4096 | |
979d38be | 56 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
57 | |
58 | /* Disk configuration info. */ | |
59 | #define IMSM_MAX_DEVICES 255 | |
60 | struct imsm_disk { | |
61 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
62 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
63 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
f2f27e63 DW |
64 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
65 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
66 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 67 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb DW |
68 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
69 | #define IMSM_DISK_FILLERS 4 | |
cdddbdbc DW |
70 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ |
71 | }; | |
72 | ||
73 | /* RAID map configuration infos. */ | |
74 | struct imsm_map { | |
75 | __u32 pba_of_lba0; /* start address of partition */ | |
76 | __u32 blocks_per_member;/* blocks per member */ | |
77 | __u32 num_data_stripes; /* number of data stripes */ | |
78 | __u16 blocks_per_strip; | |
79 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
80 | #define IMSM_T_STATE_NORMAL 0 | |
81 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
82 | #define IMSM_T_STATE_DEGRADED 2 |
83 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
84 | __u8 raid_level; |
85 | #define IMSM_T_RAID0 0 | |
86 | #define IMSM_T_RAID1 1 | |
87 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
88 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
89 | __u8 num_domains; /* number of parity domains */ |
90 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 91 | __u8 ddf; |
cdddbdbc | 92 | __u32 filler[7]; /* expansion area */ |
7eef0453 | 93 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 94 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
95 | * top byte contains some flags |
96 | */ | |
cdddbdbc DW |
97 | } __attribute__ ((packed)); |
98 | ||
99 | struct imsm_vol { | |
f8f603f1 | 100 | __u32 curr_migr_unit; |
fe7ed8cb | 101 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 102 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
103 | #define MIGR_INIT 0 |
104 | #define MIGR_REBUILD 1 | |
105 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
106 | #define MIGR_GEN_MIGR 3 | |
107 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 108 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
109 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
110 | __u8 dirty; | |
fe7ed8cb DW |
111 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
112 | __u16 verify_errors; /* number of mismatches */ | |
113 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
114 | __u32 filler[4]; | |
cdddbdbc DW |
115 | struct imsm_map map[1]; |
116 | /* here comes another one if migr_state */ | |
117 | } __attribute__ ((packed)); | |
118 | ||
119 | struct imsm_dev { | |
fe7ed8cb | 120 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
121 | __u32 size_low; |
122 | __u32 size_high; | |
fe7ed8cb DW |
123 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
124 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
125 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
126 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
127 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
128 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
129 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
130 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
131 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
132 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
133 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
134 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
135 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
136 | __u32 status; /* Persistent RaidDev status */ |
137 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
138 | __u8 migr_priority; |
139 | __u8 num_sub_vols; | |
140 | __u8 tid; | |
141 | __u8 cng_master_disk; | |
142 | __u16 cache_policy; | |
143 | __u8 cng_state; | |
144 | __u8 cng_sub_state; | |
145 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
146 | __u32 filler[IMSM_DEV_FILLERS]; |
147 | struct imsm_vol vol; | |
148 | } __attribute__ ((packed)); | |
149 | ||
150 | struct imsm_super { | |
151 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
152 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
153 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
154 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
155 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
156 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
157 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
158 | __u8 num_disks; /* 0x38 Number of configured disks */ |
159 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
160 | __u8 error_log_pos; /* 0x3A */ |
161 | __u8 fill[1]; /* 0x3B */ | |
162 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
163 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
164 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
165 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
166 | #define IMSM_FILLERS 35 | |
167 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
168 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
169 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 170 | /* here comes BBM logs */ |
cdddbdbc DW |
171 | } __attribute__ ((packed)); |
172 | ||
604b746f JD |
173 | #define BBM_LOG_MAX_ENTRIES 254 |
174 | ||
175 | struct bbm_log_entry { | |
176 | __u64 defective_block_start; | |
177 | #define UNREADABLE 0xFFFFFFFF | |
178 | __u32 spare_block_offset; | |
179 | __u16 remapped_marked_count; | |
180 | __u16 disk_ordinal; | |
181 | } __attribute__ ((__packed__)); | |
182 | ||
183 | struct bbm_log { | |
184 | __u32 signature; /* 0xABADB10C */ | |
185 | __u32 entry_count; | |
186 | __u32 reserved_spare_block_count; /* 0 */ | |
187 | __u32 reserved; /* 0xFFFF */ | |
188 | __u64 first_spare_lba; | |
189 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
190 | } __attribute__ ((__packed__)); | |
191 | ||
192 | ||
cdddbdbc DW |
193 | #ifndef MDASSEMBLE |
194 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
195 | #endif | |
196 | ||
1484e727 DW |
197 | static __u8 migr_type(struct imsm_dev *dev) |
198 | { | |
199 | if (dev->vol.migr_type == MIGR_VERIFY && | |
200 | dev->status & DEV_VERIFY_AND_FIX) | |
201 | return MIGR_REPAIR; | |
202 | else | |
203 | return dev->vol.migr_type; | |
204 | } | |
205 | ||
206 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
207 | { | |
208 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
209 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
210 | */ | |
211 | if (migr_type == MIGR_REPAIR) { | |
212 | dev->vol.migr_type = MIGR_VERIFY; | |
213 | dev->status |= DEV_VERIFY_AND_FIX; | |
214 | } else { | |
215 | dev->vol.migr_type = migr_type; | |
216 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
217 | } | |
218 | } | |
219 | ||
87eb16df | 220 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 221 | { |
87eb16df DW |
222 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
223 | } | |
cdddbdbc | 224 | |
87eb16df DW |
225 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
226 | { | |
227 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
228 | } |
229 | ||
ba2de7ba DW |
230 | struct intel_dev { |
231 | struct imsm_dev *dev; | |
232 | struct intel_dev *next; | |
f21e18ca | 233 | unsigned index; |
ba2de7ba DW |
234 | }; |
235 | ||
1a64be56 LM |
236 | enum action { |
237 | DISK_REMOVE = 1, | |
238 | DISK_ADD | |
239 | }; | |
cdddbdbc DW |
240 | /* internal representation of IMSM metadata */ |
241 | struct intel_super { | |
242 | union { | |
949c47a0 DW |
243 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
244 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 245 | }; |
949c47a0 | 246 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
247 | void *next_buf; /* for realloc'ing buf from the manager */ |
248 | size_t next_len; | |
c2c087e6 | 249 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 250 | int current_vol; /* index of raid device undergoing creation */ |
0dcecb2e | 251 | __u32 create_offset; /* common start for 'current_vol' */ |
148acb7b | 252 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 253 | struct intel_dev *devlist; |
cdddbdbc DW |
254 | struct dl { |
255 | struct dl *next; | |
256 | int index; | |
257 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
258 | int major, minor; | |
259 | char *devname; | |
b9f594fe | 260 | struct imsm_disk disk; |
cdddbdbc | 261 | int fd; |
0dcecb2e DW |
262 | int extent_cnt; |
263 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 264 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 265 | enum action action; |
cdddbdbc | 266 | } *disks; |
1a64be56 LM |
267 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
268 | active */ | |
47ee5a45 | 269 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 270 | struct bbm_log *bbm_log; |
88c32bb1 DW |
271 | const char *hba; /* device path of the raid controller for this metadata */ |
272 | const struct imsm_orom *orom; /* platform firmware support */ | |
a2b97981 DW |
273 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
274 | }; | |
275 | ||
276 | struct intel_disk { | |
277 | struct imsm_disk disk; | |
278 | #define IMSM_UNKNOWN_OWNER (-1) | |
279 | int owner; | |
280 | struct intel_disk *next; | |
cdddbdbc DW |
281 | }; |
282 | ||
c2c087e6 DW |
283 | struct extent { |
284 | unsigned long long start, size; | |
285 | }; | |
286 | ||
694575e7 KW |
287 | /* definitions of reshape process types */ |
288 | enum imsm_reshape_type { | |
289 | CH_TAKEOVER, | |
290 | CH_CHUNK_MIGR, | |
291 | CH_LEVEL_MIGRATION | |
292 | }; | |
293 | ||
88758e9d DW |
294 | /* definition of messages passed to imsm_process_update */ |
295 | enum imsm_update_type { | |
296 | update_activate_spare, | |
8273f55e | 297 | update_create_array, |
33414a01 | 298 | update_kill_array, |
aa534678 | 299 | update_rename_array, |
1a64be56 | 300 | update_add_remove_disk, |
78b10e66 | 301 | update_reshape_container_disks, |
88758e9d DW |
302 | }; |
303 | ||
304 | struct imsm_update_activate_spare { | |
305 | enum imsm_update_type type; | |
d23fe947 | 306 | struct dl *dl; |
88758e9d DW |
307 | int slot; |
308 | int array; | |
309 | struct imsm_update_activate_spare *next; | |
310 | }; | |
311 | ||
78b10e66 N |
312 | struct geo_params { |
313 | int dev_id; | |
314 | char *dev_name; | |
315 | long long size; | |
316 | int level; | |
317 | int layout; | |
318 | int chunksize; | |
319 | int raid_disks; | |
320 | }; | |
321 | ||
322 | ||
323 | struct imsm_update_reshape { | |
324 | enum imsm_update_type type; | |
325 | int old_raid_disks; | |
326 | int new_raid_disks; | |
d195167d | 327 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
328 | }; |
329 | ||
54c2c1ea DW |
330 | struct disk_info { |
331 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
332 | }; | |
333 | ||
8273f55e DW |
334 | struct imsm_update_create_array { |
335 | enum imsm_update_type type; | |
8273f55e | 336 | int dev_idx; |
6a3e913e | 337 | struct imsm_dev dev; |
8273f55e DW |
338 | }; |
339 | ||
33414a01 DW |
340 | struct imsm_update_kill_array { |
341 | enum imsm_update_type type; | |
342 | int dev_idx; | |
343 | }; | |
344 | ||
aa534678 DW |
345 | struct imsm_update_rename_array { |
346 | enum imsm_update_type type; | |
347 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
348 | int dev_idx; | |
349 | }; | |
350 | ||
1a64be56 | 351 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
352 | enum imsm_update_type type; |
353 | }; | |
354 | ||
cdddbdbc DW |
355 | static struct supertype *match_metadata_desc_imsm(char *arg) |
356 | { | |
357 | struct supertype *st; | |
358 | ||
359 | if (strcmp(arg, "imsm") != 0 && | |
360 | strcmp(arg, "default") != 0 | |
361 | ) | |
362 | return NULL; | |
363 | ||
364 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
365 | if (!st) |
366 | return NULL; | |
ef609477 | 367 | memset(st, 0, sizeof(*st)); |
d1d599ea | 368 | st->container_dev = NoMdDev; |
cdddbdbc DW |
369 | st->ss = &super_imsm; |
370 | st->max_devs = IMSM_MAX_DEVICES; | |
371 | st->minor_version = 0; | |
372 | st->sb = NULL; | |
373 | return st; | |
374 | } | |
375 | ||
0e600426 | 376 | #ifndef MDASSEMBLE |
cdddbdbc DW |
377 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
378 | { | |
379 | return &mpb->sig[MPB_SIG_LEN]; | |
380 | } | |
0e600426 | 381 | #endif |
cdddbdbc | 382 | |
949c47a0 DW |
383 | /* retrieve a disk directly from the anchor when the anchor is known to be |
384 | * up-to-date, currently only at load time | |
385 | */ | |
386 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 387 | { |
949c47a0 | 388 | if (index >= mpb->num_disks) |
cdddbdbc DW |
389 | return NULL; |
390 | return &mpb->disk[index]; | |
391 | } | |
392 | ||
95d07a2c LM |
393 | /* retrieve the disk description based on a index of the disk |
394 | * in the sub-array | |
395 | */ | |
396 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 397 | { |
b9f594fe DW |
398 | struct dl *d; |
399 | ||
400 | for (d = super->disks; d; d = d->next) | |
401 | if (d->index == index) | |
95d07a2c LM |
402 | return d; |
403 | ||
404 | return NULL; | |
405 | } | |
406 | /* retrieve a disk from the parsed metadata */ | |
407 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
408 | { | |
409 | struct dl *dl; | |
410 | ||
411 | dl = get_imsm_dl_disk(super, index); | |
412 | if (dl) | |
413 | return &dl->disk; | |
414 | ||
b9f594fe | 415 | return NULL; |
949c47a0 DW |
416 | } |
417 | ||
418 | /* generate a checksum directly from the anchor when the anchor is known to be | |
419 | * up-to-date, currently only at load or write_super after coalescing | |
420 | */ | |
421 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
422 | { |
423 | __u32 end = mpb->mpb_size / sizeof(end); | |
424 | __u32 *p = (__u32 *) mpb; | |
425 | __u32 sum = 0; | |
426 | ||
97f734fd N |
427 | while (end--) { |
428 | sum += __le32_to_cpu(*p); | |
429 | p++; | |
430 | } | |
cdddbdbc DW |
431 | |
432 | return sum - __le32_to_cpu(mpb->check_sum); | |
433 | } | |
434 | ||
a965f303 DW |
435 | static size_t sizeof_imsm_map(struct imsm_map *map) |
436 | { | |
437 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
438 | } | |
439 | ||
440 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 441 | { |
a965f303 DW |
442 | struct imsm_map *map = &dev->vol.map[0]; |
443 | ||
444 | if (second_map && !dev->vol.migr_state) | |
445 | return NULL; | |
446 | else if (second_map) { | |
447 | void *ptr = map; | |
448 | ||
449 | return ptr + sizeof_imsm_map(map); | |
450 | } else | |
451 | return map; | |
452 | ||
453 | } | |
cdddbdbc | 454 | |
3393c6af DW |
455 | /* return the size of the device. |
456 | * migr_state increases the returned size if map[0] were to be duplicated | |
457 | */ | |
458 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
459 | { |
460 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
461 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
462 | |
463 | /* migrating means an additional map */ | |
a965f303 DW |
464 | if (dev->vol.migr_state) |
465 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
466 | else if (migr_state) |
467 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
468 | |
469 | return size; | |
470 | } | |
471 | ||
54c2c1ea DW |
472 | #ifndef MDASSEMBLE |
473 | /* retrieve disk serial number list from a metadata update */ | |
474 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
475 | { | |
476 | void *u = update; | |
477 | struct disk_info *inf; | |
478 | ||
479 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
480 | sizeof_imsm_dev(&update->dev, 0); | |
481 | ||
482 | return inf; | |
483 | } | |
484 | #endif | |
485 | ||
949c47a0 | 486 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
487 | { |
488 | int offset; | |
489 | int i; | |
490 | void *_mpb = mpb; | |
491 | ||
949c47a0 | 492 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
493 | return NULL; |
494 | ||
495 | /* devices start after all disks */ | |
496 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
497 | ||
498 | for (i = 0; i <= index; i++) | |
499 | if (i == index) | |
500 | return _mpb + offset; | |
501 | else | |
3393c6af | 502 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
503 | |
504 | return NULL; | |
505 | } | |
506 | ||
949c47a0 DW |
507 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
508 | { | |
ba2de7ba DW |
509 | struct intel_dev *dv; |
510 | ||
949c47a0 DW |
511 | if (index >= super->anchor->num_raid_devs) |
512 | return NULL; | |
ba2de7ba DW |
513 | for (dv = super->devlist; dv; dv = dv->next) |
514 | if (dv->index == index) | |
515 | return dv->dev; | |
516 | return NULL; | |
949c47a0 DW |
517 | } |
518 | ||
98130f40 AK |
519 | /* |
520 | * for second_map: | |
521 | * == 0 get first map | |
522 | * == 1 get second map | |
523 | * == -1 than get map according to the current migr_state | |
524 | */ | |
525 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
526 | int slot, | |
527 | int second_map) | |
7eef0453 DW |
528 | { |
529 | struct imsm_map *map; | |
530 | ||
98130f40 AK |
531 | if (second_map == -1) { |
532 | if (dev->vol.migr_state) | |
533 | map = get_imsm_map(dev, 1); | |
534 | else | |
535 | map = get_imsm_map(dev, 0); | |
536 | } else { | |
537 | map = get_imsm_map(dev, second_map); | |
538 | } | |
7eef0453 | 539 | |
ff077194 DW |
540 | /* top byte identifies disk under rebuild */ |
541 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
542 | } | |
543 | ||
544 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 545 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 546 | { |
98130f40 | 547 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
548 | |
549 | return ord_to_idx(ord); | |
7eef0453 DW |
550 | } |
551 | ||
be73972f DW |
552 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
553 | { | |
554 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
555 | } | |
556 | ||
f21e18ca | 557 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
558 | { |
559 | int slot; | |
560 | __u32 ord; | |
561 | ||
562 | for (slot = 0; slot < map->num_members; slot++) { | |
563 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
564 | if (ord_to_idx(ord) == idx) | |
565 | return slot; | |
566 | } | |
567 | ||
568 | return -1; | |
569 | } | |
570 | ||
cdddbdbc DW |
571 | static int get_imsm_raid_level(struct imsm_map *map) |
572 | { | |
573 | if (map->raid_level == 1) { | |
574 | if (map->num_members == 2) | |
575 | return 1; | |
576 | else | |
577 | return 10; | |
578 | } | |
579 | ||
580 | return map->raid_level; | |
581 | } | |
582 | ||
c2c087e6 DW |
583 | static int cmp_extent(const void *av, const void *bv) |
584 | { | |
585 | const struct extent *a = av; | |
586 | const struct extent *b = bv; | |
587 | if (a->start < b->start) | |
588 | return -1; | |
589 | if (a->start > b->start) | |
590 | return 1; | |
591 | return 0; | |
592 | } | |
593 | ||
0dcecb2e | 594 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 595 | { |
c2c087e6 | 596 | int memberships = 0; |
620b1713 | 597 | int i; |
c2c087e6 | 598 | |
949c47a0 DW |
599 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
600 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 601 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 602 | |
620b1713 DW |
603 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
604 | memberships++; | |
c2c087e6 | 605 | } |
0dcecb2e DW |
606 | |
607 | return memberships; | |
608 | } | |
609 | ||
610 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
611 | { | |
612 | /* find a list of used extents on the given physical device */ | |
613 | struct extent *rv, *e; | |
620b1713 | 614 | int i; |
0dcecb2e DW |
615 | int memberships = count_memberships(dl, super); |
616 | __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
617 | ||
c2c087e6 DW |
618 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
619 | if (!rv) | |
620 | return NULL; | |
621 | e = rv; | |
622 | ||
949c47a0 DW |
623 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
624 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 625 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 626 | |
620b1713 DW |
627 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
628 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
629 | e->size = __le32_to_cpu(map->blocks_per_member); | |
630 | e++; | |
c2c087e6 DW |
631 | } |
632 | } | |
633 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
634 | ||
14e8215b DW |
635 | /* determine the start of the metadata |
636 | * when no raid devices are defined use the default | |
637 | * ...otherwise allow the metadata to truncate the value | |
638 | * as is the case with older versions of imsm | |
639 | */ | |
640 | if (memberships) { | |
641 | struct extent *last = &rv[memberships - 1]; | |
642 | __u32 remainder; | |
643 | ||
644 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
645 | (last->start + last->size); | |
dda5855f DW |
646 | /* round down to 1k block to satisfy precision of the kernel |
647 | * 'size' interface | |
648 | */ | |
649 | remainder &= ~1UL; | |
650 | /* make sure remainder is still sane */ | |
f21e18ca | 651 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 652 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
653 | if (reservation > remainder) |
654 | reservation = remainder; | |
655 | } | |
656 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
657 | e->size = 0; |
658 | return rv; | |
659 | } | |
660 | ||
14e8215b DW |
661 | /* try to determine how much space is reserved for metadata from |
662 | * the last get_extents() entry, otherwise fallback to the | |
663 | * default | |
664 | */ | |
665 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
666 | { | |
667 | struct extent *e; | |
668 | int i; | |
669 | __u32 rv; | |
670 | ||
671 | /* for spares just return a minimal reservation which will grow | |
672 | * once the spare is picked up by an array | |
673 | */ | |
674 | if (dl->index == -1) | |
675 | return MPB_SECTOR_CNT; | |
676 | ||
677 | e = get_extents(super, dl); | |
678 | if (!e) | |
679 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
680 | ||
681 | /* scroll to last entry */ | |
682 | for (i = 0; e[i].size; i++) | |
683 | continue; | |
684 | ||
685 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
686 | ||
687 | free(e); | |
688 | ||
689 | return rv; | |
690 | } | |
691 | ||
25ed7e59 DW |
692 | static int is_spare(struct imsm_disk *disk) |
693 | { | |
694 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
695 | } | |
696 | ||
697 | static int is_configured(struct imsm_disk *disk) | |
698 | { | |
699 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
700 | } | |
701 | ||
702 | static int is_failed(struct imsm_disk *disk) | |
703 | { | |
704 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
705 | } | |
706 | ||
80e7f8c3 AC |
707 | /* Return minimum size of a spare that can be used in this array*/ |
708 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
709 | { | |
710 | struct intel_super *super = st->sb; | |
711 | struct dl *dl; | |
712 | struct extent *e; | |
713 | int i; | |
714 | unsigned long long rv = 0; | |
715 | ||
716 | if (!super) | |
717 | return rv; | |
718 | /* find first active disk in array */ | |
719 | dl = super->disks; | |
720 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
721 | dl = dl->next; | |
722 | if (!dl) | |
723 | return rv; | |
724 | /* find last lba used by subarrays */ | |
725 | e = get_extents(super, dl); | |
726 | if (!e) | |
727 | return rv; | |
728 | for (i = 0; e[i].size; i++) | |
729 | continue; | |
730 | if (i > 0) | |
731 | rv = e[i-1].start + e[i-1].size; | |
732 | free(e); | |
733 | /* add the amount of space needed for metadata */ | |
734 | rv = rv + MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
735 | return rv * 512; | |
736 | } | |
737 | ||
1799c9e8 | 738 | #ifndef MDASSEMBLE |
1e5c6983 DW |
739 | static __u64 blocks_per_migr_unit(struct imsm_dev *dev); |
740 | ||
44470971 | 741 | static void print_imsm_dev(struct imsm_dev *dev, char *uuid, int disk_idx) |
cdddbdbc DW |
742 | { |
743 | __u64 sz; | |
0d80bb2f | 744 | int slot, i; |
a965f303 | 745 | struct imsm_map *map = get_imsm_map(dev, 0); |
dd8bcb3b | 746 | struct imsm_map *map2 = get_imsm_map(dev, 1); |
b10b37b8 | 747 | __u32 ord; |
cdddbdbc DW |
748 | |
749 | printf("\n"); | |
1e7bc0ed | 750 | printf("[%.16s]:\n", dev->volume); |
44470971 | 751 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
752 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
753 | if (map2) | |
754 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
755 | printf("\n"); | |
756 | printf(" Members : %d", map->num_members); | |
757 | if (map2) | |
758 | printf(" <-- %d", map2->num_members); | |
759 | printf("\n"); | |
0d80bb2f DW |
760 | printf(" Slots : ["); |
761 | for (i = 0; i < map->num_members; i++) { | |
dd8bcb3b | 762 | ord = get_imsm_ord_tbl_ent(dev, i, 0); |
0d80bb2f DW |
763 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
764 | } | |
dd8bcb3b AK |
765 | printf("]"); |
766 | if (map2) { | |
767 | printf(" <-- ["); | |
768 | for (i = 0; i < map2->num_members; i++) { | |
769 | ord = get_imsm_ord_tbl_ent(dev, i, 1); | |
770 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); | |
771 | } | |
772 | printf("]"); | |
773 | } | |
774 | printf("\n"); | |
620b1713 DW |
775 | slot = get_imsm_disk_slot(map, disk_idx); |
776 | if (slot >= 0) { | |
98130f40 | 777 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
b10b37b8 DW |
778 | printf(" This Slot : %d%s\n", slot, |
779 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
780 | } else | |
cdddbdbc DW |
781 | printf(" This Slot : ?\n"); |
782 | sz = __le32_to_cpu(dev->size_high); | |
783 | sz <<= 32; | |
784 | sz += __le32_to_cpu(dev->size_low); | |
785 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
786 | human_size(sz * 512)); | |
787 | sz = __le32_to_cpu(map->blocks_per_member); | |
788 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
789 | human_size(sz * 512)); | |
790 | printf(" Sector Offset : %u\n", | |
791 | __le32_to_cpu(map->pba_of_lba0)); | |
792 | printf(" Num Stripes : %u\n", | |
793 | __le32_to_cpu(map->num_data_stripes)); | |
dd8bcb3b | 794 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 795 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
796 | if (map2) |
797 | printf(" <-- %u KiB", | |
798 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
799 | printf("\n"); | |
cdddbdbc | 800 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 801 | printf(" Migrate State : "); |
1484e727 DW |
802 | if (dev->vol.migr_state) { |
803 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 804 | printf("initialize\n"); |
1484e727 | 805 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 806 | printf("rebuild\n"); |
1484e727 | 807 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 808 | printf("check\n"); |
1484e727 | 809 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 810 | printf("general migration\n"); |
1484e727 | 811 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 812 | printf("state change\n"); |
1484e727 | 813 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 814 | printf("repair\n"); |
1484e727 | 815 | else |
8655a7b1 DW |
816 | printf("<unknown:%d>\n", migr_type(dev)); |
817 | } else | |
818 | printf("idle\n"); | |
3393c6af DW |
819 | printf(" Map State : %s", map_state_str[map->map_state]); |
820 | if (dev->vol.migr_state) { | |
821 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1e5c6983 | 822 | |
b10b37b8 | 823 | printf(" <-- %s", map_state_str[map->map_state]); |
1e5c6983 DW |
824 | printf("\n Checkpoint : %u (%llu)", |
825 | __le32_to_cpu(dev->vol.curr_migr_unit), | |
94fcb80a | 826 | (unsigned long long)blocks_per_migr_unit(dev)); |
3393c6af DW |
827 | } |
828 | printf("\n"); | |
cdddbdbc | 829 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
830 | } |
831 | ||
14e8215b | 832 | static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved) |
cdddbdbc | 833 | { |
949c47a0 | 834 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); |
1f24f035 | 835 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
836 | __u64 sz; |
837 | ||
d362da3d | 838 | if (index < 0 || !disk) |
e9d82038 DW |
839 | return; |
840 | ||
cdddbdbc | 841 | printf("\n"); |
1f24f035 | 842 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
cdddbdbc | 843 | printf(" Disk%02d Serial : %s\n", index, str); |
25ed7e59 DW |
844 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
845 | is_configured(disk) ? " active" : "", | |
846 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 847 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
14e8215b | 848 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
849 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
850 | human_size(sz * 512)); | |
851 | } | |
852 | ||
a5d85af7 | 853 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 854 | |
cdddbdbc DW |
855 | static void examine_super_imsm(struct supertype *st, char *homehost) |
856 | { | |
857 | struct intel_super *super = st->sb; | |
949c47a0 | 858 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
859 | char str[MAX_SIGNATURE_LENGTH]; |
860 | int i; | |
27fd6274 DW |
861 | struct mdinfo info; |
862 | char nbuf[64]; | |
cdddbdbc | 863 | __u32 sum; |
14e8215b | 864 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 865 | struct dl *dl; |
27fd6274 | 866 | |
cdddbdbc DW |
867 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
868 | printf(" Magic : %s\n", str); | |
869 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
870 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 871 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
872 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
873 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
a5d85af7 | 874 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 875 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 876 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
877 | sum = __le32_to_cpu(mpb->check_sum); |
