]>
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" |
cdddbdbc DW |
24 | #include <values.h> |
25 | #include <scsi/sg.h> | |
26 | #include <ctype.h> | |
27 | ||
28 | /* MPB == Metadata Parameter Block */ | |
29 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
30 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
31 | #define MPB_VERSION_RAID0 "1.0.00" | |
32 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
33 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
34 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 35 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
36 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
37 | #define MPB_VERSION_CNG "1.2.06" | |
38 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
39 | #define MAX_SIGNATURE_LENGTH 32 |
40 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb DW |
41 | |
42 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
43 | #define MPB_ATTRIB_PM __cpu_to_le32(0x40000000) | |
44 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
47 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
48 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
49 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
50 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
51 | ||
c2c087e6 DW |
52 | #define MPB_SECTOR_CNT 418 |
53 | #define IMSM_RESERVED_SECTORS 4096 | |
cdddbdbc DW |
54 | |
55 | /* Disk configuration info. */ | |
56 | #define IMSM_MAX_DEVICES 255 | |
57 | struct imsm_disk { | |
58 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
59 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
60 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
f2f27e63 DW |
61 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
62 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
63 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
64 | #define USABLE_DISK __cpu_to_le32(0x08) /* Fully usable unless FAILED_DISK is set */ | |
cdddbdbc | 65 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb DW |
66 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
67 | #define IMSM_DISK_FILLERS 4 | |
cdddbdbc DW |
68 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ |
69 | }; | |
70 | ||
71 | /* RAID map configuration infos. */ | |
72 | struct imsm_map { | |
73 | __u32 pba_of_lba0; /* start address of partition */ | |
74 | __u32 blocks_per_member;/* blocks per member */ | |
75 | __u32 num_data_stripes; /* number of data stripes */ | |
76 | __u16 blocks_per_strip; | |
77 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
78 | #define IMSM_T_STATE_NORMAL 0 | |
79 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
80 | #define IMSM_T_STATE_DEGRADED 2 |
81 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
82 | __u8 raid_level; |
83 | #define IMSM_T_RAID0 0 | |
84 | #define IMSM_T_RAID1 1 | |
85 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
86 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
87 | __u8 num_domains; /* number of parity domains */ |
88 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
89 | __u8 reserved[1]; | |
cdddbdbc | 90 | __u32 filler[7]; /* expansion area */ |
7eef0453 | 91 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 92 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
93 | * top byte contains some flags |
94 | */ | |
cdddbdbc DW |
95 | } __attribute__ ((packed)); |
96 | ||
97 | struct imsm_vol { | |
f8f603f1 | 98 | __u32 curr_migr_unit; |
fe7ed8cb | 99 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 100 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
101 | #define MIGR_INIT 0 |
102 | #define MIGR_REBUILD 1 | |
103 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
104 | #define MIGR_GEN_MIGR 3 | |
105 | #define MIGR_STATE_CHANGE 4 | |
cdddbdbc DW |
106 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
107 | __u8 dirty; | |
fe7ed8cb DW |
108 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
109 | __u16 verify_errors; /* number of mismatches */ | |
110 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
111 | __u32 filler[4]; | |
cdddbdbc DW |
112 | struct imsm_map map[1]; |
113 | /* here comes another one if migr_state */ | |
114 | } __attribute__ ((packed)); | |
115 | ||
116 | struct imsm_dev { | |
fe7ed8cb | 117 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
118 | __u32 size_low; |
119 | __u32 size_high; | |
fe7ed8cb DW |
120 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
121 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
122 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
123 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
124 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
125 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
126 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
127 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
128 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
129 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
130 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
131 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
132 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
133 | __u32 status; /* Persistent RaidDev status */ |
134 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
135 | __u8 migr_priority; |
136 | __u8 num_sub_vols; | |
137 | __u8 tid; | |
138 | __u8 cng_master_disk; | |
139 | __u16 cache_policy; | |
140 | __u8 cng_state; | |
141 | __u8 cng_sub_state; | |
142 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
143 | __u32 filler[IMSM_DEV_FILLERS]; |
144 | struct imsm_vol vol; | |
145 | } __attribute__ ((packed)); | |
146 | ||
147 | struct imsm_super { | |
148 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
149 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
150 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
151 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
152 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
153 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
154 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
155 | __u8 num_disks; /* 0x38 Number of configured disks */ |
156 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
157 | __u8 error_log_pos; /* 0x3A */ |
158 | __u8 fill[1]; /* 0x3B */ | |
159 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
160 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
161 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
162 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
163 | #define IMSM_FILLERS 35 | |
164 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
165 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
166 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 167 | /* here comes BBM logs */ |
cdddbdbc DW |
168 | } __attribute__ ((packed)); |
169 | ||
604b746f JD |
170 | #define BBM_LOG_MAX_ENTRIES 254 |
171 | ||
172 | struct bbm_log_entry { | |
173 | __u64 defective_block_start; | |
174 | #define UNREADABLE 0xFFFFFFFF | |
175 | __u32 spare_block_offset; | |
176 | __u16 remapped_marked_count; | |
177 | __u16 disk_ordinal; | |
178 | } __attribute__ ((__packed__)); | |
179 | ||
180 | struct bbm_log { | |
181 | __u32 signature; /* 0xABADB10C */ | |
182 | __u32 entry_count; | |
183 | __u32 reserved_spare_block_count; /* 0 */ | |
184 | __u32 reserved; /* 0xFFFF */ | |
185 | __u64 first_spare_lba; | |
186 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
187 | } __attribute__ ((__packed__)); | |
188 | ||
189 | ||
cdddbdbc DW |
190 | #ifndef MDASSEMBLE |
191 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
192 | #endif | |
193 | ||
87eb16df | 194 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 195 | { |
87eb16df DW |
196 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
197 | } | |
cdddbdbc | 198 | |
87eb16df DW |
199 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
200 | { | |
201 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
202 | } |
203 | ||
204 | /* internal representation of IMSM metadata */ | |
205 | struct intel_super { | |
206 | union { | |
949c47a0 DW |
207 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
208 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 209 | }; |
949c47a0 | 210 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
211 | void *next_buf; /* for realloc'ing buf from the manager */ |
212 | size_t next_len; | |
c2c087e6 DW |
213 | int updates_pending; /* count of pending updates for mdmon */ |
214 | int creating_imsm; /* flag to indicate container creation */ | |
bf5a934a | 215 | int current_vol; /* index of raid device undergoing creation */ |
949c47a0 DW |
216 | #define IMSM_MAX_RAID_DEVS 2 |
217 | struct imsm_dev *dev_tbl[IMSM_MAX_RAID_DEVS]; | |
cdddbdbc DW |
218 | struct dl { |
219 | struct dl *next; | |
220 | int index; | |
221 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
222 | int major, minor; | |
223 | char *devname; | |
b9f594fe | 224 | struct imsm_disk disk; |
cdddbdbc DW |
225 | int fd; |
226 | } *disks; | |
43dad3d6 | 227 | struct dl *add; /* list of disks to add while mdmon active */ |
47ee5a45 | 228 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 229 | struct bbm_log *bbm_log; |
cdddbdbc DW |
230 | }; |
231 | ||
c2c087e6 DW |
232 | struct extent { |
233 | unsigned long long start, size; | |
234 | }; | |
235 | ||
88758e9d DW |
236 | /* definition of messages passed to imsm_process_update */ |
237 | enum imsm_update_type { | |
238 | update_activate_spare, | |
8273f55e | 239 | update_create_array, |
43dad3d6 | 240 | update_add_disk, |
88758e9d DW |
241 | }; |
242 | ||
243 | struct imsm_update_activate_spare { | |
244 | enum imsm_update_type type; | |
d23fe947 | 245 | struct dl *dl; |
88758e9d DW |
246 | int slot; |
247 | int array; | |
248 | struct imsm_update_activate_spare *next; | |
249 | }; | |
250 | ||
8273f55e DW |
251 | struct imsm_update_create_array { |
252 | enum imsm_update_type type; | |
8273f55e | 253 | int dev_idx; |
6a3e913e | 254 | struct imsm_dev dev; |
8273f55e DW |
255 | }; |
256 | ||
43dad3d6 DW |
257 | struct imsm_update_add_disk { |
258 | enum imsm_update_type type; | |
259 | }; | |
260 | ||
cdddbdbc DW |
261 | static struct supertype *match_metadata_desc_imsm(char *arg) |
262 | { | |
263 | struct supertype *st; | |
264 | ||
265 | if (strcmp(arg, "imsm") != 0 && | |
266 | strcmp(arg, "default") != 0 | |
267 | ) | |
268 | return NULL; | |
269 | ||
270 | st = malloc(sizeof(*st)); | |
ef609477 | 271 | memset(st, 0, sizeof(*st)); |
cdddbdbc DW |
272 | st->ss = &super_imsm; |
273 | st->max_devs = IMSM_MAX_DEVICES; | |
274 | st->minor_version = 0; | |
275 | st->sb = NULL; | |
276 | return st; | |
277 | } | |
278 | ||
0e600426 | 279 | #ifndef MDASSEMBLE |
cdddbdbc DW |
280 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
281 | { | |
282 | return &mpb->sig[MPB_SIG_LEN]; | |
283 | } | |
0e600426 | 284 | #endif |
cdddbdbc | 285 | |
949c47a0 DW |
286 | /* retrieve a disk directly from the anchor when the anchor is known to be |
287 | * up-to-date, currently only at load time | |
288 | */ | |
289 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 290 | { |
949c47a0 | 291 | if (index >= mpb->num_disks) |
cdddbdbc DW |
292 | return NULL; |
293 | return &mpb->disk[index]; | |
294 | } | |
295 | ||
0e600426 | 296 | #ifndef MDASSEMBLE |
b9f594fe | 297 | /* retrieve a disk from the parsed metadata */ |
949c47a0 DW |
298 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) |
299 | { | |
b9f594fe DW |
300 | struct dl *d; |
301 | ||
302 | for (d = super->disks; d; d = d->next) | |
303 | if (d->index == index) | |
304 | return &d->disk; | |
305 | ||
306 | return NULL; | |
949c47a0 | 307 | } |
0e600426 | 308 | #endif |
949c47a0 DW |
309 | |
310 | /* generate a checksum directly from the anchor when the anchor is known to be | |
311 | * up-to-date, currently only at load or write_super after coalescing | |
312 | */ | |
313 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
314 | { |
315 | __u32 end = mpb->mpb_size / sizeof(end); | |
316 | __u32 *p = (__u32 *) mpb; | |
317 | __u32 sum = 0; | |
318 | ||
97f734fd N |
319 | while (end--) { |
320 | sum += __le32_to_cpu(*p); | |
321 | p++; | |
322 | } | |
cdddbdbc DW |
323 | |
324 | return sum - __le32_to_cpu(mpb->check_sum); | |
325 | } | |
326 | ||
a965f303 DW |
327 | static size_t sizeof_imsm_map(struct imsm_map *map) |
328 | { | |
329 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
330 | } | |
331 | ||
332 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 333 | { |
a965f303 DW |
334 | struct imsm_map *map = &dev->vol.map[0]; |
335 | ||
336 | if (second_map && !dev->vol.migr_state) | |
337 | return NULL; | |
338 | else if (second_map) { | |
339 | void *ptr = map; | |
340 | ||
341 | return ptr + sizeof_imsm_map(map); | |
342 | } else | |
343 | return map; | |
344 | ||
345 | } | |
cdddbdbc | 346 | |
3393c6af DW |
347 | /* return the size of the device. |
348 | * migr_state increases the returned size if map[0] were to be duplicated | |
349 | */ | |
350 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
351 | { |
352 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
353 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
354 | |
355 | /* migrating means an additional map */ | |
a965f303 DW |
356 | if (dev->vol.migr_state) |
357 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
358 | else if (migr_state) |
359 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
360 | |
361 | return size; | |
362 | } | |
363 | ||
949c47a0 | 364 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
365 | { |
366 | int offset; | |
367 | int i; | |
368 | void *_mpb = mpb; | |
369 | ||
949c47a0 | 370 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
371 | return NULL; |
372 | ||
373 | /* devices start after all disks */ | |
374 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
375 | ||
376 | for (i = 0; i <= index; i++) | |
377 | if (i == index) | |
378 | return _mpb + offset; | |
379 | else | |
3393c6af | 380 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
381 | |
382 | return NULL; | |
383 | } | |
384 | ||
949c47a0 DW |
385 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
386 | { | |
387 | if (index >= super->anchor->num_raid_devs) | |
388 | return NULL; | |
389 | return super->dev_tbl[index]; | |
390 | } | |
391 | ||
7eef0453 DW |
392 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, int slot) |
393 | { | |
394 | struct imsm_map *map; | |
395 | ||
396 | if (dev->vol.migr_state) | |
7eef0453 | 397 | map = get_imsm_map(dev, 1); |
fb9bf0d3 DW |
398 | else |
399 | map = get_imsm_map(dev, 0); | |
7eef0453 | 400 | |
ff077194 DW |
401 | /* top byte identifies disk under rebuild */ |
402 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
403 | } | |
404 | ||
405 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
406 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot) | |
407 | { | |
408 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot); | |
409 | ||
410 | return ord_to_idx(ord); | |
7eef0453 DW |
411 | } |
412 | ||
be73972f DW |
413 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
414 | { | |
415 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
416 | } | |
417 | ||
cdddbdbc DW |
418 | static int get_imsm_raid_level(struct imsm_map *map) |
419 | { | |
420 | if (map->raid_level == 1) { | |
421 | if (map->num_members == 2) | |
422 | return 1; | |
423 | else | |
424 | return 10; | |
425 | } | |
426 | ||
427 | return map->raid_level; | |
428 | } | |
429 | ||
c2c087e6 DW |
430 | static int cmp_extent(const void *av, const void *bv) |
431 | { | |
432 | const struct extent *a = av; | |
433 | const struct extent *b = bv; | |
434 | if (a->start < b->start) | |
435 | return -1; | |
436 | if (a->start > b->start) | |
437 | return 1; | |
