]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb DW |
43 | |
44 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
45 | #define MPB_ATTRIB_PM __cpu_to_le32(0x40000000) | |
46 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
47 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
48 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
51 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
52 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
53 | ||
c2c087e6 DW |
54 | #define MPB_SECTOR_CNT 418 |
55 | #define IMSM_RESERVED_SECTORS 4096 | |
979d38be | 56 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
57 | |
58 | /* Disk configuration info. */ | |
59 | #define IMSM_MAX_DEVICES 255 | |
60 | struct imsm_disk { | |
61 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
62 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
63 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
f2f27e63 DW |
64 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
65 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
66 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 67 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb DW |
68 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
69 | #define IMSM_DISK_FILLERS 4 | |
cdddbdbc DW |
70 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ |
71 | }; | |
72 | ||
73 | /* RAID map configuration infos. */ | |
74 | struct imsm_map { | |
75 | __u32 pba_of_lba0; /* start address of partition */ | |
76 | __u32 blocks_per_member;/* blocks per member */ | |
77 | __u32 num_data_stripes; /* number of data stripes */ | |
78 | __u16 blocks_per_strip; | |
79 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
80 | #define IMSM_T_STATE_NORMAL 0 | |
81 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
82 | #define IMSM_T_STATE_DEGRADED 2 |
83 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
84 | __u8 raid_level; |
85 | #define IMSM_T_RAID0 0 | |
86 | #define IMSM_T_RAID1 1 | |
87 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
88 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
89 | __u8 num_domains; /* number of parity domains */ |
90 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 91 | __u8 ddf; |
cdddbdbc | 92 | __u32 filler[7]; /* expansion area */ |
7eef0453 | 93 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 94 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
95 | * top byte contains some flags |
96 | */ | |
cdddbdbc DW |
97 | } __attribute__ ((packed)); |
98 | ||
99 | struct imsm_vol { | |
f8f603f1 | 100 | __u32 curr_migr_unit; |
fe7ed8cb | 101 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 102 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
103 | #define MIGR_INIT 0 |
104 | #define MIGR_REBUILD 1 | |
105 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
106 | #define MIGR_GEN_MIGR 3 | |
107 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 108 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
109 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
110 | __u8 dirty; | |
fe7ed8cb DW |
111 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
112 | __u16 verify_errors; /* number of mismatches */ | |
113 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
114 | __u32 filler[4]; | |
cdddbdbc DW |
115 | struct imsm_map map[1]; |
116 | /* here comes another one if migr_state */ | |
117 | } __attribute__ ((packed)); | |
118 | ||
119 | struct imsm_dev { | |
fe7ed8cb | 120 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
121 | __u32 size_low; |
122 | __u32 size_high; | |
fe7ed8cb DW |
123 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
124 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
125 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
126 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
127 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
128 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
129 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
130 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
131 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
132 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
133 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
134 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
135 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
136 | __u32 status; /* Persistent RaidDev status */ |
137 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
138 | __u8 migr_priority; |
139 | __u8 num_sub_vols; | |
140 | __u8 tid; | |
141 | __u8 cng_master_disk; | |
142 | __u16 cache_policy; | |
143 | __u8 cng_state; | |
144 | __u8 cng_sub_state; | |
145 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
146 | __u32 filler[IMSM_DEV_FILLERS]; |
147 | struct imsm_vol vol; | |