878 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 879 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 880 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
881 | printf(" Disks : %d\n", mpb->num_disks); |
882 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
14e8215b | 883 | print_imsm_disk(mpb, super->disks->index, reserved); |
604b746f JD |
884 | if (super->bbm_log) { |
885 | struct bbm_log *log = super->bbm_log; | |
886 | ||
887 | printf("\n"); | |
888 | printf("Bad Block Management Log:\n"); | |
889 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
890 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
891 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
892 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
893 | printf(" First Spare : %llx\n", |
894 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 895 | } |
44470971 DW |
896 | for (i = 0; i < mpb->num_raid_devs; i++) { |
897 | struct mdinfo info; | |
898 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
899 | ||
900 | super->current_vol = i; | |
a5d85af7 | 901 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 902 | fname_from_uuid(st, &info, nbuf, ':'); |
44470971 DW |
903 | print_imsm_dev(dev, nbuf + 5, super->disks->index); |
904 | } | |
cdddbdbc DW |
905 | for (i = 0; i < mpb->num_disks; i++) { |
906 | if (i == super->disks->index) | |
907 | continue; | |
14e8215b | 908 | print_imsm_disk(mpb, i, reserved); |
cdddbdbc | 909 | } |
94827db3 N |
910 | for (dl = super->disks ; dl; dl = dl->next) { |
911 | struct imsm_disk *disk; | |
912 | char str[MAX_RAID_SERIAL_LEN + 1]; | |
913 | __u64 sz; | |
914 | ||
915 | if (dl->index >= 0) | |
916 | continue; | |
917 | ||
918 | disk = &dl->disk; | |
919 | printf("\n"); | |
920 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); | |
921 | printf(" Disk Serial : %s\n", str); | |
922 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", | |
923 | is_configured(disk) ? " active" : "", | |
924 | is_failed(disk) ? " failed" : ""); | |
925 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
926 | sz = __le32_to_cpu(disk->total_blocks) - reserved; | |
927 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, | |
928 | human_size(sz * 512)); | |
929 | } | |
cdddbdbc DW |
930 | } |
931 | ||
061f2c6a | 932 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 933 | { |
27fd6274 | 934 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
935 | struct mdinfo info; |
936 | char nbuf[64]; | |
1e7bc0ed | 937 | struct intel_super *super = st->sb; |
1e7bc0ed | 938 | |
0d5a423f DW |
939 | if (!super->anchor->num_raid_devs) { |
940 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 941 | return; |
0d5a423f | 942 | } |
ff54de6e | 943 | |
a5d85af7 | 944 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
945 | fname_from_uuid(st, &info, nbuf, ':'); |
946 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
947 | } | |
948 | ||
949 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
950 | { | |
951 | /* We just write a generic IMSM ARRAY entry */ | |
952 | struct mdinfo info; | |
953 | char nbuf[64]; | |
954 | char nbuf1[64]; | |
955 | struct intel_super *super = st->sb; | |
956 | int i; | |
957 | ||
958 | if (!super->anchor->num_raid_devs) | |
959 | return; | |
960 | ||
a5d85af7 | 961 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 962 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
963 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
964 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
965 | ||
966 | super->current_vol = i; | |
a5d85af7 | 967 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 968 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 969 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 970 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 971 | } |
cdddbdbc DW |
972 | } |
973 | ||
9d84c8ea DW |
974 | static void export_examine_super_imsm(struct supertype *st) |
975 | { | |
976 | struct intel_super *super = st->sb; | |
977 | struct imsm_super *mpb = super->anchor; | |
978 | struct mdinfo info; | |
979 | char nbuf[64]; | |
980 | ||
a5d85af7 | 981 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
982 | fname_from_uuid(st, &info, nbuf, ':'); |
983 | printf("MD_METADATA=imsm\n"); | |
984 | printf("MD_LEVEL=container\n"); | |
985 | printf("MD_UUID=%s\n", nbuf+5); | |
986 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
987 | } | |
988 | ||
cdddbdbc DW |
989 | static void detail_super_imsm(struct supertype *st, char *homehost) |
990 | { | |
3ebe00a1 DW |
991 | struct mdinfo info; |
992 | char nbuf[64]; | |
993 | ||
a5d85af7 | 994 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 995 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 996 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
997 | } |
998 | ||
999 | static void brief_detail_super_imsm(struct supertype *st) | |
1000 | { | |
ff54de6e N |
1001 | struct mdinfo info; |
1002 | char nbuf[64]; | |
a5d85af7 | 1003 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1004 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1005 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1006 | } |
d665cc31 DW |
1007 | |
1008 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1009 | static void fd2devname(int fd, char *name); | |
1010 | ||
1011 | static int imsm_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) | |
1012 | { | |
1013 | /* dump an unsorted list of devices attached to ahci, as well as | |
1014 | * non-connected ports | |
1015 | */ | |
1016 | int hba_len = strlen(hba_path) + 1; | |
1017 | struct dirent *ent; | |
1018 | DIR *dir; | |
1019 | char *path = NULL; | |
1020 | int err = 0; | |
1021 | unsigned long port_mask = (1 << port_count) - 1; | |
1022 | ||
f21e18ca | 1023 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1024 | if (verbose) |
1025 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1026 | return 2; | |
1027 | } | |
1028 | ||
1029 | /* scroll through /sys/dev/block looking for devices attached to | |
1030 | * this hba | |
1031 | */ | |
1032 | dir = opendir("/sys/dev/block"); | |
1033 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1034 | int fd; | |
1035 | char model[64]; | |
1036 | char vendor[64]; | |
1037 | char buf[1024]; | |
1038 | int major, minor; | |
1039 | char *device; | |
1040 | char *c; | |
1041 | int port; | |
1042 | int type; | |
1043 | ||
1044 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1045 | continue; | |
1046 | path = devt_to_devpath(makedev(major, minor)); | |
1047 | if (!path) | |
1048 | continue; | |
1049 | if (!path_attached_to_hba(path, hba_path)) { | |
1050 | free(path); | |
1051 | path = NULL; | |
1052 | continue; | |
1053 | } | |
1054 | ||
1055 | /* retrieve the scsi device type */ | |
1056 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1057 | if (verbose) | |
1058 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1059 | err = 2; | |
1060 | break; | |
1061 | } | |
1062 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1063 | if (load_sys(device, buf) != 0) { | |
1064 | if (verbose) | |
1065 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1066 | path); | |
1067 | err = 2; | |
1068 | free(device); | |
1069 | break; | |
1070 | } | |
1071 | type = strtoul(buf, NULL, 10); | |
1072 | ||
1073 | /* if it's not a disk print the vendor and model */ | |
1074 | if (!(type == 0 || type == 7 || type == 14)) { | |
1075 | vendor[0] = '\0'; | |
1076 | model[0] = '\0'; | |
1077 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1078 | if (load_sys(device, buf) == 0) { | |
1079 | strncpy(vendor, buf, sizeof(vendor)); | |
1080 | vendor[sizeof(vendor) - 1] = '\0'; | |
1081 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1082 | while (isspace(*c) || *c == '\0') | |
1083 | *c-- = '\0'; | |
1084 | ||
1085 | } | |
1086 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1087 | if (load_sys(device, buf) == 0) { | |
1088 | strncpy(model, buf, sizeof(model)); | |
1089 | model[sizeof(model) - 1] = '\0'; | |
1090 | c = (char *) &model[sizeof(model) - 1]; | |
1091 | while (isspace(*c) || *c == '\0') | |
1092 | *c-- = '\0'; | |
1093 | } | |
1094 | ||
1095 | if (vendor[0] && model[0]) | |
1096 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1097 | else | |
1098 | switch (type) { /* numbers from hald/linux/device.c */ | |
1099 | case 1: sprintf(buf, "tape"); break; | |
1100 | case 2: sprintf(buf, "printer"); break; | |
1101 | case 3: sprintf(buf, "processor"); break; | |
1102 | case 4: | |
1103 | case 5: sprintf(buf, "cdrom"); break; | |
1104 | case 6: sprintf(buf, "scanner"); break; | |
1105 | case 8: sprintf(buf, "media_changer"); break; | |
1106 | case 9: sprintf(buf, "comm"); break; | |
1107 | case 12: sprintf(buf, "raid"); break; | |
1108 | default: sprintf(buf, "unknown"); | |
1109 | } | |
1110 | } else | |
1111 | buf[0] = '\0'; | |
1112 | free(device); | |
1113 | ||
1114 | /* chop device path to 'host%d' and calculate the port number */ | |
1115 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1116 | if (!c) { |
1117 | if (verbose) | |
1118 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1119 | err = 2; | |
1120 | break; | |
1121 | } | |
d665cc31 DW |
1122 | *c = '\0'; |
1123 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1124 | port -= host_base; | |
1125 | else { | |
1126 | if (verbose) { | |
1127 | *c = '/'; /* repair the full string */ | |
1128 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1129 | path); | |
1130 | } | |
1131 | err = 2; | |
1132 | break; | |
1133 | } | |
1134 | ||
1135 | /* mark this port as used */ | |
1136 | port_mask &= ~(1 << port); | |
1137 | ||
1138 | /* print out the device information */ | |
1139 | if (buf[0]) { | |
1140 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1141 | continue; | |
1142 | } | |
1143 | ||
1144 | fd = dev_open(ent->d_name, O_RDONLY); | |
1145 | if (fd < 0) | |
1146 | printf(" Port%d : - disk info unavailable -\n", port); | |
1147 | else { | |
1148 | fd2devname(fd, buf); | |
1149 | printf(" Port%d : %s", port, buf); | |
1150 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
1151 | printf(" (%s)\n", buf); | |
1152 | else | |
1153 | printf("()\n"); | |
1154 | } | |
1155 | close(fd); | |
1156 | free(path); | |
1157 | path = NULL; | |
1158 | } | |
1159 | if (path) | |
1160 | free(path); | |
1161 | if (dir) | |
1162 | closedir(dir); | |
1163 | if (err == 0) { | |
1164 | int i; | |
1165 | ||
1166 | for (i = 0; i < port_count; i++) | |
1167 | if (port_mask & (1 << i)) | |
1168 | printf(" Port%d : - no device attached -\n", i); | |
1169 | } | |
1170 | ||
1171 | return err; | |
1172 | } | |
1173 | ||
5615172f | 1174 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1175 | { |
1176 | /* There are two components to imsm platform support, the ahci SATA | |
1177 | * controller and the option-rom. To find the SATA controller we | |
1178 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1179 | * controller with the Intel vendor id is present. This approach | |
1180 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1181 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1182 | * detect platform raid capabilities. The option-rom resides in a | |
1183 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1184 | * platform capabilities. If raid support is disabled in the BIOS the | |
1185 | * option-rom capability structure will not be available. | |
1186 | */ | |
1187 | const struct imsm_orom *orom; | |
1188 | struct sys_dev *list, *hba; | |
1189 | DIR *dir; | |
1190 | struct dirent *ent; | |
1191 | const char *hba_path; | |
1192 | int host_base = 0; | |
1193 | int port_count = 0; | |
1194 | ||
5615172f DW |
1195 | if (enumerate_only) { |
1196 | if (check_env("IMSM_NO_PLATFORM") || find_imsm_orom()) | |
1197 | return 0; | |
1198 | return 2; | |
1199 | } | |
1200 | ||
d665cc31 DW |
1201 | list = find_driver_devices("pci", "ahci"); |
1202 | for (hba = list; hba; hba = hba->next) | |
1203 | if (devpath_to_vendor(hba->path) == 0x8086) | |
1204 | break; | |
1205 | ||
1206 | if (!hba) { | |
1207 | if (verbose) | |
1208 | fprintf(stderr, Name ": unable to find active ahci controller\n"); | |
1209 | free_sys_dev(&list); | |
1210 | return 2; | |
1211 | } else if (verbose) | |
1212 | fprintf(stderr, Name ": found Intel SATA AHCI Controller\n"); | |
1213 | hba_path = hba->path; | |
1214 | hba->path = NULL; | |
1215 | free_sys_dev(&list); | |
1216 | ||
1217 | orom = find_imsm_orom(); | |
1218 | if (!orom) { | |
1219 | if (verbose) | |
1220 | fprintf(stderr, Name ": imsm option-rom not found\n"); | |
1221 | return 2; | |
1222 | } | |
1223 | ||
1224 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1225 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1226 | orom->hotfix_ver, orom->build); | |
1227 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1228 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1229 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1230 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1231 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1232 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
8be094f0 DW |
1233 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", |
1234 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1235 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1236 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1237 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1238 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1239 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1240 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1241 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1242 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1243 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1244 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1245 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1246 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1247 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1248 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1249 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
d665cc31 DW |
1250 | printf(" Max Disks : %d\n", orom->tds); |
1251 | printf(" Max Volumes : %d\n", orom->vpa); | |
1252 | printf(" I/O Controller : %s\n", hba_path); | |
1253 | ||
1254 | /* find the smallest scsi host number to determine a port number base */ | |
1255 | dir = opendir(hba_path); | |
1256 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1257 | int host; | |
1258 | ||
1259 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1260 | continue; | |
1261 | if (port_count == 0) | |
1262 | host_base = host; | |
1263 | else if (host < host_base) | |
1264 | host_base = host; | |
1265 | ||
1266 | if (host + 1 > port_count + host_base) | |
1267 | port_count = host + 1 - host_base; | |
1268 | ||
1269 | } | |
1270 | if (dir) | |
1271 | closedir(dir); | |
1272 | ||
1273 | if (!port_count || imsm_enumerate_ports(hba_path, port_count, | |
1274 | host_base, verbose) != 0) { | |
1275 | if (verbose) | |
1276 | fprintf(stderr, Name ": failed to enumerate ports\n"); | |
1277 | return 2; | |
1278 | } | |
1279 | ||
1280 | return 0; | |
1281 | } | |
cdddbdbc DW |
1282 | #endif |
1283 | ||
1284 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1285 | { | |
5115ca67 DW |
1286 | /* the imsm metadata format does not specify any host |
1287 | * identification information. We return -1 since we can never | |
1288 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1289 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1290 | * exclude member disks that do not belong, and we rely on |
1291 | * mdadm.conf to specify the arrays that should be assembled. | |
1292 | * Auto-assembly may still pick up "foreign" arrays. | |
1293 | */ | |
cdddbdbc | 1294 | |
9362c1c8 | 1295 | return -1; |
cdddbdbc DW |
1296 | } |
1297 | ||
1298 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1299 | { | |
51006d85 N |
1300 | /* The uuid returned here is used for: |
1301 | * uuid to put into bitmap file (Create, Grow) | |
1302 | * uuid for backup header when saving critical section (Grow) | |
1303 | * comparing uuids when re-adding a device into an array | |
1304 | * In these cases the uuid required is that of the data-array, | |
1305 | * not the device-set. | |
1306 | * uuid to recognise same set when adding a missing device back | |
1307 | * to an array. This is a uuid for the device-set. | |
1308 | * | |
1309 | * For each of these we can make do with a truncated | |
1310 | * or hashed uuid rather than the original, as long as | |
1311 | * everyone agrees. | |
1312 | * In each case the uuid required is that of the data-array, | |
1313 | * not the device-set. | |
43dad3d6 | 1314 | */ |
51006d85 N |
1315 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1316 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1317 | * - the orig_family_num of the container |
51006d85 N |
1318 | * - the index number of the volume |
1319 | * - the 'serial' number of the volume. | |
1320 | * Hopefully these are all constant. | |
1321 | */ | |
1322 | struct intel_super *super = st->sb; | |
43dad3d6 | 1323 | |
51006d85 N |
1324 | char buf[20]; |
1325 | struct sha1_ctx ctx; | |
1326 | struct imsm_dev *dev = NULL; | |
148acb7b | 1327 | __u32 family_num; |
51006d85 | 1328 | |
148acb7b DW |
1329 | /* some mdadm versions failed to set ->orig_family_num, in which |
1330 | * case fall back to ->family_num. orig_family_num will be | |
1331 | * fixed up with the first metadata update. | |
1332 | */ | |
1333 | family_num = super->anchor->orig_family_num; | |
1334 | if (family_num == 0) | |
1335 | family_num = super->anchor->family_num; | |
51006d85 | 1336 | sha1_init_ctx(&ctx); |
92bd8f8d | 1337 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1338 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1339 | if (super->current_vol >= 0) |
1340 | dev = get_imsm_dev(super, super->current_vol); | |
1341 | if (dev) { | |
1342 | __u32 vol = super->current_vol; | |
1343 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1344 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1345 | } | |
1346 | sha1_finish_ctx(&ctx, buf); | |
1347 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1348 | } |
1349 | ||
0d481d37 | 1350 | #if 0 |
4f5bc454 DW |
1351 | static void |
1352 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1353 | { |
cdddbdbc DW |
1354 | __u8 *v = get_imsm_version(mpb); |
1355 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1356 | char major[] = { 0, 0, 0 }; | |
1357 | char minor[] = { 0 ,0, 0 }; | |
1358 | char patch[] = { 0, 0, 0 }; | |
1359 | char *ver_parse[] = { major, minor, patch }; | |
1360 | int i, j; | |
1361 | ||
1362 | i = j = 0; | |
1363 | while (*v != '\0' && v < end) { | |
1364 | if (*v != '.' && j < 2) | |
1365 | ver_parse[i][j++] = *v; | |
1366 | else { | |
1367 | i++; | |
1368 | j = 0; | |
1369 | } | |
1370 | v++; | |
1371 | } | |
1372 | ||
4f5bc454 DW |
1373 | *m = strtol(minor, NULL, 0); |
1374 | *p = strtol(patch, NULL, 0); | |
1375 | } | |
0d481d37 | 1376 | #endif |
4f5bc454 | 1377 | |
1e5c6983 DW |
1378 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1379 | { | |
1380 | /* migr_strip_size when repairing or initializing parity */ | |
1381 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1382 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1383 | ||
1384 | switch (get_imsm_raid_level(map)) { | |
1385 | case 5: | |
1386 | case 10: | |
1387 | return chunk; | |
1388 | default: | |
1389 | return 128*1024 >> 9; | |
1390 | } | |
1391 | } | |
1392 | ||
1393 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1394 | { | |
1395 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1396 | * this is different than migr_strip_size_resync(), but it's good | |
1397 | * to be compatible | |
1398 | */ | |
1399 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1400 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1401 | ||
1402 | switch (get_imsm_raid_level(map)) { | |
1403 | case 1: | |
1404 | case 10: | |
1405 | if (map->num_members % map->num_domains == 0) | |
1406 | return 128*1024 >> 9; | |
1407 | else | |
1408 | return chunk; | |
1409 | case 5: | |
1410 | return max((__u32) 64*1024 >> 9, chunk); | |
1411 | default: | |
1412 | return 128*1024 >> 9; | |
1413 | } | |
1414 | } | |
1415 | ||
1416 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
1417 | { | |
1418 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1419 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1420 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); | |
1421 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
1422 | ||
1423 | return max((__u32) 1, hi_chunk / lo_chunk); | |
1424 | } | |
1425 | ||
1426 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
1427 | { | |
1428 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1429 | int level = get_imsm_raid_level(lo); | |
1430 | ||
1431 | if (level == 1 || level == 10) { | |
1432 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1433 | ||
1434 | return hi->num_domains; | |
1435 | } else | |
1436 | return num_stripes_per_unit_resync(dev); | |
1437 | } | |
1438 | ||
98130f40 | 1439 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
1440 | { |
1441 | /* named 'imsm_' because raid0, raid1 and raid10 | |
1442 | * counter-intuitively have the same number of data disks | |
1443 | */ | |
98130f40 | 1444 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
1445 | |
1446 | switch (get_imsm_raid_level(map)) { | |
1447 | case 0: | |
1448 | case 1: | |
1449 | case 10: | |
1450 | return map->num_members; | |
1451 | case 5: | |
1452 | return map->num_members - 1; | |
1453 | default: | |
1454 | dprintf("%s: unsupported raid level\n", __func__); | |
1455 | return 0; | |
1456 | } | |
1457 | } | |
1458 | ||
1459 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
1460 | { | |
1461 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1462 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1463 | ||
1464 | switch(get_imsm_raid_level(map)) { | |
1465 | case 1: | |
1466 | case 10: | |
1467 | return chunk * map->num_domains; | |
1468 | case 5: | |
1469 | return chunk * map->num_members; | |
1470 | default: | |
1471 | return chunk; | |
1472 | } | |
1473 | } | |
1474 | ||
1475 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
1476 | { | |
1477 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1478 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1479 | __u32 strip = block / chunk; | |
1480 | ||
1481 | switch (get_imsm_raid_level(map)) { | |
1482 | case 1: | |
1483 | case 10: { | |
1484 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
1485 | __u32 vol_stripe = vol_strip / map->num_members; | |
1486 | ||
1487 | return vol_stripe * chunk + block % chunk; | |
1488 | } case 5: { | |
1489 | __u32 stripe = strip / (map->num_members - 1); | |
1490 | ||
1491 | return stripe * chunk + block % chunk; | |
1492 | } | |
1493 | default: | |
1494 | return 0; | |
1495 | } | |
1496 | } | |
1497 | ||
1498 | static __u64 blocks_per_migr_unit(struct imsm_dev *dev) | |
1499 | { | |
1500 | /* calculate the conversion factor between per member 'blocks' | |
1501 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
1502 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
1503 | */ | |
1504 | if (!dev->vol.migr_state) | |
1505 | return 0; | |
1506 | ||
1507 | switch (migr_type(dev)) { | |
6345120e | 1508 | case MIGR_GEN_MIGR: |
1e5c6983 DW |
1509 | case MIGR_VERIFY: |
1510 | case MIGR_REPAIR: | |
1511 | case MIGR_INIT: { | |
1512 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1513 | __u32 stripes_per_unit; | |
1514 | __u32 blocks_per_unit; | |
1515 | __u32 parity_depth; | |
1516 | __u32 migr_chunk; | |
1517 | __u32 block_map; | |
1518 | __u32 block_rel; | |
1519 | __u32 segment; | |
1520 | __u32 stripe; | |
1521 | __u8 disks; | |
1522 | ||
1523 | /* yes, this is really the translation of migr_units to | |
1524 | * per-member blocks in the 'resync' case | |
1525 | */ | |
1526 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
1527 | migr_chunk = migr_strip_blocks_resync(dev); | |
98130f40 | 1528 | disks = imsm_num_data_members(dev, 0); |
1e5c6983 DW |
1529 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
1530 | stripe = __le32_to_cpu(map->blocks_per_strip) * disks; | |
1531 | segment = blocks_per_unit / stripe; | |
1532 | block_rel = blocks_per_unit - segment * stripe; | |
1533 | parity_depth = parity_segment_depth(dev); | |
1534 | block_map = map_migr_block(dev, block_rel); | |
1535 | return block_map + parity_depth * segment; | |
1536 | } | |
1537 | case MIGR_REBUILD: { | |
1538 | __u32 stripes_per_unit; | |
1539 | __u32 migr_chunk; | |
1540 | ||
1541 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
1542 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
1543 | return migr_chunk * stripes_per_unit; | |
1544 | } | |
1e5c6983 DW |
1545 | case MIGR_STATE_CHANGE: |
1546 | default: | |
1547 | return 0; | |
1548 | } | |
1549 | } | |
1550 | ||
c2c087e6 DW |
1551 | static int imsm_level_to_layout(int level) |
1552 | { | |
1553 | switch (level) { | |
1554 | case 0: | |
1555 | case 1: | |
1556 | return 0; | |
1557 | case 5: | |
1558 | case 6: | |
a380c027 | 1559 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 1560 | case 10: |
c92a2527 | 1561 | return 0x102; |
c2c087e6 | 1562 | } |
a18a888e | 1563 | return UnSet; |
c2c087e6 DW |
1564 | } |
1565 | ||
a5d85af7 | 1566 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
1567 | { |
1568 | struct intel_super *super = st->sb; | |
949c47a0 | 1569 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 1570 | struct imsm_map *map = get_imsm_map(dev, 0); |
81ac8b4d | 1571 | struct imsm_map *prev_map = get_imsm_map(dev, 1); |
b335e593 | 1572 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 1573 | struct dl *dl; |
e207da2f | 1574 | char *devname; |
a5d85af7 | 1575 | int map_disks = info->array.raid_disks; |
bf5a934a | 1576 | |
b335e593 AK |
1577 | if (prev_map) |
1578 | map_to_analyse = prev_map; | |
1579 | ||
efb30e7f DW |
1580 | for (dl = super->disks; dl; dl = dl->next) |
1581 | if (dl->raiddisk == info->disk.raid_disk) | |
1582 | break; | |
bf5a934a | 1583 | info->container_member = super->current_vol; |
b335e593 AK |
1584 | info->array.raid_disks = map_to_analyse->num_members; |
1585 | info->array.level = get_imsm_raid_level(map_to_analyse); | |
bf5a934a DW |
1586 | info->array.layout = imsm_level_to_layout(info->array.level); |
1587 | info->array.md_minor = -1; | |
1588 | info->array.ctime = 0; | |
1589 | info->array.utime = 0; | |
b335e593 AK |
1590 | info->array.chunk_size = |
1591 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 1592 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
1593 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
1594 | info->custom_array_size <<= 32; | |
1595 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
b335e593 AK |
1596 | if (prev_map) { |
1597 | info->new_level = get_imsm_raid_level(map); | |
1598 | info->new_layout = imsm_level_to_layout(info->new_level); | |
1599 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
1600 | } else { | |
1601 | info->new_level = UnSet; | |
1602 | info->new_layout = UnSet; | |
1603 | info->new_chunk = info->array.chunk_size; | |
1604 | } | |
301406c9 DW |
1605 | info->disk.major = 0; |
1606 | info->disk.minor = 0; | |
efb30e7f DW |
1607 | if (dl) { |
1608 | info->disk.major = dl->major; | |
1609 | info->disk.minor = dl->minor; | |
1610 | } | |
bf5a934a | 1611 | |
b335e593 AK |
1612 | info->data_offset = __le32_to_cpu(map_to_analyse->pba_of_lba0); |
1613 | info->component_size = | |
1614 | __le32_to_cpu(map_to_analyse->blocks_per_member); | |
301406c9 | 1615 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 1616 | info->recovery_start = MaxSector; |
81ac8b4d AK |
1617 | info->reshape_active = (prev_map != NULL); |
1618 | if (info->reshape_active) | |
1619 | info->delta_disks = map->num_members - prev_map->num_members; | |
1620 | else | |
1621 | info->delta_disks = 0; | |
bf5a934a | 1622 | |
b335e593 AK |
1623 | if (map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
1624 | dev->vol.dirty) { | |
301406c9 | 1625 | info->resync_start = 0; |
1e5c6983 DW |
1626 | } else if (dev->vol.migr_state) { |
1627 | switch (migr_type(dev)) { | |
1628 | case MIGR_REPAIR: | |
1629 | case MIGR_INIT: { | |
1630 | __u64 blocks_per_unit = blocks_per_migr_unit(dev); | |
1631 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); | |
1632 | ||
1633 | info->resync_start = blocks_per_unit * units; | |
1634 | break; | |
1635 | } | |
1636 | case MIGR_VERIFY: | |
1637 | /* we could emulate the checkpointing of | |
1638 | * 'sync_action=check' migrations, but for now | |
1639 | * we just immediately complete them | |
1640 | */ | |
1641 | case MIGR_REBUILD: | |
1642 | /* this is handled by container_content_imsm() */ | |
1643 | case MIGR_GEN_MIGR: | |
1644 | case MIGR_STATE_CHANGE: | |
1645 | /* FIXME handle other migrations */ | |
1646 | default: | |
1647 | /* we are not dirty, so... */ | |
1648 | info->resync_start = MaxSector; | |
1649 | } | |
1650 | } else | |
b7528a20 | 1651 | info->resync_start = MaxSector; |
301406c9 DW |
1652 | |
1653 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
1654 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 1655 | |
f35f2525 N |
1656 | info->array.major_version = -1; |
1657 | info->array.minor_version = -2; | |
e207da2f AW |
1658 | devname = devnum2devname(st->container_dev); |
1659 | *info->text_version = '\0'; | |
1660 | if (devname) | |
1661 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
1662 | free(devname); | |
a67dd8cc | 1663 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 1664 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
1665 | |
1666 | if (dmap) { | |
1667 | int i, j; | |
1668 | for (i=0; i<map_disks; i++) { | |
1669 | dmap[i] = 0; | |
1670 | if (i < info->array.raid_disks) { | |
1671 | struct imsm_disk *dsk; | |
98130f40 | 1672 | j = get_imsm_disk_idx(dev, i, -1); |
a5d85af7 N |
1673 | dsk = get_imsm_disk(super, j); |
1674 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
1675 | dmap[i] = 1; | |
1676 | } | |