438 | return 0; | |
439 | } | |
440 | ||
441 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
442 | { | |
443 | /* find a list of used extents on the given physical device */ | |
c2c087e6 DW |
444 | struct extent *rv, *e; |
445 | int i, j; | |
446 | int memberships = 0; | |
14e8215b | 447 | __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
c2c087e6 | 448 | |
949c47a0 DW |
449 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
450 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 451 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 DW |
452 | |
453 | for (j = 0; j < map->num_members; j++) { | |
ff077194 | 454 | __u32 index = get_imsm_disk_idx(dev, j); |
c2c087e6 DW |
455 | |
456 | if (index == dl->index) | |
457 | memberships++; | |
458 | } | |
459 | } | |
460 | rv = malloc(sizeof(struct extent) * (memberships + 1)); | |
461 | if (!rv) | |
462 | return NULL; | |
463 | e = rv; | |
464 | ||
949c47a0 DW |
465 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
466 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 467 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 DW |
468 | |
469 | for (j = 0; j < map->num_members; j++) { | |
ff077194 | 470 | __u32 index = get_imsm_disk_idx(dev, j); |
c2c087e6 DW |
471 | |
472 | if (index == dl->index) { | |
473 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
474 | e->size = __le32_to_cpu(map->blocks_per_member); | |
475 | e++; | |
476 | } | |
477 | } | |
478 | } | |
479 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
480 | ||
14e8215b DW |
481 | /* determine the start of the metadata |
482 | * when no raid devices are defined use the default | |
483 | * ...otherwise allow the metadata to truncate the value | |
484 | * as is the case with older versions of imsm | |
485 | */ | |
486 | if (memberships) { | |
487 | struct extent *last = &rv[memberships - 1]; | |
488 | __u32 remainder; | |
489 | ||
490 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
491 | (last->start + last->size); | |
492 | if (reservation > remainder) | |
493 | reservation = remainder; | |
494 | } | |
495 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
496 | e->size = 0; |
497 | return rv; | |
498 | } | |
499 | ||
14e8215b DW |
500 | /* try to determine how much space is reserved for metadata from |
501 | * the last get_extents() entry, otherwise fallback to the | |
502 | * default | |
503 | */ | |
504 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
505 | { | |
506 | struct extent *e; | |
507 | int i; | |
508 | __u32 rv; | |
509 | ||
510 | /* for spares just return a minimal reservation which will grow | |
511 | * once the spare is picked up by an array | |
512 | */ | |
513 | if (dl->index == -1) | |
514 | return MPB_SECTOR_CNT; | |
515 | ||
516 | e = get_extents(super, dl); | |
517 | if (!e) | |
518 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
519 | ||
520 | /* scroll to last entry */ | |
521 | for (i = 0; e[i].size; i++) | |
522 | continue; | |
523 | ||
524 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
525 | ||
526 | free(e); | |
527 | ||
528 | return rv; | |
529 | } | |
530 | ||
531 | #ifndef MDASSEMBLE | |
44470971 | 532 | static void print_imsm_dev(struct imsm_dev *dev, char *uuid, int disk_idx) |
cdddbdbc DW |
533 | { |
534 | __u64 sz; | |
535 | int slot; | |
a965f303 | 536 | struct imsm_map *map = get_imsm_map(dev, 0); |
b10b37b8 | 537 | __u32 ord; |
cdddbdbc DW |
538 | |
539 | printf("\n"); | |
1e7bc0ed | 540 | printf("[%.16s]:\n", dev->volume); |
44470971 | 541 | printf(" UUID : %s\n", uuid); |
cdddbdbc DW |
542 | printf(" RAID Level : %d\n", get_imsm_raid_level(map)); |
543 | printf(" Members : %d\n", map->num_members); | |
544 | for (slot = 0; slot < map->num_members; slot++) | |
44470971 | 545 | if (disk_idx== get_imsm_disk_idx(dev, slot)) |
cdddbdbc | 546 | break; |
b10b37b8 DW |
547 | if (slot < map->num_members) { |
548 | ord = get_imsm_ord_tbl_ent(dev, slot); | |
549 | printf(" This Slot : %d%s\n", slot, | |
550 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
551 | } else | |
cdddbdbc DW |
552 | printf(" This Slot : ?\n"); |
553 | sz = __le32_to_cpu(dev->size_high); | |
554 | sz <<= 32; | |
555 | sz += __le32_to_cpu(dev->size_low); | |
556 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
557 | human_size(sz * 512)); | |
558 | sz = __le32_to_cpu(map->blocks_per_member); | |
559 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
560 | human_size(sz * 512)); | |
561 | printf(" Sector Offset : %u\n", | |
562 | __le32_to_cpu(map->pba_of_lba0)); | |
563 | printf(" Num Stripes : %u\n", | |
564 | __le32_to_cpu(map->num_data_stripes)); | |
565 | printf(" Chunk Size : %u KiB\n", | |
566 | __le16_to_cpu(map->blocks_per_strip) / 2); | |
567 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); | |
3393c6af DW |
568 | printf(" Migrate State : %s", dev->vol.migr_state ? "migrating" : "idle"); |
569 | if (dev->vol.migr_state) | |
570 | printf(": %s", dev->vol.migr_type ? "rebuilding" : "initializing"); | |
571 | printf("\n"); | |
572 | printf(" Map State : %s", map_state_str[map->map_state]); | |
573 | if (dev->vol.migr_state) { | |
574 | struct imsm_map *map = get_imsm_map(dev, 1); | |
b10b37b8 | 575 | printf(" <-- %s", map_state_str[map->map_state]); |
3393c6af DW |
576 | } |
577 | printf("\n"); | |
cdddbdbc | 578 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
579 | } |
580 | ||
14e8215b | 581 | static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved) |
cdddbdbc | 582 | { |
949c47a0 | 583 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); |
1f24f035 | 584 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
585 | __u32 s; |
586 | __u64 sz; | |
587 | ||
e9d82038 DW |
588 | if (index < 0) |
589 | return; | |
590 | ||
cdddbdbc | 591 | printf("\n"); |
1f24f035 | 592 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
cdddbdbc | 593 | printf(" Disk%02d Serial : %s\n", index, str); |
f2f27e63 | 594 | s = disk->status; |
cdddbdbc DW |
595 | printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "", |
596 | s&CONFIGURED_DISK ? " active" : "", | |
597 | s&FAILED_DISK ? " failed" : "", | |
598 | s&USABLE_DISK ? " usable" : ""); | |
599 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
14e8215b | 600 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
601 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
602 | human_size(sz * 512)); | |
603 | } | |
604 | ||
44470971 DW |
605 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info); |
606 | ||
cdddbdbc DW |
607 | static void examine_super_imsm(struct supertype *st, char *homehost) |
608 | { | |
609 | struct intel_super *super = st->sb; | |
949c47a0 | 610 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
611 | char str[MAX_SIGNATURE_LENGTH]; |
612 | int i; | |
27fd6274 DW |
613 | struct mdinfo info; |
614 | char nbuf[64]; | |
cdddbdbc | 615 | __u32 sum; |
14e8215b | 616 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
cdddbdbc | 617 | |
27fd6274 | 618 | |
cdddbdbc DW |
619 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
620 | printf(" Magic : %s\n", str); | |
621 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
622 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
623 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); | |
624 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
27fd6274 DW |
625 | getinfo_super_imsm(st, &info); |
626 | fname_from_uuid(st, &info, nbuf,'-'); | |
627 | printf(" UUID : %s\n", nbuf + 5); | |
cdddbdbc DW |
628 | sum = __le32_to_cpu(mpb->check_sum); |
629 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 630 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 631 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
632 | printf(" Disks : %d\n", mpb->num_disks); |
633 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
14e8215b | 634 | print_imsm_disk(mpb, super->disks->index, reserved); |
604b746f JD |
635 | if (super->bbm_log) { |
636 | struct bbm_log *log = super->bbm_log; | |
637 | ||
638 | printf("\n"); | |
639 | printf("Bad Block Management Log:\n"); | |
640 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
641 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
642 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
643 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
644 | printf(" First Spare : %llx\n", __le64_to_cpu(log->first_spare_lba)); | |
645 | } | |
44470971 DW |
646 | for (i = 0; i < mpb->num_raid_devs; i++) { |
647 | struct mdinfo info; | |
648 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
649 | ||
650 | super->current_vol = i; | |
651 | getinfo_super_imsm(st, &info); | |
652 | fname_from_uuid(st, &info, nbuf, '-'); | |
653 | print_imsm_dev(dev, nbuf + 5, super->disks->index); | |
654 | } | |
cdddbdbc DW |
655 | for (i = 0; i < mpb->num_disks; i++) { |
656 | if (i == super->disks->index) | |
657 | continue; | |
14e8215b | 658 | print_imsm_disk(mpb, i, reserved); |
cdddbdbc DW |
659 | } |
660 | } | |
661 | ||
662 | static void brief_examine_super_imsm(struct supertype *st) | |
663 | { | |
27fd6274 | 664 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
665 | struct mdinfo info; |
666 | char nbuf[64]; | |
cf8de691 | 667 | char nbuf1[64]; |
1e7bc0ed DW |
668 | struct intel_super *super = st->sb; |
669 | int i; | |
670 | ||
671 | if (!super->anchor->num_raid_devs) | |
672 | return; | |
ff54de6e N |
673 | |
674 | getinfo_super_imsm(st, &info); | |
675 | fname_from_uuid(st, &info, nbuf,'-'); | |
cf8de691 | 676 | printf("ARRAY metadata=imsm auto=md UUID=%s\n", nbuf + 5); |
1e7bc0ed DW |
677 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
678 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
679 | ||
680 | super->current_vol = i; | |
681 | getinfo_super_imsm(st, &info); | |
cf8de691 | 682 | fname_from_uuid(st, &info, nbuf1,'-'); |
44470971 DW |
683 | printf("ARRAY /dev/md/%.16s container=%s\n" |
684 | " member=%d auto=mdp UUID=%s\n", | |
cf8de691 | 685 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 686 | } |
cdddbdbc DW |
687 | } |
688 | ||
689 | static void detail_super_imsm(struct supertype *st, char *homehost) | |
690 | { | |
3ebe00a1 DW |
691 | struct mdinfo info; |
692 | char nbuf[64]; | |
693 | ||
694 | getinfo_super_imsm(st, &info); | |
695 | fname_from_uuid(st, &info, nbuf,'-'); | |
696 | printf("\n UUID : %s\n", nbuf + 5); | |
cdddbdbc DW |
697 | } |
698 | ||
699 | static void brief_detail_super_imsm(struct supertype *st) | |
700 | { | |
ff54de6e N |
701 | struct mdinfo info; |
702 | char nbuf[64]; | |
703 | getinfo_super_imsm(st, &info); | |
704 | fname_from_uuid(st, &info, nbuf,'-'); | |
705 | printf(" UUID=%s", nbuf + 5); | |
cdddbdbc DW |
706 | } |
707 | #endif | |
708 | ||
709 | static int match_home_imsm(struct supertype *st, char *homehost) | |
710 | { | |
5115ca67 DW |
711 | /* the imsm metadata format does not specify any host |
712 | * identification information. We return -1 since we can never | |
713 | * confirm nor deny whether a given array is "meant" for this | |
714 | * host. We rely on compare_super and the 'family_num' field to | |
715 | * exclude member disks that do not belong, and we rely on | |
716 | * mdadm.conf to specify the arrays that should be assembled. | |
717 | * Auto-assembly may still pick up "foreign" arrays. | |
718 | */ | |
cdddbdbc | 719 | |
9362c1c8 | 720 | return -1; |
cdddbdbc DW |
721 | } |
722 | ||
723 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
724 | { | |
51006d85 N |
725 | /* The uuid returned here is used for: |
726 | * uuid to put into bitmap file (Create, Grow) | |
727 | * uuid for backup header when saving critical section (Grow) | |
728 | * comparing uuids when re-adding a device into an array | |
729 | * In these cases the uuid required is that of the data-array, | |
730 | * not the device-set. | |
731 | * uuid to recognise same set when adding a missing device back | |
732 | * to an array. This is a uuid for the device-set. | |
733 | * | |
734 | * For each of these we can make do with a truncated | |
735 | * or hashed uuid rather than the original, as long as | |
736 | * everyone agrees. | |
737 | * In each case the uuid required is that of the data-array, | |
738 | * not the device-set. | |
43dad3d6 | 739 | */ |
51006d85 N |
740 | /* imsm does not track uuid's so we synthesis one using sha1 on |
741 | * - The signature (Which is constant for all imsm array, but no matter) | |
742 | * - the family_num of the container | |
743 | * - the index number of the volume | |
744 | * - the 'serial' number of the volume. | |
745 | * Hopefully these are all constant. | |
746 | */ | |
747 | struct intel_super *super = st->sb; | |
43dad3d6 | 748 | |
51006d85 N |
749 | char buf[20]; |
750 | struct sha1_ctx ctx; | |
751 | struct imsm_dev *dev = NULL; | |
752 | ||
753 | sha1_init_ctx(&ctx); | |
754 | sha1_process_bytes(super->anchor->sig, MAX_SIGNATURE_LENGTH, &ctx); | |
755 | sha1_process_bytes(&super->anchor->family_num, sizeof(__u32), &ctx); | |
756 | if (super->current_vol >= 0) | |
757 | dev = get_imsm_dev(super, super->current_vol); | |
758 | if (dev) { | |
759 | __u32 vol = super->current_vol; | |
760 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
761 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
762 | } | |
763 | sha1_finish_ctx(&ctx, buf); | |
764 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
765 | } |
766 | ||
0d481d37 | 767 | #if 0 |
4f5bc454 DW |
768 | static void |
769 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 770 | { |
cdddbdbc DW |
771 | __u8 *v = get_imsm_version(mpb); |
772 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
773 | char major[] = { 0, 0, 0 }; | |
774 | char minor[] = { 0 ,0, 0 }; | |
775 | char patch[] = { 0, 0, 0 }; | |
776 | char *ver_parse[] = { major, minor, patch }; | |
777 | int i, j; | |
778 | ||
779 | i = j = 0; | |
780 | while (*v != '\0' && v < end) { | |
781 | if (*v != '.' && j < 2) | |
782 | ver_parse[i][j++] = *v; | |
783 | else { | |
784 | i++; | |
785 | j = 0; | |
786 | } | |
787 | v++; | |
788 | } | |
789 | ||
4f5bc454 DW |
790 | *m = strtol(minor, NULL, 0); |
791 | *p = strtol(patch, NULL, 0); | |
792 | } | |
0d481d37 | 793 | #endif |
4f5bc454 | 794 | |
c2c087e6 DW |
795 | static int imsm_level_to_layout(int level) |
796 | { | |
797 | switch (level) { | |
798 | case 0: | |
799 | case 1: | |
800 | return 0; | |
801 | case 5: | |
802 | case 6: | |
a380c027 | 803 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 804 | case 10: |
c92a2527 | 805 | return 0x102; |
c2c087e6 DW |
806 | } |
807 | return -1; | |
808 | } | |
809 | ||
bf5a934a DW |
810 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info) |
811 | { | |
812 | struct intel_super *super = st->sb; | |
949c47a0 | 813 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 814 | struct imsm_map *map = get_imsm_map(dev, 0); |
bf5a934a DW |
815 | |
816 | info->container_member = super->current_vol; | |
817 | info->array.raid_disks = map->num_members; | |
818 | info->array.level = get_imsm_raid_level(map); | |
819 | info->array.layout = imsm_level_to_layout(info->array.level); | |
820 | info->array.md_minor = -1; | |
821 | info->array.ctime = 0; | |
822 | info->array.utime = 0; | |
301406c9 DW |
823 | info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9; |
824 | info->array.state = !dev->vol.dirty; | |
825 | ||
826 | info->disk.major = 0; | |
827 | info->disk.minor = 0; | |
bf5a934a DW |
828 | |
829 | info->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
830 | info->component_size = __le32_to_cpu(map->blocks_per_member); | |
301406c9 | 831 | memset(info->uuid, 0, sizeof(info->uuid)); |
bf5a934a | 832 | |
f8f603f1 | 833 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty) |
301406c9 | 834 | info->resync_start = 0; |
f8f603f1 DW |