148 | } __attribute__ ((packed)); | |
149 | ||
150 | struct imsm_super { | |
151 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
152 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
153 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
154 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
155 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
156 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
157 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
158 | __u8 num_disks; /* 0x38 Number of configured disks */ |
159 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
160 | __u8 error_log_pos; /* 0x3A */ |
161 | __u8 fill[1]; /* 0x3B */ | |
162 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
163 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
164 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
165 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
166 | #define IMSM_FILLERS 35 | |
167 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
168 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
169 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 170 | /* here comes BBM logs */ |
cdddbdbc DW |
171 | } __attribute__ ((packed)); |
172 | ||
604b746f JD |
173 | #define BBM_LOG_MAX_ENTRIES 254 |
174 | ||
175 | struct bbm_log_entry { | |
176 | __u64 defective_block_start; | |
177 | #define UNREADABLE 0xFFFFFFFF | |
178 | __u32 spare_block_offset; | |
179 | __u16 remapped_marked_count; | |
180 | __u16 disk_ordinal; | |
181 | } __attribute__ ((__packed__)); | |
182 | ||
183 | struct bbm_log { | |
184 | __u32 signature; /* 0xABADB10C */ | |
185 | __u32 entry_count; | |
186 | __u32 reserved_spare_block_count; /* 0 */ | |
187 | __u32 reserved; /* 0xFFFF */ | |
188 | __u64 first_spare_lba; | |
189 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
190 | } __attribute__ ((__packed__)); | |
191 | ||
192 | ||
cdddbdbc DW |
193 | #ifndef MDASSEMBLE |
194 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
195 | #endif | |
196 | ||
1484e727 DW |
197 | static __u8 migr_type(struct imsm_dev *dev) |
198 | { | |
199 | if (dev->vol.migr_type == MIGR_VERIFY && | |
200 | dev->status & DEV_VERIFY_AND_FIX) | |
201 | return MIGR_REPAIR; | |
202 | else | |
203 | return dev->vol.migr_type; | |
204 | } | |
205 | ||
206 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
207 | { | |
208 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
209 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
210 | */ | |
211 | if (migr_type == MIGR_REPAIR) { | |
212 | dev->vol.migr_type = MIGR_VERIFY; | |
213 | dev->status |= DEV_VERIFY_AND_FIX; | |
214 | } else { | |
215 | dev->vol.migr_type = migr_type; | |
216 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
217 | } | |
218 | } | |
219 | ||
87eb16df | 220 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 221 | { |
87eb16df DW |
222 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
223 | } | |
cdddbdbc | 224 | |
87eb16df DW |
225 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
226 | { | |
227 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
228 | } |
229 | ||
ba2de7ba DW |
230 | struct intel_dev { |
231 | struct imsm_dev *dev; | |
232 | struct intel_dev *next; | |
f21e18ca | 233 | unsigned index; |
ba2de7ba DW |
234 | }; |
235 | ||
88654014 LM |
236 | struct intel_hba { |
237 | enum sys_dev_type type; | |
238 | char *path; | |
239 | char *pci_id; | |
240 | struct intel_hba *next; | |
241 | }; | |
242 | ||
1a64be56 LM |
243 | enum action { |
244 | DISK_REMOVE = 1, | |
245 | DISK_ADD | |
246 | }; | |
cdddbdbc DW |
247 | /* internal representation of IMSM metadata */ |
248 | struct intel_super { | |
249 | union { | |
949c47a0 DW |
250 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
251 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 252 | }; |
949c47a0 | 253 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
254 | void *next_buf; /* for realloc'ing buf from the manager */ |
255 | size_t next_len; | |
c2c087e6 | 256 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 257 | int current_vol; /* index of raid device undergoing creation */ |
0dcecb2e | 258 | __u32 create_offset; /* common start for 'current_vol' */ |
148acb7b | 259 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 260 | struct intel_dev *devlist; |
cdddbdbc DW |
261 | struct dl { |
262 | struct dl *next; | |
263 | int index; | |
264 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
265 | int major, minor; | |
266 | char *devname; | |
b9f594fe | 267 | struct imsm_disk disk; |
cdddbdbc | 268 | int fd; |
0dcecb2e DW |
269 | int extent_cnt; |
270 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 271 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 272 | enum action action; |