1677 | } | |
1678 | } | |
81ac8b4d | 1679 | } |
bf5a934a | 1680 | |
97b4d0e9 DW |
1681 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed); |
1682 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev); | |
1683 | ||
1684 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
1685 | { | |
1686 | struct dl *d; | |
1687 | ||
1688 | for (d = super->missing; d; d = d->next) | |
1689 | if (d->index == index) | |
1690 | return &d->disk; | |
1691 | return NULL; | |
1692 | } | |
1693 | ||
a5d85af7 | 1694 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
1695 | { |
1696 | struct intel_super *super = st->sb; | |
4f5bc454 | 1697 | struct imsm_disk *disk; |
a5d85af7 | 1698 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
1699 | int max_enough = -1; |
1700 | int i; | |
1701 | struct imsm_super *mpb; | |
4f5bc454 | 1702 | |
bf5a934a | 1703 | if (super->current_vol >= 0) { |
a5d85af7 | 1704 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
1705 | return; |
1706 | } | |
d23fe947 DW |
1707 | |
1708 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
1709 | * spares | |
1710 | */ | |
1711 | info->array.raid_disks = 0; | |
cdddbdbc DW |
1712 | info->array.level = LEVEL_CONTAINER; |
1713 | info->array.layout = 0; | |
1714 | info->array.md_minor = -1; | |
c2c087e6 | 1715 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
1716 | info->array.utime = 0; |
1717 | info->array.chunk_size = 0; | |
1718 | ||
1719 | info->disk.major = 0; | |
1720 | info->disk.minor = 0; | |
cdddbdbc | 1721 | info->disk.raid_disk = -1; |
c2c087e6 | 1722 | info->reshape_active = 0; |
f35f2525 N |
1723 | info->array.major_version = -1; |
1724 | info->array.minor_version = -2; | |
c2c087e6 | 1725 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 1726 | info->safe_mode_delay = 0; |
c2c087e6 DW |
1727 | info->disk.number = -1; |
1728 | info->disk.state = 0; | |
c5afc314 | 1729 | info->name[0] = 0; |
921d9e16 | 1730 | info->recovery_start = MaxSector; |
c2c087e6 | 1731 | |
97b4d0e9 | 1732 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 1733 | mpb = super->anchor; |
97b4d0e9 | 1734 | |
ab3cb6b3 N |
1735 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1736 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1737 | int failed, enough, j, missing = 0; | |
1738 | struct imsm_map *map; | |
1739 | __u8 state; | |
97b4d0e9 | 1740 | |
ab3cb6b3 N |
1741 | failed = imsm_count_failed(super, dev); |
1742 | state = imsm_check_degraded(super, dev, failed); | |
1743 | map = get_imsm_map(dev, dev->vol.migr_state); | |
1744 | ||
1745 | /* any newly missing disks? | |
1746 | * (catches single-degraded vs double-degraded) | |
1747 | */ | |
1748 | for (j = 0; j < map->num_members; j++) { | |
98130f40 | 1749 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
ab3cb6b3 N |
1750 | __u32 idx = ord_to_idx(ord); |
1751 | ||
1752 | if (!(ord & IMSM_ORD_REBUILD) && | |
1753 | get_imsm_missing(super, idx)) { | |
1754 | missing = 1; | |
1755 | break; | |
1756 | } | |
97b4d0e9 | 1757 | } |
ab3cb6b3 N |
1758 | |
1759 | if (state == IMSM_T_STATE_FAILED) | |
1760 | enough = -1; | |
1761 | else if (state == IMSM_T_STATE_DEGRADED && | |
1762 | (state != map->map_state || missing)) | |
1763 | enough = 0; | |
1764 | else /* we're normal, or already degraded */ | |
1765 | enough = 1; | |
1766 | ||
1767 | /* in the missing/failed disk case check to see | |
1768 | * if at least one array is runnable | |
1769 | */ | |
1770 | max_enough = max(max_enough, enough); | |
1771 | } | |
1772 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
1773 | info->container_enough = max_enough; | |
97b4d0e9 | 1774 | |
4a04ec6c | 1775 | if (super->disks) { |
14e8215b DW |
1776 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
1777 | ||
b9f594fe | 1778 | disk = &super->disks->disk; |
14e8215b DW |
1779 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
1780 | info->component_size = reserved; | |
25ed7e59 | 1781 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
1782 | /* we don't change info->disk.raid_disk here because |
1783 | * this state will be finalized in mdmon after we have | |
1784 | * found the 'most fresh' version of the metadata | |
1785 | */ | |
25ed7e59 DW |
1786 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
1787 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 1788 | } |
a575e2a7 DW |
1789 | |
1790 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
1791 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
1792 | */ | |
1793 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 1794 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
1795 | else |
1796 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
1797 | |
1798 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
1799 | if (map) { | |
1800 | int i; | |
1801 | for (i = 0; i < map_disks; i++) | |
1802 | map[i] = 1; | |
1803 | } | |
1804 | ||
cdddbdbc DW |
1805 | } |
1806 | ||
5c4cd5da AC |
1807 | /* allocates memory and fills disk in mdinfo structure |
1808 | * for each disk in array */ | |
1809 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
1810 | { | |
1811 | struct mdinfo *mddev = NULL; | |
1812 | struct intel_super *super = st->sb; | |
1813 | struct imsm_disk *disk; | |
1814 | int count = 0; | |
1815 | struct dl *dl; | |
1816 | if (!super || !super->disks) | |
1817 | return NULL; | |
1818 | dl = super->disks; | |
1819 | mddev = malloc(sizeof(*mddev)); | |
1820 | if (!mddev) { | |
1821 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
1822 | return NULL; | |
1823 | } | |
1824 | memset(mddev, 0, sizeof(*mddev)); | |
1825 | while (dl) { | |
1826 | struct mdinfo *tmp; | |
1827 | disk = &dl->disk; | |
1828 | tmp = malloc(sizeof(*tmp)); | |
1829 | if (!tmp) { | |
1830 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
1831 | if (mddev) | |
1832 | sysfs_free(mddev); | |
1833 | return NULL; | |
1834 | } | |
1835 | memset(tmp, 0, sizeof(*tmp)); | |
1836 | if (mddev->devs) | |
1837 | tmp->next = mddev->devs; | |
1838 | mddev->devs = tmp; | |
1839 | tmp->disk.number = count++; | |
1840 | tmp->disk.major = dl->major; | |
1841 | tmp->disk.minor = dl->minor; | |
1842 | tmp->disk.state = is_configured(disk) ? | |
1843 | (1 << MD_DISK_ACTIVE) : 0; | |
1844 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
1845 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
1846 | tmp->disk.raid_disk = -1; | |
1847 | dl = dl->next; | |
1848 | } | |
1849 | return mddev; | |
1850 | } | |
1851 | ||
cdddbdbc DW |
1852 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
1853 | char *update, char *devname, int verbose, | |
1854 | int uuid_set, char *homehost) | |
1855 | { | |
f352c545 DW |
1856 | /* For 'assemble' and 'force' we need to return non-zero if any |
1857 | * change was made. For others, the return value is ignored. | |
1858 | * Update options are: | |
1859 | * force-one : This device looks a bit old but needs to be included, | |
1860 | * update age info appropriately. | |
1861 | * assemble: clear any 'faulty' flag to allow this device to | |
1862 | * be assembled. | |
1863 | * force-array: Array is degraded but being forced, mark it clean | |
1864 | * if that will be needed to assemble it. | |
1865 | * | |
1866 | * newdev: not used ???? | |
1867 | * grow: Array has gained a new device - this is currently for | |
1868 | * linear only | |
1869 | * resync: mark as dirty so a resync will happen. | |
1870 | * name: update the name - preserving the homehost | |
6e46bf34 | 1871 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
1872 | * |
1873 | * Following are not relevant for this imsm: | |
1874 | * sparc2.2 : update from old dodgey metadata | |
1875 | * super-minor: change the preferred_minor number | |
1876 | * summaries: update redundant counters. | |
f352c545 DW |
1877 | * homehost: update the recorded homehost |
1878 | * _reshape_progress: record new reshape_progress position. | |
1879 | */ | |
6e46bf34 DW |
1880 | int rv = 1; |
1881 | struct intel_super *super = st->sb; | |
1882 | struct imsm_super *mpb; | |
f352c545 | 1883 | |
6e46bf34 DW |
1884 | /* we can only update container info */ |
1885 | if (!super || super->current_vol >= 0 || !super->anchor) | |
1886 | return 1; | |
1887 | ||
1888 | mpb = super->anchor; | |
1889 | ||
1890 | if (strcmp(update, "uuid") == 0 && uuid_set && !info->update_private) | |
1e2b2765 | 1891 | rv = -1; |
6e46bf34 DW |
1892 | else if (strcmp(update, "uuid") == 0 && uuid_set && info->update_private) { |
1893 | mpb->orig_family_num = *((__u32 *) info->update_private); | |
1894 | rv = 0; | |
1895 | } else if (strcmp(update, "uuid") == 0) { | |
1896 | __u32 *new_family = malloc(sizeof(*new_family)); | |
1897 | ||
1898 | /* update orig_family_number with the incoming random | |
1899 | * data, report the new effective uuid, and store the | |
1900 | * new orig_family_num for future updates. | |
1901 | */ | |
1902 | if (new_family) { | |
1903 | memcpy(&mpb->orig_family_num, info->uuid, sizeof(__u32)); | |
1904 | uuid_from_super_imsm(st, info->uuid); | |
1905 | *new_family = mpb->orig_family_num; | |
1906 | info->update_private = new_family; | |
1907 | rv = 0; | |
1908 | } | |
1909 | } else if (strcmp(update, "assemble") == 0) | |
1910 | rv = 0; | |
1911 | else | |
1e2b2765 | 1912 | rv = -1; |
f352c545 | 1913 | |
6e46bf34 DW |
1914 | /* successful update? recompute checksum */ |
1915 | if (rv == 0) | |
1916 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
1917 | |
1918 | return rv; | |
cdddbdbc DW |
1919 | } |
1920 | ||
c2c087e6 | 1921 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 1922 | { |
c2c087e6 | 1923 | size_t size; |
cdddbdbc | 1924 | |
c2c087e6 DW |
1925 | size = sizeof(struct imsm_super); |
1926 | size += (disks - 1) * sizeof(struct imsm_disk); | |
1927 | size += 2 * sizeof(struct imsm_dev); | |
1928 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
1929 | size += (4 - 2) * sizeof(struct imsm_map); | |
1930 | /* 4 possible disk_ord_tbl's */ | |
1931 | size += 4 * (disks - 1) * sizeof(__u32); | |
1932 | ||
1933 | return size; | |
1934 | } | |
1935 | ||
1936 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
1937 | { | |
1938 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
1939 | return 0; | |
1940 | ||
1941 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
1942 | } |
1943 | ||
ba2de7ba DW |
1944 | static void free_devlist(struct intel_super *super) |
1945 | { | |
1946 | struct intel_dev *dv; | |
1947 | ||
1948 | while (super->devlist) { | |
1949 | dv = super->devlist->next; | |
1950 | free(super->devlist->dev); | |
1951 | free(super->devlist); | |
1952 | super->devlist = dv; | |
1953 | } | |
1954 | } | |
1955 | ||
1956 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
1957 | { | |
1958 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
1959 | } | |
1960 | ||
cdddbdbc DW |
1961 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
1962 | { | |
1963 | /* | |
1964 | * return: | |
1965 | * 0 same, or first was empty, and second was copied | |
1966 | * 1 second had wrong number | |
1967 | * 2 wrong uuid | |
1968 | * 3 wrong other info | |
1969 | */ | |
1970 | struct intel_super *first = st->sb; | |
1971 | struct intel_super *sec = tst->sb; | |
1972 | ||
1973 | if (!first) { | |
1974 | st->sb = tst->sb; | |
1975 | tst->sb = NULL; | |
1976 | return 0; | |
1977 | } | |
1978 | ||
d23fe947 DW |
1979 | /* if an anchor does not have num_raid_devs set then it is a free |
1980 | * floating spare | |
1981 | */ | |
1982 | if (first->anchor->num_raid_devs > 0 && | |
1983 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
1984 | /* Determine if these disks might ever have been |
1985 | * related. Further disambiguation can only take place | |
1986 | * in load_super_imsm_all | |
1987 | */ | |
1988 | __u32 first_family = first->anchor->orig_family_num; | |
1989 | __u32 sec_family = sec->anchor->orig_family_num; | |
1990 | ||
f796af5d DW |
1991 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
1992 | MAX_SIGNATURE_LENGTH) != 0) | |
1993 | return 3; | |
1994 | ||
a2b97981 DW |
1995 | if (first_family == 0) |
1996 | first_family = first->anchor->family_num; | |
1997 | if (sec_family == 0) | |
1998 | sec_family = sec->anchor->family_num; | |
1999 | ||
2000 | if (first_family != sec_family) | |
d23fe947 | 2001 | return 3; |
f796af5d | 2002 | |
d23fe947 | 2003 | } |
cdddbdbc | 2004 | |
f796af5d | 2005 | |
3e372e5a DW |
2006 | /* if 'first' is a spare promote it to a populated mpb with sec's |
2007 | * family number | |
2008 | */ | |
2009 | if (first->anchor->num_raid_devs == 0 && | |
2010 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 2011 | int i; |
ba2de7ba DW |
2012 | struct intel_dev *dv; |
2013 | struct imsm_dev *dev; | |
78d30f94 DW |
2014 | |
2015 | /* we need to copy raid device info from sec if an allocation | |
2016 | * fails here we don't associate the spare | |
2017 | */ | |
2018 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
2019 | dv = malloc(sizeof(*dv)); |
2020 | if (!dv) | |
2021 | break; | |
2022 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
2023 | if (!dev) { | |
2024 | free(dv); | |
2025 | break; | |
78d30f94 | 2026 | } |
ba2de7ba DW |
2027 | dv->dev = dev; |
2028 | dv->index = i; | |
2029 | dv->next = first->devlist; | |
2030 | first->devlist = dv; | |
78d30f94 | 2031 | } |
709743c5 | 2032 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
2033 | /* allocation failure */ |
2034 | free_devlist(first); | |
2035 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
2036 | return 3; | |
78d30f94 | 2037 | } |
3e372e5a | 2038 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 2039 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 2040 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 2041 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
2042 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
2043 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
2044 | } |
2045 | ||
cdddbdbc DW |
2046 | return 0; |
2047 | } | |
2048 | ||
0030e8d6 DW |
2049 | static void fd2devname(int fd, char *name) |
2050 | { | |
2051 | struct stat st; | |
2052 | char path[256]; | |
33a6535d | 2053 | char dname[PATH_MAX]; |
0030e8d6 DW |
2054 | char *nm; |
2055 | int rv; | |
2056 | ||
2057 | name[0] = '\0'; | |
2058 | if (fstat(fd, &st) != 0) | |
2059 | return; | |
2060 | sprintf(path, "/sys/dev/block/%d:%d", | |
2061 | major(st.st_rdev), minor(st.st_rdev)); | |
2062 | ||
2063 | rv = readlink(path, dname, sizeof(dname)); | |
2064 | if (rv <= 0) | |
2065 | return; | |
2066 | ||
2067 | dname[rv] = '\0'; | |
2068 | nm = strrchr(dname, '/'); | |
2069 | nm++; | |
2070 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
2071 | } | |
2072 | ||
cdddbdbc DW |
2073 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
2074 | ||
2075 | static int imsm_read_serial(int fd, char *devname, | |
2076 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
2077 | { | |
2078 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
2079 | int rv; |
2080 | int rsp_len; | |
1f24f035 | 2081 | int len; |
316e2bf4 DW |
2082 | char *dest; |
2083 | char *src; | |
2084 | char *rsp_buf; | |
2085 | int i; | |
cdddbdbc DW |
2086 | |
2087 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 2088 | |
f9ba0ff1 DW |
2089 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
2090 | ||
40ebbb9c | 2091 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
2092 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
2093 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
2094 | return 0; |
2095 | } | |
2096 | ||
cdddbdbc DW |
2097 | if (rv != 0) { |
2098 | if (devname) | |
2099 | fprintf(stderr, | |
2100 | Name ": Failed to retrieve serial for %s\n", | |
2101 | devname); | |
2102 | return rv; | |
2103 | } | |
2104 | ||
2105 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
2106 | if (!rsp_len) { |
2107 | if (devname) | |
2108 | fprintf(stderr, | |
2109 | Name ": Failed to retrieve serial for %s\n", | |
2110 | devname); | |
2111 | return 2; | |
2112 | } | |
1f24f035 | 2113 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 2114 | |
316e2bf4 DW |
2115 | /* trim all whitespace and non-printable characters and convert |
2116 | * ':' to ';' | |
2117 | */ | |
2118 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
2119 | src = &rsp_buf[i]; | |
2120 | if (*src > 0x20) { | |
2121 | /* ':' is reserved for use in placeholder serial | |
2122 | * numbers for missing disks | |
2123 | */ | |
2124 | if (*src == ':') | |
2125 | *dest++ = ';'; | |
2126 | else | |
2127 | *dest++ = *src; | |
2128 | } | |
2129 | } | |
2130 | len = dest - rsp_buf; | |
2131 | dest = rsp_buf; | |
2132 | ||
2133 | /* truncate leading characters */ | |
2134 | if (len > MAX_RAID_SERIAL_LEN) { | |
2135 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 2136 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 2137 | } |
5c3db629 | 2138 | |
5c3db629 | 2139 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 2140 | memcpy(serial, dest, len); |
cdddbdbc DW |
2141 | |
2142 | return 0; | |
2143 | } | |
2144 | ||
1f24f035 DW |
2145 | static int serialcmp(__u8 *s1, __u8 *s2) |
2146 | { | |
2147 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
2148 | } | |
2149 | ||
2150 | static void serialcpy(__u8 *dest, __u8 *src) | |
2151 | { | |
2152 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
2153 | } | |
2154 | ||
1799c9e8 | 2155 | #ifndef MDASSEMBLE |
54c2c1ea DW |
2156 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
2157 | { | |
2158 | struct dl *dl; | |
2159 | ||
2160 | for (dl = super->disks; dl; dl = dl->next) | |
2161 | if (serialcmp(dl->serial, serial) == 0) | |
2162 | break; | |
2163 | ||
2164 | return dl; | |
2165 | } | |
1799c9e8 | 2166 | #endif |
54c2c1ea | 2167 | |
a2b97981 DW |
2168 | static struct imsm_disk * |
2169 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
2170 | { | |
2171 | int i; | |
2172 | ||
2173 | for (i = 0; i < mpb->num_disks; i++) { | |
2174 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
2175 | ||
2176 | if (serialcmp(disk->serial, serial) == 0) { | |
2177 | if (idx) | |
2178 | *idx = i; | |
2179 | return disk; | |
2180 | } | |
2181 | } | |
2182 | ||
2183 | return NULL; | |
2184 | } | |
2185 | ||
cdddbdbc DW |
2186 | static int |
2187 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
2188 | { | |
a2b97981 | 2189 | struct imsm_disk *disk; |
cdddbdbc DW |
2190 | struct dl *dl; |
2191 | struct stat stb; | |
cdddbdbc | 2192 | int rv; |
a2b97981 | 2193 | char name[40]; |
d23fe947 DW |
2194 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
2195 | ||
2196 | rv = imsm_read_serial(fd, devname, serial); | |
2197 | ||
2198 | if (rv != 0) | |
2199 | return 2; | |
2200 | ||
a2b97981 | 2201 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 2202 | if (!dl) { |
cdddbdbc DW |
2203 | if (devname) |
2204 | fprintf(stderr, | |
2205 | Name ": failed to allocate disk buffer for %s\n", | |
2206 | devname); | |
2207 | return 2; | |
2208 | } | |
cdddbdbc | 2209 | |
a2b97981 DW |
2210 | fstat(fd, &stb); |
2211 | dl->major = major(stb.st_rdev); | |
2212 | dl->minor = minor(stb.st_rdev); | |
2213 | dl->next = super->disks; | |
2214 | dl->fd = keep_fd ? fd : -1; | |
2215 | assert(super->disks == NULL); | |
2216 | super->disks = dl; | |
2217 | serialcpy(dl->serial, serial); | |
2218 | dl->index = -2; | |
2219 | dl->e = NULL; | |
2220 | fd2devname(fd, name); | |
2221 | if (devname) | |
2222 | dl->devname = strdup(devname); | |
2223 | else | |
2224 | dl->devname = strdup(name); | |
cdddbdbc | 2225 | |
d23fe947 | 2226 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
2227 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
2228 | if (disk) { | |
2229 | dl->disk = *disk; | |
2230 | /* only set index on disks that are a member of a | |
2231 | * populated contianer, i.e. one with raid_devs | |
2232 | */ | |
2233 | if (is_failed(&dl->disk)) | |
3f6efecc | 2234 | dl->index = -2; |
a2b97981 DW |
2235 | else if (is_spare(&dl->disk)) |
2236 | dl->index = -1; | |
3f6efecc DW |
2237 | } |
2238 | ||
949c47a0 DW |
2239 | return 0; |
2240 | } | |
2241 | ||
0e600426 | 2242 | #ifndef MDASSEMBLE |
0c046afd DW |
2243 | /* When migrating map0 contains the 'destination' state while map1 |
2244 | * contains the current state. When not migrating map0 contains the | |
2245 | * current state. This routine assumes that map[0].map_state is set to | |
2246 | * the current array state before being called. | |
2247 | * | |
2248 | * Migration is indicated by one of the following states | |
2249 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 2250 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 2251 | * map1state=unitialized) |
1484e727 | 2252 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 2253 | * map1state=normal) |
e3bba0e0 | 2254 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd DW |
2255 | * map1state=degraded) |
2256 | */ | |
0556e1a2 | 2257 | static void migrate(struct imsm_dev *dev, __u8 to_state, int migr_type) |
3393c6af | 2258 | { |
0c046afd | 2259 | struct imsm_map *dest; |
3393c6af DW |
2260 | struct imsm_map *src = get_imsm_map(dev, 0); |
2261 | ||
0c046afd | 2262 | dev->vol.migr_state = 1; |
1484e727 | 2263 | set_migr_type(dev, migr_type); |
f8f603f1 | 2264 | dev->vol.curr_migr_unit = 0; |
0c046afd DW |
2265 | dest = get_imsm_map(dev, 1); |
2266 | ||
0556e1a2 | 2267 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 2268 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
2269 | if ((migr_type == MIGR_REBUILD) || |
2270 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
2271 | __u32 ord; |
2272 | int i; | |
2273 | ||
2274 | for (i = 0; i < src->num_members; i++) { | |
2275 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
2276 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
2277 | } | |
2278 | } | |
2279 | ||
0c046afd | 2280 | src->map_state = to_state; |
949c47a0 | 2281 | } |
f8f603f1 DW |
2282 | |
2283 | static void end_migration(struct imsm_dev *dev, __u8 map_state) | |
2284 | { | |
2285 | struct imsm_map *map = get_imsm_map(dev, 0); | |
0556e1a2 | 2286 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
28bce06f | 2287 | int i, j; |
0556e1a2 DW |
2288 | |
2289 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
2290 | * completed in the last migration. | |
2291 | * | |
28bce06f | 2292 | * FIXME add support for raid-level-migration |
0556e1a2 DW |
2293 | */ |
2294 | for (i = 0; i < prev->num_members; i++) | |
28bce06f AK |
2295 | for (j = 0; j < map->num_members; j++) |
2296 | /* during online capacity expansion | |
2297 | * disks position can be changed if takeover is used | |
2298 | */ | |
2299 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
2300 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
2301 | map->disk_ord_tbl[j] |= prev->disk_ord_tbl[i]; | |
2302 | break; | |
2303 | } | |
f8f603f1 DW |
2304 | |
2305 | dev->vol.migr_state = 0; | |
28bce06f | 2306 | dev->vol.migr_type = 0; |
f8f603f1 DW |
2307 | dev->vol.curr_migr_unit = 0; |
2308 | map->map_state = map_state; | |
2309 | } | |
0e600426 | 2310 | #endif |
949c47a0 DW |
2311 | |
2312 | static int parse_raid_devices(struct intel_super *super) | |
2313 | { | |
2314 | int i; | |
2315 | struct imsm_dev *dev_new; | |
4d7b1503 DW |
2316 | size_t len, len_migr; |
2317 | size_t space_needed = 0; | |
2318 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
2319 | |
2320 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
2321 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 2322 | struct intel_dev *dv; |
949c47a0 | 2323 | |
4d7b1503 DW |
2324 | len = sizeof_imsm_dev(dev_iter, 0); |
2325 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
2326 | if (len_migr > len) | |
2327 | space_needed += len_migr - len; | |
2328 | ||
ba2de7ba DW |
2329 | dv = malloc(sizeof(*dv)); |
2330 | if (!dv) | |
2331 | return 1; | |
4d7b1503 | 2332 | dev_new = malloc(len_migr); |
ba2de7ba DW |
2333 | if (!dev_new) { |
2334 | free(dv); | |
949c47a0 | 2335 | return 1; |
ba2de7ba | 2336 | } |
949c47a0 | 2337 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
2338 | dv->dev = dev_new; |
2339 | dv->index = i; | |
2340 | dv->next = super->devlist; | |
2341 | super->devlist = dv; | |
949c47a0 | 2342 | } |
cdddbdbc | 2343 | |
4d7b1503 DW |
2344 | /* ensure that super->buf is large enough when all raid devices |
2345 | * are migrating | |
2346 | */ | |
2347 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
2348 | void *buf; | |
2349 | ||
2350 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
2351 | if (posix_memalign(&buf, 512, len) != 0) | |
2352 | return 1; | |
2353 | ||
1f45a8ad DW |
2354 | memcpy(buf, super->buf, super->len); |
2355 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
2356 | free(super->buf); |
2357 | super->buf = buf; | |
2358 | super->len = len; | |
2359 | } | |
2360 | ||
cdddbdbc DW |
2361 | return 0; |
2362 | } | |
2363 | ||
604b746f JD |
2364 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
2365 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
2366 | { | |
2367 | void *ptr = NULL; | |
2368 | ||
2369 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
2370 | ptr = mpb; | |
2371 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
2372 | } | |
2373 | ||
2374 | return ptr; | |
2375 | } | |
2376 | ||
d23fe947 | 2377 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 2378 | |
cdddbdbc DW |
2379 | /* load_imsm_mpb - read matrix metadata |
2380 | * allocates super->mpb to be freed by free_super | |
2381 | */ | |
2382 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
2383 | { | |
2384 | unsigned long long dsize; | |
cdddbdbc DW |
2385 | unsigned long long sectors; |
2386 | struct stat; | |
6416d527 | 2387 | struct imsm_super *anchor; |
cdddbdbc DW |
2388 | __u32 check_sum; |
2389 | ||
cdddbdbc | 2390 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
2391 | if (dsize < 1024) { |
2392 | if (devname) | |
2393 | fprintf(stderr, | |
2394 | Name ": %s: device to small for imsm\n", | |
2395 | devname); | |
2396 | return 1; | |
2397 | } | |
cdddbdbc DW |
2398 | |
2399 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
2400 | if (devname) | |
2401 | fprintf(stderr, | |
2402 | Name ": Cannot seek to anchor block on %s: %s\n", | |
2403 | devname, strerror(errno)); | |
2404 | return 1; | |
2405 | } | |
2406 | ||
949c47a0 | 2407 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
2408 | if (devname) |
2409 | fprintf(stderr, | |
2410 | Name ": Failed to allocate imsm anchor buffer" | |
2411 | " on %s\n", devname); | |
2412 | return 1; | |
2413 | } | |
949c47a0 | 2414 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
2415 | if (devname) |
2416 | fprintf(stderr, | |
2417 | Name ": Cannot read anchor block on %s: %s\n", | |
2418 | devname, strerror(errno)); | |
6416d527 | 2419 | free(anchor); |
cdddbdbc DW |
2420 | return 1; |
2421 | } | |
2422 | ||
6416d527 | 2423 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
2424 | if (devname) |
2425 | fprintf(stderr, | |
2426 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 2427 | free(anchor); |
cdddbdbc DW |
2428 | return 2; |
2429 | } | |
2430 | ||
d23fe947 | 2431 | __free_imsm(super, 0); |
949c47a0 DW |
2432 | super->len = ROUND_UP(anchor->mpb_size, 512); |
2433 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
2434 | if (devname) |
2435 | fprintf(stderr, | |
2436 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 2437 | super->len); |
6416d527 | 2438 | free(anchor); |
cdddbdbc DW |
2439 | return 2; |
2440 | } | |
949c47a0 | 2441 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 2442 | |
6416d527 NB |
2443 | sectors = mpb_sectors(anchor) - 1; |
2444 | free(anchor); | |
949c47a0 | 2445 | if (!sectors) { |
ecf45690 DW |
2446 | check_sum = __gen_imsm_checksum(super->anchor); |
2447 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
2448 | if (devname) | |
2449 | fprintf(stderr, | |
2450 | Name ": IMSM checksum %x != %x on %s\n", | |
2451 | check_sum, | |
2452 | __le32_to_cpu(super->anchor->check_sum), | |
2453 | devname); | |
2454 | return 2; | |
2455 | } | |
2456 | ||
a2b97981 | 2457 | return 0; |
949c47a0 | 2458 | } |
cdddbdbc DW |
2459 | |
2460 | /* read the extended mpb */ | |
2461 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
2462 | if (devname) | |
2463 | fprintf(stderr, | |
2464 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
2465 | devname, strerror(errno)); | |
2466 | return 1; | |
2467 | } | |
2468 | ||
f21e18ca | 2469 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
2470 | if (devname) |
2471 | fprintf(stderr, | |
2472 | Name ": Cannot read extended mpb on %s: %s\n", | |
2473 | devname, strerror(errno)); | |
2474 | return 2; | |
2475 | } | |
2476 | ||
949c47a0 DW |
2477 | check_sum = __gen_imsm_checksum(super->anchor); |
2478 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
2479 | if (devname) |
2480 | fprintf(stderr, | |
2481 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 2482 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 2483 | devname); |
db575f3b | 2484 | return 3; |
cdddbdbc DW |
2485 | } |
2486 | ||
604b746f JD |
2487 | /* FIXME the BBM log is disk specific so we cannot use this global |
2488 | * buffer for all disks. Ok for now since we only look at the global | |
2489 | * bbm_log_size parameter to gate assembly | |
2490 | */ | |
2491 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
2492 | ||
a2b97981 DW |
2493 | return 0; |
2494 | } | |
2495 | ||
2496 | static int | |
2497 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
2498 | { | |
2499 | int err; | |
2500 | ||
2501 | err = load_imsm_mpb(fd, super, devname); | |
2502 | if (err) | |
2503 | return err; | |
2504 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
2505 | if (err) | |
2506 | return err; | |
2507 | err = parse_raid_devices(super); | |
4d7b1503 | 2508 | |
a2b97981 | 2509 | return err; |
cdddbdbc DW |
2510 | } |
2511 | ||
ae6aad82 DW |
2512 | static void __free_imsm_disk(struct dl *d) |
2513 | { | |
2514 | if (d->fd >= 0) | |
2515 | close(d->fd); | |
2516 | if (d->devname) | |
2517 | free(d->devname); | |
0dcecb2e DW |
2518 | if (d->e) |
2519 | free(d->e); | |
ae6aad82 DW |
2520 | free(d); |
2521 | ||
2522 | } | |
1a64be56 | 2523 | |
cdddbdbc DW |