835 | else if (dev->vol.migr_state) |
836 | info->resync_start = __le32_to_cpu(dev->vol.curr_migr_unit); | |
301406c9 DW |
837 | else |
838 | info->resync_start = ~0ULL; | |
839 | ||
840 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
841 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 842 | |
f35f2525 N |
843 | info->array.major_version = -1; |
844 | info->array.minor_version = -2; | |
bf5a934a DW |
845 | sprintf(info->text_version, "/%s/%d", |
846 | devnum2devname(st->container_dev), | |
847 | info->container_member); | |
a67dd8cc | 848 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 849 | uuid_from_super_imsm(st, info->uuid); |
bf5a934a DW |
850 | } |
851 | ||
852 | ||
4f5bc454 DW |
853 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info) |
854 | { | |
855 | struct intel_super *super = st->sb; | |
4f5bc454 DW |
856 | struct imsm_disk *disk; |
857 | __u32 s; | |
4f5bc454 | 858 | |
bf5a934a DW |
859 | if (super->current_vol >= 0) { |
860 | getinfo_super_imsm_volume(st, info); | |
861 | return; | |
862 | } | |
d23fe947 DW |
863 | |
864 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
865 | * spares | |
866 | */ | |
867 | info->array.raid_disks = 0; | |
cdddbdbc DW |
868 | info->array.level = LEVEL_CONTAINER; |
869 | info->array.layout = 0; | |
870 | info->array.md_minor = -1; | |
c2c087e6 | 871 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
872 | info->array.utime = 0; |
873 | info->array.chunk_size = 0; | |
874 | ||
875 | info->disk.major = 0; | |
876 | info->disk.minor = 0; | |
cdddbdbc | 877 | info->disk.raid_disk = -1; |
c2c087e6 | 878 | info->reshape_active = 0; |
f35f2525 N |
879 | info->array.major_version = -1; |
880 | info->array.minor_version = -2; | |
c2c087e6 | 881 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 882 | info->safe_mode_delay = 0; |
c2c087e6 DW |
883 | info->disk.number = -1; |
884 | info->disk.state = 0; | |
c5afc314 | 885 | info->name[0] = 0; |
c2c087e6 | 886 | |
4a04ec6c | 887 | if (super->disks) { |
14e8215b DW |
888 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
889 | ||
b9f594fe | 890 | disk = &super->disks->disk; |
14e8215b DW |
891 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
892 | info->component_size = reserved; | |
f2f27e63 | 893 | s = disk->status; |
4a04ec6c DW |
894 | info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0; |
895 | info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0; | |
032e9e29 | 896 | info->disk.state |= s & SPARE_DISK ? 0 : (1 << MD_DISK_SYNC); |
cdddbdbc | 897 | } |
a575e2a7 DW |
898 | |
899 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
900 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
901 | */ | |
902 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 903 | uuid_from_super_imsm(st, info->uuid); |
032e9e29 DW |
904 | else |
905 | memcpy(info->uuid, uuid_match_any, sizeof(int[4])); | |
cdddbdbc DW |
906 | } |
907 | ||
cdddbdbc DW |
908 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
909 | char *update, char *devname, int verbose, | |
910 | int uuid_set, char *homehost) | |
911 | { | |
f352c545 DW |
912 | /* FIXME */ |
913 | ||
914 | /* For 'assemble' and 'force' we need to return non-zero if any | |
915 | * change was made. For others, the return value is ignored. | |
916 | * Update options are: | |
917 | * force-one : This device looks a bit old but needs to be included, | |
918 | * update age info appropriately. | |
919 | * assemble: clear any 'faulty' flag to allow this device to | |
920 | * be assembled. | |
921 | * force-array: Array is degraded but being forced, mark it clean | |
922 | * if that will be needed to assemble it. | |
923 | * | |
924 | * newdev: not used ???? | |
925 | * grow: Array has gained a new device - this is currently for | |
926 | * linear only | |
927 | * resync: mark as dirty so a resync will happen. | |
928 | * name: update the name - preserving the homehost | |
929 | * | |
930 | * Following are not relevant for this imsm: | |
931 | * sparc2.2 : update from old dodgey metadata | |
932 | * super-minor: change the preferred_minor number | |
933 | * summaries: update redundant counters. | |
934 | * uuid: Change the uuid of the array to match watch is given | |
935 | * homehost: update the recorded homehost | |
936 | * _reshape_progress: record new reshape_progress position. | |
937 | */ | |
938 | int rv = 0; | |
939 | //struct intel_super *super = st->sb; | |
940 | //struct imsm_super *mpb = super->mpb; | |
941 | ||
942 | if (strcmp(update, "grow") == 0) { | |
943 | } | |
944 | if (strcmp(update, "resync") == 0) { | |
945 | /* dev->vol.dirty = 1; */ | |
946 | } | |
947 | ||
948 | /* IMSM has no concept of UUID or homehost */ | |
949 | ||
950 | return rv; | |
cdddbdbc DW |
951 | } |
952 | ||
c2c087e6 | 953 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 954 | { |
c2c087e6 | 955 | size_t size; |
cdddbdbc | 956 | |
c2c087e6 DW |
957 | size = sizeof(struct imsm_super); |
958 | size += (disks - 1) * sizeof(struct imsm_disk); | |
959 | size += 2 * sizeof(struct imsm_dev); | |
960 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
961 | size += (4 - 2) * sizeof(struct imsm_map); | |
962 | /* 4 possible disk_ord_tbl's */ | |
963 | size += 4 * (disks - 1) * sizeof(__u32); | |
964 | ||
965 | return size; | |
966 | } | |
967 | ||
968 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
969 | { | |
970 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
971 | return 0; | |
972 | ||
973 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
974 | } |
975 | ||
976 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) | |
977 | { | |
978 | /* | |
979 | * return: | |
980 | * 0 same, or first was empty, and second was copied | |
981 | * 1 second had wrong number | |
982 | * 2 wrong uuid | |
983 | * 3 wrong other info | |
984 | */ | |
985 | struct intel_super *first = st->sb; | |
986 | struct intel_super *sec = tst->sb; | |
987 | ||
988 | if (!first) { | |
989 | st->sb = tst->sb; | |
990 | tst->sb = NULL; | |
991 | return 0; | |
992 | } | |
993 | ||
949c47a0 | 994 | if (memcmp(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH) != 0) |
cdddbdbc | 995 | return 3; |
d23fe947 DW |
996 | |
997 | /* if an anchor does not have num_raid_devs set then it is a free | |
998 | * floating spare | |
999 | */ | |
1000 | if (first->anchor->num_raid_devs > 0 && | |
1001 | sec->anchor->num_raid_devs > 0) { | |
1002 | if (first->anchor->family_num != sec->anchor->family_num) | |
1003 | return 3; | |
d23fe947 | 1004 | } |
cdddbdbc | 1005 | |
3e372e5a DW |
1006 | /* if 'first' is a spare promote it to a populated mpb with sec's |
1007 | * family number | |
1008 | */ | |
1009 | if (first->anchor->num_raid_devs == 0 && | |
1010 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 DW |
1011 | int i; |
1012 | ||
1013 | /* we need to copy raid device info from sec if an allocation | |
1014 | * fails here we don't associate the spare | |
1015 | */ | |
1016 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
1017 | first->dev_tbl[i] = malloc(sizeof(struct imsm_dev)); | |
1018 | if (!first->dev_tbl) { | |
1019 | while (--i >= 0) { | |
1020 | free(first->dev_tbl[i]); | |
1021 | first->dev_tbl[i] = NULL; | |
1022 | } | |
1023 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
1024 | return 3; | |
1025 | } | |
1026 | *first->dev_tbl[i] = *sec->dev_tbl[i]; | |
1027 | } | |
1028 | ||
3e372e5a DW |
1029 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
1030 | first->anchor->family_num = sec->anchor->family_num; | |
1031 | } | |
1032 | ||
cdddbdbc DW |
1033 | return 0; |
1034 | } | |
1035 | ||
0030e8d6 DW |
1036 | static void fd2devname(int fd, char *name) |
1037 | { | |
1038 | struct stat st; | |
1039 | char path[256]; | |
1040 | char dname[100]; | |
1041 | char *nm; | |
1042 | int rv; | |
1043 | ||
1044 | name[0] = '\0'; | |
1045 | if (fstat(fd, &st) != 0) | |
1046 | return; | |
1047 | sprintf(path, "/sys/dev/block/%d:%d", | |
1048 | major(st.st_rdev), minor(st.st_rdev)); | |
1049 | ||
1050 | rv = readlink(path, dname, sizeof(dname)); | |
1051 | if (rv <= 0) | |
1052 | return; | |
1053 | ||
1054 | dname[rv] = '\0'; | |
1055 | nm = strrchr(dname, '/'); | |
1056 | nm++; | |
1057 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
1058 | } | |
1059 | ||
1060 | ||
cdddbdbc DW |
1061 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
1062 | ||
1063 | static int imsm_read_serial(int fd, char *devname, | |
1064 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
1065 | { | |
1066 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
1067 | int rv; |
1068 | int rsp_len; | |
1f24f035 DW |
1069 | int len; |
1070 | char *c, *rsp_buf; | |
cdddbdbc DW |
1071 | |
1072 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 1073 | |
f9ba0ff1 DW |
1074 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
1075 | ||
40ebbb9c | 1076 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
1077 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
1078 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
1079 | return 0; |
1080 | } | |
1081 | ||
cdddbdbc DW |
1082 | if (rv != 0) { |
1083 | if (devname) | |
1084 | fprintf(stderr, | |
1085 | Name ": Failed to retrieve serial for %s\n", | |
1086 | devname); | |
1087 | return rv; | |
1088 | } | |
1089 | ||
5c3db629 | 1090 | /* trim leading whitespace */ |
cdddbdbc | 1091 | rsp_len = scsi_serial[3]; |
1f24f035 DW |
1092 | rsp_buf = (char *) &scsi_serial[4]; |
1093 | c = rsp_buf; | |
1094 | while (isspace(*c)) | |
1095 | c++; | |
5c3db629 DW |
1096 | |
1097 | /* truncate len to the end of rsp_buf if necessary */ | |
1f24f035 DW |
1098 | if (c + MAX_RAID_SERIAL_LEN > rsp_buf + rsp_len) |
1099 | len = rsp_len - (c - rsp_buf); | |
1100 | else | |
1101 | len = MAX_RAID_SERIAL_LEN; | |
5c3db629 DW |
1102 | |
1103 | /* initialize the buffer and copy rsp_buf characters */ | |
1104 | memset(serial, 0, MAX_RAID_SERIAL_LEN); | |
1f24f035 | 1105 | memcpy(serial, c, len); |
5c3db629 DW |
1106 | |
1107 | /* trim trailing whitespace starting with the last character copied */ | |
1f24f035 DW |
1108 | c = (char *) &serial[len - 1]; |
1109 | while (isspace(*c) || *c == '\0') | |
1110 | *c-- = '\0'; | |
cdddbdbc DW |
1111 | |
1112 | return 0; | |
1113 | } | |
1114 | ||
1f24f035 DW |
1115 | static int serialcmp(__u8 *s1, __u8 *s2) |
1116 | { | |
1117 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
1118 | } | |
1119 | ||
1120 | static void serialcpy(__u8 *dest, __u8 *src) | |
1121 | { | |
1122 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
1123 | } | |
1124 | ||
cdddbdbc DW |
1125 | static int |
1126 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
1127 | { | |
cdddbdbc DW |
1128 | struct dl *dl; |
1129 | struct stat stb; | |
cdddbdbc DW |
1130 | int rv; |
1131 | int i; | |
d23fe947 DW |
1132 | int alloc = 1; |
1133 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
1134 | ||
1135 | rv = imsm_read_serial(fd, devname, serial); | |
1136 | ||
1137 | if (rv != 0) | |
1138 | return 2; | |
1139 | ||
1140 | /* check if this is a disk we have seen before. it may be a spare in | |
1141 | * super->disks while the current anchor believes it is a raid member, | |
1142 | * check if we need to update dl->index | |
1143 | */ | |
1144 | for (dl = super->disks; dl; dl = dl->next) | |
1f24f035 | 1145 | if (serialcmp(dl->serial, serial) == 0) |
d23fe947 DW |
1146 | break; |
1147 | ||
1148 | if (!dl) | |
1149 | dl = malloc(sizeof(*dl)); | |
1150 | else | |
1151 | alloc = 0; | |
cdddbdbc | 1152 | |
b9f594fe | 1153 | if (!dl) { |
cdddbdbc DW |
1154 | if (devname) |
1155 | fprintf(stderr, | |
1156 | Name ": failed to allocate disk buffer for %s\n", | |
1157 | devname); | |
1158 | return 2; | |
1159 | } | |
cdddbdbc | 1160 | |
d23fe947 DW |
1161 | if (alloc) { |
1162 | fstat(fd, &stb); | |
1163 | dl->major = major(stb.st_rdev); | |
1164 | dl->minor = minor(stb.st_rdev); | |
1165 | dl->next = super->disks; | |
1166 | dl->fd = keep_fd ? fd : -1; | |
1167 | dl->devname = devname ? strdup(devname) : NULL; | |
1f24f035 | 1168 | serialcpy(dl->serial, serial); |
8796fdc4 | 1169 | dl->index = -2; |
d23fe947 DW |
1170 | } else if (keep_fd) { |
1171 | close(dl->fd); | |
1172 | dl->fd = fd; | |
1173 | } | |
cdddbdbc | 1174 | |
d23fe947 | 1175 | /* look up this disk's index in the current anchor */ |
949c47a0 DW |
1176 | for (i = 0; i < super->anchor->num_disks; i++) { |
1177 | struct imsm_disk *disk_iter; | |
1178 | ||
1179 | disk_iter = __get_imsm_disk(super->anchor, i); | |
cdddbdbc | 1180 | |
1f24f035 | 1181 | if (serialcmp(disk_iter->serial, dl->serial) == 0) { |
b9f594fe | 1182 | dl->disk = *disk_iter; |
d23fe947 DW |
1183 | /* only set index on disks that are a member of a |
1184 | * populated contianer, i.e. one with raid_devs | |
1185 | */ | |
f2f27e63 | 1186 | if (dl->disk.status & FAILED_DISK) |
6c386dd3 | 1187 | dl->index = -2; |
f2f27e63 | 1188 | else if (dl->disk.status & SPARE_DISK) |
d23fe947 DW |
1189 | dl->index = -1; |
1190 | else | |
1191 | dl->index = i; | |
8796fdc4 | 1192 | |
cdddbdbc | 1193 | break; |
949c47a0 | 1194 | } |
cdddbdbc DW |
1195 | } |
1196 | ||
3f6efecc DW |
1197 | /* no match, maybe a stale failed drive */ |
1198 | if (i == super->anchor->num_disks && dl->index >= 0) { | |
1199 | dl->disk = *__get_imsm_disk(super->anchor, dl->index); | |
f2f27e63 | 1200 | if (dl->disk.status & FAILED_DISK) |
3f6efecc DW |
1201 | dl->index = -2; |
1202 | } | |
1203 | ||
d23fe947 DW |
1204 | if (alloc) |
1205 | super->disks = dl; | |
6c386dd3 | 1206 | |
949c47a0 DW |
1207 | return 0; |
1208 | } | |
1209 | ||
1210 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
1211 | { | |
3393c6af DW |
1212 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); |
1213 | } | |
1214 | ||
0e600426 | 1215 | #ifndef MDASSEMBLE |
0c046afd DW |
1216 | /* When migrating map0 contains the 'destination' state while map1 |
1217 | * contains the current state. When not migrating map0 contains the | |
1218 | * current state. This routine assumes that map[0].map_state is set to | |
1219 | * the current array state before being called. | |
1220 | * | |
1221 | * Migration is indicated by one of the following states | |
1222 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 1223 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 1224 | * map1state=unitialized) |
e3bba0e0 | 1225 | * 3/ Verify (Resync) (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 1226 | * map1state=normal) |
e3bba0e0 | 1227 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd DW |
1228 | * map1state=degraded) |
1229 | */ | |
1230 | static void migrate(struct imsm_dev *dev, __u8 to_state, int rebuild_resync) | |
3393c6af | 1231 | { |
0c046afd | 1232 | struct imsm_map *dest; |
3393c6af DW |
1233 | struct imsm_map *src = get_imsm_map(dev, 0); |
1234 | ||
0c046afd DW |
1235 | dev->vol.migr_state = 1; |
1236 | dev->vol.migr_type = rebuild_resync; | |
f8f603f1 | 1237 | dev->vol.curr_migr_unit = 0; |
0c046afd DW |
1238 | dest = get_imsm_map(dev, 1); |
1239 | ||
3393c6af | 1240 | memcpy(dest, src, sizeof_imsm_map(src)); |
0c046afd | 1241 | src->map_state = to_state; |
949c47a0 | 1242 | } |
f8f603f1 DW |
1243 | |
1244 | static void end_migration(struct imsm_dev *dev, __u8 map_state) | |
1245 | { | |
1246 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1247 | ||
1248 | dev->vol.migr_state = 0; | |
1249 | dev->vol.curr_migr_unit = 0; | |
1250 | map->map_state = map_state; | |
1251 | } | |
0e600426 | 1252 | #endif |
949c47a0 DW |
1253 | |
1254 | static int parse_raid_devices(struct intel_super *super) | |
1255 | { | |
1256 | int i; | |
1257 | struct imsm_dev *dev_new; | |
4d7b1503 DW |
1258 | size_t len, len_migr; |
1259 | size_t space_needed = 0; | |
1260 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
1261 | |
1262 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
1263 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
1264 | ||
4d7b1503 DW |
1265 | len = sizeof_imsm_dev(dev_iter, 0); |
1266 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
1267 | if (len_migr > len) | |
1268 | space_needed += len_migr - len; | |
1269 | ||
1270 | dev_new = malloc(len_migr); | |
949c47a0 DW |
1271 | if (!dev_new) |
1272 | return 1; | |
1273 | imsm_copy_dev(dev_new, dev_iter); | |
1274 | super->dev_tbl[i] = dev_new; | |
1275 | } | |
cdddbdbc | 1276 | |
4d7b1503 DW |
1277 | /* ensure that super->buf is large enough when all raid devices |
1278 | * are migrating | |
1279 | */ | |
1280 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
1281 | void *buf; | |
1282 | ||
1283 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
1284 | if (posix_memalign(&buf, 512, len) != 0) | |
1285 | return 1; | |
1286 | ||
1287 | memcpy(buf, super->buf, len); | |
1288 | free(super->buf); | |
1289 | super->buf = buf; | |
1290 | super->len = len; | |
1291 | } | |
1292 | ||
cdddbdbc DW |
1293 | return 0; |
1294 | } | |
1295 | ||
604b746f JD |
1296 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
1297 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
1298 | { | |
1299 | void *ptr = NULL; | |
1300 | ||
1301 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
1302 | ptr = mpb; | |
1303 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
1304 | } | |
1305 | ||
1306 | return ptr; | |
1307 | } | |
1308 | ||
d23fe947 | 1309 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 1310 | |
cdddbdbc DW |
1311 | /* load_imsm_mpb - read matrix metadata |
1312 | * allocates super->mpb to be freed by free_super | |
1313 | */ | |
1314 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
1315 | { | |
1316 | unsigned long long dsize; | |
cdddbdbc DW |
1317 | unsigned long long sectors; |
1318 | struct stat; | |
6416d527 | 1319 | struct imsm_super *anchor; |
cdddbdbc | 1320 | __u32 check_sum; |
949c47a0 | 1321 | int rc; |
cdddbdbc | 1322 | |
cdddbdbc DW |
1323 | get_dev_size(fd, NULL, &dsize); |
1324 | ||
1325 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
1326 | if (devname) | |
1327 | fprintf(stderr, | |
1328 | Name ": Cannot seek to anchor block on %s: %s\n", | |
1329 | devname, strerror(errno)); | |
1330 | return 1; | |
1331 | } | |
1332 | ||
949c47a0 | 1333 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
1334 | if (devname) |
1335 | fprintf(stderr, | |
1336 | Name ": Failed to allocate imsm anchor buffer" | |
1337 | " on %s\n", devname); | |
1338 | return 1; | |
1339 | } | |
949c47a0 | 1340 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
1341 | if (devname) |
1342 | fprintf(stderr, | |
1343 | Name ": Cannot read anchor block on %s: %s\n", | |
1344 | devname, strerror(errno)); | |
6416d527 | 1345 | free(anchor); |
cdddbdbc DW |
1346 | return 1; |
1347 | } | |
1348 | ||
6416d527 | 1349 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
1350 | if (devname) |
1351 | fprintf(stderr, | |
1352 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 1353 | free(anchor); |
cdddbdbc DW |
1354 | return 2; |
1355 | } | |
1356 | ||
d23fe947 | 1357 | __free_imsm(super, 0); |
949c47a0 DW |
1358 | super->len = ROUND_UP(anchor->mpb_size, 512); |
1359 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
1360 | if (devname) |
1361 | fprintf(stderr, | |
1362 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 1363 | super->len); |
6416d527 | 1364 | free(anchor); |
cdddbdbc DW |
1365 | return 2; |
1366 | } | |
949c47a0 | 1367 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 1368 | |
6416d527 NB |
1369 | sectors = mpb_sectors(anchor) - 1; |
1370 | free(anchor); | |
949c47a0 DW |
1371 | if (!sectors) { |
1372 | rc = load_imsm_disk(fd, super, devname, 0); | |
1373 | if (rc == 0) | |
1374 | rc = parse_raid_devices(super); | |
1375 | return rc; | |
1376 | } | |
cdddbdbc DW |
1377 | |
1378 | /* read the extended mpb */ | |
1379 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
1380 | if (devname) | |
1381 | fprintf(stderr, | |
1382 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
1383 | devname, strerror(errno)); | |
1384 | return 1; | |
1385 | } | |
1386 | ||
949c47a0 | 1387 | if (read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
1388 | if (devname) |
1389 | fprintf(stderr, | |
1390 | Name ": Cannot read extended mpb on %s: %s\n", | |
1391 | devname, strerror(errno)); | |
1392 | return 2; | |
1393 | } | |
1394 | ||
949c47a0 DW |
1395 | check_sum = __gen_imsm_checksum(super->anchor); |
1396 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
1397 | if (devname) |
1398 | fprintf(stderr, | |
1399 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 1400 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc DW |
1401 | devname); |
1402 | return 2; | |
1403 | } | |
1404 | ||
604b746f JD |
1405 | /* FIXME the BBM log is disk specific so we cannot use this global |
1406 | * buffer for all disks. Ok for now since we only look at the global | |
1407 | * bbm_log_size parameter to gate assembly | |
1408 | */ | |
1409 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
1410 | ||
949c47a0 DW |
1411 | rc = load_imsm_disk(fd, super, devname, 0); |
1412 | if (rc == 0) | |
1413 | rc = parse_raid_devices(super); | |
4d7b1503 | 1414 | |
949c47a0 | 1415 | return rc; |
cdddbdbc DW |
1416 | } |
1417 | ||
ae6aad82 DW |
1418 | static void __free_imsm_disk(struct dl *d) |
1419 | { | |
1420 | if (d->fd >= 0) | |
1421 | close(d->fd); | |
1422 | if (d->devname) | |
1423 | free(d->devname); | |
1424 | free(d); | |
1425 | ||
1426 | } | |
cdddbdbc DW |
1427 | static void free_imsm_disks(struct intel_super *super) |
1428 | { | |
47ee5a45 | 1429 | struct dl *d; |
cdddbdbc | 1430 | |
47ee5a45 DW |
1431 | while (super->disks) { |
1432 | d = super->disks; | |
cdddbdbc | 1433 | super->disks = d->next; |
ae6aad82 | 1434 | __free_imsm_disk(d); |
cdddbdbc | 1435 | } |
47ee5a45 DW |
1436 | while (super->missing) { |
1437 | d = super->missing; | |
1438 | super->missing = d->next; | |
1439 | __free_imsm_disk(d); | |
1440 | } | |
1441 | ||
cdddbdbc DW |
1442 | } |
1443 | ||
9ca2c81c | 1444 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 1445 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 1446 | { |
949c47a0 DW |
1447 | int i; |
1448 | ||
9ca2c81c | 1449 | if (super->buf) { |
949c47a0 | 1450 | free(super->buf); |
9ca2c81c DW |
1451 | super->buf = NULL; |
1452 | } | |
d23fe947 DW |
1453 | if (free_disks) |
1454 | free_imsm_disks(super); | |
949c47a0 | 1455 | for (i = 0; i < IMSM_MAX_RAID_DEVS; i++) |
9ca2c81c | 1456 | if (super->dev_tbl[i]) { |
949c47a0 | 1457 | free(super->dev_tbl[i]); |
9ca2c81c DW |
1458 | super->dev_tbl[i] = NULL; |
1459 | } | |
cdddbdbc DW |
1460 | } |
1461 | ||
9ca2c81c DW |
1462 | static void free_imsm(struct intel_super *super) |
1463 | { | |
d23fe947 | 1464 | __free_imsm(super, 1); |
9ca2c81c DW |
1465 | free(super); |
1466 | } | |
cdddbdbc DW |
1467 | |
1468 | static void free_super_imsm(struct supertype *st) | |
1469 | { | |
1470 | struct intel_super *super = st->sb; | |
1471 | ||
1472 | if (!super) | |
1473 | return; | |
1474 | ||
1475 | free_imsm(super); | |
1476 | st->sb = NULL; | |
1477 | } | |
1478 | ||
c2c087e6 DW |
1479 | static struct intel_super *alloc_super(int creating_imsm) |
1480 | { | |
1481 | struct intel_super *super = malloc(sizeof(*super)); | |
1482 | ||
1483 | if (super) { | |
1484 | memset(super, 0, sizeof(*super)); | |
1485 | super->creating_imsm = creating_imsm; | |
bf5a934a | 1486 | super->current_vol = -1; |
c2c087e6 DW |
1487 | } |
1488 | ||
1489 | return super; | |
1490 | } | |
1491 | ||
cdddbdbc | 1492 | #ifndef MDASSEMBLE |
47ee5a45 DW |
1493 | /* find_missing - helper routine for load_super_imsm_all that identifies |
1494 | * disks that have disappeared from the system. This routine relies on | |
1495 | * the mpb being uptodate, which it is at load time. | |
1496 | */ | |
1497 | static int find_missing(struct intel_super *super) | |
1498 | { | |
1499 | int i; | |
1500 | struct imsm_super *mpb = super->anchor; | |
1501 | struct dl *dl; | |
1502 | struct imsm_disk *disk; | |
47ee5a45 DW |
1503 | |
1504 | for (i = 0; i < mpb->num_disks; i++) { | |
1505 | disk = __get_imsm_disk(mpb, i); | |
1506 | for (dl = super->disks; dl; dl = dl->next) | |
1507 | if (serialcmp(dl->disk.serial, disk->serial) == 0) | |
1508 | break; | |
1509 | if (dl) | |
1510 | continue; | |
1511 | /* ok we have a 'disk' without a live entry in | |
1512 | * super->disks | |
1513 | */ | |
f2f27e63 | 1514 | if (disk->status & FAILED_DISK || !(disk->status & USABLE_DISK)) |
47ee5a45 DW |
1515 | continue; /* never mind, already marked */ |
1516 | ||
1517 | dl = malloc(sizeof(*dl)); | |
1518 | if (!dl) | |
1519 | return 1; | |
1520 | dl->major = 0; | |
1521 | dl->minor = 0; | |
1522 | dl->fd = -1; | |
1523 | dl->devname = strdup("missing"); | |
1524 | dl->index = i; | |
1525 | serialcpy(dl->serial, disk->serial); | |
1526 | dl->disk = *disk; | |
1527 | dl->next = super->missing; | |
1528 | super->missing = dl; | |
1529 | } | |
1530 | ||
1531 | return 0; | |
1532 | } | |
1533 | ||
cdddbdbc DW |
1534 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
1535 | char *devname, int keep_fd) | |
1536 | { | |
1537 | struct mdinfo *sra; | |
1538 | struct intel_super *super; | |
1539 | struct mdinfo *sd, *best = NULL; | |
1540 | __u32 bestgen = 0; | |
1541 | __u32 gen; | |
1542 | char nm[20]; | |
1543 | int dfd; | |
1544 | int rv; | |
1545 | ||
1546 | /* check if this disk is a member of an active array */ | |
1547 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
1548 | if (!sra) | |
1549 | return 1; | |
1550 | ||
1551 | if (sra->array.major_version != -1 || | |
1552 | sra->array.minor_version != -2 || | |
1553 | strcmp(sra->text_version, "imsm") != 0) | |
1554 | return 1; | |
1555 | ||
c2c087e6 | 1556 | super = alloc_super(0); |
cdddbdbc DW |
1557 | if (!super) |
1558 | return 1; | |
1559 | ||
d23fe947 | 1560 | /* find the most up to date disk in this array, skipping spares */ |
cdddbdbc DW |
1561 | for (sd = sra->devs; sd; sd = sd->next) { |
1562 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1563 | dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY); | |
1564 | if (!dfd) { | |
1565 | free_imsm(super); | |
1566 | return 2; | |
1567 | } | |
1568 | rv = load_imsm_mpb(dfd, super, NULL); | |
1569 | if (!keep_fd) | |
1570 | close(dfd); | |
1571 | if (rv == 0) { | |
d23fe947 DW |
1572 | if (super->anchor->num_raid_devs == 0) |
1573 | gen = 0; | |
1574 | else | |
1575 | gen = __le32_to_cpu(super->anchor->generation_num); | |
cdddbdbc DW |
1576 | if (!best || gen > bestgen) { |
1577 | bestgen = gen; | |
1578 | best = sd; | |
1579 | } | |
1580 | } else { | |
1581 | free_imsm(super); | |
1582 | return 2; | |
1583 | } | |
1584 | } | |
1585 | ||
1586 | if (!best) { | |
1587 | free_imsm(super); | |
1588 | return 1; | |
1589 | } | |
1590 | ||
1591 | /* load the most up to date anchor */ | |
1592 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
1593 | dfd = dev_open(nm, O_RDONLY); | |
1594 | if (!dfd) { | |
1595 | free_imsm(super); | |
1596 | return 1; | |
1597 | } | |
1598 | rv = load_imsm_mpb(dfd, super, NULL); | |
1599 | close(dfd); | |
1600 | if (rv != 0) { | |
1601 | free_imsm(super); | |
1602 | return 2; | |
1603 | } | |
1604 | ||
d23fe947 | 1605 | /* re-parse the disk list with the current anchor */ |
cdddbdbc DW |
1606 | for (sd = sra->devs ; sd ; sd = sd->next) { |
1607 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1608 | dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY); | |
1609 | if (!dfd) { | |
1610 | free_imsm(super); | |
1611 | return 2; | |
1612 | } | |
1613 | load_imsm_disk(dfd, super, NULL, keep_fd); | |
1614 | if (!keep_fd) | |
1615 | close(dfd); | |
1616 | } | |
1617 | ||
47ee5a45 DW |
1618 | |
1619 | if (find_missing(super) != 0) { | |
1620 | free_imsm(super); | |
1621 | return 2; | |
1622 | } | |
1623 | ||
f7e7067b | 1624 | if (st->subarray[0]) { |
949c47a0 | 1625 | if (atoi(st->subarray) <= super->anchor->num_raid_devs) |
bf5a934a DW |
1626 | super->current_vol = atoi(st->subarray); |
1627 | else | |
1628 | return 1; | |
f7e7067b NB |
1629 | } |
1630 | ||
cdddbdbc | 1631 | *sbp = super; |
43dad3d6 | 1632 | st->container_dev = fd2devnum(fd); |
cdddbdbc | 1633 | if (st->ss == NULL) { |
bf5a934a | 1634 | st->ss = &super_imsm; |
cdddbdbc DW |
1635 | st->minor_version = 0; |
1636 | st->max_devs = IMSM_MAX_DEVICES; | |
1637 | } | |
352452c3 | 1638 | st->loaded_container = 1; |
cdddbdbc DW |
1639 | |
1640 | return 0; | |
1641 | } | |
1642 | #endif | |
1643 | ||
1644 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
1645 | { | |
1646 | struct intel_super *super; | |
1647 | int rv; | |
1648 | ||
1649 | #ifndef MDASSEMBLE | |
3dbccbcf | 1650 | if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0) |
cdddbdbc DW |
1651 | return 0; |
1652 | #endif | |
f7e7067b NB |
1653 | if (st->subarray[0]) |
1654 | return 1; /* FIXME */ | |
cdddbdbc | 1655 | |
c2c087e6 | 1656 | super = alloc_super(0); |
cdddbdbc DW |
1657 | if (!super) { |
1658 | fprintf(stderr, | |
1659 | Name ": malloc of %zu failed.\n", | |
1660 | sizeof(*super)); | |
1661 | return 1; | |
1662 | } | |
1663 | ||
1664 | rv = load_imsm_mpb(fd, super, devname); | |
1665 | ||
1666 | if (rv) { | |
1667 | if (devname) | |
1668 | fprintf(stderr, | |
1669 | Name ": Failed to load all information " | |
1670 | "sections on %s\n", devname); | |
1671 | free_imsm(super); | |
1672 | return rv; | |
1673 | } | |
1674 | ||
1675 | st->sb = super; | |
1676 | if (st->ss == NULL) { | |
1677 | st->ss = &super_imsm; | |
1678 | st->minor_version = 0; | |
1679 | st->max_devs = IMSM_MAX_DEVICES; | |
1680 | } | |
352452c3 | 1681 | st->loaded_container = 0; |
cdddbdbc DW |
1682 | |
1683 | return 0; | |
1684 | } | |
1685 | ||
ef6ffade DW |
1686 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
1687 | { | |
1688 | if (info->level == 1) | |
1689 | return 128; | |
1690 | return info->chunk_size >> 9; | |
1691 | } | |
1692 | ||
1693 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info) | |
1694 | { | |
1695 | __u32 num_stripes; | |
1696 | ||
1697 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
1698 | if (info->level == 1) | |
1699 | num_stripes /= 2; | |
1700 | ||
1701 | return num_stripes; | |
1702 | } | |
1703 | ||
fcfd9599 DW |
1704 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
1705 | { | |
1706 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
1707 | } | |
1708 | ||
8b353278 DW |
1709 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
1710 | unsigned long long size, char *name, | |
1711 | char *homehost, int *uuid) | |
cdddbdbc | 1712 | { |
c2c087e6 DW |
1713 | /* We are creating a volume inside a pre-existing container. |
1714 | * so st->sb is already set. | |
1715 | */ | |
1716 | struct intel_super *super = st->sb; | |
949c47a0 | 1717 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
1718 | struct imsm_dev *dev; |
1719 | struct imsm_vol *vol; | |
1720 | struct imsm_map *map; | |
1721 | int idx = mpb->num_raid_devs; | |
1722 | int i; | |
1723 | unsigned long long array_blocks; | |
c2c087e6 | 1724 | __u32 offset = 0; |
2c092cad | 1725 | size_t size_old, size_new; |
cdddbdbc | 1726 | |
c2c087e6 DW |
1727 | if (mpb->num_raid_devs >= 2) { |
1728 | fprintf(stderr, Name": This imsm-container already has the " | |
1729 | "maximum of 2 volumes\n"); | |
1730 | return 0; | |
1731 | } | |
1732 | ||
2c092cad DW |
1733 | /* ensure the mpb is large enough for the new data */ |
1734 | size_old = __le32_to_cpu(mpb->mpb_size); | |
1735 | size_new = disks_to_mpb_size(info->nr_disks); | |
1736 | if (size_new > size_old) { | |