2524 | static void free_imsm_disks(struct intel_super *super) |
2525 | { | |
47ee5a45 | 2526 | struct dl *d; |
cdddbdbc | 2527 | |
47ee5a45 DW |
2528 | while (super->disks) { |
2529 | d = super->disks; | |
cdddbdbc | 2530 | super->disks = d->next; |
ae6aad82 | 2531 | __free_imsm_disk(d); |
cdddbdbc | 2532 | } |
47ee5a45 DW |
2533 | while (super->missing) { |
2534 | d = super->missing; | |
2535 | super->missing = d->next; | |
2536 | __free_imsm_disk(d); | |
2537 | } | |
2538 | ||
cdddbdbc DW |
2539 | } |
2540 | ||
9ca2c81c | 2541 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 2542 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 2543 | { |
9ca2c81c | 2544 | if (super->buf) { |
949c47a0 | 2545 | free(super->buf); |
9ca2c81c DW |
2546 | super->buf = NULL; |
2547 | } | |
d23fe947 DW |
2548 | if (free_disks) |
2549 | free_imsm_disks(super); | |
ba2de7ba | 2550 | free_devlist(super); |
88c32bb1 DW |
2551 | if (super->hba) { |
2552 | free((void *) super->hba); | |
2553 | super->hba = NULL; | |
2554 | } | |
cdddbdbc DW |
2555 | } |
2556 | ||
9ca2c81c DW |
2557 | static void free_imsm(struct intel_super *super) |
2558 | { | |
d23fe947 | 2559 | __free_imsm(super, 1); |
9ca2c81c DW |
2560 | free(super); |
2561 | } | |
cdddbdbc DW |
2562 | |
2563 | static void free_super_imsm(struct supertype *st) | |
2564 | { | |
2565 | struct intel_super *super = st->sb; | |
2566 | ||
2567 | if (!super) | |
2568 | return; | |
2569 | ||
2570 | free_imsm(super); | |
2571 | st->sb = NULL; | |
2572 | } | |
2573 | ||
49133e57 | 2574 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
2575 | { |
2576 | struct intel_super *super = malloc(sizeof(*super)); | |
2577 | ||
2578 | if (super) { | |
2579 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 2580 | super->current_vol = -1; |
0dcecb2e | 2581 | super->create_offset = ~((__u32 ) 0); |
88c32bb1 DW |
2582 | if (!check_env("IMSM_NO_PLATFORM")) |
2583 | super->orom = find_imsm_orom(); | |
cceebc67 | 2584 | if (super->orom && !check_env("IMSM_TEST_OROM")) { |
88c32bb1 DW |
2585 | struct sys_dev *list, *ent; |
2586 | ||
2587 | /* find the first intel ahci controller */ | |
2588 | list = find_driver_devices("pci", "ahci"); | |
2589 | for (ent = list; ent; ent = ent->next) | |
2590 | if (devpath_to_vendor(ent->path) == 0x8086) | |
2591 | break; | |
2592 | if (ent) { | |
2593 | super->hba = ent->path; | |
2594 | ent->path = NULL; | |
2595 | } | |
2596 | free_sys_dev(&list); | |
2597 | } | |
c2c087e6 DW |
2598 | } |
2599 | ||
2600 | return super; | |
2601 | } | |
2602 | ||
cdddbdbc | 2603 | #ifndef MDASSEMBLE |
47ee5a45 DW |
2604 | /* find_missing - helper routine for load_super_imsm_all that identifies |
2605 | * disks that have disappeared from the system. This routine relies on | |
2606 | * the mpb being uptodate, which it is at load time. | |
2607 | */ | |
2608 | static int find_missing(struct intel_super *super) | |
2609 | { | |
2610 | int i; | |
2611 | struct imsm_super *mpb = super->anchor; | |
2612 | struct dl *dl; | |
2613 | struct imsm_disk *disk; | |
47ee5a45 DW |
2614 | |
2615 | for (i = 0; i < mpb->num_disks; i++) { | |
2616 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 2617 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
2618 | if (dl) |
2619 | continue; | |
47ee5a45 DW |
2620 | |
2621 | dl = malloc(sizeof(*dl)); | |
2622 | if (!dl) | |
2623 | return 1; | |
2624 | dl->major = 0; | |
2625 | dl->minor = 0; | |
2626 | dl->fd = -1; | |
2627 | dl->devname = strdup("missing"); | |
2628 | dl->index = i; | |
2629 | serialcpy(dl->serial, disk->serial); | |
2630 | dl->disk = *disk; | |
689c9bf3 | 2631 | dl->e = NULL; |
47ee5a45 DW |
2632 | dl->next = super->missing; |
2633 | super->missing = dl; | |
2634 | } | |
2635 | ||
2636 | return 0; | |
2637 | } | |
2638 | ||
a2b97981 DW |
2639 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
2640 | { | |
2641 | struct intel_disk *idisk = disk_list; | |
2642 | ||
2643 | while (idisk) { | |
2644 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
2645 | break; | |
2646 | idisk = idisk->next; | |
2647 | } | |
2648 | ||
2649 | return idisk; | |
2650 | } | |
2651 | ||
2652 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
2653 | struct intel_super *super, | |
2654 | struct intel_disk **disk_list) | |
2655 | { | |
2656 | struct imsm_disk *d = &super->disks->disk; | |
2657 | struct imsm_super *mpb = super->anchor; | |
2658 | int i, j; | |
2659 | ||
2660 | for (i = 0; i < tbl_size; i++) { | |
2661 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
2662 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
2663 | ||
2664 | if (tbl_mpb->family_num == mpb->family_num) { | |
2665 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
2666 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
2667 | __func__, super->disks->major, | |
2668 | super->disks->minor, | |
2669 | table[i]->disks->major, | |
2670 | table[i]->disks->minor); | |
2671 | break; | |
2672 | } | |
2673 | ||
2674 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
2675 | is_configured(d) == is_configured(tbl_d)) && | |
2676 | tbl_mpb->generation_num < mpb->generation_num) { | |
2677 | /* current version of the mpb is a | |
2678 | * better candidate than the one in | |
2679 | * super_table, but copy over "cross | |
2680 | * generational" status | |
2681 | */ | |
2682 | struct intel_disk *idisk; | |
2683 | ||
2684 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
2685 | __func__, super->disks->major, | |
2686 | super->disks->minor, | |
2687 | table[i]->disks->major, | |
2688 | table[i]->disks->minor); | |
2689 | ||
2690 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
2691 | if (idisk && is_failed(&idisk->disk)) | |
2692 | tbl_d->status |= FAILED_DISK; | |
2693 | break; | |
2694 | } else { | |
2695 | struct intel_disk *idisk; | |
2696 | struct imsm_disk *disk; | |
2697 | ||
2698 | /* tbl_mpb is more up to date, but copy | |
2699 | * over cross generational status before | |
2700 | * returning | |
2701 | */ | |
2702 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
2703 | if (disk && is_failed(disk)) | |
2704 | d->status |= FAILED_DISK; | |
2705 | ||
2706 | idisk = disk_list_get(d->serial, *disk_list); | |
2707 | if (idisk) { | |
2708 | idisk->owner = i; | |
2709 | if (disk && is_configured(disk)) | |
2710 | idisk->disk.status |= CONFIGURED_DISK; | |
2711 | } | |
2712 | ||
2713 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
2714 | __func__, super->disks->major, | |
2715 | super->disks->minor, | |
2716 | table[i]->disks->major, | |
2717 | table[i]->disks->minor); | |
2718 | ||
2719 | return tbl_size; | |
2720 | } | |
2721 | } | |
2722 | } | |
2723 | ||
2724 | if (i >= tbl_size) | |
2725 | table[tbl_size++] = super; | |
2726 | else | |
2727 | table[i] = super; | |
2728 | ||
2729 | /* update/extend the merged list of imsm_disk records */ | |
2730 | for (j = 0; j < mpb->num_disks; j++) { | |
2731 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
2732 | struct intel_disk *idisk; | |
2733 | ||
2734 | idisk = disk_list_get(disk->serial, *disk_list); | |
2735 | if (idisk) { | |
2736 | idisk->disk.status |= disk->status; | |
2737 | if (is_configured(&idisk->disk) || | |
2738 | is_failed(&idisk->disk)) | |
2739 | idisk->disk.status &= ~(SPARE_DISK); | |
2740 | } else { | |
2741 | idisk = calloc(1, sizeof(*idisk)); | |
2742 | if (!idisk) | |
2743 | return -1; | |
2744 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
2745 | idisk->disk = *disk; | |
2746 | idisk->next = *disk_list; | |
2747 | *disk_list = idisk; | |
2748 | } | |
2749 | ||
2750 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
2751 | idisk->owner = i; | |
2752 | } | |
2753 | ||
2754 | return tbl_size; | |
2755 | } | |
2756 | ||
2757 | static struct intel_super * | |
2758 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
2759 | const int owner) | |
2760 | { | |
2761 | struct imsm_super *mpb = super->anchor; | |
2762 | int ok_count = 0; | |
2763 | int i; | |
2764 | ||
2765 | for (i = 0; i < mpb->num_disks; i++) { | |
2766 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
2767 | struct intel_disk *idisk; | |
2768 | ||
2769 | idisk = disk_list_get(disk->serial, disk_list); | |
2770 | if (idisk) { | |
2771 | if (idisk->owner == owner || | |
2772 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
2773 | ok_count++; | |
2774 | else | |
2775 | dprintf("%s: '%.16s' owner %d != %d\n", | |
2776 | __func__, disk->serial, idisk->owner, | |
2777 | owner); | |
2778 | } else { | |
2779 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
2780 | __func__, __le32_to_cpu(mpb->family_num), i, | |
2781 | disk->serial); | |
2782 | break; | |
2783 | } | |
2784 | } | |
2785 | ||
2786 | if (ok_count == mpb->num_disks) | |
2787 | return super; | |
2788 | return NULL; | |
2789 | } | |
2790 | ||
2791 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
2792 | { | |
2793 | struct intel_super *s; | |
2794 | ||
2795 | for (s = super_list; s; s = s->next) { | |
2796 | if (family_num != s->anchor->family_num) | |
2797 | continue; | |
2798 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
2799 | __le32_to_cpu(family_num), s->disks->devname); | |
2800 | } | |
2801 | } | |
2802 | ||
2803 | static struct intel_super * | |
2804 | imsm_thunderdome(struct intel_super **super_list, int len) | |
2805 | { | |
2806 | struct intel_super *super_table[len]; | |
2807 | struct intel_disk *disk_list = NULL; | |
2808 | struct intel_super *champion, *spare; | |
2809 | struct intel_super *s, **del; | |
2810 | int tbl_size = 0; | |
2811 | int conflict; | |
2812 | int i; | |
2813 | ||
2814 | memset(super_table, 0, sizeof(super_table)); | |
2815 | for (s = *super_list; s; s = s->next) | |
2816 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
2817 | ||
2818 | for (i = 0; i < tbl_size; i++) { | |
2819 | struct imsm_disk *d; | |
2820 | struct intel_disk *idisk; | |
2821 | struct imsm_super *mpb = super_table[i]->anchor; | |
2822 | ||
2823 | s = super_table[i]; | |
2824 | d = &s->disks->disk; | |
2825 | ||
2826 | /* 'd' must appear in merged disk list for its | |
2827 | * configuration to be valid | |
2828 | */ | |
2829 | idisk = disk_list_get(d->serial, disk_list); | |
2830 | if (idisk && idisk->owner == i) | |
2831 | s = validate_members(s, disk_list, i); | |
2832 | else | |
2833 | s = NULL; | |
2834 | ||
2835 | if (!s) | |
2836 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
2837 | __func__, mpb->family_num, | |
2838 | super_table[i]->disks->major, | |
2839 | super_table[i]->disks->minor); | |
2840 | super_table[i] = s; | |
2841 | } | |
2842 | ||
2843 | /* This is where the mdadm implementation differs from the Windows | |
2844 | * driver which has no strict concept of a container. We can only | |
2845 | * assemble one family from a container, so when returning a prodigal | |
2846 | * array member to this system the code will not be able to disambiguate | |
2847 | * the container contents that should be assembled ("foreign" versus | |
2848 | * "local"). It requires user intervention to set the orig_family_num | |
2849 | * to a new value to establish a new container. The Windows driver in | |
2850 | * this situation fixes up the volume name in place and manages the | |
2851 | * foreign array as an independent entity. | |
2852 | */ | |
2853 | s = NULL; | |
2854 | spare = NULL; | |
2855 | conflict = 0; | |
2856 | for (i = 0; i < tbl_size; i++) { | |
2857 | struct intel_super *tbl_ent = super_table[i]; | |
2858 | int is_spare = 0; | |
2859 | ||
2860 | if (!tbl_ent) | |
2861 | continue; | |
2862 | ||
2863 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
2864 | spare = tbl_ent; | |
2865 | is_spare = 1; | |
2866 | } | |
2867 | ||
2868 | if (s && !is_spare) { | |
2869 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
2870 | conflict++; | |
2871 | } else if (!s && !is_spare) | |
2872 | s = tbl_ent; | |
2873 | } | |
2874 | ||
2875 | if (!s) | |
2876 | s = spare; | |
2877 | if (!s) { | |
2878 | champion = NULL; | |
2879 | goto out; | |
2880 | } | |
2881 | champion = s; | |
2882 | ||
2883 | if (conflict) | |
2884 | fprintf(stderr, "Chose family %#x on '%s', " | |
2885 | "assemble conflicts to new container with '--update=uuid'\n", | |
2886 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
2887 | ||
2888 | /* collect all dl's onto 'champion', and update them to | |
2889 | * champion's version of the status | |
2890 | */ | |
2891 | for (s = *super_list; s; s = s->next) { | |
2892 | struct imsm_super *mpb = champion->anchor; | |
2893 | struct dl *dl = s->disks; | |
2894 | ||
2895 | if (s == champion) | |
2896 | continue; | |
2897 | ||
2898 | for (i = 0; i < mpb->num_disks; i++) { | |
2899 | struct imsm_disk *disk; | |
2900 | ||
2901 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
2902 | if (disk) { | |
2903 | dl->disk = *disk; | |
2904 | /* only set index on disks that are a member of | |
2905 | * a populated contianer, i.e. one with | |
2906 | * raid_devs | |
2907 | */ | |
2908 | if (is_failed(&dl->disk)) | |
2909 | dl->index = -2; | |
2910 | else if (is_spare(&dl->disk)) | |
2911 | dl->index = -1; | |
2912 | break; | |
2913 | } | |
2914 | } | |
2915 | ||
2916 | if (i >= mpb->num_disks) { | |
2917 | struct intel_disk *idisk; | |
2918 | ||
2919 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 2920 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
2921 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
2922 | dl->index = -1; | |
2923 | else { | |
2924 | dl->index = -2; | |
2925 | continue; | |
2926 | } | |
2927 | } | |
2928 | ||
2929 | dl->next = champion->disks; | |
2930 | champion->disks = dl; | |
2931 | s->disks = NULL; | |
2932 | } | |
2933 | ||
2934 | /* delete 'champion' from super_list */ | |
2935 | for (del = super_list; *del; ) { | |
2936 | if (*del == champion) { | |
2937 | *del = (*del)->next; | |
2938 | break; | |
2939 | } else | |
2940 | del = &(*del)->next; | |
2941 | } | |
2942 | champion->next = NULL; | |
2943 | ||
2944 | out: | |
2945 | while (disk_list) { | |
2946 | struct intel_disk *idisk = disk_list; | |
2947 | ||
2948 | disk_list = disk_list->next; | |
2949 | free(idisk); | |
2950 | } | |
2951 | ||
2952 | return champion; | |
2953 | } | |
2954 | ||
cdddbdbc | 2955 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
e1902a7b | 2956 | char *devname) |
cdddbdbc DW |
2957 | { |
2958 | struct mdinfo *sra; | |
a2b97981 DW |
2959 | struct intel_super *super_list = NULL; |
2960 | struct intel_super *super = NULL; | |
db575f3b | 2961 | int devnum = fd2devnum(fd); |
a2b97981 | 2962 | struct mdinfo *sd; |
db575f3b | 2963 | int retry; |
a2b97981 DW |
2964 | int err = 0; |
2965 | int i; | |
dab4a513 DW |
2966 | |
2967 | /* check if 'fd' an opened container */ | |
b526e52d | 2968 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
cdddbdbc DW |
2969 | if (!sra) |
2970 | return 1; | |
2971 | ||
2972 | if (sra->array.major_version != -1 || | |
2973 | sra->array.minor_version != -2 || | |
1602d52c AW |
2974 | strcmp(sra->text_version, "imsm") != 0) { |
2975 | err = 1; | |
2976 | goto error; | |
2977 | } | |
a2b97981 DW |
2978 | /* load all mpbs */ |
2979 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
49133e57 | 2980 | struct intel_super *s = alloc_super(); |
7a6ecd55 | 2981 | char nm[32]; |
a2b97981 DW |
2982 | int dfd; |
2983 | ||
2984 | err = 1; | |
2985 | if (!s) | |
2986 | goto error; | |
2987 | s->next = super_list; | |
2988 | super_list = s; | |
cdddbdbc | 2989 | |
a2b97981 | 2990 | err = 2; |
cdddbdbc | 2991 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 2992 | dfd = dev_open(nm, O_RDWR); |
a2b97981 DW |
2993 | if (dfd < 0) |
2994 | goto error; | |
2995 | ||
e1902a7b | 2996 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
db575f3b DW |
2997 | |
2998 | /* retry the load if we might have raced against mdmon */ | |
a2b97981 | 2999 | if (err == 3 && mdmon_running(devnum)) |
db575f3b DW |
3000 | for (retry = 0; retry < 3; retry++) { |
3001 | usleep(3000); | |
e1902a7b | 3002 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
a2b97981 | 3003 | if (err != 3) |
db575f3b DW |
3004 | break; |
3005 | } | |
a2b97981 DW |
3006 | if (err) |
3007 | goto error; | |
cdddbdbc DW |
3008 | } |
3009 | ||
a2b97981 DW |
3010 | /* all mpbs enter, maybe one leaves */ |
3011 | super = imsm_thunderdome(&super_list, i); | |
3012 | if (!super) { | |
3013 | err = 1; | |
3014 | goto error; | |
cdddbdbc DW |
3015 | } |
3016 | ||
47ee5a45 DW |
3017 | if (find_missing(super) != 0) { |
3018 | free_imsm(super); | |
a2b97981 DW |
3019 | err = 2; |
3020 | goto error; | |
47ee5a45 | 3021 | } |
a2b97981 DW |
3022 | err = 0; |
3023 | ||
3024 | error: | |
3025 | while (super_list) { | |
3026 | struct intel_super *s = super_list; | |
3027 | ||
3028 | super_list = super_list->next; | |
3029 | free_imsm(s); | |
3030 | } | |
1602d52c | 3031 | sysfs_free(sra); |
a2b97981 DW |
3032 | |
3033 | if (err) | |
3034 | return err; | |
f7e7067b | 3035 | |
cdddbdbc | 3036 | *sbp = super; |
db575f3b | 3037 | st->container_dev = devnum; |
a2b97981 | 3038 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 3039 | st->ss = &super_imsm; |
cdddbdbc DW |
3040 | st->minor_version = 0; |
3041 | st->max_devs = IMSM_MAX_DEVICES; | |
3042 | } | |
cdddbdbc DW |
3043 | return 0; |
3044 | } | |
2b959fbf N |
3045 | |
3046 | static int load_container_imsm(struct supertype *st, int fd, char *devname) | |
3047 | { | |
3048 | return load_super_imsm_all(st, fd, &st->sb, devname); | |
3049 | } | |
cdddbdbc DW |
3050 | #endif |
3051 | ||
3052 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
3053 | { | |
3054 | struct intel_super *super; | |
3055 | int rv; | |
3056 | ||
691c6ee1 N |
3057 | if (test_partition(fd)) |
3058 | /* IMSM not allowed on partitions */ | |
3059 | return 1; | |
3060 | ||
37424f13 DW |
3061 | free_super_imsm(st); |
3062 | ||
49133e57 | 3063 | super = alloc_super(); |
cdddbdbc DW |
3064 | if (!super) { |
3065 | fprintf(stderr, | |
3066 | Name ": malloc of %zu failed.\n", | |
3067 | sizeof(*super)); | |
3068 | return 1; | |
3069 | } | |
3070 | ||
a2b97981 | 3071 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
3072 | |
3073 | if (rv) { | |
3074 | if (devname) | |
3075 | fprintf(stderr, | |
3076 | Name ": Failed to load all information " | |
3077 | "sections on %s\n", devname); | |
3078 | free_imsm(super); | |
3079 | return rv; | |
3080 | } | |
3081 | ||
3082 | st->sb = super; | |
3083 | if (st->ss == NULL) { | |
3084 | st->ss = &super_imsm; | |
3085 | st->minor_version = 0; | |
3086 | st->max_devs = IMSM_MAX_DEVICES; | |
3087 | } | |
cdddbdbc DW |
3088 | return 0; |
3089 | } | |
3090 | ||
ef6ffade DW |
3091 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
3092 | { | |
3093 | if (info->level == 1) | |
3094 | return 128; | |
3095 | return info->chunk_size >> 9; | |
3096 | } | |
3097 | ||
ff596308 | 3098 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains) |
ef6ffade DW |
3099 | { |
3100 | __u32 num_stripes; | |
3101 | ||
3102 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
ff596308 | 3103 | num_stripes /= num_domains; |
ef6ffade DW |
3104 | |
3105 | return num_stripes; | |
3106 | } | |
3107 | ||
fcfd9599 DW |
3108 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
3109 | { | |
4025c288 DW |
3110 | if (info->level == 1) |
3111 | return info->size * 2; | |
3112 | else | |
3113 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
fcfd9599 DW |
3114 | } |
3115 | ||
4d1313e9 DW |
3116 | static void imsm_update_version_info(struct intel_super *super) |
3117 | { | |
3118 | /* update the version and attributes */ | |
3119 | struct imsm_super *mpb = super->anchor; | |
3120 | char *version; | |
3121 | struct imsm_dev *dev; | |
3122 | struct imsm_map *map; | |
3123 | int i; | |
3124 | ||
3125 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
3126 | dev = get_imsm_dev(super, i); | |
3127 | map = get_imsm_map(dev, 0); | |
3128 | if (__le32_to_cpu(dev->size_high) > 0) | |
3129 | mpb->attributes |= MPB_ATTRIB_2TB; | |
3130 | ||
3131 | /* FIXME detect when an array spans a port multiplier */ | |
3132 | #if 0 | |
3133 | mpb->attributes |= MPB_ATTRIB_PM; | |
3134 | #endif | |
3135 | ||
3136 | if (mpb->num_raid_devs > 1 || | |
3137 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
3138 | version = MPB_VERSION_ATTRIBS; | |
3139 | switch (get_imsm_raid_level(map)) { | |
3140 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
3141 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
3142 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
3143 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
3144 | } | |
3145 | } else { | |
3146 | if (map->num_members >= 5) | |
3147 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
3148 | else if (dev->status == DEV_CLONE_N_GO) | |
3149 | version = MPB_VERSION_CNG; | |
3150 | else if (get_imsm_raid_level(map) == 5) | |
3151 | version = MPB_VERSION_RAID5; | |
3152 | else if (map->num_members >= 3) | |
3153 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
3154 | else if (get_imsm_raid_level(map) == 1) | |
3155 | version = MPB_VERSION_RAID1; | |
3156 | else | |
3157 | version = MPB_VERSION_RAID0; | |
3158 | } | |
3159 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
3160 | } | |
3161 | } | |
3162 | ||
aa534678 DW |
3163 | static int check_name(struct intel_super *super, char *name, int quiet) |
3164 | { | |
3165 | struct imsm_super *mpb = super->anchor; | |
3166 | char *reason = NULL; | |
3167 | int i; | |
3168 | ||
3169 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
3170 | reason = "must be 16 characters or less"; | |
3171 | ||
3172 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
3173 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3174 | ||
3175 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
3176 | reason = "already exists"; | |
3177 | break; | |
3178 | } | |
3179 | } | |
3180 | ||
3181 | if (reason && !quiet) | |
3182 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
3183 | ||
3184 | return !reason; | |
3185 | } | |
3186 | ||
8b353278 DW |
3187 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
3188 | unsigned long long size, char *name, | |
3189 | char *homehost, int *uuid) | |
cdddbdbc | 3190 | { |
c2c087e6 DW |
3191 | /* We are creating a volume inside a pre-existing container. |
3192 | * so st->sb is already set. | |
3193 | */ | |
3194 | struct intel_super *super = st->sb; | |
949c47a0 | 3195 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 3196 | struct intel_dev *dv; |
c2c087e6 DW |
3197 | struct imsm_dev *dev; |
3198 | struct imsm_vol *vol; | |
3199 | struct imsm_map *map; | |
3200 | int idx = mpb->num_raid_devs; | |
3201 | int i; | |
3202 | unsigned long long array_blocks; | |
2c092cad | 3203 | size_t size_old, size_new; |
ff596308 | 3204 | __u32 num_data_stripes; |
cdddbdbc | 3205 | |
88c32bb1 | 3206 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 3207 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 3208 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
3209 | return 0; |
3210 | } | |
3211 | ||
2c092cad DW |
3212 | /* ensure the mpb is large enough for the new data */ |
3213 | size_old = __le32_to_cpu(mpb->mpb_size); | |
3214 | size_new = disks_to_mpb_size(info->nr_disks); | |
3215 | if (size_new > size_old) { | |
3216 | void *mpb_new; | |
3217 | size_t size_round = ROUND_UP(size_new, 512); | |
3218 | ||
3219 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
3220 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
3221 | return 0; | |
3222 | } | |
3223 | memcpy(mpb_new, mpb, size_old); | |
3224 | free(mpb); | |
3225 | mpb = mpb_new; | |
949c47a0 | 3226 | super->anchor = mpb_new; |
2c092cad DW |
3227 | mpb->mpb_size = __cpu_to_le32(size_new); |
3228 | memset(mpb_new + size_old, 0, size_round - size_old); | |
3229 | } | |
bf5a934a | 3230 | super->current_vol = idx; |
d23fe947 DW |
3231 | /* when creating the first raid device in this container set num_disks |
3232 | * to zero, i.e. delete this spare and add raid member devices in | |
3233 | * add_to_super_imsm_volume() | |
3234 | */ | |
3235 | if (super->current_vol == 0) | |
3236 | mpb->num_disks = 0; | |
5a038140 | 3237 | |
aa534678 DW |
3238 | if (!check_name(super, name, 0)) |
3239 | return 0; | |
ba2de7ba DW |
3240 | dv = malloc(sizeof(*dv)); |
3241 | if (!dv) { | |
3242 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
3243 | return 0; | |
3244 | } | |
949c47a0 DW |
3245 | dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
3246 | if (!dev) { | |
ba2de7ba | 3247 | free(dv); |
949c47a0 DW |
3248 | fprintf(stderr, Name": could not allocate raid device\n"); |
3249 | return 0; | |
3250 | } | |
c2c087e6 | 3251 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
03bcbc65 DW |
3252 | if (info->level == 1) |
3253 | array_blocks = info_to_blocks_per_member(info); | |
3254 | else | |
3255 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
3256 | info->layout, info->chunk_size, | |
3257 | info->size*2); | |
979d38be DW |
3258 | /* round array size down to closest MB */ |
3259 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
3260 | ||
c2c087e6 DW |
3261 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
3262 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
3263 | dev->status = __cpu_to_le32(0); | |
3264 | dev->reserved_blocks = __cpu_to_le32(0); | |
3265 | vol = &dev->vol; | |
3266 | vol->migr_state = 0; | |
1484e727 | 3267 | set_migr_type(dev, MIGR_INIT); |
c2c087e6 | 3268 | vol->dirty = 0; |
f8f603f1 | 3269 | vol->curr_migr_unit = 0; |
a965f303 | 3270 | map = get_imsm_map(dev, 0); |
0dcecb2e | 3271 | map->pba_of_lba0 = __cpu_to_le32(super->create_offset); |
fcfd9599 | 3272 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade | 3273 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 3274 | map->failed_disk_num = ~0; |
c2c087e6 DW |
3275 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
3276 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 3277 | map->ddf = 1; |
ef6ffade DW |
3278 | |
3279 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
3280 | free(dev); |
3281 | free(dv); | |
ef6ffade DW |
3282 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
3283 | "in a raid1 volume\n"); | |
3284 | return 0; | |
3285 | } | |
81062a36 DW |
3286 | |
3287 | map->raid_level = info->level; | |
4d1313e9 | 3288 | if (info->level == 10) { |
c2c087e6 | 3289 | map->raid_level = 1; |
4d1313e9 | 3290 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
3291 | } else if (info->level == 1) |
3292 | map->num_domains = info->raid_disks; | |
3293 | else | |
ff596308 | 3294 | map->num_domains = 1; |
81062a36 | 3295 | |
ff596308 DW |
3296 | num_data_stripes = info_to_num_data_stripes(info, map->num_domains); |
3297 | map->num_data_stripes = __cpu_to_le32(num_data_stripes); | |
ef6ffade | 3298 | |
c2c087e6 DW |
3299 | map->num_members = info->raid_disks; |
3300 | for (i = 0; i < map->num_members; i++) { | |
3301 | /* initialized in add_to_super */ | |
4eb26970 | 3302 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 3303 | } |
949c47a0 | 3304 | mpb->num_raid_devs++; |
ba2de7ba DW |
3305 | |
3306 | dv->dev = dev; | |
3307 | dv->index = super->current_vol; | |
3308 | dv->next = super->devlist; | |
3309 | super->devlist = dv; | |
c2c087e6 | 3310 | |
4d1313e9 DW |
3311 | imsm_update_version_info(super); |
3312 | ||
c2c087e6 | 3313 | return 1; |
cdddbdbc DW |
3314 | } |
3315 | ||
bf5a934a DW |
3316 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
3317 | unsigned long long size, char *name, | |
3318 | char *homehost, int *uuid) | |
3319 | { | |
3320 | /* This is primarily called by Create when creating a new array. | |
3321 | * We will then get add_to_super called for each component, and then | |
3322 | * write_init_super called to write it out to each device. | |
3323 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
3324 | * array. | |
3325 | * To create on a pre-existing array a different method will be called. | |
3326 | * This one is just for fresh drives. | |
3327 | */ | |
3328 | struct intel_super *super; | |
3329 | struct imsm_super *mpb; | |
3330 | size_t mpb_size; | |
4d1313e9 | 3331 | char *version; |
bf5a934a | 3332 | |
bf5a934a | 3333 | if (st->sb) |
e683ca88 DW |
3334 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
3335 | ||
3336 | if (info) | |
3337 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
3338 | else | |
3339 | mpb_size = 512; | |
bf5a934a | 3340 | |
49133e57 | 3341 | super = alloc_super(); |
e683ca88 | 3342 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 3343 | free(super); |
e683ca88 DW |
3344 | super = NULL; |
3345 | } | |
3346 | if (!super) { | |
3347 | fprintf(stderr, Name | |
3348 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
3349 | return 0; |
3350 | } | |
e683ca88 | 3351 | memset(super->buf, 0, mpb_size); |
ef649044 | 3352 | mpb = super->buf; |
e683ca88 DW |
3353 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
3354 | st->sb = super; | |
3355 | ||
3356 | if (info == NULL) { | |
3357 | /* zeroing superblock */ | |
3358 | return 0; | |
3359 | } | |
bf5a934a | 3360 | |
4d1313e9 DW |
3361 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
3362 | ||
3363 | version = (char *) mpb->sig; | |
3364 | strcpy(version, MPB_SIGNATURE); | |
3365 | version += strlen(MPB_SIGNATURE); | |
3366 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 3367 | |
bf5a934a DW |
3368 | return 1; |
3369 | } | |
3370 | ||
0e600426 | 3371 | #ifndef MDASSEMBLE |
f20c3968 | 3372 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
3373 | int fd, char *devname) |
3374 | { | |
3375 | struct intel_super *super = st->sb; | |
d23fe947 | 3376 | struct imsm_super *mpb = super->anchor; |
bf5a934a DW |
3377 | struct dl *dl; |
3378 | struct imsm_dev *dev; | |
3379 | struct imsm_map *map; | |
4eb26970 | 3380 | int slot; |
bf5a934a | 3381 | |
949c47a0 | 3382 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 3383 | map = get_imsm_map(dev, 0); |
bf5a934a | 3384 | |
208933a7 N |
3385 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
3386 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
3387 | devname); | |
3388 | return 1; | |
3389 | } | |
3390 | ||
efb30e7f DW |
3391 | if (fd == -1) { |