1737 | void *mpb_new; | |
1738 | size_t size_round = ROUND_UP(size_new, 512); | |
1739 | ||
1740 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
1741 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
1742 | return 0; | |
1743 | } | |
1744 | memcpy(mpb_new, mpb, size_old); | |
1745 | free(mpb); | |
1746 | mpb = mpb_new; | |
949c47a0 | 1747 | super->anchor = mpb_new; |
2c092cad DW |
1748 | mpb->mpb_size = __cpu_to_le32(size_new); |
1749 | memset(mpb_new + size_old, 0, size_round - size_old); | |
1750 | } | |
bf5a934a | 1751 | super->current_vol = idx; |
d23fe947 DW |
1752 | /* when creating the first raid device in this container set num_disks |
1753 | * to zero, i.e. delete this spare and add raid member devices in | |
1754 | * add_to_super_imsm_volume() | |
1755 | */ | |
1756 | if (super->current_vol == 0) | |
1757 | mpb->num_disks = 0; | |
bf5a934a | 1758 | sprintf(st->subarray, "%d", idx); |
949c47a0 DW |
1759 | dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
1760 | if (!dev) { | |
1761 | fprintf(stderr, Name": could not allocate raid device\n"); | |
1762 | return 0; | |
1763 | } | |
c2c087e6 DW |
1764 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
1765 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
1766 | info->layout, info->chunk_size, | |
1767 | info->size*2); | |
1768 | dev->size_low = __cpu_to_le32((__u32) array_blocks); | |
1769 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1770 | dev->status = __cpu_to_le32(0); | |
1771 | dev->reserved_blocks = __cpu_to_le32(0); | |
1772 | vol = &dev->vol; | |
1773 | vol->migr_state = 0; | |
e3bba0e0 | 1774 | vol->migr_type = MIGR_INIT; |
c2c087e6 | 1775 | vol->dirty = 0; |
f8f603f1 | 1776 | vol->curr_migr_unit = 0; |
c2c087e6 | 1777 | for (i = 0; i < idx; i++) { |
949c47a0 | 1778 | struct imsm_dev *prev = get_imsm_dev(super, i); |
a965f303 | 1779 | struct imsm_map *pmap = get_imsm_map(prev, 0); |
c2c087e6 DW |
1780 | |
1781 | offset += __le32_to_cpu(pmap->blocks_per_member); | |
1782 | offset += IMSM_RESERVED_SECTORS; | |
1783 | } | |
a965f303 | 1784 | map = get_imsm_map(dev, 0); |
c2c087e6 | 1785 | map->pba_of_lba0 = __cpu_to_le32(offset); |
fcfd9599 | 1786 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade DW |
1787 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
1788 | map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info)); | |
c2c087e6 DW |
1789 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
1790 | IMSM_T_STATE_NORMAL; | |
ef6ffade DW |
1791 | |
1792 | if (info->level == 1 && info->raid_disks > 2) { | |
1793 | fprintf(stderr, Name": imsm does not support more than 2 disks" | |
1794 | "in a raid1 volume\n"); | |
1795 | return 0; | |
1796 | } | |
c2c087e6 DW |
1797 | if (info->level == 10) |
1798 | map->raid_level = 1; | |
1799 | else | |
1800 | map->raid_level = info->level; | |
ef6ffade | 1801 | |
c2c087e6 DW |
1802 | map->num_members = info->raid_disks; |
1803 | for (i = 0; i < map->num_members; i++) { | |
1804 | /* initialized in add_to_super */ | |
be73972f | 1805 | set_imsm_ord_tbl_ent(map, i, 0); |
c2c087e6 | 1806 | } |
949c47a0 DW |
1807 | mpb->num_raid_devs++; |
1808 | super->dev_tbl[super->current_vol] = dev; | |
c2c087e6 DW |
1809 | |
1810 | return 1; | |
cdddbdbc DW |
1811 | } |
1812 | ||
bf5a934a DW |
1813 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
1814 | unsigned long long size, char *name, | |
1815 | char *homehost, int *uuid) | |
1816 | { | |
1817 | /* This is primarily called by Create when creating a new array. | |
1818 | * We will then get add_to_super called for each component, and then | |
1819 | * write_init_super called to write it out to each device. | |
1820 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
1821 | * array. | |
1822 | * To create on a pre-existing array a different method will be called. | |
1823 | * This one is just for fresh drives. | |
1824 | */ | |
1825 | struct intel_super *super; | |
1826 | struct imsm_super *mpb; | |
1827 | size_t mpb_size; | |
1828 | ||
1829 | if (!info) { | |
1830 | st->sb = NULL; | |
1831 | return 0; | |
1832 | } | |
1833 | if (st->sb) | |
1834 | return init_super_imsm_volume(st, info, size, name, homehost, | |
1835 | uuid); | |
1836 | ||
1837 | super = alloc_super(1); | |
1838 | if (!super) | |
1839 | return 0; | |
1840 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
ef649044 | 1841 | if (posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a DW |
1842 | free(super); |
1843 | return 0; | |
1844 | } | |
ef649044 | 1845 | mpb = super->buf; |
bf5a934a DW |
1846 | memset(mpb, 0, mpb_size); |
1847 | ||
1848 | memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE)); | |
1849 | memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5, | |
1850 | strlen(MPB_VERSION_RAID5)); | |
1851 | mpb->mpb_size = mpb_size; | |
1852 | ||
bf5a934a DW |
1853 | st->sb = super; |
1854 | return 1; | |
1855 | } | |
1856 | ||
0e600426 | 1857 | #ifndef MDASSEMBLE |
bf5a934a DW |
1858 | static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
1859 | int fd, char *devname) | |
1860 | { | |
1861 | struct intel_super *super = st->sb; | |
d23fe947 | 1862 | struct imsm_super *mpb = super->anchor; |
bf5a934a DW |
1863 | struct dl *dl; |
1864 | struct imsm_dev *dev; | |
1865 | struct imsm_map *map; | |
bf5a934a | 1866 | |
949c47a0 | 1867 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 1868 | map = get_imsm_map(dev, 0); |
bf5a934a DW |
1869 | |
1870 | for (dl = super->disks; dl ; dl = dl->next) | |
1871 | if (dl->major == dk->major && | |
1872 | dl->minor == dk->minor) | |
1873 | break; | |
d23fe947 | 1874 | |
bf5a934a DW |
1875 | if (!dl || ! (dk->state & (1<<MD_DISK_SYNC))) |
1876 | return; | |
1877 | ||
d23fe947 DW |
1878 | /* add a pristine spare to the metadata */ |
1879 | if (dl->index < 0) { | |
1880 | dl->index = super->anchor->num_disks; | |
1881 | super->anchor->num_disks++; | |
1882 | } | |
be73972f | 1883 | set_imsm_ord_tbl_ent(map, dk->number, dl->index); |
f2f27e63 | 1884 | dl->disk.status = CONFIGURED_DISK | USABLE_DISK; |
d23fe947 DW |
1885 | |
1886 | /* if we are creating the first raid device update the family number */ | |
1887 | if (super->current_vol == 0) { | |
1888 | __u32 sum; | |
1889 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
1890 | struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index); | |
1891 | ||
1892 | *_dev = *dev; | |
1893 | *_disk = dl->disk; | |
1894 | sum = __gen_imsm_checksum(mpb); | |
1895 | mpb->family_num = __cpu_to_le32(sum); | |
1896 | } | |
bf5a934a DW |
1897 | } |
1898 | ||
c2c087e6 | 1899 | static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
cdddbdbc DW |
1900 | int fd, char *devname) |
1901 | { | |
c2c087e6 | 1902 | struct intel_super *super = st->sb; |
c2c087e6 DW |
1903 | struct dl *dd; |
1904 | unsigned long long size; | |
f2f27e63 | 1905 | __u32 id; |
c2c087e6 DW |
1906 | int rv; |
1907 | struct stat stb; | |
1908 | ||
bf5a934a DW |
1909 | if (super->current_vol >= 0) { |
1910 | add_to_super_imsm_volume(st, dk, fd, devname); | |
1911 | return; | |
1912 | } | |
1913 | ||
c2c087e6 DW |
1914 | fstat(fd, &stb); |
1915 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 1916 | if (!dd) { |
c2c087e6 DW |
1917 | fprintf(stderr, |
1918 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
1919 | abort(); | |
1920 | } | |
1921 | memset(dd, 0, sizeof(*dd)); | |
1922 | dd->major = major(stb.st_rdev); | |
1923 | dd->minor = minor(stb.st_rdev); | |
b9f594fe | 1924 | dd->index = -1; |
c2c087e6 | 1925 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 DW |
1926 | dd->fd = fd; |
1927 | rv = imsm_read_serial(fd, devname, dd->serial); | |
1928 | if (rv) { | |
1929 | fprintf(stderr, | |
0030e8d6 | 1930 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 1931 | free(dd); |
0030e8d6 | 1932 | abort(); |
c2c087e6 DW |
1933 | } |
1934 | ||
c2c087e6 DW |
1935 | get_dev_size(fd, NULL, &size); |
1936 | size /= 512; | |
1f24f035 | 1937 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 1938 | dd->disk.total_blocks = __cpu_to_le32(size); |
f2f27e63 | 1939 | dd->disk.status = USABLE_DISK | SPARE_DISK; |
c2c087e6 | 1940 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 1941 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 1942 | else |
b9f594fe | 1943 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
1944 | |
1945 | if (st->update_tail) { | |
1946 | dd->next = super->add; | |
1947 | super->add = dd; | |
1948 | } else { | |
1949 | dd->next = super->disks; | |
1950 | super->disks = dd; | |
1951 | } | |
cdddbdbc DW |
1952 | } |
1953 | ||
c2c087e6 DW |
1954 | static int store_imsm_mpb(int fd, struct intel_super *super); |
1955 | ||
d23fe947 DW |
1956 | /* spare records have their own family number and do not have any defined raid |
1957 | * devices | |
1958 | */ | |
1959 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
1960 | { | |
1961 | struct imsm_super mpb_save; | |
1962 | struct imsm_super *mpb = super->anchor; | |
1963 | __u32 sum; | |
1964 | struct dl *d; | |
1965 | ||
1966 | mpb_save = *mpb; | |
1967 | mpb->num_raid_devs = 0; | |
1968 | mpb->num_disks = 1; | |
1969 | mpb->mpb_size = sizeof(struct imsm_super); | |
1970 | mpb->generation_num = __cpu_to_le32(1UL); | |
1971 | ||
1972 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 1973 | if (d->index != -1) |
d23fe947 DW |
1974 | continue; |
1975 | ||
1976 | mpb->disk[0] = d->disk; | |
1977 | sum = __gen_imsm_checksum(mpb); | |
1978 | mpb->family_num = __cpu_to_le32(sum); | |
1979 | sum = __gen_imsm_checksum(mpb); | |
1980 | mpb->check_sum = __cpu_to_le32(sum); | |
1981 | ||
1982 | if (store_imsm_mpb(d->fd, super)) { | |
1983 | fprintf(stderr, "%s: failed for device %d:%d %s\n", | |
1984 | __func__, d->major, d->minor, strerror(errno)); | |
1985 | *mpb = mpb_save; | |
e74255d9 | 1986 | return 1; |
d23fe947 DW |
1987 | } |
1988 | if (doclose) { | |
1989 | close(d->fd); | |
1990 | d->fd = -1; | |
1991 | } | |
1992 | } | |
1993 | ||
1994 | *mpb = mpb_save; | |
e74255d9 | 1995 | return 0; |
d23fe947 DW |
1996 | } |
1997 | ||
c2c087e6 | 1998 | static int write_super_imsm(struct intel_super *super, int doclose) |
cdddbdbc | 1999 | { |
949c47a0 | 2000 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
2001 | struct dl *d; |
2002 | __u32 generation; | |
2003 | __u32 sum; | |
d23fe947 | 2004 | int spares = 0; |
949c47a0 | 2005 | int i; |
a48ac0a8 | 2006 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
cdddbdbc | 2007 | |
c2c087e6 DW |
2008 | /* 'generation' is incremented everytime the metadata is written */ |
2009 | generation = __le32_to_cpu(mpb->generation_num); | |
2010 | generation++; | |
2011 | mpb->generation_num = __cpu_to_le32(generation); | |
2012 | ||
1ee1e9fc | 2013 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; |
d23fe947 | 2014 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 2015 | if (d->index == -1) |
d23fe947 | 2016 | spares++; |
1ee1e9fc | 2017 | else |
d23fe947 | 2018 | mpb->disk[d->index] = d->disk; |
d23fe947 | 2019 | } |
47ee5a45 DW |
2020 | for (d = super->missing; d; d = d->next) |
2021 | mpb->disk[d->index] = d->disk; | |
b9f594fe | 2022 | |
949c47a0 DW |
2023 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2024 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
2025 | ||
2026 | imsm_copy_dev(dev, super->dev_tbl[i]); | |
a48ac0a8 | 2027 | mpb_size += sizeof_imsm_dev(dev, 0); |
949c47a0 | 2028 | } |
a48ac0a8 DW |
2029 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
2030 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 2031 | |
c2c087e6 | 2032 | /* recalculate checksum */ |
949c47a0 | 2033 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
2034 | mpb->check_sum = __cpu_to_le32(sum); |
2035 | ||
d23fe947 | 2036 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 2037 | for (d = super->disks; d ; d = d->next) { |
d23fe947 DW |
2038 | if (d->index < 0) |
2039 | continue; | |
8796fdc4 | 2040 | if (store_imsm_mpb(d->fd, super)) |
c2c087e6 DW |
2041 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
2042 | __func__, d->major, d->minor, strerror(errno)); | |
c2c087e6 DW |
2043 | if (doclose) { |
2044 | close(d->fd); | |
2045 | d->fd = -1; | |
2046 | } | |
2047 | } | |
2048 | ||
d23fe947 DW |
2049 | if (spares) |
2050 | return write_super_imsm_spares(super, doclose); | |
2051 | ||
e74255d9 | 2052 | return 0; |
c2c087e6 DW |
2053 | } |
2054 | ||
0e600426 | 2055 | |
43dad3d6 DW |
2056 | static int create_array(struct supertype *st) |
2057 | { | |
2058 | size_t len; | |
2059 | struct imsm_update_create_array *u; | |
2060 | struct intel_super *super = st->sb; | |
2061 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
2062 | ||
2063 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0); | |
2064 | u = malloc(len); | |
2065 | if (!u) { | |
2066 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
2067 | __func__); | |
2068 | return 1; | |
2069 | } | |
2070 | ||
2071 | u->type = update_create_array; | |
2072 | u->dev_idx = super->current_vol; | |
2073 | imsm_copy_dev(&u->dev, dev); | |
2074 | append_metadata_update(st, u, len); | |
2075 | ||
2076 | return 0; | |
2077 | } | |
2078 | ||
7801ac20 | 2079 | static int _add_disk(struct supertype *st) |
43dad3d6 DW |
2080 | { |
2081 | struct intel_super *super = st->sb; | |
2082 | size_t len; | |
2083 | struct imsm_update_add_disk *u; | |
2084 | ||
2085 | if (!super->add) | |
2086 | return 0; | |
2087 | ||
2088 | len = sizeof(*u); | |
2089 | u = malloc(len); | |
2090 | if (!u) { | |
2091 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
2092 | __func__); | |
2093 | return 1; | |
2094 | } | |
2095 | ||
2096 | u->type = update_add_disk; | |
2097 | append_metadata_update(st, u, len); | |
2098 | ||
2099 | return 0; | |
2100 | } | |
2101 | ||
c2c087e6 DW |
2102 | static int write_init_super_imsm(struct supertype *st) |
2103 | { | |
8273f55e | 2104 | if (st->update_tail) { |
43dad3d6 DW |
2105 | /* queue the recently created array / added disk |
2106 | * as a metadata update */ | |
8273f55e | 2107 | struct intel_super *super = st->sb; |
8273f55e | 2108 | struct dl *d; |
43dad3d6 | 2109 | int rv; |
8273f55e | 2110 | |
43dad3d6 DW |
2111 | /* determine if we are creating a volume or adding a disk */ |
2112 | if (super->current_vol < 0) { | |
2113 | /* in the add disk case we are running in mdmon | |
2114 | * context, so don't close fd's | |
2115 | */ | |
7801ac20 | 2116 | return _add_disk(st); |
43dad3d6 DW |
2117 | } else |
2118 | rv = create_array(st); | |
8273f55e DW |
2119 | |
2120 | for (d = super->disks; d ; d = d->next) { | |
2121 | close(d->fd); | |
2122 | d->fd = -1; | |
2123 | } | |
2124 | ||
43dad3d6 | 2125 | return rv; |
8273f55e DW |
2126 | } else |
2127 | return write_super_imsm(st->sb, 1); | |
cdddbdbc | 2128 | } |
0e600426 | 2129 | #endif |
cdddbdbc DW |
2130 | |
2131 | static int store_zero_imsm(struct supertype *st, int fd) | |
2132 | { | |
551c80c1 | 2133 | unsigned long long dsize; |
6416d527 | 2134 | void *buf; |
551c80c1 DW |
2135 | |
2136 | get_dev_size(fd, NULL, &dsize); | |
2137 | ||
2138 | /* first block is stored on second to last sector of the disk */ | |
2139 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
2140 | return 1; | |
2141 | ||
ad97895e DW |
2142 | if (posix_memalign(&buf, 512, 512) != 0) |
2143 | return 1; | |
2144 | ||
eb7ea463 DW |
2145 | memset(buf, 0, 512); |
2146 | if (write(fd, buf, 512) != 512) | |
551c80c1 | 2147 | return 1; |
cdddbdbc DW |
2148 | return 0; |
2149 | } | |
2150 | ||
0e600426 N |
2151 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
2152 | { | |
2153 | return __le32_to_cpu(mpb->bbm_log_size); | |
2154 | } | |
2155 | ||
2156 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
2157 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
2158 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 2159 | unsigned long long size, char *dev, |
2c514b71 NB |
2160 | unsigned long long *freesize, |
2161 | int verbose) | |
cdddbdbc | 2162 | { |
c2c087e6 DW |
2163 | int fd; |
2164 | unsigned long long ldsize; | |
cdddbdbc | 2165 | |
c2c087e6 DW |
2166 | if (level != LEVEL_CONTAINER) |
2167 | return 0; | |
2168 | if (!dev) | |
2169 | return 1; | |
2170 | ||
2171 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
2172 | if (fd < 0) { | |
2c514b71 NB |
2173 | if (verbose) |
2174 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
2175 | dev, strerror(errno)); | |
c2c087e6 DW |
2176 | return 0; |
2177 | } | |
2178 | if (!get_dev_size(fd, dev, &ldsize)) { | |
2179 | close(fd); | |
2180 | return 0; | |
2181 | } | |
2182 | close(fd); | |
2183 | ||
2184 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
2185 | ||
2186 | return 1; | |
cdddbdbc DW |
2187 | } |
2188 | ||
c2c087e6 DW |
2189 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
2190 | * FIX ME add ahci details | |
2191 | */ | |
8b353278 DW |
2192 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
2193 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 2194 | unsigned long long size, char *dev, |
2c514b71 NB |
2195 | unsigned long long *freesize, |
2196 | int verbose) | |
cdddbdbc | 2197 | { |
c2c087e6 DW |
2198 | struct stat stb; |
2199 | struct intel_super *super = st->sb; | |
2200 | struct dl *dl; | |
2201 | unsigned long long pos = 0; | |
2202 | unsigned long long maxsize; | |
2203 | struct extent *e; | |
2204 | int i; | |
cdddbdbc | 2205 | |
c2c087e6 DW |
2206 | if (level == LEVEL_CONTAINER) |
2207 | return 0; | |
2208 | ||
2209 | if (level == 1 && raiddisks > 2) { | |
2c514b71 NB |
2210 | if (verbose) |
2211 | fprintf(stderr, Name ": imsm does not support more " | |
2212 | "than 2 in a raid1 configuration\n"); | |
c2c087e6 DW |
2213 | return 0; |
2214 | } | |
2215 | ||
2216 | /* We must have the container info already read in. */ | |
2217 | if (!super) | |
2218 | return 0; | |
2219 | ||
2220 | if (!dev) { | |
2221 | /* General test: make sure there is space for | |
2da8544a DW |
2222 | * 'raiddisks' device extents of size 'size' at a given |
2223 | * offset | |
c2c087e6 DW |
2224 | */ |
2225 | unsigned long long minsize = size*2 /* convert to blocks */; | |
2da8544a | 2226 | unsigned long long start_offset = ~0ULL; |
c2c087e6 DW |
2227 | int dcnt = 0; |
2228 | if (minsize == 0) | |
2229 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
2230 | for (dl = super->disks; dl ; dl = dl->next) { | |
2231 | int found = 0; | |
2232 | ||
bf5a934a | 2233 | pos = 0; |
c2c087e6 DW |
2234 | i = 0; |
2235 | e = get_extents(super, dl); | |
2236 | if (!e) continue; | |
2237 | do { | |
2238 | unsigned long long esize; | |
2239 | esize = e[i].start - pos; | |
2240 | if (esize >= minsize) | |
2241 | found = 1; | |
2da8544a DW |
2242 | if (found && start_offset == ~0ULL) { |
2243 | start_offset = pos; | |
2244 | break; | |
2245 | } else if (found && pos != start_offset) { | |
2246 | found = 0; | |
2247 | break; | |
2248 | } | |
c2c087e6 DW |
2249 | pos = e[i].start + e[i].size; |
2250 | i++; | |
2251 | } while (e[i-1].size); | |
2252 | if (found) | |
2253 | dcnt++; | |
2254 | free(e); | |
2255 | } | |
2256 | if (dcnt < raiddisks) { | |
2c514b71 NB |
2257 | if (verbose) |
2258 | fprintf(stderr, Name ": imsm: Not enough " | |
2259 | "devices with space for this array " | |
2260 | "(%d < %d)\n", | |
2261 | dcnt, raiddisks); | |
c2c087e6 DW |
2262 | return 0; |
2263 | } | |
2264 | return 1; | |
2265 | } | |
2266 | /* This device must be a member of the set */ | |
2267 | if (stat(dev, &stb) < 0) | |
2268 | return 0; | |
2269 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
2270 | return 0; | |
2271 | for (dl = super->disks ; dl ; dl = dl->next) { | |
2272 | if (dl->major == major(stb.st_rdev) && | |
2273 | dl->minor == minor(stb.st_rdev)) | |
2274 | break; | |
2275 | } | |
2276 | if (!dl) { | |
2c514b71 NB |
2277 | if (verbose) |
2278 | fprintf(stderr, Name ": %s is not in the " | |
2279 | "same imsm set\n", dev); | |
c2c087e6 DW |
2280 | return 0; |
2281 | } | |
2282 | e = get_extents(super, dl); | |
2283 | maxsize = 0; | |
2284 | i = 0; | |
2285 | if (e) do { | |
2286 | unsigned long long esize; | |
2287 | esize = e[i].start - pos; | |
2288 | if (esize >= maxsize) | |
2289 | maxsize = esize; | |
2290 | pos = e[i].start + e[i].size; | |
2291 | i++; | |
2292 | } while (e[i-1].size); | |
2293 | *freesize = maxsize; | |
2294 | ||
2295 | return 1; | |
cdddbdbc DW |
2296 | } |
2297 | ||
bf5a934a DW |
2298 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
2299 | int raiddisks, int chunk, unsigned long long size, | |
2300 | char *dev, unsigned long long *freesize, | |
2301 | int verbose) | |
2302 | { | |
2303 | int fd, cfd; | |
2304 | struct mdinfo *sra; | |
2305 | ||
2306 | /* if given unused devices create a container | |
2307 | * if given given devices in a container create a member volume | |
2308 | */ | |
2309 | if (level == LEVEL_CONTAINER) { | |
2310 | /* Must be a fresh device to add to a container */ | |
2311 | return validate_geometry_imsm_container(st, level, layout, | |
2312 | raiddisks, chunk, size, | |
2313 | dev, freesize, | |
2314 | verbose); | |
2315 | } | |
2316 | ||
2317 | if (st->sb) { | |
2318 | /* creating in a given container */ | |
2319 | return validate_geometry_imsm_volume(st, level, layout, | |
2320 | raiddisks, chunk, size, | |
2321 | dev, freesize, verbose); | |
2322 | } | |
2323 | ||
2324 | /* limit creation to the following levels */ | |
2325 | if (!dev) | |
2326 | switch (level) { | |
2327 | case 0: | |
2328 | case 1: | |
2329 | case 10: | |
2330 | case 5: | |
2331 | break; | |
2332 | default: | |
2333 | return 1; | |
2334 | } | |
2335 | ||
2336 | /* This device needs to be a device in an 'imsm' container */ | |
2337 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
2338 | if (fd >= 0) { | |
2339 | if (verbose) | |
2340 | fprintf(stderr, | |
2341 | Name ": Cannot create this array on device %s\n", | |
2342 | dev); | |
2343 | close(fd); | |
2344 | return 0; | |
2345 | } | |
2346 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
2347 | if (verbose) | |
2348 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
2349 | dev, strerror(errno)); | |
2350 | return 0; | |
2351 | } | |
2352 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
2353 | cfd = open_container(fd); | |
2354 | if (cfd < 0) { | |
2355 | close(fd); | |
2356 | if (verbose) | |
2357 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
2358 | dev); | |
2359 | return 0; | |
2360 | } | |
2361 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
2362 | close(fd); | |
2363 | if (sra && sra->array.major_version == -1 && | |
2364 | strcmp(sra->text_version, "imsm") == 0) { | |
2365 | /* This is a member of a imsm container. Load the container | |
2366 | * and try to create a volume | |
2367 | */ | |
2368 | struct intel_super *super; | |
2369 | ||
2370 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) { | |
2371 | st->sb = super; | |
2372 | st->container_dev = fd2devnum(cfd); | |
2373 | close(cfd); | |
2374 | return validate_geometry_imsm_volume(st, level, layout, | |
2375 | raiddisks, chunk, | |
2376 | size, dev, | |
2377 | freesize, verbose); | |
2378 | } | |
2379 | close(cfd); | |
2380 | } else /* may belong to another container */ | |
2381 | return 0; | |
2382 | ||
2383 | return 1; | |
2384 | } | |
0e600426 | 2385 | #endif /* MDASSEMBLE */ |
bf5a934a | 2386 | |
cdddbdbc DW |
2387 | static struct mdinfo *container_content_imsm(struct supertype *st) |
2388 | { | |
4f5bc454 DW |
2389 | /* Given a container loaded by load_super_imsm_all, |
2390 | * extract information about all the arrays into | |
2391 | * an mdinfo tree. | |
2392 | * | |
2393 | * For each imsm_dev create an mdinfo, fill it in, | |
2394 | * then look for matching devices in super->disks | |
2395 | * and create appropriate device mdinfo. | |
2396 | */ | |
2397 | struct intel_super *super = st->sb; | |
949c47a0 | 2398 | struct imsm_super *mpb = super->anchor; |
4f5bc454 DW |
2399 | struct mdinfo *rest = NULL; |
2400 | int i; | |
cdddbdbc | 2401 | |
604b746f JD |
2402 | /* do not assemble arrays that might have bad blocks */ |
2403 | if (imsm_bbm_log_size(super->anchor)) { | |
2404 | fprintf(stderr, Name ": BBM log found in metadata. " | |
2405 | "Cannot activate array(s).\n"); | |
2406 | return NULL; | |
2407 | } | |
2408 | ||
4f5bc454 | 2409 | for (i = 0; i < mpb->num_raid_devs; i++) { |
949c47a0 | 2410 | struct imsm_dev *dev = get_imsm_dev(super, i); |
a965f303 | 2411 | struct imsm_map *map = get_imsm_map(dev, 0); |
4f5bc454 | 2412 | struct mdinfo *this; |
4f5bc454 DW |
2413 | int slot; |
2414 | ||
2415 | this = malloc(sizeof(*this)); | |
2416 | memset(this, 0, sizeof(*this)); | |
2417 | this->next = rest; | |
4f5bc454 | 2418 | |
301406c9 DW |
2419 | super->current_vol = i; |
2420 | getinfo_super_imsm_volume(st, this); | |
4f5bc454 | 2421 | for (slot = 0 ; slot < map->num_members; slot++) { |
4f5bc454 DW |
2422 | struct mdinfo *info_d; |
2423 | struct dl *d; | |
2424 | int idx; | |
9a1608e5 | 2425 | int skip; |
4f5bc454 | 2426 | __u32 s; |
7eef0453 | 2427 | __u32 ord; |
4f5bc454 | 2428 | |
9a1608e5 | 2429 | skip = 0; |
ff077194 | 2430 | idx = get_imsm_disk_idx(dev, slot); |
7eef0453 | 2431 | ord = get_imsm_ord_tbl_ent(dev, slot); |
4f5bc454 DW |
2432 | for (d = super->disks; d ; d = d->next) |
2433 | if (d->index == idx) | |
2434 | break; | |
2435 | ||
2436 | if (d == NULL) | |
9a1608e5 DW |
2437 | skip = 1; |
2438 | ||
f2f27e63 | 2439 | s = d ? d->disk.status : 0; |
9a1608e5 DW |
2440 | if (s & FAILED_DISK) |
2441 | skip = 1; | |
2442 | if (!(s & USABLE_DISK)) | |
2443 | skip = 1; | |
7eef0453 DW |
2444 | if (ord & IMSM_ORD_REBUILD) |
2445 | skip = 1; | |
9a1608e5 DW |
2446 | |
2447 | /* | |
2448 | * if we skip some disks the array will be assmebled degraded; | |
2449 | * reset resync start to avoid a dirty-degraded situation | |
2450 | * | |
2451 | * FIXME handle dirty degraded | |
2452 | */ | |
2453 | if (skip && !dev->vol.dirty) | |
2454 | this->resync_start = ~0ULL; | |
2455 | if (skip) | |
2456 | continue; | |
4f5bc454 DW |
2457 | |
2458 | info_d = malloc(sizeof(*info_d)); | |
9a1608e5 DW |
2459 | if (!info_d) { |
2460 | fprintf(stderr, Name ": failed to allocate disk" | |
2461 | " for volume %s\n", (char *) dev->volume); | |
2462 | free(this); | |
2463 | this = rest; | |
2464 | break; | |
2465 | } | |
4f5bc454 DW |
2466 | memset(info_d, 0, sizeof(*info_d)); |
2467 | info_d->next = this->devs; | |
2468 | this->devs = info_d; | |
2469 | ||
4f5bc454 DW |
2470 | info_d->disk.number = d->index; |
2471 | info_d->disk.major = d->major; | |
2472 | info_d->disk.minor = d->minor; | |
2473 | info_d->disk.raid_disk = slot; | |
4f5bc454 DW |
2474 | |
2475 | this->array.working_disks++; | |
2476 | ||
2477 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
2478 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
2479 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
2480 | if (d->devname) | |
2481 | strcpy(info_d->name, d->devname); | |
2482 | } | |
9a1608e5 | 2483 | rest = this; |
4f5bc454 DW |
2484 | } |
2485 | ||
2486 | return rest; | |
cdddbdbc DW |
2487 | } |
2488 | ||
845dea95 | 2489 | |
0e600426 | 2490 | #ifndef MDASSEMBLE |
cba0191b NB |
2491 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
2492 | char *inst) | |
845dea95 | 2493 | { |
0372d5a2 | 2494 | struct intel_super *super = c->sb; |
949c47a0 | 2495 | struct imsm_super *mpb = super->anchor; |
0372d5a2 | 2496 | |
949c47a0 | 2497 | if (atoi(inst) >= mpb->num_raid_devs) { |
0372d5a2 DW |
2498 | fprintf(stderr, "%s: subarry index %d, out of range\n", |
2499 | __func__, atoi(inst)); | |
2500 | return -ENODEV; | |
2501 | } | |
2502 | ||
4e6e574a | 2503 | dprintf("imsm: open_new %s\n", inst); |
cba0191b | 2504 | a->info.container_member = atoi(inst); |
845dea95 NB |
2505 | return 0; |
2506 | } | |
2507 | ||
fb49eef2 | 2508 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed) |
c2a1e7da | 2509 | { |
a965f303 | 2510 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
2511 | |
2512 | if (!failed) | |
3393c6af DW |
2513 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
2514 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
2515 | |
2516 | switch (get_imsm_raid_level(map)) { | |
2517 | case 0: | |
2518 | return IMSM_T_STATE_FAILED; | |
2519 | break; | |
2520 | case 1: | |
2521 | if (failed < map->num_members) | |
2522 | return IMSM_T_STATE_DEGRADED; | |
2523 | else | |
2524 | return IMSM_T_STATE_FAILED; | |
2525 | break; | |
2526 | case 10: | |
2527 | { | |
2528 | /** | |
c92a2527 DW |
2529 | * check to see if any mirrors have failed, otherwise we |
2530 | * are degraded. Even numbered slots are mirrored on | |
2531 | * slot+1 | |
c2a1e7da | 2532 | */ |
c2a1e7da | 2533 | int i; |
d9b420a5 N |
2534 | /* gcc -Os complains that this is unused */ |
2535 | int insync = insync; | |
c2a1e7da DW |
2536 | |
2537 | for (i = 0; i < map->num_members; i++) { | |
c92a2527 DW |
2538 | __u32 ord = get_imsm_ord_tbl_ent(dev, i); |
2539 | int idx = ord_to_idx(ord); | |
2540 | struct imsm_disk *disk; | |
c2a1e7da | 2541 | |
c92a2527 DW |
2542 | /* reset the potential in-sync count on even-numbered |
2543 | * slots. num_copies is always 2 for imsm raid10 | |
2544 | */ | |
2545 | if ((i & 1) == 0) | |
2546 | insync = 2; | |
c2a1e7da | 2547 | |
c92a2527 | 2548 | disk = get_imsm_disk(super, idx); |
f2f27e63 | 2549 | if (!disk || disk->status & FAILED_DISK || |
c92a2527 DW |
2550 | ord & IMSM_ORD_REBUILD) |
2551 | insync--; | |
c2a1e7da | 2552 | |
c92a2527 DW |
2553 | /* no in-sync disks left in this mirror the |
2554 | * array has failed | |
2555 | */ | |
2556 | if (insync == 0) | |
2557 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
2558 | } |
2559 | ||
2560 | return IMSM_T_STATE_DEGRADED; | |
2561 | } | |
2562 | case 5: | |
2563 | if (failed < 2) | |
2564 | return IMSM_T_STATE_DEGRADED; | |
2565 | else | |
2566 | return IMSM_T_STATE_FAILED; | |
2567 | break; | |
2568 | default: | |
2569 | break; | |
2570 | } | |
2571 | ||
2572 | return map->map_state; | |
2573 | } | |
2574 | ||
ff077194 | 2575 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev) |
c2a1e7da DW |
2576 | { |
2577 | int i; | |
2578 | int failed = 0; | |
2579 | struct imsm_disk *disk; | |
ff077194 | 2580 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
2581 | |
2582 | for (i = 0; i < map->num_members; i++) { | |
b10b37b8 DW |
2583 | __u32 ord = get_imsm_ord_tbl_ent(dev, i); |
2584 | int idx = ord_to_idx(ord); | |
c2a1e7da | 2585 | |
949c47a0 | 2586 | disk = get_imsm_disk(super, idx); |
f2f27e63 | 2587 | if (!disk || disk->status & FAILED_DISK || |
b10b37b8 | 2588 | ord & IMSM_ORD_REBUILD) |
fcb84475 | 2589 | failed++; |
c2a1e7da DW |
2590 | } |
2591 | ||
2592 | return failed; | |
845dea95 NB |
2593 | } |
2594 | ||
0c046afd DW |
2595 | static int is_resyncing(struct imsm_dev *dev) |
2596 | { | |
2597 | struct imsm_map *migr_map; | |
2598 | ||
2599 | if (!dev->vol.migr_state) | |
2600 | return 0; | |
2601 | ||
e3bba0e0 | 2602 | if (dev->vol.migr_type == MIGR_INIT) |
0c046afd DW |
2603 | return 1; |
2604 | ||
2605 | migr_map = get_imsm_map(dev, 1); | |
2606 | ||
2607 | if (migr_map->map_state == IMSM_T_STATE_NORMAL) | |
2608 | return 1; | |
2609 | else | |
2610 | return 0; | |
2611 | } | |
2612 | ||
2613 | static int is_rebuilding(struct imsm_dev *dev) | |
2614 | { | |
2615 | struct imsm_map *migr_map; | |
2616 | ||
2617 | if (!dev->vol.migr_state) | |
2618 | return 0; | |
2619 | ||
e3bba0e0 | 2620 | if (dev->vol.migr_type != MIGR_REBUILD) |
0c046afd DW |
2621 | return 0; |
2622 | ||
2623 | migr_map = get_imsm_map(dev, 1); | |