3392 | /* we're doing autolayout so grab the pre-marked (in | |
3393 | * validate_geometry) raid_disk | |
3394 | */ | |
3395 | for (dl = super->disks; dl; dl = dl->next) | |
3396 | if (dl->raiddisk == dk->raid_disk) | |
3397 | break; | |
3398 | } else { | |
3399 | for (dl = super->disks; dl ; dl = dl->next) | |
3400 | if (dl->major == dk->major && | |
3401 | dl->minor == dk->minor) | |
3402 | break; | |
3403 | } | |
d23fe947 | 3404 | |
208933a7 N |
3405 | if (!dl) { |
3406 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 3407 | return 1; |
208933a7 | 3408 | } |
bf5a934a | 3409 | |
d23fe947 DW |
3410 | /* add a pristine spare to the metadata */ |
3411 | if (dl->index < 0) { | |
3412 | dl->index = super->anchor->num_disks; | |
3413 | super->anchor->num_disks++; | |
3414 | } | |
4eb26970 DW |
3415 | /* Check the device has not already been added */ |
3416 | slot = get_imsm_disk_slot(map, dl->index); | |
3417 | if (slot >= 0 && | |
98130f40 | 3418 | (get_imsm_ord_tbl_ent(dev, slot, -1) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
3419 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
3420 | devname); | |
3421 | return 1; | |
3422 | } | |
be73972f | 3423 | set_imsm_ord_tbl_ent(map, dk->number, dl->index); |
ee5aad5a | 3424 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 DW |
3425 | |
3426 | /* if we are creating the first raid device update the family number */ | |
3427 | if (super->current_vol == 0) { | |
3428 | __u32 sum; | |
3429 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
3430 | struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index); | |
3431 | ||
791b666a AW |
3432 | if (!_dev || !_disk) { |
3433 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
3434 | return 1; | |
3435 | } | |
d23fe947 DW |
3436 | *_dev = *dev; |
3437 | *_disk = dl->disk; | |
148acb7b DW |
3438 | sum = random32(); |
3439 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 3440 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 3441 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 3442 | } |
f20c3968 DW |
3443 | |
3444 | return 0; | |
bf5a934a DW |
3445 | } |
3446 | ||
f20c3968 | 3447 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
cdddbdbc DW |
3448 | int fd, char *devname) |
3449 | { | |
c2c087e6 | 3450 | struct intel_super *super = st->sb; |
c2c087e6 DW |
3451 | struct dl *dd; |
3452 | unsigned long long size; | |
f2f27e63 | 3453 | __u32 id; |
c2c087e6 DW |
3454 | int rv; |
3455 | struct stat stb; | |
3456 | ||
88c32bb1 DW |
3457 | /* if we are on an RAID enabled platform check that the disk is |
3458 | * attached to the raid controller | |
3459 | */ | |
3460 | if (super->hba && !disk_attached_to_hba(fd, super->hba)) { | |
3461 | fprintf(stderr, | |
3462 | Name ": %s is not attached to the raid controller: %s\n", | |
3463 | devname ? : "disk", super->hba); | |
3464 | return 1; | |
3465 | } | |
3466 | ||
f20c3968 DW |
3467 | if (super->current_vol >= 0) |
3468 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 3469 | |
c2c087e6 DW |
3470 | fstat(fd, &stb); |
3471 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 3472 | if (!dd) { |
c2c087e6 DW |
3473 | fprintf(stderr, |
3474 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 3475 | return 1; |
c2c087e6 DW |
3476 | } |
3477 | memset(dd, 0, sizeof(*dd)); | |
3478 | dd->major = major(stb.st_rdev); | |
3479 | dd->minor = minor(stb.st_rdev); | |
b9f594fe | 3480 | dd->index = -1; |
c2c087e6 | 3481 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 3482 | dd->fd = fd; |
689c9bf3 | 3483 | dd->e = NULL; |
1a64be56 | 3484 | dd->action = DISK_ADD; |
c2c087e6 | 3485 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 3486 | if (rv) { |
c2c087e6 | 3487 | fprintf(stderr, |
0030e8d6 | 3488 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 3489 | free(dd); |
0030e8d6 | 3490 | abort(); |
c2c087e6 DW |
3491 | } |
3492 | ||
c2c087e6 DW |
3493 | get_dev_size(fd, NULL, &size); |
3494 | size /= 512; | |
1f24f035 | 3495 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 3496 | dd->disk.total_blocks = __cpu_to_le32(size); |
ee5aad5a | 3497 | dd->disk.status = SPARE_DISK; |
c2c087e6 | 3498 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 3499 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 3500 | else |
b9f594fe | 3501 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
3502 | |
3503 | if (st->update_tail) { | |
1a64be56 LM |
3504 | dd->next = super->disk_mgmt_list; |
3505 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
3506 | } else { |
3507 | dd->next = super->disks; | |
3508 | super->disks = dd; | |
3509 | } | |
f20c3968 DW |
3510 | |
3511 | return 0; | |
cdddbdbc DW |
3512 | } |
3513 | ||
1a64be56 LM |
3514 | |
3515 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
3516 | { | |
3517 | struct intel_super *super = st->sb; | |
3518 | struct dl *dd; | |
3519 | ||
3520 | /* remove from super works only in mdmon - for communication | |
3521 | * manager - monitor. Check if communication memory buffer | |
3522 | * is prepared. | |
3523 | */ | |
3524 | if (!st->update_tail) { | |
3525 | fprintf(stderr, | |
3526 | Name ": %s shall be used in mdmon context only" | |
3527 | "(line %d).\n", __func__, __LINE__); | |
3528 | return 1; | |
3529 | } | |
3530 | dd = malloc(sizeof(*dd)); | |
3531 | if (!dd) { | |
3532 | fprintf(stderr, | |
3533 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
3534 | return 1; | |
3535 | } | |
3536 | memset(dd, 0, sizeof(*dd)); | |
3537 | dd->major = dk->major; | |
3538 | dd->minor = dk->minor; | |
3539 | dd->index = -1; | |
3540 | dd->fd = -1; | |
3541 | dd->disk.status = SPARE_DISK; | |
3542 | dd->action = DISK_REMOVE; | |
3543 | ||
3544 | dd->next = super->disk_mgmt_list; | |
3545 | super->disk_mgmt_list = dd; | |
3546 | ||
3547 | ||
3548 | return 0; | |
3549 | } | |
3550 | ||
f796af5d DW |
3551 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
3552 | ||
3553 | static union { | |
3554 | char buf[512]; | |
3555 | struct imsm_super anchor; | |
3556 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 3557 | |
d23fe947 DW |
3558 | /* spare records have their own family number and do not have any defined raid |
3559 | * devices | |
3560 | */ | |
3561 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
3562 | { | |
d23fe947 | 3563 | struct imsm_super *mpb = super->anchor; |
f796af5d | 3564 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
3565 | __u32 sum; |
3566 | struct dl *d; | |
3567 | ||
f796af5d DW |
3568 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
3569 | spare->generation_num = __cpu_to_le32(1UL), | |
3570 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
3571 | spare->num_disks = 1, | |
3572 | spare->num_raid_devs = 0, | |
3573 | spare->cache_size = mpb->cache_size, | |
3574 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
3575 | ||
3576 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
3577 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
3578 | |
3579 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 3580 | if (d->index != -1) |
d23fe947 DW |
3581 | continue; |
3582 | ||
f796af5d DW |
3583 | spare->disk[0] = d->disk; |
3584 | sum = __gen_imsm_checksum(spare); | |
3585 | spare->family_num = __cpu_to_le32(sum); | |
3586 | spare->orig_family_num = 0; | |
3587 | sum = __gen_imsm_checksum(spare); | |
3588 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 3589 | |
f796af5d | 3590 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
3591 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
3592 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 3593 | return 1; |
d23fe947 DW |
3594 | } |
3595 | if (doclose) { | |
3596 | close(d->fd); | |
3597 | d->fd = -1; | |
3598 | } | |
3599 | } | |
3600 | ||
e74255d9 | 3601 | return 0; |
d23fe947 DW |
3602 | } |
3603 | ||
36988a3d | 3604 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 3605 | { |
36988a3d | 3606 | struct intel_super *super = st->sb; |
949c47a0 | 3607 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
3608 | struct dl *d; |
3609 | __u32 generation; | |
3610 | __u32 sum; | |
d23fe947 | 3611 | int spares = 0; |
949c47a0 | 3612 | int i; |
a48ac0a8 | 3613 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 3614 | int num_disks = 0; |
cdddbdbc | 3615 | |
c2c087e6 DW |
3616 | /* 'generation' is incremented everytime the metadata is written */ |
3617 | generation = __le32_to_cpu(mpb->generation_num); | |
3618 | generation++; | |
3619 | mpb->generation_num = __cpu_to_le32(generation); | |
3620 | ||
148acb7b DW |
3621 | /* fix up cases where previous mdadm releases failed to set |
3622 | * orig_family_num | |
3623 | */ | |
3624 | if (mpb->orig_family_num == 0) | |
3625 | mpb->orig_family_num = mpb->family_num; | |
3626 | ||
d23fe947 | 3627 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 3628 | if (d->index == -1) |
d23fe947 | 3629 | spares++; |
36988a3d | 3630 | else { |
d23fe947 | 3631 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
3632 | num_disks++; |
3633 | } | |
d23fe947 | 3634 | } |
36988a3d | 3635 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 3636 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
3637 | num_disks++; |
3638 | } | |
3639 | mpb->num_disks = num_disks; | |
3640 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 3641 | |
949c47a0 DW |
3642 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3643 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
3644 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
3645 | if (dev && dev2) { | |
3646 | imsm_copy_dev(dev, dev2); | |
3647 | mpb_size += sizeof_imsm_dev(dev, 0); | |
3648 | } | |
949c47a0 | 3649 | } |
a48ac0a8 DW |
3650 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
3651 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 3652 | |
c2c087e6 | 3653 | /* recalculate checksum */ |
949c47a0 | 3654 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
3655 | mpb->check_sum = __cpu_to_le32(sum); |
3656 | ||
d23fe947 | 3657 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 3658 | for (d = super->disks; d ; d = d->next) { |
d23fe947 DW |
3659 | if (d->index < 0) |
3660 | continue; | |
f796af5d | 3661 | if (store_imsm_mpb(d->fd, mpb)) |
c2c087e6 DW |
3662 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
3663 | __func__, d->major, d->minor, strerror(errno)); | |
c2c087e6 DW |
3664 | if (doclose) { |
3665 | close(d->fd); | |
3666 | d->fd = -1; | |
3667 | } | |
3668 | } | |
3669 | ||
d23fe947 DW |
3670 | if (spares) |
3671 | return write_super_imsm_spares(super, doclose); | |
3672 | ||
e74255d9 | 3673 | return 0; |
c2c087e6 DW |
3674 | } |
3675 | ||
0e600426 | 3676 | |
9b1fb677 | 3677 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
3678 | { |
3679 | size_t len; | |
3680 | struct imsm_update_create_array *u; | |
3681 | struct intel_super *super = st->sb; | |
9b1fb677 | 3682 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
54c2c1ea DW |
3683 | struct imsm_map *map = get_imsm_map(dev, 0); |
3684 | struct disk_info *inf; | |
3685 | struct imsm_disk *disk; | |
3686 | int i; | |
43dad3d6 | 3687 | |
54c2c1ea DW |
3688 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
3689 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
3690 | u = malloc(len); |
3691 | if (!u) { | |
3692 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
3693 | __func__); | |
3694 | return 1; | |
3695 | } | |
3696 | ||
3697 | u->type = update_create_array; | |
9b1fb677 | 3698 | u->dev_idx = dev_idx; |
43dad3d6 | 3699 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
3700 | inf = get_disk_info(u); |
3701 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 3702 | int idx = get_imsm_disk_idx(dev, i, -1); |
9b1fb677 | 3703 | |
54c2c1ea DW |
3704 | disk = get_imsm_disk(super, idx); |
3705 | serialcpy(inf[i].serial, disk->serial); | |
3706 | } | |
43dad3d6 DW |
3707 | append_metadata_update(st, u, len); |
3708 | ||
3709 | return 0; | |
3710 | } | |
3711 | ||
1a64be56 | 3712 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
3713 | { |
3714 | struct intel_super *super = st->sb; | |
3715 | size_t len; | |
1a64be56 | 3716 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 3717 | |
1a64be56 | 3718 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
3719 | return 0; |
3720 | ||
3721 | len = sizeof(*u); | |
3722 | u = malloc(len); | |
3723 | if (!u) { | |
3724 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
3725 | __func__); | |
3726 | return 1; | |
3727 | } | |
3728 | ||
1a64be56 | 3729 | u->type = update_add_remove_disk; |
43dad3d6 DW |
3730 | append_metadata_update(st, u, len); |
3731 | ||
3732 | return 0; | |
3733 | } | |
3734 | ||
c2c087e6 DW |
3735 | static int write_init_super_imsm(struct supertype *st) |
3736 | { | |
9b1fb677 DW |
3737 | struct intel_super *super = st->sb; |
3738 | int current_vol = super->current_vol; | |
3739 | ||
3740 | /* we are done with current_vol reset it to point st at the container */ | |
3741 | super->current_vol = -1; | |
3742 | ||
8273f55e | 3743 | if (st->update_tail) { |
43dad3d6 DW |
3744 | /* queue the recently created array / added disk |
3745 | * as a metadata update */ | |
8273f55e | 3746 | struct dl *d; |
43dad3d6 | 3747 | int rv; |
8273f55e | 3748 | |
43dad3d6 | 3749 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 3750 | if (current_vol < 0) { |
1a64be56 LM |
3751 | /* in the mgmt (add/remove) disk case we are running |
3752 | * in mdmon context, so don't close fd's | |
43dad3d6 | 3753 | */ |
1a64be56 | 3754 | return mgmt_disk(st); |
43dad3d6 | 3755 | } else |
9b1fb677 | 3756 | rv = create_array(st, current_vol); |
8273f55e DW |
3757 | |
3758 | for (d = super->disks; d ; d = d->next) { | |
3759 | close(d->fd); | |
3760 | d->fd = -1; | |
3761 | } | |
3762 | ||
43dad3d6 | 3763 | return rv; |
d682f344 N |
3764 | } else { |
3765 | struct dl *d; | |
3766 | for (d = super->disks; d; d = d->next) | |
3767 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 3768 | return write_super_imsm(st, 1); |
d682f344 | 3769 | } |
cdddbdbc | 3770 | } |
0e600426 | 3771 | #endif |
cdddbdbc | 3772 | |
e683ca88 | 3773 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 3774 | { |
e683ca88 DW |
3775 | struct intel_super *super = st->sb; |
3776 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 3777 | |
e683ca88 | 3778 | if (!mpb) |
ad97895e DW |
3779 | return 1; |
3780 | ||
1799c9e8 | 3781 | #ifndef MDASSEMBLE |
e683ca88 | 3782 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
3783 | #else |
3784 | return 1; | |
3785 | #endif | |
cdddbdbc DW |
3786 | } |
3787 | ||
0e600426 N |
3788 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
3789 | { | |
3790 | return __le32_to_cpu(mpb->bbm_log_size); | |
3791 | } | |
3792 | ||
3793 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
3794 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
3795 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 3796 | unsigned long long size, char *dev, |
2c514b71 NB |
3797 | unsigned long long *freesize, |
3798 | int verbose) | |
cdddbdbc | 3799 | { |
c2c087e6 DW |
3800 | int fd; |
3801 | unsigned long long ldsize; | |
88c32bb1 | 3802 | const struct imsm_orom *orom; |
cdddbdbc | 3803 | |
c2c087e6 DW |
3804 | if (level != LEVEL_CONTAINER) |
3805 | return 0; | |
3806 | if (!dev) | |
3807 | return 1; | |
3808 | ||
88c32bb1 DW |
3809 | if (check_env("IMSM_NO_PLATFORM")) |
3810 | orom = NULL; | |
3811 | else | |
3812 | orom = find_imsm_orom(); | |
3813 | if (orom && raiddisks > orom->tds) { | |
3814 | if (verbose) | |
3815 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
3816 | " platform supported disks: %d\n", | |
3817 | raiddisks, orom->tds); | |
3818 | return 0; | |
3819 | } | |
3820 | ||
c2c087e6 DW |
3821 | fd = open(dev, O_RDONLY|O_EXCL, 0); |
3822 | if (fd < 0) { | |
2c514b71 NB |
3823 | if (verbose) |
3824 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
3825 | dev, strerror(errno)); | |
c2c087e6 DW |
3826 | return 0; |
3827 | } | |
3828 | if (!get_dev_size(fd, dev, &ldsize)) { | |
3829 | close(fd); | |
3830 | return 0; | |
3831 | } | |
3832 | close(fd); | |
3833 | ||
3834 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
3835 | ||
3836 | return 1; | |
cdddbdbc DW |
3837 | } |
3838 | ||
0dcecb2e DW |
3839 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
3840 | { | |
3841 | const unsigned long long base_start = e[*idx].start; | |
3842 | unsigned long long end = base_start + e[*idx].size; | |
3843 | int i; | |
3844 | ||
3845 | if (base_start == end) | |
3846 | return 0; | |
3847 | ||
3848 | *idx = *idx + 1; | |
3849 | for (i = *idx; i < num_extents; i++) { | |
3850 | /* extend overlapping extents */ | |
3851 | if (e[i].start >= base_start && | |
3852 | e[i].start <= end) { | |
3853 | if (e[i].size == 0) | |
3854 | return 0; | |
3855 | if (e[i].start + e[i].size > end) | |
3856 | end = e[i].start + e[i].size; | |
3857 | } else if (e[i].start > end) { | |
3858 | *idx = i; | |
3859 | break; | |
3860 | } | |
3861 | } | |
3862 | ||
3863 | return end - base_start; | |
3864 | } | |
3865 | ||
3866 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
3867 | { | |
3868 | /* build a composite disk with all known extents and generate a new | |
3869 | * 'maxsize' given the "all disks in an array must share a common start | |
3870 | * offset" constraint | |
3871 | */ | |
3872 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
3873 | struct dl *dl; | |
3874 | int i, j; | |
3875 | int start_extent; | |
3876 | unsigned long long pos; | |
b9d77223 | 3877 | unsigned long long start = 0; |
0dcecb2e DW |
3878 | unsigned long long maxsize; |
3879 | unsigned long reserve; | |
3880 | ||
3881 | if (!e) | |
a7dd165b | 3882 | return 0; |
0dcecb2e DW |
3883 | |
3884 | /* coalesce and sort all extents. also, check to see if we need to | |
3885 | * reserve space between member arrays | |
3886 | */ | |
3887 | j = 0; | |
3888 | for (dl = super->disks; dl; dl = dl->next) { | |
3889 | if (!dl->e) | |
3890 | continue; | |
3891 | for (i = 0; i < dl->extent_cnt; i++) | |
3892 | e[j++] = dl->e[i]; | |
3893 | } | |
3894 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
3895 | ||
3896 | /* merge extents */ | |
3897 | i = 0; | |
3898 | j = 0; | |
3899 | while (i < sum_extents) { | |
3900 | e[j].start = e[i].start; | |
3901 | e[j].size = find_size(e, &i, sum_extents); | |
3902 | j++; | |
3903 | if (e[j-1].size == 0) | |
3904 | break; | |
3905 | } | |
3906 | ||
3907 | pos = 0; | |
3908 | maxsize = 0; | |
3909 | start_extent = 0; | |
3910 | i = 0; | |
3911 | do { | |
3912 | unsigned long long esize; | |
3913 | ||
3914 | esize = e[i].start - pos; | |
3915 | if (esize >= maxsize) { | |
3916 | maxsize = esize; | |
3917 | start = pos; | |
3918 | start_extent = i; | |
3919 | } | |
3920 | pos = e[i].start + e[i].size; | |
3921 | i++; | |
3922 | } while (e[i-1].size); | |
3923 | free(e); | |
3924 | ||
a7dd165b DW |
3925 | if (maxsize == 0) |
3926 | return 0; | |
3927 | ||
3928 | /* FIXME assumes volume at offset 0 is the first volume in a | |
3929 | * container | |
3930 | */ | |
0dcecb2e DW |
3931 | if (start_extent > 0) |
3932 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
3933 | else | |
3934 | reserve = 0; | |
3935 | ||
3936 | if (maxsize < reserve) | |
a7dd165b | 3937 | return 0; |
0dcecb2e DW |
3938 | |
3939 | super->create_offset = ~((__u32) 0); | |
3940 | if (start + reserve > super->create_offset) | |
a7dd165b | 3941 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
3942 | super->create_offset = start + reserve; |
3943 | ||
3944 | return maxsize - reserve; | |
3945 | } | |
3946 | ||
88c32bb1 DW |
3947 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
3948 | { | |
3949 | if (level < 0 || level == 6 || level == 4) | |
3950 | return 0; | |
3951 | ||
3952 | /* if we have an orom prevent invalid raid levels */ | |
3953 | if (orom) | |
3954 | switch (level) { | |
3955 | case 0: return imsm_orom_has_raid0(orom); | |
3956 | case 1: | |
3957 | if (raiddisks > 2) | |
3958 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
3959 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
3960 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
3961 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
3962 | } |
3963 | else | |
3964 | return 1; /* not on an Intel RAID platform so anything goes */ | |
3965 | ||
3966 | return 0; | |
3967 | } | |
3968 | ||
35f81cbb | 3969 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) |
6592ce37 DW |
3970 | static int |
3971 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
3972 | int raiddisks, int chunk, int verbose) | |
3973 | { | |
3974 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { | |
3975 | pr_vrb(": platform does not support raid%d with %d disk%s\n", | |
3976 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
3977 | return 0; | |
3978 | } | |
3979 | if (super->orom && level != 1 && | |
3980 | !imsm_orom_has_chunk(super->orom, chunk)) { | |
3981 | pr_vrb(": platform does not support a chunk size of: %d\n", chunk); | |
3982 | return 0; | |
3983 | } | |
3984 | if (layout != imsm_level_to_layout(level)) { | |
3985 | if (level == 5) | |
3986 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
3987 | else if (level == 10) | |
3988 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
3989 | else | |
3990 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
3991 | layout, level); | |
3992 | return 0; | |
3993 | } | |
3994 | ||
3995 | return 1; | |
3996 | } | |
3997 | ||
c2c087e6 DW |
3998 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
3999 | * FIX ME add ahci details | |
4000 | */ | |
8b353278 DW |
4001 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
4002 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 4003 | unsigned long long size, char *dev, |
2c514b71 NB |
4004 | unsigned long long *freesize, |
4005 | int verbose) | |
cdddbdbc | 4006 | { |
c2c087e6 DW |
4007 | struct stat stb; |
4008 | struct intel_super *super = st->sb; | |
a20d2ba5 | 4009 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4010 | struct dl *dl; |
4011 | unsigned long long pos = 0; | |
4012 | unsigned long long maxsize; | |
4013 | struct extent *e; | |
4014 | int i; | |
cdddbdbc | 4015 | |
88c32bb1 DW |
4016 | /* We must have the container info already read in. */ |
4017 | if (!super) | |
c2c087e6 DW |
4018 | return 0; |
4019 | ||
6592ce37 | 4020 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, verbose)) |
c2c087e6 | 4021 | return 0; |
c2c087e6 DW |
4022 | |
4023 | if (!dev) { | |
4024 | /* General test: make sure there is space for | |
2da8544a DW |
4025 | * 'raiddisks' device extents of size 'size' at a given |
4026 | * offset | |
c2c087e6 | 4027 | */ |
e46273eb | 4028 | unsigned long long minsize = size; |
b7528a20 | 4029 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
4030 | int dcnt = 0; |
4031 | if (minsize == 0) | |
4032 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
4033 | for (dl = super->disks; dl ; dl = dl->next) { | |
4034 | int found = 0; | |
4035 | ||
bf5a934a | 4036 | pos = 0; |
c2c087e6 DW |
4037 | i = 0; |
4038 | e = get_extents(super, dl); | |
4039 | if (!e) continue; | |
4040 | do { | |
4041 | unsigned long long esize; | |
4042 | esize = e[i].start - pos; | |
4043 | if (esize >= minsize) | |
4044 | found = 1; | |
b7528a20 | 4045 | if (found && start_offset == MaxSector) { |
2da8544a DW |
4046 | start_offset = pos; |
4047 | break; | |
4048 | } else if (found && pos != start_offset) { | |
4049 | found = 0; | |
4050 | break; | |
4051 | } | |
c2c087e6 DW |
4052 | pos = e[i].start + e[i].size; |
4053 | i++; | |
4054 | } while (e[i-1].size); | |
4055 | if (found) | |
4056 | dcnt++; | |
4057 | free(e); | |
4058 | } | |
4059 | if (dcnt < raiddisks) { | |
2c514b71 NB |
4060 | if (verbose) |
4061 | fprintf(stderr, Name ": imsm: Not enough " | |
4062 | "devices with space for this array " | |
4063 | "(%d < %d)\n", | |
4064 | dcnt, raiddisks); | |
c2c087e6 DW |
4065 | return 0; |
4066 | } | |
4067 | return 1; | |
4068 | } | |
0dcecb2e | 4069 | |
c2c087e6 DW |
4070 | /* This device must be a member of the set */ |
4071 | if (stat(dev, &stb) < 0) | |
4072 | return 0; | |
4073 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
4074 | return 0; | |
4075 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
4076 | if (dl->major == (int)major(stb.st_rdev) && |
4077 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
4078 | break; |
4079 | } | |
4080 | if (!dl) { | |
2c514b71 NB |
4081 | if (verbose) |
4082 | fprintf(stderr, Name ": %s is not in the " | |
4083 | "same imsm set\n", dev); | |
c2c087e6 | 4084 | return 0; |
a20d2ba5 DW |
4085 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
4086 | /* If a volume is present then the current creation attempt | |
4087 | * cannot incorporate new spares because the orom may not | |
4088 | * understand this configuration (all member disks must be | |
4089 | * members of each array in the container). | |
4090 | */ | |
4091 | fprintf(stderr, Name ": %s is a spare and a volume" | |
4092 | " is already defined for this container\n", dev); | |
4093 | fprintf(stderr, Name ": The option-rom requires all member" | |
4094 | " disks to be a member of all volumes\n"); | |
4095 | return 0; | |
c2c087e6 | 4096 | } |
0dcecb2e DW |
4097 | |
4098 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
4099 | e = get_extents(super, dl); |
4100 | maxsize = 0; | |
4101 | i = 0; | |
0dcecb2e DW |
4102 | if (e) { |
4103 | do { | |
4104 | unsigned long long esize; | |
4105 | ||
4106 | esize = e[i].start - pos; | |
4107 | if (esize >= maxsize) | |
4108 | maxsize = esize; | |
4109 | pos = e[i].start + e[i].size; | |
4110 | i++; | |
4111 | } while (e[i-1].size); | |
4112 | dl->e = e; | |
4113 | dl->extent_cnt = i; | |
4114 | } else { | |
4115 | if (verbose) | |
4116 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
4117 | dev); | |
4118 | return 0; | |
4119 | } | |
4120 | if (maxsize < size) { | |
4121 | if (verbose) | |
4122 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
4123 | dev, maxsize, size); | |
4124 | return 0; | |
4125 | } | |
4126 | ||
4127 | /* count total number of extents for merge */ | |
4128 | i = 0; | |
4129 | for (dl = super->disks; dl; dl = dl->next) | |
4130 | if (dl->e) | |
4131 | i += dl->extent_cnt; | |
4132 | ||
4133 | maxsize = merge_extents(super, i); | |
a7dd165b | 4134 | if (maxsize < size || maxsize == 0) { |
0dcecb2e DW |
4135 | if (verbose) |
4136 | fprintf(stderr, Name ": not enough space after merge (%llu < %llu)\n", | |
4137 | maxsize, size); | |
4138 | return 0; | |
0dcecb2e DW |
4139 | } |
4140 | ||
c2c087e6 DW |
4141 | *freesize = maxsize; |
4142 | ||
4143 | return 1; | |
cdddbdbc DW |
4144 | } |
4145 | ||
efb30e7f DW |
4146 | static int reserve_space(struct supertype *st, int raiddisks, |
4147 | unsigned long long size, int chunk, | |
4148 | unsigned long long *freesize) | |
4149 | { | |
4150 | struct intel_super *super = st->sb; | |
4151 | struct imsm_super *mpb = super->anchor; | |
4152 | struct dl *dl; | |
4153 | int i; | |
4154 | int extent_cnt; | |
4155 | struct extent *e; | |
4156 | unsigned long long maxsize; | |
4157 | unsigned long long minsize; | |
4158 | int cnt; | |
4159 | int used; | |
4160 | ||
4161 | /* find the largest common start free region of the possible disks */ | |
4162 | used = 0; | |
4163 | extent_cnt = 0; | |
4164 | cnt = 0; | |
4165 | for (dl = super->disks; dl; dl = dl->next) { | |
4166 | dl->raiddisk = -1; | |
4167 | ||
4168 | if (dl->index >= 0) | |
4169 | used++; | |
4170 | ||
4171 | /* don't activate new spares if we are orom constrained | |
4172 | * and there is already a volume active in the container | |
4173 | */ | |
4174 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
4175 | continue; | |
4176 | ||
4177 | e = get_extents(super, dl); | |
4178 | if (!e) | |
4179 | continue; | |
4180 | for (i = 1; e[i-1].size; i++) | |
4181 | ; | |
4182 | dl->e = e; | |
4183 | dl->extent_cnt = i; | |
4184 | extent_cnt += i; | |
4185 | cnt++; | |
4186 | } | |
4187 | ||
4188 | maxsize = merge_extents(super, extent_cnt); | |
4189 | minsize = size; | |
4190 | if (size == 0) | |
4191 | minsize = chunk; | |
4192 | ||
4193 | if (cnt < raiddisks || | |
4194 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
4195 | maxsize < minsize || |
4196 | maxsize == 0) { | |
efb30e7f DW |
4197 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
4198 | return 0; /* No enough free spaces large enough */ | |
4199 | } | |
4200 | ||
4201 | if (size == 0) { | |
4202 | size = maxsize; | |
4203 | if (chunk) { | |
4204 | size /= chunk; | |
4205 | size *= chunk; | |
4206 | } | |
4207 | } | |
4208 | ||
4209 | cnt = 0; | |
4210 | for (dl = super->disks; dl; dl = dl->next) | |
4211 | if (dl->e) | |
4212 | dl->raiddisk = cnt++; | |
4213 | ||
4214 | *freesize = size; | |
4215 | ||
4216 | return 1; | |
4217 | } | |
4218 | ||
bf5a934a DW |
4219 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
4220 | int raiddisks, int chunk, unsigned long long size, | |
4221 | char *dev, unsigned long long *freesize, | |
4222 | int verbose) | |
4223 | { | |
4224 | int fd, cfd; | |
4225 | struct mdinfo *sra; | |
20cbe8d2 | 4226 | int is_member = 0; |
bf5a934a DW |
4227 | |
4228 | /* if given unused devices create a container | |
4229 | * if given given devices in a container create a member volume | |
4230 | */ | |
4231 | if (level == LEVEL_CONTAINER) { | |
4232 | /* Must be a fresh device to add to a container */ | |
4233 | return validate_geometry_imsm_container(st, level, layout, | |
4234 | raiddisks, chunk, size, | |
4235 | dev, freesize, | |
4236 | verbose); | |
4237 | } | |
4238 | ||
8592f29d N |
4239 | if (!dev) { |
4240 | if (st->sb && freesize) { | |
efb30e7f DW |
4241 | /* we are being asked to automatically layout a |
4242 | * new volume based on the current contents of | |
4243 | * the container. If the the parameters can be | |
4244 | * satisfied reserve_space will record the disks, | |
4245 | * start offset, and size of the volume to be | |
4246 | * created. add_to_super and getinfo_super | |
4247 | * detect when autolayout is in progress. | |
4248 | */ | |
6592ce37 DW |
4249 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
4250 | raiddisks, chunk, | |
4251 | verbose)) | |