2624 | ||
2625 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
2626 | return 1; | |
2627 | else | |
2628 | return 0; | |
2629 | } | |
2630 | ||
47ee5a45 DW |
2631 | static void mark_failure(struct imsm_disk *disk) |
2632 | { | |
f2f27e63 | 2633 | if (disk->status & FAILED_DISK) |
47ee5a45 | 2634 | return; |
f2f27e63 | 2635 | disk->status |= FAILED_DISK; |
47ee5a45 DW |
2636 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
2637 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
2638 | } | |
2639 | ||
0c046afd DW |
2640 | /* Handle dirty -> clean transititions and resync. Degraded and rebuild |
2641 | * states are handled in imsm_set_disk() with one exception, when a | |
2642 | * resync is stopped due to a new failure this routine will set the | |
2643 | * 'degraded' state for the array. | |
2644 | */ | |
01f157d7 | 2645 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
2646 | { |
2647 | int inst = a->info.container_member; | |
2648 | struct intel_super *super = a->container->sb; | |
949c47a0 | 2649 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2650 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd DW |
2651 | int failed = imsm_count_failed(super, dev); |
2652 | __u8 map_state = imsm_check_degraded(super, dev, failed); | |
a862209d | 2653 | |
47ee5a45 DW |
2654 | /* before we activate this array handle any missing disks */ |
2655 | if (consistent == 2 && super->missing) { | |
2656 | struct dl *dl; | |
2657 | ||
2658 | dprintf("imsm: mark missing\n"); | |
2659 | end_migration(dev, map_state); | |
2660 | for (dl = super->missing; dl; dl = dl->next) | |
2661 | mark_failure(&dl->disk); | |
2662 | super->updates_pending++; | |
2663 | } | |
2664 | ||
0c046afd | 2665 | if (consistent == 2 && |
593add1b | 2666 | (!is_resync_complete(a) || |
0c046afd DW |
2667 | map_state != IMSM_T_STATE_NORMAL || |
2668 | dev->vol.migr_state)) | |
01f157d7 | 2669 | consistent = 0; |
272906ef | 2670 | |
593add1b | 2671 | if (is_resync_complete(a)) { |
0c046afd DW |
2672 | /* complete intialization / resync, |
2673 | * recovery is completed in ->set_disk | |
2674 | */ | |
2675 | if (is_resyncing(dev)) { | |
2676 | dprintf("imsm: mark resync done\n"); | |
f8f603f1 | 2677 | end_migration(dev, map_state); |
115c3803 | 2678 | super->updates_pending++; |
115c3803 | 2679 | } |
0c046afd DW |
2680 | } else if (!is_resyncing(dev) && !failed) { |
2681 | /* mark the start of the init process if nothing is failed */ | |
2682 | dprintf("imsm: mark resync start (%llu)\n", a->resync_start); | |
e3bba0e0 DW |
2683 | if (map->map_state == IMSM_T_STATE_NORMAL) |
2684 | migrate(dev, IMSM_T_STATE_NORMAL, MIGR_REBUILD); | |
2685 | else | |
2686 | migrate(dev, IMSM_T_STATE_NORMAL, MIGR_INIT); | |
3393c6af | 2687 | super->updates_pending++; |
115c3803 | 2688 | } |
a862209d | 2689 | |
f8f603f1 DW |
2690 | /* check if we can update the migration checkpoint */ |
2691 | if (dev->vol.migr_state && | |
2692 | __le32_to_cpu(dev->vol.curr_migr_unit) != a->resync_start) { | |
2693 | dprintf("imsm: checkpoint migration (%llu)\n", a->resync_start); | |
2694 | dev->vol.curr_migr_unit = __cpu_to_le32(a->resync_start); | |
2695 | super->updates_pending++; | |
2696 | } | |
2697 | ||
3393c6af | 2698 | /* mark dirty / clean */ |
0c046afd | 2699 | if (dev->vol.dirty != !consistent) { |
3393c6af | 2700 | dprintf("imsm: mark '%s' (%llu)\n", |
0c046afd DW |
2701 | consistent ? "clean" : "dirty", a->resync_start); |
2702 | if (consistent) | |
2703 | dev->vol.dirty = 0; | |
2704 | else | |
2705 | dev->vol.dirty = 1; | |
a862209d DW |
2706 | super->updates_pending++; |
2707 | } | |
01f157d7 | 2708 | return consistent; |
a862209d DW |
2709 | } |
2710 | ||
8d45d196 | 2711 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 2712 | { |
8d45d196 DW |
2713 | int inst = a->info.container_member; |
2714 | struct intel_super *super = a->container->sb; | |
949c47a0 | 2715 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2716 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 | 2717 | struct imsm_disk *disk; |
0c046afd | 2718 | int failed; |
b10b37b8 | 2719 | __u32 ord; |
0c046afd | 2720 | __u8 map_state; |
8d45d196 DW |
2721 | |
2722 | if (n > map->num_members) | |
2723 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
2724 | n, map->num_members - 1); | |
2725 | ||
2726 | if (n < 0) | |
2727 | return; | |
2728 | ||
4e6e574a | 2729 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 2730 | |
b10b37b8 DW |
2731 | ord = get_imsm_ord_tbl_ent(dev, n); |
2732 | disk = get_imsm_disk(super, ord_to_idx(ord)); | |
8d45d196 | 2733 | |
5802a811 | 2734 | /* check for new failures */ |
f2f27e63 | 2735 | if ((state & DS_FAULTY) && !(disk->status & FAILED_DISK)) { |
47ee5a45 | 2736 | mark_failure(disk); |
5802a811 | 2737 | super->updates_pending++; |
8d45d196 | 2738 | } |
47ee5a45 | 2739 | |
19859edc | 2740 | /* check if in_sync */ |
b10b37b8 DW |
2741 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD) { |
2742 | struct imsm_map *migr_map = get_imsm_map(dev, 1); | |
2743 | ||
2744 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
2745 | super->updates_pending++; |
2746 | } | |
8d45d196 | 2747 | |
0c046afd DW |
2748 | failed = imsm_count_failed(super, dev); |
2749 | map_state = imsm_check_degraded(super, dev, failed); | |
5802a811 | 2750 | |
0c046afd DW |
2751 | /* check if recovery complete, newly degraded, or failed */ |
2752 | if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) { | |
f8f603f1 | 2753 | end_migration(dev, map_state); |
0c046afd DW |
2754 | super->updates_pending++; |
2755 | } else if (map_state == IMSM_T_STATE_DEGRADED && | |
2756 | map->map_state != map_state && | |
2757 | !dev->vol.migr_state) { | |
2758 | dprintf("imsm: mark degraded\n"); | |
2759 | map->map_state = map_state; | |
2760 | super->updates_pending++; | |
2761 | } else if (map_state == IMSM_T_STATE_FAILED && | |
2762 | map->map_state != map_state) { | |
2763 | dprintf("imsm: mark failed\n"); | |
f8f603f1 | 2764 | end_migration(dev, map_state); |
0c046afd | 2765 | super->updates_pending++; |
5802a811 | 2766 | } |
845dea95 NB |
2767 | } |
2768 | ||
c2a1e7da DW |
2769 | static int store_imsm_mpb(int fd, struct intel_super *super) |
2770 | { | |
949c47a0 | 2771 | struct imsm_super *mpb = super->anchor; |
c2a1e7da DW |
2772 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
2773 | unsigned long long dsize; | |
2774 | unsigned long long sectors; | |
2775 | ||
2776 | get_dev_size(fd, NULL, &dsize); | |
2777 | ||
272f648f DW |
2778 | if (mpb_size > 512) { |
2779 | /* -1 to account for anchor */ | |
2780 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 2781 | |
272f648f DW |
2782 | /* write the extended mpb to the sectors preceeding the anchor */ |
2783 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
2784 | return 1; | |
c2a1e7da | 2785 | |
99e29264 | 2786 | if (write(fd, super->buf + 512, 512 * sectors) != 512 * sectors) |
272f648f DW |
2787 | return 1; |
2788 | } | |
c2a1e7da | 2789 | |
272f648f DW |
2790 | /* first block is stored on second to last sector of the disk */ |
2791 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
2792 | return 1; |
2793 | ||
272f648f | 2794 | if (write(fd, super->buf, 512) != 512) |
c2a1e7da DW |
2795 | return 1; |
2796 | ||
c2a1e7da DW |
2797 | return 0; |
2798 | } | |
2799 | ||
2e735d19 | 2800 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 2801 | { |
2e735d19 | 2802 | struct intel_super *super = container->sb; |
c2a1e7da DW |
2803 | |
2804 | if (!super->updates_pending) | |
2805 | return; | |
2806 | ||
c2c087e6 | 2807 | write_super_imsm(super, 0); |
c2a1e7da DW |
2808 | |
2809 | super->updates_pending = 0; | |
845dea95 NB |
2810 | } |
2811 | ||
272906ef DW |
2812 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
2813 | { | |
2814 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
ff077194 | 2815 | int i = get_imsm_disk_idx(dev, idx); |
272906ef DW |
2816 | struct dl *dl; |
2817 | ||
2818 | for (dl = super->disks; dl; dl = dl->next) | |
2819 | if (dl->index == i) | |
2820 | break; | |
2821 | ||
f2f27e63 | 2822 | if (dl && dl->disk.status & FAILED_DISK) |
272906ef DW |
2823 | dl = NULL; |
2824 | ||
2825 | if (dl) | |
2826 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
2827 | ||
2828 | return dl; | |
2829 | } | |
2830 | ||
e553d2a4 | 2831 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, struct active_array *a) |
272906ef DW |
2832 | { |
2833 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
e553d2a4 | 2834 | int idx = get_imsm_disk_idx(dev, slot); |
272906ef DW |
2835 | struct imsm_map *map = get_imsm_map(dev, 0); |
2836 | unsigned long long esize; | |
2837 | unsigned long long pos; | |
2838 | struct mdinfo *d; | |
2839 | struct extent *ex; | |
2840 | int j; | |
2841 | int found; | |
2842 | __u32 array_start; | |
2843 | struct dl *dl; | |
2844 | ||
2845 | for (dl = super->disks; dl; dl = dl->next) { | |
2846 | /* If in this array, skip */ | |
2847 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
2848 | if (d->state_fd >= 0 && |
2849 | d->disk.major == dl->major && | |
272906ef DW |
2850 | d->disk.minor == dl->minor) { |
2851 | dprintf("%x:%x already in array\n", dl->major, dl->minor); | |
2852 | break; | |
2853 | } | |
2854 | if (d) | |
2855 | continue; | |
2856 | ||
e553d2a4 | 2857 | /* skip in use or failed drives */ |
f2f27e63 | 2858 | if (dl->disk.status & FAILED_DISK || idx == dl->index) { |
9a1608e5 DW |
2859 | dprintf("%x:%x status ( %s%s)\n", |
2860 | dl->major, dl->minor, | |
f2f27e63 | 2861 | dl->disk.status & FAILED_DISK ? "failed " : "", |
e553d2a4 | 2862 | idx == dl->index ? "in use " : ""); |
9a1608e5 DW |
2863 | continue; |
2864 | } | |
2865 | ||
272906ef DW |
2866 | /* Does this unused device have the requisite free space? |
2867 | * We need a->info.component_size sectors | |
2868 | */ | |
2869 | ex = get_extents(super, dl); | |
2870 | if (!ex) { | |
2871 | dprintf("cannot get extents\n"); | |
2872 | continue; | |
2873 | } | |
2874 | found = 0; | |
2875 | j = 0; | |
2876 | pos = 0; | |
2877 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
2878 | ||
2879 | do { | |
2880 | /* check that we can start at pba_of_lba0 with | |
2881 | * a->info.component_size of space | |
2882 | */ | |
2883 | esize = ex[j].start - pos; | |
2884 | if (array_start >= pos && | |
2885 | array_start + a->info.component_size < ex[j].start) { | |
2886 | found = 1; | |
2887 | break; | |
2888 | } | |
2889 | pos = ex[j].start + ex[j].size; | |
2890 | j++; | |
2891 | ||
2892 | } while (ex[j-1].size); | |
2893 | ||
2894 | free(ex); | |
2895 | if (!found) { | |
2896 | dprintf("%x:%x does not have %llu at %d\n", | |
2897 | dl->major, dl->minor, | |
2898 | a->info.component_size, | |
2899 | __le32_to_cpu(map->pba_of_lba0)); | |
2900 | /* No room */ | |
2901 | continue; | |
2902 | } else | |
2903 | break; | |
2904 | } | |
2905 | ||
2906 | return dl; | |
2907 | } | |
2908 | ||
88758e9d DW |
2909 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
2910 | struct metadata_update **updates) | |
2911 | { | |
2912 | /** | |
d23fe947 DW |
2913 | * Find a device with unused free space and use it to replace a |
2914 | * failed/vacant region in an array. We replace failed regions one a | |
2915 | * array at a time. The result is that a new spare disk will be added | |
2916 | * to the first failed array and after the monitor has finished | |
2917 | * propagating failures the remainder will be consumed. | |
88758e9d | 2918 | * |
d23fe947 DW |
2919 | * FIXME add a capability for mdmon to request spares from another |
2920 | * container. | |
88758e9d DW |
2921 | */ |
2922 | ||
2923 | struct intel_super *super = a->container->sb; | |
88758e9d | 2924 | int inst = a->info.container_member; |
949c47a0 | 2925 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2926 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
2927 | int failed = a->info.array.raid_disks; |
2928 | struct mdinfo *rv = NULL; | |
2929 | struct mdinfo *d; | |
2930 | struct mdinfo *di; | |
2931 | struct metadata_update *mu; | |
2932 | struct dl *dl; | |
2933 | struct imsm_update_activate_spare *u; | |
2934 | int num_spares = 0; | |
2935 | int i; | |
2936 | ||
2937 | for (d = a->info.devs ; d ; d = d->next) { | |
2938 | if ((d->curr_state & DS_FAULTY) && | |
2939 | d->state_fd >= 0) | |
2940 | /* wait for Removal to happen */ | |
2941 | return NULL; | |
2942 | if (d->state_fd >= 0) | |
2943 | failed--; | |
2944 | } | |
2945 | ||
2946 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
2947 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
fb49eef2 | 2948 | if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
2949 | return NULL; |
2950 | ||
2951 | /* For each slot, if it is not working, find a spare */ | |
88758e9d DW |
2952 | for (i = 0; i < a->info.array.raid_disks; i++) { |
2953 | for (d = a->info.devs ; d ; d = d->next) | |
2954 | if (d->disk.raid_disk == i) | |
2955 | break; | |
2956 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
2957 | if (d && (d->state_fd >= 0)) | |
2958 | continue; | |
2959 | ||
272906ef DW |
2960 | /* |
2961 | * OK, this device needs recovery. Try to re-add the previous | |
2962 | * occupant of this slot, if this fails add a new spare | |
2963 | */ | |
2964 | dl = imsm_readd(super, i, a); | |
2965 | if (!dl) | |
2966 | dl = imsm_add_spare(super, i, a); | |
2967 | if (!dl) | |
2968 | continue; | |
2969 | ||
2970 | /* found a usable disk with enough space */ | |
2971 | di = malloc(sizeof(*di)); | |
79244939 DW |
2972 | if (!di) |
2973 | continue; | |
272906ef DW |
2974 | memset(di, 0, sizeof(*di)); |
2975 | ||
2976 | /* dl->index will be -1 in the case we are activating a | |
2977 | * pristine spare. imsm_process_update() will create a | |
2978 | * new index in this case. Once a disk is found to be | |
2979 | * failed in all member arrays it is kicked from the | |
2980 | * metadata | |
2981 | */ | |
2982 | di->disk.number = dl->index; | |
d23fe947 | 2983 | |
272906ef DW |
2984 | /* (ab)use di->devs to store a pointer to the device |
2985 | * we chose | |
2986 | */ | |
2987 | di->devs = (struct mdinfo *) dl; | |
2988 | ||
2989 | di->disk.raid_disk = i; | |
2990 | di->disk.major = dl->major; | |
2991 | di->disk.minor = dl->minor; | |
2992 | di->disk.state = 0; | |
2993 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
2994 | di->component_size = a->info.component_size; | |
2995 | di->container_member = inst; | |
2996 | di->next = rv; | |
2997 | rv = di; | |
2998 | num_spares++; | |
2999 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
3000 | i, di->data_offset); | |
88758e9d | 3001 | |
272906ef | 3002 | break; |
88758e9d DW |
3003 | } |
3004 | ||
3005 | if (!rv) | |
3006 | /* No spares found */ | |
3007 | return rv; | |
3008 | /* Now 'rv' has a list of devices to return. | |
3009 | * Create a metadata_update record to update the | |
3010 | * disk_ord_tbl for the array | |
3011 | */ | |
3012 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
3013 | if (mu) { |
3014 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
3015 | if (mu->buf == NULL) { | |
3016 | free(mu); | |
3017 | mu = NULL; | |
3018 | } | |
3019 | } | |
3020 | if (!mu) { | |
3021 | while (rv) { | |
3022 | struct mdinfo *n = rv->next; | |
3023 | ||
3024 | free(rv); | |
3025 | rv = n; | |
3026 | } | |
3027 | return NULL; | |
3028 | } | |
3029 | ||
88758e9d DW |
3030 | mu->space = NULL; |
3031 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; | |
3032 | mu->next = *updates; | |
3033 | u = (struct imsm_update_activate_spare *) mu->buf; | |
3034 | ||
3035 | for (di = rv ; di ; di = di->next) { | |
3036 | u->type = update_activate_spare; | |
d23fe947 DW |
3037 | u->dl = (struct dl *) di->devs; |
3038 | di->devs = NULL; | |
88758e9d DW |