4252 | return 0; | |
efb30e7f | 4253 | return reserve_space(st, raiddisks, size, chunk, freesize); |
8592f29d N |
4254 | } |
4255 | return 1; | |
4256 | } | |
bf5a934a DW |
4257 | if (st->sb) { |
4258 | /* creating in a given container */ | |
4259 | return validate_geometry_imsm_volume(st, level, layout, | |
4260 | raiddisks, chunk, size, | |
4261 | dev, freesize, verbose); | |
4262 | } | |
4263 | ||
bf5a934a DW |
4264 | /* This device needs to be a device in an 'imsm' container */ |
4265 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4266 | if (fd >= 0) { | |
4267 | if (verbose) | |
4268 | fprintf(stderr, | |
4269 | Name ": Cannot create this array on device %s\n", | |
4270 | dev); | |
4271 | close(fd); | |
4272 | return 0; | |
4273 | } | |
4274 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
4275 | if (verbose) | |
4276 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
4277 | dev, strerror(errno)); | |
4278 | return 0; | |
4279 | } | |
4280 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
4281 | cfd = open_container(fd); | |
20cbe8d2 | 4282 | close(fd); |
bf5a934a | 4283 | if (cfd < 0) { |
bf5a934a DW |
4284 | if (verbose) |
4285 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
4286 | dev); | |
4287 | return 0; | |
4288 | } | |
4289 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 4290 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
4291 | strcmp(sra->text_version, "imsm") == 0) |
4292 | is_member = 1; | |
4293 | sysfs_free(sra); | |
4294 | if (is_member) { | |
bf5a934a DW |
4295 | /* This is a member of a imsm container. Load the container |
4296 | * and try to create a volume | |
4297 | */ | |
4298 | struct intel_super *super; | |
4299 | ||
e1902a7b | 4300 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL) == 0) { |
bf5a934a DW |
4301 | st->sb = super; |
4302 | st->container_dev = fd2devnum(cfd); | |
4303 | close(cfd); | |
4304 | return validate_geometry_imsm_volume(st, level, layout, | |
4305 | raiddisks, chunk, | |
4306 | size, dev, | |
4307 | freesize, verbose); | |
4308 | } | |
20cbe8d2 | 4309 | } |
bf5a934a | 4310 | |
20cbe8d2 AW |
4311 | if (verbose) |
4312 | fprintf(stderr, Name ": failed container membership check\n"); | |
4313 | ||
4314 | close(cfd); | |
4315 | return 0; | |
bf5a934a | 4316 | } |
0bd16cf2 | 4317 | |
30f58b22 | 4318 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
4319 | { |
4320 | struct intel_super *super = st->sb; | |
4321 | ||
30f58b22 DW |
4322 | if (level && *level == UnSet) |
4323 | *level = LEVEL_CONTAINER; | |
4324 | ||
4325 | if (level && layout && *layout == UnSet) | |
4326 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 4327 | |
1d54f286 N |
4328 | if (chunk && (*chunk == UnSet || *chunk == 0) && |
4329 | super && super->orom) | |
30f58b22 | 4330 | *chunk = imsm_orom_default_chunk(super->orom); |
0bd16cf2 DJ |
4331 | } |
4332 | ||
33414a01 DW |
4333 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
4334 | ||
4335 | static int kill_subarray_imsm(struct supertype *st) | |
4336 | { | |
4337 | /* remove the subarray currently referenced by ->current_vol */ | |
4338 | __u8 i; | |
4339 | struct intel_dev **dp; | |
4340 | struct intel_super *super = st->sb; | |
4341 | __u8 current_vol = super->current_vol; | |
4342 | struct imsm_super *mpb = super->anchor; | |
4343 | ||
4344 | if (super->current_vol < 0) | |
4345 | return 2; | |
4346 | super->current_vol = -1; /* invalidate subarray cursor */ | |
4347 | ||
4348 | /* block deletions that would change the uuid of active subarrays | |
4349 | * | |
4350 | * FIXME when immutable ids are available, but note that we'll | |
4351 | * also need to fixup the invalidated/active subarray indexes in | |
4352 | * mdstat | |
4353 | */ | |
4354 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4355 | char subarray[4]; | |
4356 | ||
4357 | if (i < current_vol) | |
4358 | continue; | |
4359 | sprintf(subarray, "%u", i); | |
4360 | if (is_subarray_active(subarray, st->devname)) { | |
4361 | fprintf(stderr, | |
4362 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
4363 | current_vol, i); | |
4364 | ||
4365 | return 2; | |
4366 | } | |
4367 | } | |
4368 | ||
4369 | if (st->update_tail) { | |
4370 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
4371 | ||
4372 | if (!u) | |
4373 | return 2; | |
4374 | u->type = update_kill_array; | |
4375 | u->dev_idx = current_vol; | |
4376 | append_metadata_update(st, u, sizeof(*u)); | |
4377 | ||
4378 | return 0; | |
4379 | } | |
4380 | ||
4381 | for (dp = &super->devlist; *dp;) | |
4382 | if ((*dp)->index == current_vol) { | |
4383 | *dp = (*dp)->next; | |
4384 | } else { | |
4385 | handle_missing(super, (*dp)->dev); | |
4386 | if ((*dp)->index > current_vol) | |
4387 | (*dp)->index--; | |
4388 | dp = &(*dp)->next; | |
4389 | } | |
4390 | ||
4391 | /* no more raid devices, all active components are now spares, | |
4392 | * but of course failed are still failed | |
4393 | */ | |
4394 | if (--mpb->num_raid_devs == 0) { | |
4395 | struct dl *d; | |
4396 | ||
4397 | for (d = super->disks; d; d = d->next) | |
4398 | if (d->index > -2) { | |
4399 | d->index = -1; | |
4400 | d->disk.status = SPARE_DISK; | |
4401 | } | |
4402 | } | |
4403 | ||
4404 | super->updates_pending++; | |
4405 | ||
4406 | return 0; | |
4407 | } | |
aa534678 | 4408 | |
a951a4f7 | 4409 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 4410 | char *update, struct mddev_ident *ident) |
aa534678 DW |
4411 | { |
4412 | /* update the subarray currently referenced by ->current_vol */ | |
4413 | struct intel_super *super = st->sb; | |
4414 | struct imsm_super *mpb = super->anchor; | |
4415 | ||
aa534678 DW |
4416 | if (strcmp(update, "name") == 0) { |
4417 | char *name = ident->name; | |
a951a4f7 N |
4418 | char *ep; |
4419 | int vol; | |
aa534678 | 4420 | |
a951a4f7 | 4421 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
4422 | fprintf(stderr, |
4423 | Name ": Unable to update name of active subarray\n"); | |
4424 | return 2; | |
4425 | } | |
4426 | ||
4427 | if (!check_name(super, name, 0)) | |
4428 | return 2; | |
4429 | ||
a951a4f7 N |
4430 | vol = strtoul(subarray, &ep, 10); |
4431 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
4432 | return 2; | |
4433 | ||
aa534678 DW |
4434 | if (st->update_tail) { |
4435 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
4436 | ||
4437 | if (!u) | |
4438 | return 2; | |
4439 | u->type = update_rename_array; | |
a951a4f7 | 4440 | u->dev_idx = vol; |
aa534678 DW |
4441 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
4442 | append_metadata_update(st, u, sizeof(*u)); | |
4443 | } else { | |
4444 | struct imsm_dev *dev; | |
4445 | int i; | |
4446 | ||
a951a4f7 | 4447 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
4448 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
4449 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4450 | dev = get_imsm_dev(super, i); | |
4451 | handle_missing(super, dev); | |
4452 | } | |
4453 | super->updates_pending++; | |
4454 | } | |
4455 | } else | |
4456 | return 2; | |
4457 | ||
4458 | return 0; | |
4459 | } | |
0e600426 | 4460 | #endif /* MDASSEMBLE */ |
bf5a934a | 4461 | |
28bce06f AK |
4462 | static int is_gen_migration(struct imsm_dev *dev) |
4463 | { | |
4464 | if (!dev->vol.migr_state) | |
4465 | return 0; | |
4466 | ||
4467 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
4468 | return 1; | |
4469 | ||
4470 | return 0; | |
4471 | } | |
4472 | ||
1e5c6983 DW |
4473 | static int is_rebuilding(struct imsm_dev *dev) |
4474 | { | |
4475 | struct imsm_map *migr_map; | |
4476 | ||
4477 | if (!dev->vol.migr_state) | |
4478 | return 0; | |
4479 | ||
4480 | if (migr_type(dev) != MIGR_REBUILD) | |
4481 | return 0; | |
4482 | ||
4483 | migr_map = get_imsm_map(dev, 1); | |
4484 | ||
4485 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
4486 | return 1; | |
4487 | else | |
4488 | return 0; | |
4489 | } | |
4490 | ||
4491 | static void update_recovery_start(struct imsm_dev *dev, struct mdinfo *array) | |
4492 | { | |
4493 | struct mdinfo *rebuild = NULL; | |
4494 | struct mdinfo *d; | |
4495 | __u32 units; | |
4496 | ||
4497 | if (!is_rebuilding(dev)) | |
4498 | return; | |
4499 | ||
4500 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
4501 | for (d = array->devs; d; d = d->next) | |
4502 | if (d->recovery_start == 0) { | |
4503 | if (rebuild) | |
4504 | return; | |
4505 | rebuild = d; | |
4506 | } | |
4507 | ||
4363fd80 DW |
4508 | if (!rebuild) { |
4509 | /* (?) none of the disks are marked with | |
4510 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
4511 | * disk_ord_tbl was not correctly updated | |
4512 | */ | |
4513 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
4514 | return; | |
4515 | } | |
4516 | ||
1e5c6983 DW |
4517 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
4518 | rebuild->recovery_start = units * blocks_per_migr_unit(dev); | |
4519 | } | |
4520 | ||
4521 | ||
00bbdbda | 4522 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 4523 | { |
4f5bc454 DW |
4524 | /* Given a container loaded by load_super_imsm_all, |
4525 | * extract information about all the arrays into | |
4526 | * an mdinfo tree. | |
00bbdbda | 4527 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
4528 | * |
4529 | * For each imsm_dev create an mdinfo, fill it in, | |
4530 | * then look for matching devices in super->disks | |
4531 | * and create appropriate device mdinfo. | |
4532 | */ | |
4533 | struct intel_super *super = st->sb; | |
949c47a0 | 4534 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 4535 | struct mdinfo *rest = NULL; |
00bbdbda | 4536 | unsigned int i; |
a06d022d | 4537 | int bbm_errors = 0; |
cdddbdbc | 4538 | |
a06d022d KW |
4539 | /* check for bad blocks */ |
4540 | if (imsm_bbm_log_size(super->anchor)) | |
4541 | bbm_errors = 1; | |
604b746f | 4542 | |
4f5bc454 | 4543 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
4544 | struct imsm_dev *dev; |
4545 | struct imsm_map *map; | |
86e3692b | 4546 | struct imsm_map *map2; |
4f5bc454 | 4547 | struct mdinfo *this; |
4f5bc454 | 4548 | int slot; |
00bbdbda N |
4549 | char *ep; |
4550 | ||
4551 | if (subarray && | |
4552 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
4553 | continue; | |
4554 | ||
4555 | dev = get_imsm_dev(super, i); | |
4556 | map = get_imsm_map(dev, 0); | |
86e3692b | 4557 | map2 = get_imsm_map(dev, 1); |
4f5bc454 | 4558 | |
1ce0101c DW |
4559 | /* do not publish arrays that are in the middle of an |
4560 | * unsupported migration | |
4561 | */ | |
4562 | if (dev->vol.migr_state && | |
28bce06f | 4563 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
4564 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
4565 | " unsupported migration in progress\n", | |
4566 | dev->volume); | |
4567 | continue; | |
4568 | } | |
4569 | ||
4f5bc454 | 4570 | this = malloc(sizeof(*this)); |
0fbd635c | 4571 | if (!this) { |
cf1be220 | 4572 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
4573 | sizeof(*this)); |
4574 | break; | |
4575 | } | |
4f5bc454 DW |
4576 | memset(this, 0, sizeof(*this)); |
4577 | this->next = rest; | |
4f5bc454 | 4578 | |
301406c9 | 4579 | super->current_vol = i; |
a5d85af7 | 4580 | getinfo_super_imsm_volume(st, this, NULL); |
4f5bc454 | 4581 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 4582 | unsigned long long recovery_start; |
4f5bc454 DW |
4583 | struct mdinfo *info_d; |
4584 | struct dl *d; | |
4585 | int idx; | |
9a1608e5 | 4586 | int skip; |
7eef0453 | 4587 | __u32 ord; |
4f5bc454 | 4588 | |
9a1608e5 | 4589 | skip = 0; |
98130f40 AK |
4590 | idx = get_imsm_disk_idx(dev, slot, 0); |
4591 | ord = get_imsm_ord_tbl_ent(dev, slot, 0); | |
4f5bc454 DW |
4592 | for (d = super->disks; d ; d = d->next) |
4593 | if (d->index == idx) | |
0fbd635c | 4594 | break; |
4f5bc454 | 4595 | |
1e5c6983 | 4596 | recovery_start = MaxSector; |
4f5bc454 | 4597 | if (d == NULL) |
9a1608e5 | 4598 | skip = 1; |
25ed7e59 | 4599 | if (d && is_failed(&d->disk)) |
9a1608e5 | 4600 | skip = 1; |
7eef0453 | 4601 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 4602 | recovery_start = 0; |
9a1608e5 DW |
4603 | |
4604 | /* | |
4605 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
4606 | * reset resync start to avoid a dirty-degraded |
4607 | * situation when performing the intial sync | |
9a1608e5 DW |
4608 | * |
4609 | * FIXME handle dirty degraded | |
4610 | */ | |
1e5c6983 | 4611 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 4612 | this->resync_start = MaxSector; |
9a1608e5 DW |
4613 | if (skip) |
4614 | continue; | |
4f5bc454 | 4615 | |
1e5c6983 | 4616 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
4617 | if (!info_d) { |
4618 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 4619 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
4620 | info_d = this->devs; |
4621 | while (info_d) { | |
4622 | struct mdinfo *d = info_d->next; | |
4623 | ||
4624 | free(info_d); | |
4625 | info_d = d; | |
4626 | } | |
9a1608e5 DW |
4627 | free(this); |
4628 | this = rest; | |
4629 | break; | |
4630 | } | |
4f5bc454 DW |
4631 | info_d->next = this->devs; |
4632 | this->devs = info_d; | |
4633 | ||
4f5bc454 DW |
4634 | info_d->disk.number = d->index; |
4635 | info_d->disk.major = d->major; | |
4636 | info_d->disk.minor = d->minor; | |
4637 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 4638 | info_d->recovery_start = recovery_start; |
86e3692b AK |
4639 | if (map2) { |
4640 | if (slot < map2->num_members) | |
4641 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
4642 | else |
4643 | this->array.spare_disks++; | |
86e3692b AK |
4644 | } else { |
4645 | if (slot < map->num_members) | |
4646 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
4647 | else |
4648 | this->array.spare_disks++; | |
86e3692b | 4649 | } |
1e5c6983 DW |
4650 | if (info_d->recovery_start == MaxSector) |
4651 | this->array.working_disks++; | |
4f5bc454 DW |
4652 | |
4653 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
4654 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
4655 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
4f5bc454 | 4656 | } |
1e5c6983 DW |
4657 | /* now that the disk list is up-to-date fixup recovery_start */ |
4658 | update_recovery_start(dev, this); | |
9a1608e5 | 4659 | rest = this; |
4f5bc454 DW |
4660 | } |
4661 | ||
a06d022d KW |
4662 | /* if array has bad blocks, set suitable bit in array status */ |
4663 | if (bbm_errors) | |
4664 | rest->array.state |= (1<<MD_SB_BBM_ERRORS); | |
4665 | ||
4f5bc454 | 4666 | return rest; |
cdddbdbc DW |
4667 | } |
4668 | ||
845dea95 | 4669 | |
fb49eef2 | 4670 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed) |
c2a1e7da | 4671 | { |
a965f303 | 4672 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
4673 | |
4674 | if (!failed) | |
3393c6af DW |
4675 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
4676 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
4677 | |
4678 | switch (get_imsm_raid_level(map)) { | |
4679 | case 0: | |
4680 | return IMSM_T_STATE_FAILED; | |
4681 | break; | |
4682 | case 1: | |
4683 | if (failed < map->num_members) | |
4684 | return IMSM_T_STATE_DEGRADED; | |
4685 | else | |
4686 | return IMSM_T_STATE_FAILED; | |
4687 | break; | |
4688 | case 10: | |
4689 | { | |
4690 | /** | |
c92a2527 DW |
4691 | * check to see if any mirrors have failed, otherwise we |
4692 | * are degraded. Even numbered slots are mirrored on | |
4693 | * slot+1 | |
c2a1e7da | 4694 | */ |
c2a1e7da | 4695 | int i; |
d9b420a5 N |
4696 | /* gcc -Os complains that this is unused */ |
4697 | int insync = insync; | |
c2a1e7da DW |
4698 | |
4699 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 4700 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
c92a2527 DW |
4701 | int idx = ord_to_idx(ord); |
4702 | struct imsm_disk *disk; | |
c2a1e7da | 4703 | |
c92a2527 DW |
4704 | /* reset the potential in-sync count on even-numbered |
4705 | * slots. num_copies is always 2 for imsm raid10 | |
4706 | */ | |
4707 | if ((i & 1) == 0) | |
4708 | insync = 2; | |
c2a1e7da | 4709 | |
c92a2527 | 4710 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 4711 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 4712 | insync--; |
c2a1e7da | 4713 | |
c92a2527 DW |
4714 | /* no in-sync disks left in this mirror the |
4715 | * array has failed | |
4716 | */ | |
4717 | if (insync == 0) | |
4718 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
4719 | } |
4720 | ||
4721 | return IMSM_T_STATE_DEGRADED; | |
4722 | } | |
4723 | case 5: | |
4724 | if (failed < 2) | |
4725 | return IMSM_T_STATE_DEGRADED; | |
4726 | else | |
4727 | return IMSM_T_STATE_FAILED; | |
4728 | break; | |
4729 | default: | |
4730 | break; | |
4731 | } | |
4732 | ||
4733 | return map->map_state; | |
4734 | } | |
4735 | ||
ff077194 | 4736 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev) |
c2a1e7da DW |
4737 | { |
4738 | int i; | |
4739 | int failed = 0; | |
4740 | struct imsm_disk *disk; | |
ff077194 | 4741 | struct imsm_map *map = get_imsm_map(dev, 0); |
0556e1a2 DW |
4742 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
4743 | __u32 ord; | |
4744 | int idx; | |
c2a1e7da | 4745 | |
0556e1a2 DW |
4746 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
4747 | * disks that are being rebuilt. New failures are recorded to | |
4748 | * map[0]. So we look through all the disks we started with and | |
4749 | * see if any failures are still present, or if any new ones | |
4750 | * have arrived | |
4751 | * | |
4752 | * FIXME add support for online capacity expansion and | |
4753 | * raid-level-migration | |
4754 | */ | |
4755 | for (i = 0; i < prev->num_members; i++) { | |
4756 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); | |
4757 | ord |= __le32_to_cpu(map->disk_ord_tbl[i]); | |
4758 | idx = ord_to_idx(ord); | |
c2a1e7da | 4759 | |
949c47a0 | 4760 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 4761 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
fcb84475 | 4762 | failed++; |
c2a1e7da DW |
4763 | } |
4764 | ||
4765 | return failed; | |
845dea95 NB |
4766 | } |
4767 | ||
97b4d0e9 DW |
4768 | #ifndef MDASSEMBLE |
4769 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
4770 | char *inst) | |
4771 | { | |
4772 | struct intel_super *super = c->sb; | |
4773 | struct imsm_super *mpb = super->anchor; | |
4774 | ||
4775 | if (atoi(inst) >= mpb->num_raid_devs) { | |
4776 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
4777 | __func__, atoi(inst)); | |
4778 | return -ENODEV; | |
4779 | } | |
4780 | ||
4781 | dprintf("imsm: open_new %s\n", inst); | |
4782 | a->info.container_member = atoi(inst); | |
4783 | return 0; | |
4784 | } | |
4785 | ||
0c046afd DW |
4786 | static int is_resyncing(struct imsm_dev *dev) |
4787 | { | |
4788 | struct imsm_map *migr_map; | |
4789 | ||
4790 | if (!dev->vol.migr_state) | |
4791 | return 0; | |
4792 | ||
1484e727 DW |
4793 | if (migr_type(dev) == MIGR_INIT || |
4794 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
4795 | return 1; |
4796 | ||
4c9bc37b AK |
4797 | if (migr_type(dev) == MIGR_GEN_MIGR) |
4798 | return 0; | |
4799 | ||
0c046afd DW |
4800 | migr_map = get_imsm_map(dev, 1); |
4801 | ||
4c9bc37b AK |
4802 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
4803 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
4804 | return 1; |
4805 | else | |
4806 | return 0; | |
4807 | } | |
4808 | ||
0556e1a2 DW |
4809 | /* return true if we recorded new information */ |
4810 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 4811 | { |
0556e1a2 DW |
4812 | __u32 ord; |
4813 | int slot; | |
4814 | struct imsm_map *map; | |
4815 | ||
4816 | /* new failures are always set in map[0] */ | |
4817 | map = get_imsm_map(dev, 0); | |
4818 | ||
4819 | slot = get_imsm_disk_slot(map, idx); | |
4820 | if (slot < 0) | |
4821 | return 0; | |
4822 | ||
4823 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 4824 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
4825 | return 0; |
4826 | ||
f2f27e63 | 4827 | disk->status |= FAILED_DISK; |
cf53434e | 4828 | disk->status &= ~CONFIGURED_DISK; |
0556e1a2 | 4829 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
f21e18ca | 4830 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
4831 | map->failed_disk_num = slot; |
4832 | return 1; | |
4833 | } | |
4834 | ||
4835 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
4836 | { | |
4837 | mark_failure(dev, disk, idx); | |
4838 | ||
4839 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
4840 | return; | |
4841 | ||
47ee5a45 DW |
4842 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
4843 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
4844 | } | |
4845 | ||
33414a01 DW |
4846 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
4847 | { | |
4848 | __u8 map_state; | |
4849 | struct dl *dl; | |
4850 | int failed; | |
4851 | ||
4852 | if (!super->missing) | |
4853 | return; | |
4854 | failed = imsm_count_failed(super, dev); | |
4855 | map_state = imsm_check_degraded(super, dev, failed); | |
4856 | ||
4857 | dprintf("imsm: mark missing\n"); | |
4858 | end_migration(dev, map_state); | |
4859 | for (dl = super->missing; dl; dl = dl->next) | |
4860 | mark_missing(dev, &dl->disk, dl->index); | |
4861 | super->updates_pending++; | |
4862 | } | |
4863 | ||
28bce06f AK |
4864 | static void imsm_set_disk(struct active_array *a, int n, int state); |
4865 | ||
0e2d1a4e AK |
4866 | static void imsm_progress_container_reshape(struct intel_super *super) |
4867 | { | |
4868 | /* if no device has a migr_state, but some device has a | |
4869 | * different number of members than the previous device, start | |
4870 | * changing the number of devices in this device to match | |
4871 | * previous. | |
4872 | */ | |
4873 | struct imsm_super *mpb = super->anchor; | |
4874 | int prev_disks = -1; | |
4875 | int i; | |
4876 | ||
4877 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4878 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4879 | struct imsm_map *map = get_imsm_map(dev, 0); | |
4880 | struct imsm_map *map2; | |
4881 | int prev_num_members; | |
4882 | int used_disks; | |
4883 | ||
4884 | if (dev->vol.migr_state) | |
4885 | return; | |
4886 | ||
4887 | if (prev_disks == -1) | |
4888 | prev_disks = map->num_members; | |
4889 | if (prev_disks == map->num_members) | |
4890 | continue; | |
4891 | ||
4892 | /* OK, this array needs to enter reshape mode. | |
4893 | * i.e it needs a migr_state | |
4894 | */ | |
4895 | ||
4896 | prev_num_members = map->num_members; | |
4897 | map->num_members = prev_disks; | |
4898 | dev->vol.migr_state = 1; | |
4899 | dev->vol.curr_migr_unit = 0; | |
4900 | dev->vol.migr_type = MIGR_GEN_MIGR; | |
4901 | for (i = prev_num_members; | |
4902 | i < map->num_members; i++) | |
4903 | set_imsm_ord_tbl_ent(map, i, i); | |
4904 | map2 = get_imsm_map(dev, 1); | |
4905 | /* Copy the current map */ | |
4906 | memcpy(map2, map, sizeof_imsm_map(map)); | |
4907 | map2->num_members = prev_num_members; | |
4908 | ||
4909 | /* calculate new size | |
4910 | */ | |
4911 | used_disks = imsm_num_data_members(dev, 0); | |
4912 | if (used_disks) { | |
4913 | unsigned long long array_blocks; | |
4914 | ||
4915 | array_blocks = | |
4916 | map->blocks_per_member | |
4917 | * used_disks; | |
4918 | /* round array size down to closest MB | |
4919 | */ | |
4920 | array_blocks = (array_blocks | |
4921 | >> SECT_PER_MB_SHIFT) | |
4922 | << SECT_PER_MB_SHIFT; | |
4923 | dev->size_low = | |
4924 | __cpu_to_le32((__u32)array_blocks); | |
4925 | dev->size_high = | |
4926 | __cpu_to_le32( | |
4927 | (__u32)(array_blocks >> 32)); | |
4928 | } | |
4929 | super->updates_pending++; | |
4930 | } | |
4931 | } | |
4932 | ||
aad6f216 | 4933 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
4934 | * states are handled in imsm_set_disk() with one exception, when a |
4935 | * resync is stopped due to a new failure this routine will set the | |
4936 | * 'degraded' state for the array. | |
4937 | */ | |
01f157d7 | 4938 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
4939 | { |
4940 | int inst = a->info.container_member; | |
4941 | struct intel_super *super = a->container->sb; | |
949c47a0 | 4942 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 4943 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd DW |
4944 | int failed = imsm_count_failed(super, dev); |
4945 | __u8 map_state = imsm_check_degraded(super, dev, failed); | |
1e5c6983 | 4946 | __u32 blocks_per_unit; |
a862209d | 4947 | |
1af97990 AK |
4948 | if (dev->vol.migr_state && |
4949 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
4950 | /* array state change is blocked due to reshape action | |
aad6f216 N |
4951 | * We might need to |
4952 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
4953 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
4954 | * - update curr_migr_unit | |
1af97990 | 4955 | */ |
aad6f216 N |
4956 | if (a->curr_action == reshape) { |
4957 | /* still reshaping, maybe update curr_migr_unit */ | |
4958 | long long blocks_per_unit = blocks_per_migr_unit(dev); | |
4959 | long long unit = a->last_checkpoint; | |
6345120e AK |
4960 | if (blocks_per_unit) { |
4961 | unit /= blocks_per_unit; | |
4962 | if (unit > | |
4963 | __le32_to_cpu(dev->vol.curr_migr_unit)) { | |
4964 | dev->vol.curr_migr_unit = | |
4965 | __cpu_to_le32(unit); | |
4966 | super->updates_pending++; | |
4967 | } | |
aad6f216 | 4968 | } |
a4546b61 | 4969 | return 0; |
aad6f216 N |
4970 | } else { |
4971 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
4972 | /* for some reason we aborted the reshape. | |
4973 | * Better clean up | |
4974 | */ | |
4975 | struct imsm_map *map2 = get_imsm_map(dev, 1); | |
4976 | dev->vol.migr_state = 0; | |
4977 | dev->vol.migr_type = 0; | |
4978 | dev->vol.curr_migr_unit = 0; | |
4979 | memcpy(map, map2, sizeof_imsm_map(map2)); | |
4980 | super->updates_pending++; | |
4981 | } | |
4982 | if (a->last_checkpoint >= a->info.component_size) { | |
4983 | unsigned long long array_blocks; | |
4984 | int used_disks; | |
4985 | /* it seems the reshape is all done */ | |
4986 | dev->vol.migr_state = 0; | |
4987 | dev->vol.migr_type = 0; | |
4988 | dev->vol.curr_migr_unit = 0; | |
4989 | ||
98130f40 | 4990 | used_disks = imsm_num_data_members(dev, -1); |
aad6f216 N |
4991 | array_blocks = map->blocks_per_member * used_disks; |
4992 | /* round array size down to closest MB */ | |
4993 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) | |
4994 | << SECT_PER_MB_SHIFT; | |
4995 | dev->size_low = __cpu_to_le32((__u32) array_blocks); | |
4996 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
4997 | a->info.custom_array_size = array_blocks; | |
4998 | a->check_reshape = 1; /* encourage manager to update | |
4999 | * array size | |
5000 | */ | |
5001 | super->updates_pending++; | |
0e2d1a4e | 5002 | imsm_progress_container_reshape(super); |
aad6f216 N |
5003 | } |
5004 | } | |
1af97990 AK |
5005 | } |
5006 | ||
47ee5a45 | 5007 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
5008 | if (consistent == 2) |
5009 | handle_missing(super, dev); | |
1e5c6983 | 5010 | |
0c046afd | 5011 | if (consistent == 2 && |
b7941fd6 | 5012 | (!is_resync_complete(&a->info) || |
0c046afd DW |
5013 | map_state != IMSM_T_STATE_NORMAL || |
5014 | dev->vol.migr_state)) | |
01f157d7 | 5015 | consistent = 0; |
272906ef | 5016 | |
b7941fd6 | 5017 | if (is_resync_complete(&a->info)) { |
0c046afd | 5018 | /* complete intialization / resync, |
0556e1a2 DW |
5019 | * recovery and interrupted recovery is completed in |
5020 | * ->set_disk | |
0c046afd DW |
5021 | */ |
5022 | if (is_resyncing(dev)) { | |
5023 | dprintf("imsm: mark resync done\n"); | |
f8f603f1 | 5024 | end_migration(dev, map_state); |
115c3803 | 5025 | super->updates_pending++; |
484240d8 | 5026 | a->last_checkpoint = 0; |
115c3803 | 5027 | } |
0c046afd DW |
5028 | } else if (!is_resyncing(dev) && !failed) { |
5029 | /* mark the start of the init process if nothing is failed */ | |
b7941fd6 | 5030 | dprintf("imsm: mark resync start\n"); |
1484e727 | 5031 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
e3bba0e0 | 5032 | migrate(dev, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 DW |
5033 | else |
5034 | migrate(dev, IMSM_T_STATE_NORMAL, MIGR_REPAIR); | |
3393c6af | 5035 | super->updates_pending++; |
115c3803 | 5036 | } |
a862209d | 5037 | |
1e5c6983 DW |
5038 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
5039 | blocks_per_unit = blocks_per_migr_unit(dev); | |
4f0a7acc | 5040 | if (blocks_per_unit) { |
1e5c6983 DW |
5041 | __u32 units32; |
5042 | __u64 units; | |
5043 | ||
4f0a7acc | 5044 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
5045 | units32 = units; |
5046 | ||
5047 | /* check that we did not overflow 32-bits, and that | |
5048 | * curr_migr_unit needs updating | |
5049 | */ | |
5050 | if (units32 == units && | |
5051 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { | |
5052 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
5053 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
5054 | super->updates_pending++; | |
5055 | } | |
5056 | } | |
f8f603f1 | 5057 | |
3393c6af | 5058 | /* mark dirty / clean */ |
0c046afd | 5059 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 5060 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
5061 | if (consistent) |
5062 | dev->vol.dirty = 0; | |
5063 | else | |
5064 | dev->vol.dirty = 1; | |
a862209d DW |
5065 | super->updates_pending++; |
5066 | } | |
28bce06f AK |
5067 | |
5068 | /* finalize online capacity expansion/reshape */ | |
5069 | if ((a->curr_action != reshape) && | |
5070 | (a->prev_action == reshape)) { | |
5071 | struct mdinfo *mdi; | |
5072 | ||
5073 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
5074 | imsm_set_disk(a, mdi->disk.raid_disk, mdi->curr_state); | |
5075 | } | |
5076 | ||
01f157d7 | 5077 | return consistent; |
a862209d DW |
5078 | } |
5079 | ||
8d45d196 | 5080 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 5081 | { |
8d45d196 DW |
5082 | int inst = a->info.container_member; |
5083 | struct intel_super *super = a->container->sb; | |
949c47a0 | 5084 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5085 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 | 5086 | struct imsm_disk *disk; |
0c046afd | 5087 | int failed; |
b10b37b8 | 5088 | __u32 ord; |
0c046afd | 5089 | __u8 map_state; |
8d45d196 DW |
5090 | |
5091 | if (n > map->num_members) | |
5092 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
5093 | n, map->num_members - 1); | |
5094 | ||
5095 | if (n < 0) | |
5096 | return; | |
5097 | ||
4e6e574a | 5098 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 5099 | |
98130f40 | 5100 | ord = get_imsm_ord_tbl_ent(dev, n, -1); |
b10b37b8 | 5101 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 5102 | |
5802a811 | 5103 | /* check for new failures */ |
0556e1a2 DW |
5104 | if (state & DS_FAULTY) { |
5105 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
5106 | super->updates_pending++; | |
8d45d196 | 5107 | } |
47ee5a45 | 5108 | |
19859edc | 5109 | /* check if in_sync */ |
0556e1a2 | 5110 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
b10b37b8 DW |
5111 | struct imsm_map *migr_map = get_imsm_map(dev, 1); |
5112 | ||
5113 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
5114 | super->updates_pending++; |
5115 | } | |
8d45d196 | 5116 | |
0c046afd DW |
5117 | failed = imsm_count_failed(super, dev); |
5118 | map_state = imsm_check_degraded(super, dev, failed); | |
5802a811 | 5119 | |
0c046afd DW |
5120 | /* check if recovery complete, newly degraded, or failed */ |
5121 | if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) { | |
f8f603f1 | 5122 | end_migration(dev, map_state); |
0556e1a2 DW |
5123 | map = get_imsm_map(dev, 0); |
5124 | map->failed_disk_num = ~0; | |
0c046afd | 5125 | super->updates_pending++; |
484240d8 | 5126 | a->last_checkpoint = 0; |
0c046afd DW |
5127 | } else if (map_state == IMSM_T_STATE_DEGRADED && |
5128 | map->map_state != map_state && | |
5129 | !dev->vol.migr_state) { | |
5130 | dprintf("imsm: mark degraded\n"); | |
5131 | map->map_state = map_state; | |
5132 | super->updates_pending++; | |
484240d8 | 5133 | a->last_checkpoint = 0; |
0c046afd DW |
5134 | } else if (map_state == IMSM_T_STATE_FAILED && |
5135 | map->map_state != map_state) { | |
5136 | dprintf("imsm: mark failed\n"); | |
f8f603f1 | 5137 | end_migration(dev, map_state); |
0c046afd | 5138 | super->updates_pending++; |
484240d8 | 5139 | a->last_checkpoint = 0; |
28bce06f AK |
5140 | } else if (is_gen_migration(dev)) { |
5141 | dprintf("imsm: Detected General Migration in state: "); | |
5142 | if (map_state == IMSM_T_STATE_NORMAL) { | |
5143 | end_migration(dev, map_state); | |
5144 | map = get_imsm_map(dev, 0); | |
5145 | map->failed_disk_num = ~0; | |
5146 | dprintf("normal\n"); | |
5147 | } else { | |
5148 | if (map_state == IMSM_T_STATE_DEGRADED) { | |
5149 | printf("degraded\n"); | |
5150 | end_migration(dev, map_state); | |
5151 | } else { | |
5152 | dprintf("failed\n"); | |
5153 | } | |
5154 | map->map_state = map_state; | |
5155 | } | |
5156 | super->updates_pending++; | |
5802a811 | 5157 | } |
845dea95 NB |
5158 | } |
5159 | ||
f796af5d | 5160 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 5161 | { |
f796af5d | 5162 | void *buf = mpb; |
c2a1e7da DW |
5163 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
5164 | unsigned long long dsize; | |
5165 | unsigned long long sectors; | |
5166 | ||
5167 | get_dev_size(fd, NULL, &dsize); | |
5168 | ||
272f648f DW |
5169 | if (mpb_size > 512) { |
5170 | /* -1 to account for anchor */ | |
5171 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 5172 | |
272f648f DW |
5173 | /* write the extended mpb to the sectors preceeding the anchor */ |
5174 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
5175 | return 1; | |
c2a1e7da | 5176 | |
f21e18ca N |
5177 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
5178 | != 512 * sectors) | |
272f648f DW |
5179 | return 1; |
5180 | } | |
c2a1e7da | 5181 | |
272f648f DW |
5182 | /* first block is stored on second to last sector of the disk */ |
5183 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
5184 | return 1; |
5185 | ||
f796af5d | 5186 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
5187 | return 1; |
5188 | ||
c2a1e7da DW |
5189 | return 0; |
5190 | } | |
5191 | ||
2e735d19 | 5192 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 5193 | { |
2e735d19 | 5194 | struct intel_super *super = container->sb; |
c2a1e7da | 5195 | |
1a64be56 | 5196 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
5197 | if (!super->updates_pending) |
5198 | return; | |
5199 | ||
36988a3d | 5200 | write_super_imsm(container, 0); |
c2a1e7da DW |
5201 | |
5202 | super->updates_pending = 0; | |
845dea95 NB |
5203 | } |
5204 | ||
272906ef DW |
5205 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
5206 | { | |
5207 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 5208 | int i = get_imsm_disk_idx(dev, idx, -1); |
272906ef DW |
5209 | struct dl *dl; |
5210 | ||
5211 | for (dl = super->disks; dl; dl = dl->next) | |
5212 | if (dl->index == i) | |
5213 | break; | |
5214 | ||
25ed7e59 | 5215 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
5216 | dl = NULL; |
5217 | ||
5218 | if (dl) | |
5219 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
5220 | ||
5221 | return dl; | |
5222 | } | |
5223 | ||
a20d2ba5 | 5224 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
5225 | struct active_array *a, int activate_new, |
5226 | struct mdinfo *additional_test_list) | |
272906ef DW |
5227 | { |
5228 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 5229 | int idx = get_imsm_disk_idx(dev, slot, -1); |
a20d2ba5 DW |
5230 | struct imsm_super *mpb = super->anchor; |
5231 | struct imsm_map *map; | |
272906ef DW |
5232 | unsigned long long pos; |
5233 | struct mdinfo *d; | |
5234 | struct extent *ex; | |
a20d2ba5 | 5235 | int i, j; |
272906ef | 5236 | int found; |
569cc43f DW |
5237 | __u32 array_start = 0; |
5238 | __u32 array_end = 0; | |
272906ef | 5239 | struct dl *dl; |
6c932028 | 5240 | struct mdinfo *test_list; |
272906ef DW |
5241 | |
5242 | for (dl = super->disks; dl; dl = dl->next) { | |
5243 | /* If in this array, skip */ | |
5244 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
5245 | if (d->state_fd >= 0 && |
5246 | d->disk.major == dl->major && | |
272906ef | 5247 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
5248 | dprintf("%x:%x already in array\n", |
5249 | dl->major, dl->minor); | |
272906ef DW |
5250 | break; |
5251 | } | |
5252 | if (d) | |
5253 | continue; | |
6c932028 AK |
5254 | test_list = additional_test_list; |
5255 | while (test_list) { | |
5256 | if (test_list->disk.major == dl->major && | |
5257 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
5258 | dprintf("%x:%x already in additional test list\n", |
5259 | dl->major, dl->minor); | |
5260 | break; | |
5261 | } | |
6c932028 | 5262 | test_list = test_list->next; |
8ba77d32 | 5263 | } |
6c932028 | 5264 | if (test_list) |
8ba77d32 | 5265 | continue; |
272906ef | 5266 | |
e553d2a4 | 5267 | /* skip in use or failed drives */ |
25ed7e59 | 5268 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
5269 | dl->index == -2) { |
5270 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 5271 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
5272 | continue; |
5273 | } | |
5274 | ||
a20d2ba5 DW |
5275 | /* skip pure spares when we are looking for partially |
5276 | * assimilated drives | |
5277 | */ | |
5278 | if (dl->index == -1 && !activate_new) | |
5279 | continue; | |
5280 | ||
272906ef | 5281 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 5282 | * It needs to be able to cover all member volumes |
272906ef DW |
5283 | */ |
5284 | ex = get_extents(super, dl); | |
5285 | if (!ex) { | |
5286 | dprintf("cannot get extents\n"); | |
5287 | continue; | |
5288 | } | |
a20d2ba5 DW |
5289 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5290 | dev = get_imsm_dev(super, i); | |
5291 | map = get_imsm_map(dev, 0); | |
272906ef | 5292 | |
a20d2ba5 DW |
5293 | /* check if this disk is already a member of |
5294 | * this array | |
272906ef | 5295 | */ |
620b1713 | 5296 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
5297 | continue; |
5298 | ||
5299 | found = 0; | |
5300 | j = 0; | |
5301 | pos = 0; | |
5302 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
329c8278 DW |
5303 | array_end = array_start + |
5304 | __le32_to_cpu(map->blocks_per_member) - 1; | |
a20d2ba5 DW |
5305 | |
5306 | do { | |
5307 | /* check that we can start at pba_of_lba0 with | |
5308 | * blocks_per_member of space | |
5309 | */ | |
329c8278 | 5310 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
5311 | found = 1; |
5312 | break; | |
5313 | } | |
5314 | pos = ex[j].start + ex[j].size; | |
5315 | j++; | |
5316 | } while (ex[j-1].size); | |
5317 | ||
5318 | if (!found) | |
272906ef | 5319 | break; |
a20d2ba5 | 5320 | } |
272906ef DW |
5321 | |
5322 | free(ex); | |
a20d2ba5 | 5323 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
5324 | dprintf("%x:%x does not have %u to %u available\n", |
5325 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
5326 | /* No room */ |
5327 | continue; | |
a20d2ba5 DW |
5328 | } |
5329 | return dl; | |
272906ef DW |
5330 | } |
5331 | ||
5332 | return dl; | |
5333 | } | |
5334 | ||
95d07a2c LM |
5335 | |
5336 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
5337 | { | |
5338 | struct imsm_dev *dev2; | |
5339 | struct imsm_map *map; | |
5340 | struct dl *idisk; | |
5341 | int slot; | |
5342 | int idx; | |
5343 | __u8 state; | |
5344 | ||
5345 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
5346 | if (dev2) { | |
5347 | state = imsm_check_degraded(cont->sb, dev2, failed); | |
5348 | if (state == IMSM_T_STATE_FAILED) { | |
5349 | map = get_imsm_map(dev2, 0); | |
5350 | if (!map) | |
5351 | return 1; | |
5352 | for (slot = 0; slot < map->num_members; slot++) { | |
5353 | /* | |
5354 | * Check if failed disks are deleted from intel | |
5355 | * disk list or are marked to be deleted | |
5356 | */ | |
98130f40 | 5357 | idx = get_imsm_disk_idx(dev2, slot, -1); |
95d07a2c LM |
5358 | idisk = get_imsm_dl_disk(cont->sb, idx); |
5359 | /* | |
5360 | * Do not rebuild the array if failed disks | |
5361 | * from failed sub-array are not removed from | |
5362 | * container. | |
5363 | */ | |
5364 | if (idisk && | |
5365 | is_failed(&idisk->disk) && | |
5366 | (idisk->action != DISK_REMOVE)) | |
5367 | return 0; | |
5368 | } | |
5369 | } | |
5370 | } | |
5371 | return 1; | |
5372 | } | |
5373 | ||
88758e9d DW |
5374 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
5375 | struct metadata_update **updates) | |
5376 | { | |
5377 | /** | |
d23fe947 DW |
5378 | * Find a device with unused free space and use it to replace a |
5379 | * failed/vacant region in an array. We replace failed regions one a | |
5380 | * array at a time. The result is that a new spare disk will be added | |
5381 | * to the first failed array and after the monitor has finished | |
5382 | * propagating failures the remainder will be consumed. | |
88758e9d | 5383 | * |
d23fe947 DW |
5384 | * FIXME add a capability for mdmon to request spares from another |
5385 | * container. | |
88758e9d DW |
5386 | */ |
5387 | ||
5388 | struct intel_super *super = a->container->sb; | |
88758e9d | 5389 | int inst = a->info.container_member; |
949c47a0 | 5390 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5391 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
5392 | int failed = a->info.array.raid_disks; |
5393 | struct mdinfo *rv = NULL; | |
5394 | struct mdinfo *d; | |
5395 | struct mdinfo *di; | |
5396 | struct metadata_update *mu; | |
5397 | struct dl *dl; | |
5398 | struct imsm_update_activate_spare *u; | |
5399 | int num_spares = 0; | |
5400 | int i; | |
95d07a2c | 5401 | int allowed; |
88758e9d DW |
5402 | |
5403 | for (d = a->info.devs ; d ; d = d->next) { | |
5404 | if ((d->curr_state & DS_FAULTY) && | |
5405 | d->state_fd >= 0) | |
5406 | /* wait for Removal to happen */ | |
5407 | return NULL; | |
5408 | if (d->state_fd >= 0) | |
5409 | failed--; | |
5410 | } | |
5411 | ||
5412 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
5413 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 AK |
5414 | |
5415 | if (dev->vol.migr_state && | |
5416 | dev->vol.migr_type == MIGR_GEN_MIGR) | |
5417 | /* No repair during migration */ | |
5418 | return NULL; | |
5419 | ||
89c67882 AK |
5420 | if (a->info.array.level == 4) |
5421 | /* No repair for takeovered array | |
5422 | * imsm doesn't support raid4 | |
5423 | */ | |
5424 | return NULL; | |
5425 | ||
fb49eef2 | 5426 | if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
5427 | return NULL; |
5428 | ||
95d07a2c LM |
5429 | /* |
5430 | * If there are any failed disks check state of the other volume. | |
5431 | * Block rebuild if the another one is failed until failed disks | |
5432 | * are removed from container. | |
5433 | */ | |
5434 | if (failed) { | |
5435 | dprintf("found failed disks in %s, check if there another" | |
5436 | "failed sub-array.\n", | |
5437 | dev->volume); | |
5438 | /* check if states of the other volumes allow for rebuild */ | |
5439 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
5440 | if (i != inst) { | |
5441 | allowed = imsm_rebuild_allowed(a->container, | |
5442 | i, failed); | |
5443 | if (!allowed) | |
5444 | return NULL; | |
5445 | } | |
5446 | } | |
5447 | } | |
5448 | ||
88758e9d | 5449 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
5450 | for (i = 0; i < a->info.array.raid_disks; i++) { |
5451 | for (d = a->info.devs ; d ; d = d->next) | |
5452 | if (d->disk.raid_disk == i) | |
5453 | break; | |
5454 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
5455 | if (d && (d->state_fd >= 0)) | |
5456 | continue; | |
5457 | ||
272906ef | 5458 | /* |
a20d2ba5 DW |
5459 | * OK, this device needs recovery. Try to re-add the |
5460 | * previous occupant of this slot, if this fails see if | |
5461 | * we can continue the assimilation of a spare that was | |
5462 | * partially assimilated, finally try to activate a new | |
5463 | * spare. | |
272906ef DW |
5464 | */ |
5465 | dl = imsm_readd(super, i, a); | |
5466 | if (!dl) | |
8ba77d32 | 5467 | dl = imsm_add_spare(super, i, a, 0, NULL); |
a20d2ba5 | 5468 | if (!dl) |
8ba77d32 | 5469 | dl = imsm_add_spare(super, i, a, 1, NULL); |
272906ef DW |
5470 | if (!dl) |
5471 | continue; | |
5472 | ||
5473 | /* found a usable disk with enough space */ | |
5474 | di = malloc(sizeof(*di)); | |
79244939 DW |
5475 | if (!di) |
5476 | continue; | |
272906ef DW |
5477 | memset(di, 0, sizeof(*di)); |
5478 | ||
5479 | /* dl->index will be -1 in the case we are activating a | |
5480 | * pristine spare. imsm_process_update() will create a | |
5481 | * new index in this case. Once a disk is found to be | |
5482 | * failed in all member arrays it is kicked from the | |
5483 | * metadata | |
5484 | */ | |
5485 | di->disk.number = dl->index; | |
d23fe947 | 5486 | |
272906ef DW |
5487 | /* (ab)use di->devs to store a pointer to the device |
5488 | * we chose | |
5489 | */ | |
5490 | di->devs = (struct mdinfo *) dl; | |
5491 | ||
5492 | di->disk.raid_disk = i; | |
5493 | di->disk.major = dl->major; | |
5494 | di->disk.minor = dl->minor; | |
5495 | di->disk.state = 0; | |
d23534e4 | 5496 | di->recovery_start = 0; |
272906ef DW |
5497 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); |
5498 | di->component_size = a->info.component_size; | |
5499 | di->container_member = inst; | |
148acb7b | 5500 | super->random = random32(); |
272906ef DW |
5501 | di->next = rv; |
5502 | rv = di; | |
5503 | num_spares++; | |
5504 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
5505 | i, di->data_offset); | |
88758e9d | 5506 | |
272906ef | 5507 | break; |
88758e9d DW |
5508 | } |
5509 | ||
5510 | if (!rv) | |
5511 | /* No spares found */ | |
5512 | return rv; | |
5513 | /* Now 'rv' has a list of devices to return. | |
5514 | * Create a metadata_update record to update the | |
5515 | * disk_ord_tbl for the array | |
5516 | */ | |
5517 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
5518 | if (mu) { |
5519 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
5520 | if (mu->buf == NULL) { | |
5521 | free(mu); | |
5522 | mu = NULL; | |
5523 | } | |
5524 | } | |
5525 | if (!mu) { | |
5526 | while (rv) { | |
5527 | struct mdinfo *n = rv->next; | |
5528 | ||
5529 | free(rv); | |
5530 | rv = n; | |
5531 | } | |
5532 | return NULL; | |
5533 | } | |
5534 | ||
88758e9d | 5535 | mu->space = NULL; |
cb23f1f4 | 5536 | mu->space_list = NULL; |
88758e9d DW |
5537 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
5538 | mu->next = *updates; | |
5539 | u = (struct imsm_update_activate_spare *) mu->buf; | |
5540 | ||
5541 | for (di = rv ; di ; di = di->next) { | |
5542 | u->type = update_activate_spare; | |
d23fe947 DW |
5543 | u->dl = (struct dl *) di->devs; |
5544 | di->devs = NULL; | |
88758e9d DW |
5545 | u->slot = di->disk.raid_disk; |
5546 | u->array = inst; | |
5547 | u->next = u + 1; | |
5548 | u++; | |
5549 | } | |
5550 | (u-1)->next = NULL; | |
5551 | *updates = mu; | |
5552 | ||
5553 | return rv; | |
5554 | } | |
5555 | ||
54c2c1ea | 5556 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 5557 | { |
54c2c1ea DW |
5558 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
5559 | struct imsm_map *map = get_imsm_map(dev, 0); | |
5560 | struct imsm_map *new_map = get_imsm_map(&u->dev, 0); | |
5561 | struct disk_info *inf = get_disk_info(u); | |
5562 | struct imsm_disk *disk; | |
8273f55e DW |
5563 | int i; |
5564 | int j; | |
8273f55e | 5565 | |
54c2c1ea | 5566 | for (i = 0; i < map->num_members; i++) { |
98130f40 | 5567 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, -1)); |
54c2c1ea DW |
5568 | for (j = 0; j < new_map->num_members; j++) |
5569 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
5570 | return 1; |
5571 | } | |
5572 | ||
5573 | return 0; | |
5574 | } | |
5575 | ||
1a64be56 LM |
5576 | |
5577 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
5578 | { | |
5579 | struct dl *dl = NULL; | |
5580 | for (dl = super->disks; dl; dl = dl->next) | |
5581 | if ((dl->major == major) && (dl->minor == minor)) | |
5582 | return dl; | |
5583 | return NULL; | |
5584 | } | |
5585 | ||
5586 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
5587 | { | |
5588 | struct dl *prev = NULL; | |
5589 | struct dl *dl; | |
5590 | ||
5591 | prev = NULL; | |
5592 | for (dl = super->disks; dl; dl = dl->next) { | |
5593 | if ((dl->major == major) && (dl->minor == minor)) { | |
5594 | /* remove */ | |
5595 | if (prev) | |
5596 | prev->next = dl->next; | |
5597 | else | |
5598 | super->disks = dl->next; | |
5599 | dl->next = NULL; | |
5600 | __free_imsm_disk(dl); | |
5601 | dprintf("%s: removed %x:%x\n", | |
5602 | __func__, major, minor); | |
5603 | break; | |
5604 | } | |
5605 | prev = dl; | |
5606 | } | |
5607 | return 0; | |
5608 | } | |
5609 | ||
f21e18ca | 5610 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 5611 | |
1a64be56 LM |
5612 | static int add_remove_disk_update(struct intel_super *super) |
5613 | { | |
5614 | int check_degraded = 0; | |
5615 | struct dl *disk = NULL; | |
5616 | /* add/remove some spares to/from the metadata/contrainer */ | |
5617 | while (super->disk_mgmt_list) { | |
5618 | struct dl *disk_cfg; | |
5619 | ||
5620 | disk_cfg = super->disk_mgmt_list; | |
5621 | super->disk_mgmt_list = disk_cfg->next; | |
5622 | disk_cfg->next = NULL; | |
5623 | ||
5624 | if (disk_cfg->action == DISK_ADD) { | |
5625 | disk_cfg->next = super->disks; | |
5626 | super->disks = disk_cfg; | |
5627 | check_degraded = 1; | |
5628 | dprintf("%s: added %x:%x\n", | |
5629 | __func__, disk_cfg->major, | |
5630 | disk_cfg->minor); | |
5631 | } else if (disk_cfg->action == DISK_REMOVE) { | |
5632 | dprintf("Disk remove action processed: %x.%x\n", | |
5633 | disk_cfg->major, disk_cfg->minor); | |
5634 | disk = get_disk_super(super, | |
5635 | disk_cfg->major, | |
5636 | disk_cfg->minor); | |
5637 | if (disk) { | |
5638 | /* store action status */ | |
5639 | disk->action = DISK_REMOVE; | |
5640 | /* remove spare disks only */ | |
5641 | if (disk->index == -1) { | |
5642 | remove_disk_super(super, | |
5643 | disk_cfg->major, | |
5644 | disk_cfg->minor); | |
5645 | } | |
5646 | } | |
5647 | /* release allocate disk structure */ | |
5648 | __free_imsm_disk(disk_cfg); | |
5649 | } | |
5650 | } | |
5651 | return check_degraded; | |
5652 | } | |
5653 | ||
2e5dc010 N |
5654 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
5655 | struct intel_super *super, | |
5656 | void ***space_list) | |
5657 | { | |
5658 | struct dl *new_disk; | |
5659 | struct intel_dev *id; | |
5660 | int i; | |
5661 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 5662 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
5663 | void **tofree = NULL; |
5664 | int devices_to_reshape = 1; | |
5665 | struct imsm_super *mpb = super->anchor; | |
5666 | int ret_val = 0; | |
5667 | ||
5668 | dprintf("imsm: imsm_process_update() for update_reshape\n"); | |
5669 | ||
5670 | /* enable spares to use in array */ | |
5671 | for (i = 0; i < delta_disks; i++) { | |
5672 | new_disk = get_disk_super(super, | |
5673 | major(u->new_disks[i]), | |
5674 | minor(u->new_disks[i])); | |
5675 | dprintf("imsm: imsm_process_update(): new disk " | |
5676 | "for reshape is: %i:%i (%p, index = %i)\n", | |
5677 | major(u->new_disks[i]), minor(u->new_disks[i]), | |
5678 | new_disk, new_disk->index); | |
5679 | if ((new_disk == NULL) || | |
5680 | ((new_disk->index >= 0) && | |
5681 | (new_disk->index < u->old_raid_disks))) | |
5682 | goto update_reshape_exit; | |
ee4beede | 5683 | new_disk->index = disk_count++; |
2e5dc010 N |
5684 | /* slot to fill in autolayout |
5685 | */ | |
5686 | new_disk->raiddisk = new_disk->index; | |
5687 | new_disk->disk.status |= | |
5688 | CONFIGURED_DISK; | |
5689 | new_disk->disk.status &= ~SPARE_DISK; | |
5690 | } | |
5691 | ||
5692 | dprintf("imsm: process_update(): update_reshape: volume set" | |
5693 | " mpb->num_raid_devs = %i\n", mpb->num_raid_devs); | |
5694 | /* manage changes in volume | |
5695 | */ | |
5696 | for (id = super->devlist ; id; id = id->next) { | |
5697 | void **sp = *space_list; | |
5698 | struct imsm_dev *newdev; | |
5699 | struct imsm_map *newmap, *oldmap; | |
5700 | ||
5701 | if (!sp) | |
5702 | continue; | |
5703 | *space_list = *sp; | |
5704 | newdev = (void*)sp; | |
5705 | /* Copy the dev, but not (all of) the map */ | |
5706 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
5707 | oldmap = get_imsm_map(id->dev, 0); | |
5708 | newmap = get_imsm_map(newdev, 0); | |
5709 | /* Copy the current map */ | |
5710 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
5711 | /* update one device only | |
5712 | */ | |
5713 | if (devices_to_reshape) { | |
5714 | int used_disks; | |
5715 | ||
5716 | dprintf("process_update(): modifying " | |
5717 | "subdev: %i\n", id->index); | |
5718 | devices_to_reshape--; | |
5719 | newdev->vol.migr_state = 1; | |
5720 | newdev->vol.curr_migr_unit = 0; | |
5721 | newdev->vol.migr_type = MIGR_GEN_MIGR; | |
5722 | newmap->num_members = u->new_raid_disks; | |
5723 | for (i = 0; i < delta_disks; i++) { | |
5724 | set_imsm_ord_tbl_ent(newmap, | |
5725 | u->old_raid_disks + i, | |
5726 | u->old_raid_disks + i); | |
5727 | } | |
5728 | /* New map is correct, now need to save old map | |
5729 | */ | |
5730 | newmap = get_imsm_map(newdev, 1); | |
5731 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
5732 | ||
5733 | /* calculate new size | |
5734 | */ | |
5735 | used_disks = imsm_num_data_members(newdev, 0); | |
5736 | if (used_disks) { | |
5737 | unsigned long long array_blocks; | |
5738 | ||
5739 | array_blocks = | |
5740 | newmap->blocks_per_member * used_disks; | |
5741 | /* round array size down to closest MB | |
5742 | */ | |
5743 | array_blocks = (array_blocks | |
5744 | >> SECT_PER_MB_SHIFT) | |
5745 | << SECT_PER_MB_SHIFT; | |
5746 | newdev->size_low = | |
5747 | __cpu_to_le32((__u32)array_blocks); | |
5748 | newdev->size_high = | |
5749 | __cpu_to_le32((__u32)(array_blocks >> 32)); | |
5750 | } | |
5751 | } | |
5752 | ||
5753 | sp = (void **)id->dev; | |
5754 | id->dev = newdev; | |
5755 | *sp = tofree; | |
5756 | tofree = sp; | |
5757 | } | |
819bc634 AK |
5758 | if (tofree) |
5759 | *space_list = tofree; | |
2e5dc010 N |
5760 | ret_val = 1; |
5761 | ||
5762 | update_reshape_exit: | |
5763 | ||
5764 | return ret_val; | |
5765 | } | |
5766 | ||
e8319a19 DW |
5767 | static void imsm_process_update(struct supertype *st, |
5768 | struct metadata_update *update) | |
5769 | { | |
5770 | /** | |
5771 | * crack open the metadata_update envelope to find the update record | |
5772 | * update can be one of: | |
d195167d AK |
5773 | * update_reshape_container_disks - all the arrays in the container |
5774 | * are being reshaped to have more devices. We need to mark | |
5775 | * the arrays for general migration and convert selected spares | |
5776 | * into active devices. | |
5777 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
5778 | * device in an array, update the disk_ord_tbl. If this disk is |
5779 | * present in all member arrays then also clear the SPARE_DISK | |
5780 | * flag | |
d195167d AK |
5781 | * update_create_array |
5782 | * update_kill_array | |
5783 | * update_rename_array | |
5784 | * update_add_remove_disk | |
e8319a19 DW |
5785 | */ |
5786 | struct intel_super *super = st->sb; | |
4d7b1503 | 5787 | struct imsm_super *mpb; |
e8319a19 DW |
5788 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
5789 | ||
4d7b1503 DW |
5790 | /* update requires a larger buf but the allocation failed */ |
5791 | if (super->next_len && !super->next_buf) { | |
5792 | super->next_len = 0; | |
5793 | return; | |
5794 | } | |
5795 | ||
5796 | if (super->next_buf) { | |
5797 | memcpy(super->next_buf, super->buf, super->len); | |
5798 | free(super->buf); | |
5799 | super->len = super->next_len; | |
5800 | super->buf = super->next_buf; | |
5801 | ||
5802 | super->next_len = 0; | |
5803 | super->next_buf = NULL; | |
5804 | } | |
5805 | ||
5806 | mpb = super->anchor; | |
5807 | ||
e8319a19 | 5808 | switch (type) { |
78b10e66 | 5809 | case update_reshape_container_disks: { |
d195167d | 5810 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
5811 | if (apply_reshape_container_disks_update( |
5812 | u, super, &update->space_list)) | |
5813 | super->updates_pending++; | |
78b10e66 N |
5814 | break; |
5815 | } | |
e8319a19 DW |
5816 | case update_activate_spare: { |
5817 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 5818 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 5819 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd | 5820 | struct imsm_map *migr_map; |
e8319a19 DW |
5821 | struct active_array *a; |
5822 | struct imsm_disk *disk; | |
0c046afd | 5823 | __u8 to_state; |
e8319a19 | 5824 | struct dl *dl; |
e8319a19 | 5825 | unsigned int found; |
0c046afd | 5826 | int failed; |
98130f40 | 5827 | int victim = get_imsm_disk_idx(dev, u->slot, -1); |
e8319a19 DW |
5828 | int i; |
5829 | ||
5830 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 5831 | if (dl == u->dl) |
e8319a19 DW |
5832 | break; |
5833 | ||
5834 | if (!dl) { | |
5835 | fprintf(stderr, "error: imsm_activate_spare passed " | |
1f24f035 DW |
5836 | "an unknown disk (index: %d)\n", |
5837 | u->dl->index); | |
e8319a19 DW |
5838 | return; |
5839 | } | |
5840 | ||
5841 | super->updates_pending++; | |
5842 | ||
0c046afd DW |
5843 | /* count failures (excluding rebuilds and the victim) |
5844 | * to determine map[0] state | |
5845 | */ | |
5846 | failed = 0; | |
5847 | for (i = 0; i < map->num_members; i++) { | |
5848 | if (i == u->slot) | |
5849 | continue; | |
98130f40 AK |
5850 | disk = get_imsm_disk(super, |
5851 | get_imsm_disk_idx(dev, i, -1)); | |
25ed7e59 | 5852 | if (!disk || is_failed(disk)) |
0c046afd DW |
5853 | failed++; |
5854 | } | |
5855 | ||
d23fe947 DW |
5856 | /* adding a pristine spare, assign a new index */ |
5857 | if (dl->index < 0) { | |
5858 | dl->index = super->anchor->num_disks; | |
5859 | super->anchor->num_disks++; | |
5860 | } | |
d23fe947 | 5861 | disk = &dl->disk; |
f2f27e63 DW |
5862 | disk->status |= CONFIGURED_DISK; |
5863 | disk->status &= ~SPARE_DISK; | |
e8319a19 | 5864 | |
0c046afd DW |
5865 | /* mark rebuild */ |
5866 | to_state = imsm_check_degraded(super, dev, failed); | |
5867 | map->map_state = IMSM_T_STATE_DEGRADED; | |
e3bba0e0 | 5868 | migrate(dev, to_state, MIGR_REBUILD); |
0c046afd DW |
5869 | migr_map = get_imsm_map(dev, 1); |
5870 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); | |
5871 | set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD); | |
5872 | ||
148acb7b DW |
5873 | /* update the family_num to mark a new container |
5874 | * generation, being careful to record the existing | |
5875 | * family_num in orig_family_num to clean up after | |
5876 | * earlier mdadm versions that neglected to set it. | |
5877 | */ | |
5878 | if (mpb->orig_family_num == 0) | |
5879 | mpb->orig_family_num = mpb->family_num; | |
5880 | mpb->family_num += super->random; | |
5881 | ||
e8319a19 DW |
5882 | /* count arrays using the victim in the metadata */ |
5883 | found = 0; | |
5884 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 5885 | dev = get_imsm_dev(super, a->info.container_member); |
620b1713 DW |
5886 | map = get_imsm_map(dev, 0); |
5887 | ||
5888 | if (get_imsm_disk_slot(map, victim) >= 0) | |
5889 | found++; | |
e8319a19 DW |
5890 | } |
5891 | ||
24565c9a | 5892 | /* delete the victim if it is no longer being |
e8319a19 DW |
5893 | * utilized anywhere |
5894 | */ | |
e8319a19 | 5895 | if (!found) { |
ae6aad82 | 5896 | struct dl **dlp; |
24565c9a | 5897 | |
47ee5a45 DW |
5898 | /* We know that 'manager' isn't touching anything, |
5899 | * so it is safe to delete | |
5900 | */ | |
24565c9a | 5901 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) |
ae6aad82 DW |
5902 | if ((*dlp)->index == victim) |
5903 | break; | |
47ee5a45 DW |
5904 | |
5905 | /* victim may be on the missing list */ | |
5906 | if (!*dlp) | |
5907 | for (dlp = &super->missing; *dlp; dlp = &(*dlp)->next) | |
5908 | if ((*dlp)->index == victim) | |
5909 | break; | |
24565c9a | 5910 | imsm_delete(super, dlp, victim); |
e8319a19 | 5911 | } |