3039 | u->slot = di->disk.raid_disk; |
3040 | u->array = inst; | |
3041 | u->next = u + 1; | |
3042 | u++; | |
3043 | } | |
3044 | (u-1)->next = NULL; | |
3045 | *updates = mu; | |
3046 | ||
3047 | return rv; | |
3048 | } | |
3049 | ||
ff077194 | 3050 | static int disks_overlap(struct imsm_dev *d1, struct imsm_dev *d2) |
8273f55e | 3051 | { |
ff077194 DW |
3052 | struct imsm_map *m1 = get_imsm_map(d1, 0); |
3053 | struct imsm_map *m2 = get_imsm_map(d2, 0); | |
8273f55e DW |
3054 | int i; |
3055 | int j; | |
3056 | int idx; | |
3057 | ||
3058 | for (i = 0; i < m1->num_members; i++) { | |
ff077194 | 3059 | idx = get_imsm_disk_idx(d1, i); |
8273f55e | 3060 | for (j = 0; j < m2->num_members; j++) |
ff077194 | 3061 | if (idx == get_imsm_disk_idx(d2, j)) |
8273f55e DW |
3062 | return 1; |
3063 | } | |
3064 | ||
3065 | return 0; | |
3066 | } | |
3067 | ||
24565c9a | 3068 | static void imsm_delete(struct intel_super *super, struct dl **dlp, int index); |
ae6aad82 | 3069 | |
e8319a19 DW |
3070 | static void imsm_process_update(struct supertype *st, |
3071 | struct metadata_update *update) | |
3072 | { | |
3073 | /** | |
3074 | * crack open the metadata_update envelope to find the update record | |
3075 | * update can be one of: | |
3076 | * update_activate_spare - a spare device has replaced a failed | |
3077 | * device in an array, update the disk_ord_tbl. If this disk is | |
3078 | * present in all member arrays then also clear the SPARE_DISK | |
3079 | * flag | |
3080 | */ | |
3081 | struct intel_super *super = st->sb; | |
4d7b1503 | 3082 | struct imsm_super *mpb; |
e8319a19 DW |
3083 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
3084 | ||
4d7b1503 DW |
3085 | /* update requires a larger buf but the allocation failed */ |
3086 | if (super->next_len && !super->next_buf) { | |
3087 | super->next_len = 0; | |
3088 | return; | |
3089 | } | |
3090 | ||
3091 | if (super->next_buf) { | |
3092 | memcpy(super->next_buf, super->buf, super->len); | |
3093 | free(super->buf); | |
3094 | super->len = super->next_len; | |
3095 | super->buf = super->next_buf; | |
3096 | ||
3097 | super->next_len = 0; | |
3098 | super->next_buf = NULL; | |
3099 | } | |
3100 | ||
3101 | mpb = super->anchor; | |
3102 | ||
e8319a19 DW |
3103 | switch (type) { |
3104 | case update_activate_spare: { | |
3105 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 3106 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 3107 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd | 3108 | struct imsm_map *migr_map; |
e8319a19 DW |
3109 | struct active_array *a; |
3110 | struct imsm_disk *disk; | |
0c046afd | 3111 | __u8 to_state; |
e8319a19 | 3112 | struct dl *dl; |
e8319a19 | 3113 | unsigned int found; |
0c046afd DW |
3114 | int failed; |
3115 | int victim = get_imsm_disk_idx(dev, u->slot); | |
e8319a19 DW |
3116 | int i; |
3117 | ||
3118 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 3119 | if (dl == u->dl) |
e8319a19 DW |
3120 | break; |
3121 | ||
3122 | if (!dl) { | |
3123 | fprintf(stderr, "error: imsm_activate_spare passed " | |
1f24f035 DW |
3124 | "an unknown disk (index: %d)\n", |
3125 | u->dl->index); | |
e8319a19 DW |
3126 | return; |
3127 | } | |
3128 | ||
3129 | super->updates_pending++; | |
3130 | ||
0c046afd DW |
3131 | /* count failures (excluding rebuilds and the victim) |
3132 | * to determine map[0] state | |
3133 | */ | |
3134 | failed = 0; | |
3135 | for (i = 0; i < map->num_members; i++) { | |
3136 | if (i == u->slot) | |
3137 | continue; | |
3138 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i)); | |
f2f27e63 | 3139 | if (!disk || disk->status & FAILED_DISK) |
0c046afd DW |
3140 | failed++; |
3141 | } | |
3142 | ||
d23fe947 DW |
3143 | /* adding a pristine spare, assign a new index */ |
3144 | if (dl->index < 0) { | |
3145 | dl->index = super->anchor->num_disks; | |
3146 | super->anchor->num_disks++; | |
3147 | } | |
d23fe947 | 3148 | disk = &dl->disk; |
f2f27e63 DW |
3149 | disk->status |= CONFIGURED_DISK; |
3150 | disk->status &= ~SPARE_DISK; | |
e8319a19 | 3151 | |
0c046afd DW |
3152 | /* mark rebuild */ |
3153 | to_state = imsm_check_degraded(super, dev, failed); | |
3154 | map->map_state = IMSM_T_STATE_DEGRADED; | |
e3bba0e0 | 3155 | migrate(dev, to_state, MIGR_REBUILD); |
0c046afd DW |
3156 | migr_map = get_imsm_map(dev, 1); |
3157 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); | |
3158 | set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD); | |
3159 | ||
e8319a19 DW |
3160 | /* count arrays using the victim in the metadata */ |
3161 | found = 0; | |
3162 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 3163 | dev = get_imsm_dev(super, a->info.container_member); |
e8319a19 | 3164 | for (i = 0; i < map->num_members; i++) |
ff077194 | 3165 | if (victim == get_imsm_disk_idx(dev, i)) |
e8319a19 DW |
3166 | found++; |
3167 | } | |
3168 | ||
24565c9a | 3169 | /* delete the victim if it is no longer being |
e8319a19 DW |
3170 | * utilized anywhere |
3171 | */ | |
e8319a19 | 3172 | if (!found) { |
ae6aad82 | 3173 | struct dl **dlp; |
24565c9a | 3174 | |
47ee5a45 DW |
3175 | /* We know that 'manager' isn't touching anything, |
3176 | * so it is safe to delete | |
3177 | */ | |
24565c9a | 3178 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) |
ae6aad82 DW |
3179 | if ((*dlp)->index == victim) |
3180 | break; | |
47ee5a45 DW |
3181 | |
3182 | /* victim may be on the missing list */ | |
3183 | if (!*dlp) | |
3184 | for (dlp = &super->missing; *dlp; dlp = &(*dlp)->next) | |
3185 | if ((*dlp)->index == victim) | |
3186 | break; | |
24565c9a | 3187 | imsm_delete(super, dlp, victim); |
e8319a19 | 3188 | } |
8273f55e DW |
3189 | break; |
3190 | } | |
3191 | case update_create_array: { | |
3192 | /* someone wants to create a new array, we need to be aware of | |
3193 | * a few races/collisions: | |
3194 | * 1/ 'Create' called by two separate instances of mdadm | |
3195 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
3196 | * devices that have since been assimilated via | |
3197 | * activate_spare. | |
3198 | * In the event this update can not be carried out mdadm will | |
3199 | * (FIX ME) notice that its update did not take hold. | |
3200 | */ | |
3201 | struct imsm_update_create_array *u = (void *) update->buf; | |
3202 | struct imsm_dev *dev; | |
3203 | struct imsm_map *map, *new_map; | |
3204 | unsigned long long start, end; | |
3205 | unsigned long long new_start, new_end; | |
3206 | int i; | |
3207 | int overlap = 0; | |
3208 | ||
3209 | /* handle racing creates: first come first serve */ | |
3210 | if (u->dev_idx < mpb->num_raid_devs) { | |
3211 | dprintf("%s: subarray %d already defined\n", | |
3212 | __func__, u->dev_idx); | |
3213 | return; | |
3214 | } | |
3215 | ||
3216 | /* check update is next in sequence */ | |
3217 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
3218 | dprintf("%s: can not create array %d expected index %d\n", |
3219 | __func__, u->dev_idx, mpb->num_raid_devs); | |
8273f55e DW |
3220 | return; |
3221 | } | |
3222 | ||
a965f303 | 3223 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
3224 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
3225 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
3226 | ||
3227 | /* handle activate_spare versus create race: | |
3228 | * check to make sure that overlapping arrays do not include | |
3229 | * overalpping disks | |
3230 | */ | |
3231 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 3232 | dev = get_imsm_dev(super, i); |
a965f303 | 3233 | map = get_imsm_map(dev, 0); |
8273f55e DW |
3234 | start = __le32_to_cpu(map->pba_of_lba0); |
3235 | end = start + __le32_to_cpu(map->blocks_per_member); | |
3236 | if ((new_start >= start && new_start <= end) || | |
3237 | (start >= new_start && start <= new_end)) | |
3238 | overlap = 1; | |
ff077194 | 3239 | if (overlap && disks_overlap(dev, &u->dev)) { |
8273f55e DW |
3240 | dprintf("%s: arrays overlap\n", __func__); |
3241 | return; | |
3242 | } | |
3243 | } | |
3244 | /* check num_members sanity */ | |
3245 | if (new_map->num_members > mpb->num_disks) { | |
3246 | dprintf("%s: num_disks out of range\n", __func__); | |
3247 | return; | |
3248 | } | |
3249 | ||
949c47a0 DW |
3250 | /* check that prepare update was successful */ |
3251 | if (!update->space) { | |
3252 | dprintf("%s: prepare update failed\n", __func__); | |
3253 | return; | |
3254 | } | |
3255 | ||
8273f55e | 3256 | super->updates_pending++; |
949c47a0 | 3257 | dev = update->space; |
ff077194 | 3258 | map = get_imsm_map(dev, 0); |
949c47a0 DW |
3259 | update->space = NULL; |
3260 | imsm_copy_dev(dev, &u->dev); | |
e0783b41 | 3261 | map = get_imsm_map(dev, 0); |
949c47a0 | 3262 | super->dev_tbl[u->dev_idx] = dev; |
8273f55e | 3263 | mpb->num_raid_devs++; |
8273f55e | 3264 | |
e0783b41 | 3265 | /* fix up flags */ |
8273f55e DW |
3266 | for (i = 0; i < map->num_members; i++) { |
3267 | struct imsm_disk *disk; | |
8273f55e | 3268 | |
ff077194 | 3269 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i)); |
f2f27e63 DW |
3270 | disk->status |= CONFIGURED_DISK; |
3271 | disk->status &= ~SPARE_DISK; | |
8273f55e DW |
3272 | } |
3273 | break; | |
e8319a19 | 3274 | } |
43dad3d6 DW |
3275 | case update_add_disk: |
3276 | ||
3277 | /* we may be able to repair some arrays if disks are | |
3278 | * being added */ | |
3279 | if (super->add) { | |
3280 | struct active_array *a; | |
072b727f DW |
3281 | |
3282 | super->updates_pending++; | |
43dad3d6 DW |
3283 | for (a = st->arrays; a; a = a->next) |
3284 | a->check_degraded = 1; | |
3285 | } | |
e553d2a4 | 3286 | /* add some spares to the metadata */ |
43dad3d6 | 3287 | while (super->add) { |
e553d2a4 DW |
3288 | struct dl *al; |
3289 | ||
43dad3d6 DW |
3290 | al = super->add; |
3291 | super->add = al->next; | |
43dad3d6 DW |
3292 | al->next = super->disks; |
3293 | super->disks = al; | |
e553d2a4 DW |
3294 | dprintf("%s: added %x:%x\n", |
3295 | __func__, al->major, al->minor); | |
43dad3d6 DW |
3296 | } |
3297 | ||
3298 | break; | |
e8319a19 DW |
3299 | } |
3300 | } | |
88758e9d | 3301 | |
8273f55e DW |
3302 | static void imsm_prepare_update(struct supertype *st, |
3303 | struct metadata_update *update) | |
3304 | { | |
949c47a0 | 3305 | /** |
4d7b1503 DW |
3306 | * Allocate space to hold new disk entries, raid-device entries or a new |
3307 | * mpb if necessary. The manager synchronously waits for updates to | |
3308 | * complete in the monitor, so new mpb buffers allocated here can be | |
3309 | * integrated by the monitor thread without worrying about live pointers | |
3310 | * in the manager thread. | |
8273f55e | 3311 | */ |
949c47a0 | 3312 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
3313 | struct intel_super *super = st->sb; |
3314 | struct imsm_super *mpb = super->anchor; | |
3315 | size_t buf_len; | |
3316 | size_t len = 0; | |
949c47a0 DW |
3317 | |
3318 | switch (type) { | |
3319 | case update_create_array: { | |
3320 | struct imsm_update_create_array *u = (void *) update->buf; | |
949c47a0 | 3321 | |
4d7b1503 | 3322 | len = sizeof_imsm_dev(&u->dev, 1); |
949c47a0 DW |
3323 | update->space = malloc(len); |
3324 | break; | |
3325 | default: | |
3326 | break; | |
3327 | } | |
3328 | } | |
8273f55e | 3329 | |
4d7b1503 DW |
3330 | /* check if we need a larger metadata buffer */ |
3331 | if (super->next_buf) | |
3332 | buf_len = super->next_len; | |
3333 | else | |
3334 | buf_len = super->len; | |
3335 | ||
3336 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
3337 | /* ok we need a larger buf than what is currently allocated | |
3338 | * if this allocation fails process_update will notice that | |
3339 | * ->next_len is set and ->next_buf is NULL | |
3340 | */ | |
3341 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
3342 | if (super->next_buf) | |
3343 | free(super->next_buf); | |
3344 | ||
3345 | super->next_len = buf_len; | |
3346 | if (posix_memalign(&super->next_buf, buf_len, 512) != 0) | |
3347 | super->next_buf = NULL; | |
3348 | } | |
8273f55e DW |
3349 | } |
3350 | ||
ae6aad82 | 3351 | /* must be called while manager is quiesced */ |
24565c9a | 3352 | static void imsm_delete(struct intel_super *super, struct dl **dlp, int index) |
ae6aad82 DW |
3353 | { |
3354 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
3355 | struct dl *iter; |
3356 | struct imsm_dev *dev; | |
3357 | struct imsm_map *map; | |
24565c9a DW |
3358 | int i, j, num_members; |
3359 | __u32 ord; | |
ae6aad82 | 3360 | |
24565c9a DW |
3361 | dprintf("%s: deleting device[%d] from imsm_super\n", |
3362 | __func__, index); | |
ae6aad82 DW |
3363 | |
3364 | /* shift all indexes down one */ | |
3365 | for (iter = super->disks; iter; iter = iter->next) | |
24565c9a | 3366 | if (iter->index > index) |
ae6aad82 | 3367 | iter->index--; |
47ee5a45 DW |
3368 | for (iter = super->missing; iter; iter = iter->next) |
3369 | if (iter->index > index) | |
3370 | iter->index--; | |
ae6aad82 DW |
3371 | |
3372 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
3373 | dev = get_imsm_dev(super, i); | |
3374 | map = get_imsm_map(dev, 0); | |
24565c9a DW |
3375 | num_members = map->num_members; |
3376 | for (j = 0; j < num_members; j++) { | |
3377 | /* update ord entries being careful not to propagate | |
3378 | * ord-flags to the first map | |
3379 | */ | |
3380 | ord = get_imsm_ord_tbl_ent(dev, j); | |
ae6aad82 | 3381 | |
24565c9a DW |
3382 | if (ord_to_idx(ord) <= index) |
3383 | continue; | |
ae6aad82 | 3384 | |
24565c9a DW |
3385 | map = get_imsm_map(dev, 0); |
3386 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); | |
3387 | map = get_imsm_map(dev, 1); | |
3388 | if (map) | |
3389 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
3390 | } |
3391 | } | |
3392 | ||
3393 | mpb->num_disks--; | |
3394 | super->updates_pending++; | |
24565c9a DW |
3395 | if (*dlp) { |
3396 | struct dl *dl = *dlp; | |
3397 | ||
3398 | *dlp = (*dlp)->next; | |
3399 | __free_imsm_disk(dl); | |
3400 | } | |
ae6aad82 | 3401 | } |
0e600426 | 3402 | #endif /* MDASSEMBLE */ |
ae6aad82 | 3403 | |
cdddbdbc DW |
3404 | struct superswitch super_imsm = { |
3405 | #ifndef MDASSEMBLE | |
3406 | .examine_super = examine_super_imsm, | |
3407 | .brief_examine_super = brief_examine_super_imsm, | |
3408 | .detail_super = detail_super_imsm, | |
3409 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 3410 | .write_init_super = write_init_super_imsm, |
0e600426 N |
3411 | .validate_geometry = validate_geometry_imsm, |
3412 | .add_to_super = add_to_super_imsm, | |
cdddbdbc DW |
3413 | #endif |
3414 | .match_home = match_home_imsm, | |
3415 | .uuid_from_super= uuid_from_super_imsm, | |
3416 | .getinfo_super = getinfo_super_imsm, | |
3417 | .update_super = update_super_imsm, | |
3418 | ||
3419 | .avail_size = avail_size_imsm, | |
3420 | ||
3421 | .compare_super = compare_super_imsm, | |
3422 | ||
3423 | .load_super = load_super_imsm, | |
bf5a934a | 3424 | .init_super = init_super_imsm, |
cdddbdbc DW |
3425 | .store_super = store_zero_imsm, |
3426 | .free_super = free_super_imsm, | |
3427 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 3428 | .container_content = container_content_imsm, |
cdddbdbc | 3429 | |
cdddbdbc | 3430 | .external = 1, |
845dea95 | 3431 | |
0e600426 | 3432 | #ifndef MDASSEMBLE |
845dea95 NB |
3433 | /* for mdmon */ |
3434 | .open_new = imsm_open_new, | |
3435 | .load_super = load_super_imsm, | |
ed9d66aa | 3436 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
3437 | .set_disk = imsm_set_disk, |
3438 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 3439 | .activate_spare = imsm_activate_spare, |
e8319a19 | 3440 | .process_update = imsm_process_update, |
8273f55e | 3441 | .prepare_update = imsm_prepare_update, |
0e600426 | 3442 | #endif /* MDASSEMBLE */ |
cdddbdbc | 3443 | }; |