8273f55e DW |
5912 | break; |
5913 | } | |
5914 | case update_create_array: { | |
5915 | /* someone wants to create a new array, we need to be aware of | |
5916 | * a few races/collisions: | |
5917 | * 1/ 'Create' called by two separate instances of mdadm | |
5918 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
5919 | * devices that have since been assimilated via | |
5920 | * activate_spare. | |
5921 | * In the event this update can not be carried out mdadm will | |
5922 | * (FIX ME) notice that its update did not take hold. | |
5923 | */ | |
5924 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 5925 | struct intel_dev *dv; |
8273f55e DW |
5926 | struct imsm_dev *dev; |
5927 | struct imsm_map *map, *new_map; | |
5928 | unsigned long long start, end; | |
5929 | unsigned long long new_start, new_end; | |
5930 | int i; | |
54c2c1ea DW |
5931 | struct disk_info *inf; |
5932 | struct dl *dl; | |
8273f55e DW |
5933 | |
5934 | /* handle racing creates: first come first serve */ | |
5935 | if (u->dev_idx < mpb->num_raid_devs) { | |
5936 | dprintf("%s: subarray %d already defined\n", | |
5937 | __func__, u->dev_idx); | |
ba2de7ba | 5938 | goto create_error; |
8273f55e DW |
5939 | } |
5940 | ||
5941 | /* check update is next in sequence */ | |
5942 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
5943 | dprintf("%s: can not create array %d expected index %d\n", |
5944 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 5945 | goto create_error; |
8273f55e DW |
5946 | } |
5947 | ||
a965f303 | 5948 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
5949 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
5950 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
54c2c1ea | 5951 | inf = get_disk_info(u); |
8273f55e DW |
5952 | |
5953 | /* handle activate_spare versus create race: | |
5954 | * check to make sure that overlapping arrays do not include | |
5955 | * overalpping disks | |
5956 | */ | |
5957 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 5958 | dev = get_imsm_dev(super, i); |
a965f303 | 5959 | map = get_imsm_map(dev, 0); |
8273f55e DW |
5960 | start = __le32_to_cpu(map->pba_of_lba0); |
5961 | end = start + __le32_to_cpu(map->blocks_per_member); | |
5962 | if ((new_start >= start && new_start <= end) || | |
5963 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
5964 | /* overlap */; |
5965 | else | |
5966 | continue; | |
5967 | ||
5968 | if (disks_overlap(super, i, u)) { | |
8273f55e | 5969 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 5970 | goto create_error; |
8273f55e DW |
5971 | } |
5972 | } | |
8273f55e | 5973 | |
949c47a0 DW |
5974 | /* check that prepare update was successful */ |
5975 | if (!update->space) { | |
5976 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 5977 | goto create_error; |
949c47a0 DW |
5978 | } |
5979 | ||
54c2c1ea DW |
5980 | /* check that all disks are still active before committing |
5981 | * changes. FIXME: could we instead handle this by creating a | |
5982 | * degraded array? That's probably not what the user expects, | |
5983 | * so better to drop this update on the floor. | |
5984 | */ | |
5985 | for (i = 0; i < new_map->num_members; i++) { | |
5986 | dl = serial_to_dl(inf[i].serial, super); | |
5987 | if (!dl) { | |
5988 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 5989 | goto create_error; |
54c2c1ea | 5990 | } |
949c47a0 DW |
5991 | } |
5992 | ||
8273f55e | 5993 | super->updates_pending++; |
54c2c1ea DW |
5994 | |
5995 | /* convert spares to members and fixup ord_tbl */ | |
5996 | for (i = 0; i < new_map->num_members; i++) { | |
5997 | dl = serial_to_dl(inf[i].serial, super); | |
5998 | if (dl->index == -1) { | |
5999 | dl->index = mpb->num_disks; | |
6000 | mpb->num_disks++; | |
6001 | dl->disk.status |= CONFIGURED_DISK; | |
6002 | dl->disk.status &= ~SPARE_DISK; | |
6003 | } | |
6004 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
6005 | } | |
6006 | ||
ba2de7ba DW |
6007 | dv = update->space; |
6008 | dev = dv->dev; | |
949c47a0 DW |
6009 | update->space = NULL; |
6010 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
6011 | dv->index = u->dev_idx; |
6012 | dv->next = super->devlist; | |
6013 | super->devlist = dv; | |
8273f55e | 6014 | mpb->num_raid_devs++; |
8273f55e | 6015 | |
4d1313e9 | 6016 | imsm_update_version_info(super); |
8273f55e | 6017 | break; |
ba2de7ba DW |
6018 | create_error: |
6019 | /* mdmon knows how to release update->space, but not | |
6020 | * ((struct intel_dev *) update->space)->dev | |
6021 | */ | |
6022 | if (update->space) { | |
6023 | dv = update->space; | |
6024 | free(dv->dev); | |
6025 | } | |
8273f55e | 6026 | break; |
e8319a19 | 6027 | } |
33414a01 DW |
6028 | case update_kill_array: { |
6029 | struct imsm_update_kill_array *u = (void *) update->buf; | |
6030 | int victim = u->dev_idx; | |
6031 | struct active_array *a; | |
6032 | struct intel_dev **dp; | |
6033 | struct imsm_dev *dev; | |
6034 | ||
6035 | /* sanity check that we are not affecting the uuid of | |
6036 | * active arrays, or deleting an active array | |
6037 | * | |
6038 | * FIXME when immutable ids are available, but note that | |
6039 | * we'll also need to fixup the invalidated/active | |
6040 | * subarray indexes in mdstat | |
6041 | */ | |
6042 | for (a = st->arrays; a; a = a->next) | |
6043 | if (a->info.container_member >= victim) | |
6044 | break; | |
6045 | /* by definition if mdmon is running at least one array | |
6046 | * is active in the container, so checking | |
6047 | * mpb->num_raid_devs is just extra paranoia | |
6048 | */ | |
6049 | dev = get_imsm_dev(super, victim); | |
6050 | if (a || !dev || mpb->num_raid_devs == 1) { | |
6051 | dprintf("failed to delete subarray-%d\n", victim); | |
6052 | break; | |
6053 | } | |
6054 | ||
6055 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 6056 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
6057 | *dp = (*dp)->next; |
6058 | } else { | |
f21e18ca | 6059 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
6060 | (*dp)->index--; |
6061 | dp = &(*dp)->next; | |
6062 | } | |
6063 | mpb->num_raid_devs--; | |
6064 | super->updates_pending++; | |
6065 | break; | |
6066 | } | |
aa534678 DW |
6067 | case update_rename_array: { |
6068 | struct imsm_update_rename_array *u = (void *) update->buf; | |
6069 | char name[MAX_RAID_SERIAL_LEN+1]; | |
6070 | int target = u->dev_idx; | |
6071 | struct active_array *a; | |
6072 | struct imsm_dev *dev; | |
6073 | ||
6074 | /* sanity check that we are not affecting the uuid of | |
6075 | * an active array | |
6076 | */ | |
6077 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
6078 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
6079 | for (a = st->arrays; a; a = a->next) | |
6080 | if (a->info.container_member == target) | |
6081 | break; | |
6082 | dev = get_imsm_dev(super, u->dev_idx); | |
6083 | if (a || !dev || !check_name(super, name, 1)) { | |
6084 | dprintf("failed to rename subarray-%d\n", target); | |
6085 | break; | |
6086 | } | |
6087 | ||
cdbe98cd | 6088 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
6089 | super->updates_pending++; |
6090 | break; | |
6091 | } | |
1a64be56 | 6092 | case update_add_remove_disk: { |
43dad3d6 | 6093 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
6094 | * being added, check teh status of add_remove_disk |
6095 | * if discs has been added. | |
6096 | */ | |
6097 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 6098 | struct active_array *a; |
072b727f DW |
6099 | |
6100 | super->updates_pending++; | |
1a64be56 | 6101 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
6102 | a->check_degraded = 1; |
6103 | } | |
43dad3d6 | 6104 | break; |
e8319a19 | 6105 | } |
1a64be56 LM |
6106 | default: |
6107 | fprintf(stderr, "error: unsuported process update type:" | |
6108 | "(type: %d)\n", type); | |
6109 | } | |
e8319a19 | 6110 | } |
88758e9d | 6111 | |
8273f55e DW |
6112 | static void imsm_prepare_update(struct supertype *st, |
6113 | struct metadata_update *update) | |
6114 | { | |
949c47a0 | 6115 | /** |
4d7b1503 DW |
6116 | * Allocate space to hold new disk entries, raid-device entries or a new |
6117 | * mpb if necessary. The manager synchronously waits for updates to | |
6118 | * complete in the monitor, so new mpb buffers allocated here can be | |
6119 | * integrated by the monitor thread without worrying about live pointers | |
6120 | * in the manager thread. | |
8273f55e | 6121 | */ |
949c47a0 | 6122 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
6123 | struct intel_super *super = st->sb; |
6124 | struct imsm_super *mpb = super->anchor; | |
6125 | size_t buf_len; | |
6126 | size_t len = 0; | |
949c47a0 DW |
6127 | |
6128 | switch (type) { | |
78b10e66 | 6129 | case update_reshape_container_disks: { |
d195167d AK |
6130 | /* Every raid device in the container is about to |
6131 | * gain some more devices, and we will enter a | |
6132 | * reconfiguration. | |
6133 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
6134 | * will now hold 2 of them. | |
6135 | * Thus we need new 'struct imsm_dev' allocations sized | |
6136 | * as sizeof_imsm_dev but with more devices in both maps. | |
6137 | */ | |
6138 | struct imsm_update_reshape *u = (void *)update->buf; | |
6139 | struct intel_dev *dl; | |
6140 | void **space_tail = (void**)&update->space_list; | |
6141 | ||
6142 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
6143 | ||
6144 | for (dl = super->devlist; dl; dl = dl->next) { | |
6145 | int size = sizeof_imsm_dev(dl->dev, 1); | |
6146 | void *s; | |
d677e0b8 AK |
6147 | if (u->new_raid_disks > u->old_raid_disks) |
6148 | size += sizeof(__u32)*2* | |
6149 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
6150 | s = malloc(size); |
6151 | if (!s) | |
6152 | break; | |
6153 | *space_tail = s; | |
6154 | space_tail = s; | |
6155 | *space_tail = NULL; | |
6156 | } | |
6157 | ||
6158 | len = disks_to_mpb_size(u->new_raid_disks); | |
6159 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
6160 | break; |
6161 | } | |
949c47a0 DW |
6162 | case update_create_array: { |
6163 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 6164 | struct intel_dev *dv; |
54c2c1ea DW |
6165 | struct imsm_dev *dev = &u->dev; |
6166 | struct imsm_map *map = get_imsm_map(dev, 0); | |
6167 | struct dl *dl; | |
6168 | struct disk_info *inf; | |
6169 | int i; | |
6170 | int activate = 0; | |
949c47a0 | 6171 | |
54c2c1ea DW |
6172 | inf = get_disk_info(u); |
6173 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
6174 | /* allocate a new super->devlist entry */ |
6175 | dv = malloc(sizeof(*dv)); | |
6176 | if (dv) { | |
6177 | dv->dev = malloc(len); | |
6178 | if (dv->dev) | |
6179 | update->space = dv; | |
6180 | else { | |
6181 | free(dv); | |
6182 | update->space = NULL; | |
6183 | } | |
6184 | } | |
949c47a0 | 6185 | |
54c2c1ea DW |
6186 | /* count how many spares will be converted to members */ |
6187 | for (i = 0; i < map->num_members; i++) { | |
6188 | dl = serial_to_dl(inf[i].serial, super); | |
6189 | if (!dl) { | |
6190 | /* hmm maybe it failed?, nothing we can do about | |
6191 | * it here | |
6192 | */ | |
6193 | continue; | |
6194 | } | |
6195 | if (count_memberships(dl, super) == 0) | |
6196 | activate++; | |
6197 | } | |
6198 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
6199 | break; |
6200 | default: | |
6201 | break; | |
6202 | } | |
6203 | } | |
8273f55e | 6204 | |
4d7b1503 DW |
6205 | /* check if we need a larger metadata buffer */ |
6206 | if (super->next_buf) | |
6207 | buf_len = super->next_len; | |
6208 | else | |
6209 | buf_len = super->len; | |
6210 | ||
6211 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
6212 | /* ok we need a larger buf than what is currently allocated | |
6213 | * if this allocation fails process_update will notice that | |
6214 | * ->next_len is set and ->next_buf is NULL | |
6215 | */ | |
6216 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
6217 | if (super->next_buf) | |
6218 | free(super->next_buf); | |
6219 | ||
6220 | super->next_len = buf_len; | |
1f45a8ad DW |
6221 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
6222 | memset(super->next_buf, 0, buf_len); | |
6223 | else | |
4d7b1503 DW |
6224 | super->next_buf = NULL; |
6225 | } | |
8273f55e DW |
6226 | } |
6227 | ||
ae6aad82 | 6228 | /* must be called while manager is quiesced */ |
f21e18ca | 6229 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
6230 | { |
6231 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
6232 | struct dl *iter; |
6233 | struct imsm_dev *dev; | |
6234 | struct imsm_map *map; | |
24565c9a DW |
6235 | int i, j, num_members; |
6236 | __u32 ord; | |
ae6aad82 | 6237 | |
24565c9a DW |
6238 | dprintf("%s: deleting device[%d] from imsm_super\n", |
6239 | __func__, index); | |
ae6aad82 DW |
6240 | |
6241 | /* shift all indexes down one */ | |
6242 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 6243 | if (iter->index > (int)index) |
ae6aad82 | 6244 | iter->index--; |
47ee5a45 | 6245 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 6246 | if (iter->index > (int)index) |
47ee5a45 | 6247 | iter->index--; |
ae6aad82 DW |
6248 | |
6249 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6250 | dev = get_imsm_dev(super, i); | |
6251 | map = get_imsm_map(dev, 0); | |
24565c9a DW |
6252 | num_members = map->num_members; |
6253 | for (j = 0; j < num_members; j++) { | |
6254 | /* update ord entries being careful not to propagate | |
6255 | * ord-flags to the first map | |
6256 | */ | |
98130f40 | 6257 | ord = get_imsm_ord_tbl_ent(dev, j, -1); |
ae6aad82 | 6258 | |
24565c9a DW |
6259 | if (ord_to_idx(ord) <= index) |
6260 | continue; | |
ae6aad82 | 6261 | |
24565c9a DW |
6262 | map = get_imsm_map(dev, 0); |
6263 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); | |
6264 | map = get_imsm_map(dev, 1); | |
6265 | if (map) | |
6266 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
6267 | } |
6268 | } | |
6269 | ||
6270 | mpb->num_disks--; | |
6271 | super->updates_pending++; | |
24565c9a DW |
6272 | if (*dlp) { |
6273 | struct dl *dl = *dlp; | |
6274 | ||
6275 | *dlp = (*dlp)->next; | |
6276 | __free_imsm_disk(dl); | |
6277 | } | |
ae6aad82 | 6278 | } |
0e600426 | 6279 | #endif /* MDASSEMBLE */ |
ae6aad82 | 6280 | |
2cda7640 ML |
6281 | static char disk_by_path[] = "/dev/disk/by-path/"; |
6282 | ||
6283 | static const char *imsm_get_disk_controller_domain(const char *path) | |
6284 | { | |
6285 | struct sys_dev *list, *hba = NULL; | |
6286 | char disk_path[PATH_MAX]; | |
6287 | int ahci = 0; | |
6288 | char *dpath = NULL; | |
6289 | ||
6290 | list = find_driver_devices("pci", "ahci"); | |
6291 | for (hba = list; hba; hba = hba->next) | |
6292 | if (devpath_to_vendor(hba->path) == 0x8086) | |
6293 | break; | |
6294 | ||
6295 | if (hba) { | |
6296 | struct stat st; | |
6297 | ||
6298 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); | |
6299 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
6300 | if (stat(disk_path, &st) == 0) { | |
6301 | dpath = devt_to_devpath(st.st_rdev); | |
6302 | if (dpath) | |
6303 | ahci = path_attached_to_hba(dpath, hba->path); | |
6304 | } | |
6305 | } | |
6306 | dprintf("path: %s(%s) hba: %s attached: %d\n", | |
6307 | path, dpath, (hba) ? hba->path : "NULL", ahci); | |
6308 | free_sys_dev(&list); | |
6309 | if (ahci) | |
6310 | return "ahci"; | |
6311 | else | |
6312 | return NULL; | |
6313 | } | |
6314 | ||
78b10e66 N |
6315 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
6316 | { | |
6317 | char subdev_name[20]; | |
6318 | struct mdstat_ent *mdstat; | |
6319 | ||
6320 | sprintf(subdev_name, "%d", subdev); | |
6321 | mdstat = mdstat_by_subdev(subdev_name, container); | |
6322 | if (!mdstat) | |
6323 | return -1; | |
6324 | ||
6325 | *minor = mdstat->devnum; | |
6326 | free_mdstat(mdstat); | |
6327 | return 0; | |
6328 | } | |
6329 | ||
6330 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
6331 | struct geo_params *geo, | |
6332 | int *old_raid_disks) | |
6333 | { | |
694575e7 KW |
6334 | /* currently we only support increasing the number of devices |
6335 | * for a container. This increases the number of device for each | |
6336 | * member array. They must all be RAID0 or RAID5. | |
6337 | */ | |
78b10e66 N |
6338 | int ret_val = 0; |
6339 | struct mdinfo *info, *member; | |
6340 | int devices_that_can_grow = 0; | |
6341 | ||
6342 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
6343 | "st->devnum = (%i)\n", | |
6344 | st->devnum); | |
6345 | ||
6346 | if (geo->size != -1 || | |
6347 | geo->level != UnSet || | |
6348 | geo->layout != UnSet || | |
6349 | geo->chunksize != 0 || | |
6350 | geo->raid_disks == UnSet) { | |
6351 | dprintf("imsm: Container operation is allowed for " | |
6352 | "raid disks number change only.\n"); | |
6353 | return ret_val; | |
6354 | } | |
6355 | ||
6356 | info = container_content_imsm(st, NULL); | |
6357 | for (member = info; member; member = member->next) { | |
6358 | int result; | |
6359 | int minor; | |
6360 | ||
6361 | dprintf("imsm: checking device_num: %i\n", | |
6362 | member->container_member); | |
6363 | ||
6364 | if (geo->raid_disks < member->array.raid_disks) { | |
6365 | /* we work on container for Online Capacity Expansion | |
6366 | * only so raid_disks has to grow | |
6367 | */ | |
6368 | dprintf("imsm: for container operation raid disks " | |
6369 | "increase is required\n"); | |
6370 | break; | |
6371 | } | |
6372 | ||
6373 | if ((info->array.level != 0) && | |
6374 | (info->array.level != 5)) { | |
6375 | /* we cannot use this container with other raid level | |
6376 | */ | |
690aae1a | 6377 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
6378 | " raid level (%i) detected\n", |
6379 | info->array.level); | |
6380 | break; | |
6381 | } else { | |
6382 | /* check for platform support | |
6383 | * for this raid level configuration | |
6384 | */ | |
6385 | struct intel_super *super = st->sb; | |
6386 | if (!is_raid_level_supported(super->orom, | |
6387 | member->array.level, | |
6388 | geo->raid_disks)) { | |
690aae1a | 6389 | dprintf("platform does not support raid%d with" |
78b10e66 N |
6390 | " %d disk%s\n", |
6391 | info->array.level, | |
6392 | geo->raid_disks, | |
6393 | geo->raid_disks > 1 ? "s" : ""); | |
6394 | break; | |
6395 | } | |
6396 | } | |
6397 | ||
6398 | if (*old_raid_disks && | |
6399 | info->array.raid_disks != *old_raid_disks) | |
6400 | break; | |
6401 | *old_raid_disks = info->array.raid_disks; | |
6402 | ||
6403 | /* All raid5 and raid0 volumes in container | |
6404 | * have to be ready for Online Capacity Expansion | |
6405 | * so they need to be assembled. We have already | |
6406 | * checked that no recovery etc is happening. | |
6407 | */ | |
6408 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
6409 | st->container_dev, | |
6410 | &minor); | |
6411 | if (result < 0) { | |
6412 | dprintf("imsm: cannot find array\n"); | |
6413 | break; | |
6414 | } | |
6415 | devices_that_can_grow++; | |
6416 | } | |
6417 | sysfs_free(info); | |
6418 | if (!member && devices_that_can_grow) | |
6419 | ret_val = 1; | |
6420 | ||
6421 | if (ret_val) | |
6422 | dprintf("\tContainer operation allowed\n"); | |
6423 | else | |
6424 | dprintf("\tError: %i\n", ret_val); | |
6425 | ||
6426 | return ret_val; | |
6427 | } | |
6428 | ||
6429 | /* Function: get_spares_for_grow | |
6430 | * Description: Allocates memory and creates list of spare devices | |
6431 | * avaliable in container. Checks if spare drive size is acceptable. | |
6432 | * Parameters: Pointer to the supertype structure | |
6433 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
6434 | * NULL if fail | |
6435 | */ | |
6436 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
6437 | { | |
78b10e66 | 6438 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 6439 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
6440 | } |
6441 | ||
6442 | /****************************************************************************** | |
6443 | * function: imsm_create_metadata_update_for_reshape | |
6444 | * Function creates update for whole IMSM container. | |
6445 | * | |
6446 | ******************************************************************************/ | |
6447 | static int imsm_create_metadata_update_for_reshape( | |
6448 | struct supertype *st, | |
6449 | struct geo_params *geo, | |
6450 | int old_raid_disks, | |
6451 | struct imsm_update_reshape **updatep) | |
6452 | { | |
6453 | struct intel_super *super = st->sb; | |
6454 | struct imsm_super *mpb = super->anchor; | |
6455 | int update_memory_size = 0; | |
6456 | struct imsm_update_reshape *u = NULL; | |
6457 | struct mdinfo *spares = NULL; | |
6458 | int i; | |
6459 | int delta_disks = 0; | |
bbd24d86 | 6460 | struct mdinfo *dev; |
78b10e66 N |
6461 | |
6462 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
6463 | geo->raid_disks); | |
6464 | ||
6465 | delta_disks = geo->raid_disks - old_raid_disks; | |
6466 | ||
6467 | /* size of all update data without anchor */ | |
6468 | update_memory_size = sizeof(struct imsm_update_reshape); | |
6469 | ||
6470 | /* now add space for spare disks that we need to add. */ | |
6471 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
6472 | ||
6473 | u = calloc(1, update_memory_size); | |
6474 | if (u == NULL) { | |
6475 | dprintf("error: " | |
6476 | "cannot get memory for imsm_update_reshape update\n"); | |
6477 | return 0; | |
6478 | } | |
6479 | u->type = update_reshape_container_disks; | |
6480 | u->old_raid_disks = old_raid_disks; | |
6481 | u->new_raid_disks = geo->raid_disks; | |
6482 | ||
6483 | /* now get spare disks list | |
6484 | */ | |
6485 | spares = get_spares_for_grow(st); | |
6486 | ||
6487 | if (spares == NULL | |
6488 | || delta_disks > spares->array.spare_disks) { | |
6489 | dprintf("imsm: ERROR: Cannot get spare devices.\n"); | |
6490 | goto abort; | |
6491 | } | |
6492 | ||
6493 | /* we have got spares | |
6494 | * update disk list in imsm_disk list table in anchor | |
6495 | */ | |
6496 | dprintf("imsm: %i spares are available.\n\n", | |
6497 | spares->array.spare_disks); | |
6498 | ||
bbd24d86 | 6499 | dev = spares->devs; |
78b10e66 | 6500 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
6501 | struct dl *dl; |
6502 | ||
bbd24d86 AK |
6503 | if (dev == NULL) |
6504 | break; | |
78b10e66 N |
6505 | u->new_disks[i] = makedev(dev->disk.major, |
6506 | dev->disk.minor); | |
6507 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
6508 | dl->index = mpb->num_disks; |
6509 | mpb->num_disks++; | |
bbd24d86 | 6510 | dev = dev->next; |
78b10e66 | 6511 | } |
78b10e66 N |
6512 | |
6513 | abort: | |
6514 | /* free spares | |
6515 | */ | |
6516 | sysfs_free(spares); | |
6517 | ||
d677e0b8 | 6518 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 6519 | if (i == delta_disks) { |
d677e0b8 | 6520 | dprintf(" OK\n"); |
78b10e66 N |
6521 | *updatep = u; |
6522 | return update_memory_size; | |
6523 | } | |
6524 | free(u); | |
d677e0b8 | 6525 | dprintf(" Error\n"); |
78b10e66 N |
6526 | |
6527 | return 0; | |
6528 | } | |
6529 | ||
8dd70bce AK |
6530 | static void imsm_update_metadata_locally(struct supertype *st, |
6531 | void *buf, int len) | |
6532 | { | |
6533 | struct metadata_update mu; | |
6534 | ||
6535 | mu.buf = buf; | |
6536 | mu.len = len; | |
6537 | mu.space = NULL; | |
6538 | mu.space_list = NULL; | |
6539 | mu.next = NULL; | |
6540 | imsm_prepare_update(st, &mu); | |
6541 | imsm_process_update(st, &mu); | |
6542 | ||
6543 | while (mu.space_list) { | |
6544 | void **space = mu.space_list; | |
6545 | mu.space_list = *space; | |
6546 | free(space); | |
6547 | } | |
6548 | } | |
78b10e66 | 6549 | |
694575e7 KW |
6550 | /******************************************************************************* |
6551 | * Function: imsm_analyze_change | |
6552 | * Description: Function analyze and validate change for single volume migration | |
6553 | * Parameters: Geometry parameters, supertype structure | |
6554 | * Returns: Operation type code on success, -1 if fail | |
6555 | ********************************************************************************/ | |
6556 | enum imsm_reshape_type imsm_analyze_change( | |
6557 | struct supertype *st, struct geo_params *geo) | |
6558 | { | |
6559 | return -1; | |
6560 | } | |
6561 | ||
78b10e66 N |
6562 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
6563 | int layout, int chunksize, int raid_disks, | |
690aae1a | 6564 | char *backup, char *dev, int verbose) |
78b10e66 | 6565 | { |
78b10e66 N |
6566 | int ret_val = 1; |
6567 | struct geo_params geo; | |
6568 | ||
6569 | dprintf("imsm: reshape_super called.\n"); | |
6570 | ||
6571 | memset(&geo, sizeof(struct geo_params), 0); | |
6572 | ||
6573 | geo.dev_name = dev; | |
694575e7 | 6574 | geo.dev_id = st->devnum; |
78b10e66 N |
6575 | geo.size = size; |
6576 | geo.level = level; | |
6577 | geo.layout = layout; | |
6578 | geo.chunksize = chunksize; | |
6579 | geo.raid_disks = raid_disks; | |
6580 | ||
6581 | dprintf("\tfor level : %i\n", geo.level); | |
6582 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
6583 | ||
6584 | if (experimental() == 0) | |
6585 | return ret_val; | |
6586 | ||
78b10e66 | 6587 | if (st->container_dev == st->devnum) { |
694575e7 KW |
6588 | /* On container level we can only increase number of devices. */ |
6589 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 N |
6590 | int old_raid_disks = 0; |
6591 | if (imsm_reshape_is_allowed_on_container( | |
6592 | st, &geo, &old_raid_disks)) { | |
6593 | struct imsm_update_reshape *u = NULL; | |
6594 | int len; | |
6595 | ||
6596 | len = imsm_create_metadata_update_for_reshape( | |
6597 | st, &geo, old_raid_disks, &u); | |
6598 | ||
ed08d51c AK |
6599 | if (len <= 0) { |
6600 | dprintf("imsm: Cannot prepare update\n"); | |
6601 | goto exit_imsm_reshape_super; | |
6602 | } | |
6603 | ||
8dd70bce AK |
6604 | ret_val = 0; |
6605 | /* update metadata locally */ | |
6606 | imsm_update_metadata_locally(st, u, len); | |
6607 | /* and possibly remotely */ | |
6608 | if (st->update_tail) | |
6609 | append_metadata_update(st, u, len); | |
6610 | else | |
ed08d51c | 6611 | free(u); |
8dd70bce | 6612 | |
694575e7 | 6613 | } else { |
690aae1a | 6614 | fprintf(stderr, Name "imsm: Operation is not allowed " |
78b10e66 | 6615 | "on this container\n"); |
694575e7 KW |
6616 | } |
6617 | } else { | |
6618 | /* On volume level we support following operations | |
6619 | * - takeover: raid10 -> raid0; raid0 -> raid10 | |
6620 | * - chunk size migration | |
6621 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
6622 | */ | |
6623 | int change; | |
6624 | dprintf("imsm: info: Volume operation\n"); | |
6625 | change = imsm_analyze_change(st, &geo); | |
6626 | switch (change) { | |
6627 | case CH_TAKEOVER: | |
6628 | break; | |
6629 | case CH_CHUNK_MIGR: | |
6630 | break; | |
6631 | case CH_LEVEL_MIGRATION: | |
6632 | break; | |
6633 | } | |
6634 | ret_val = 0; | |
6635 | } | |
78b10e66 | 6636 | |
ed08d51c | 6637 | exit_imsm_reshape_super: |
78b10e66 N |
6638 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
6639 | return ret_val; | |
6640 | } | |
2cda7640 | 6641 | |
999b4972 N |
6642 | static int imsm_manage_reshape( |
6643 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
6644 | struct supertype *st, unsigned long stripes, | |
6645 | int *fds, unsigned long long *offsets, | |
6646 | int dests, int *destfd, unsigned long long *destoffsets) | |
6647 | { | |
6648 | /* Just use child_monitor for now */ | |
6649 | return child_monitor( | |
6650 | afd, sra, reshape, st, stripes, | |
6651 | fds, offsets, dests, destfd, destoffsets); | |
6652 | } | |
6653 | ||
cdddbdbc DW |
6654 | struct superswitch super_imsm = { |
6655 | #ifndef MDASSEMBLE | |
6656 | .examine_super = examine_super_imsm, | |
6657 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 6658 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 6659 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
6660 | .detail_super = detail_super_imsm, |
6661 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 6662 | .write_init_super = write_init_super_imsm, |
0e600426 N |
6663 | .validate_geometry = validate_geometry_imsm, |
6664 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 6665 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 6666 | .detail_platform = detail_platform_imsm, |
33414a01 | 6667 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 6668 | .update_subarray = update_subarray_imsm, |
2b959fbf | 6669 | .load_container = load_container_imsm, |
cdddbdbc DW |
6670 | #endif |
6671 | .match_home = match_home_imsm, | |
6672 | .uuid_from_super= uuid_from_super_imsm, | |
6673 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 6674 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
6675 | .update_super = update_super_imsm, |
6676 | ||
6677 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 6678 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
6679 | |
6680 | .compare_super = compare_super_imsm, | |
6681 | ||
6682 | .load_super = load_super_imsm, | |
bf5a934a | 6683 | .init_super = init_super_imsm, |
e683ca88 | 6684 | .store_super = store_super_imsm, |
cdddbdbc DW |
6685 | .free_super = free_super_imsm, |
6686 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 6687 | .container_content = container_content_imsm, |
30f58b22 | 6688 | .default_geometry = default_geometry_imsm, |
2cda7640 | 6689 | .get_disk_controller_domain = imsm_get_disk_controller_domain, |
78b10e66 | 6690 | .reshape_super = imsm_reshape_super, |
999b4972 | 6691 | .manage_reshape = imsm_manage_reshape, |
cdddbdbc | 6692 | |
cdddbdbc | 6693 | .external = 1, |
4cce4069 | 6694 | .name = "imsm", |
845dea95 | 6695 | |
0e600426 | 6696 | #ifndef MDASSEMBLE |
845dea95 NB |
6697 | /* for mdmon */ |
6698 | .open_new = imsm_open_new, | |
ed9d66aa | 6699 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
6700 | .set_disk = imsm_set_disk, |
6701 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 6702 | .activate_spare = imsm_activate_spare, |
e8319a19 | 6703 | .process_update = imsm_process_update, |
8273f55e | 6704 | .prepare_update = imsm_prepare_update, |
0e600426 | 6705 | #endif /* MDASSEMBLE */ |
cdddbdbc | 6706 | }; |