cdddbdbc | 273 | } *disks; |
1a64be56 LM |
274 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
275 | active */ | |
47ee5a45 | 276 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 277 | struct bbm_log *bbm_log; |
88654014 | 278 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 279 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
280 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
281 | }; | |
282 | ||
283 | struct intel_disk { | |
284 | struct imsm_disk disk; | |
285 | #define IMSM_UNKNOWN_OWNER (-1) | |
286 | int owner; | |
287 | struct intel_disk *next; | |
cdddbdbc DW |
288 | }; |
289 | ||
c2c087e6 DW |
290 | struct extent { |
291 | unsigned long long start, size; | |
292 | }; | |
293 | ||
694575e7 KW |
294 | /* definitions of reshape process types */ |
295 | enum imsm_reshape_type { | |
296 | CH_TAKEOVER, | |
b5347799 | 297 | CH_MIGRATION, |
694575e7 KW |
298 | }; |
299 | ||
88758e9d DW |
300 | /* definition of messages passed to imsm_process_update */ |
301 | enum imsm_update_type { | |
302 | update_activate_spare, | |
8273f55e | 303 | update_create_array, |
33414a01 | 304 | update_kill_array, |
aa534678 | 305 | update_rename_array, |
1a64be56 | 306 | update_add_remove_disk, |
78b10e66 | 307 | update_reshape_container_disks, |
bb025c2f | 308 | update_takeover |
88758e9d DW |
309 | }; |
310 | ||
311 | struct imsm_update_activate_spare { | |
312 | enum imsm_update_type type; | |
d23fe947 | 313 | struct dl *dl; |
88758e9d DW |
314 | int slot; |
315 | int array; | |
316 | struct imsm_update_activate_spare *next; | |
317 | }; | |
318 | ||
78b10e66 N |
319 | struct geo_params { |
320 | int dev_id; | |
321 | char *dev_name; | |
322 | long long size; | |
323 | int level; | |
324 | int layout; | |
325 | int chunksize; | |
326 | int raid_disks; | |
327 | }; | |
328 | ||
bb025c2f KW |
329 | enum takeover_direction { |
330 | R10_TO_R0, | |
331 | R0_TO_R10 | |
332 | }; | |
333 | struct imsm_update_takeover { | |
334 | enum imsm_update_type type; | |
335 | int subarray; | |
336 | enum takeover_direction direction; | |
337 | }; | |
78b10e66 N |
338 | |
339 | struct imsm_update_reshape { | |
340 | enum imsm_update_type type; | |
341 | int old_raid_disks; | |
342 | int new_raid_disks; | |
d195167d | 343 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
344 | }; |
345 | ||
54c2c1ea DW |
346 | struct disk_info { |
347 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
348 | }; | |
349 | ||
8273f55e DW |
350 | struct imsm_update_create_array { |
351 | enum imsm_update_type type; | |
8273f55e | 352 | int dev_idx; |
6a3e913e | 353 | struct imsm_dev dev; |
8273f55e DW |
354 | }; |
355 | ||
33414a01 DW |
356 | struct imsm_update_kill_array { |
357 | enum imsm_update_type type; | |
358 | int dev_idx; | |
359 | }; | |
360 | ||
aa534678 DW |
361 | struct imsm_update_rename_array { |
362 | enum imsm_update_type type; | |
363 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
364 | int dev_idx; | |
365 | }; | |
366 | ||
1a64be56 | 367 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
368 | enum imsm_update_type type; |
369 | }; | |
370 | ||
88654014 LM |
371 | |
372 | static const char *_sys_dev_type[] = { | |
373 | [SYS_DEV_UNKNOWN] = "Unknown", | |
374 | [SYS_DEV_SAS] = "SAS", | |
375 | [SYS_DEV_SATA] = "SATA" | |
376 | }; | |
377 | ||
378 | const char *get_sys_dev_type(enum sys_dev_type type) | |
379 | { | |
380 | if (type >= SYS_DEV_MAX) | |
381 | type = SYS_DEV_UNKNOWN; | |
382 | ||
383 | return _sys_dev_type[type]; | |
384 | } | |
385 | ||
71204a50 | 386 | #ifndef MDASSEMBLE |
88654014 LM |
387 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) |
388 | { | |
389 | struct intel_hba *result = malloc(sizeof(*result)); | |
390 | if (result) { | |
391 | result->type = device->type; | |
392 | result->path = strdup(device->path); | |
393 | result->next = NULL; | |
394 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
395 | result->pci_id++; | |
396 | } | |
397 | return result; | |
398 | } | |
399 | ||
400 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
401 | { | |
402 | struct intel_hba *result=NULL; | |
403 | for (result = hba; result; result = result->next) { | |
404 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
405 | break; | |
406 | } | |
407 | return result; | |
408 | } | |
409 | ||
410 | ||
88654014 LM |
411 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device, |
412 | const char *devname) | |
413 | { | |
414 | struct intel_hba *hba; | |
415 | ||
416 | /* check if disk attached to Intel HBA */ | |
417 | hba = find_intel_hba(super->hba, device); | |
418 | if (hba != NULL) | |
419 | return 1; | |
420 | /* Check if HBA is already attached to super */ | |
421 | if (super->hba == NULL) { | |
422 | super->hba = alloc_intel_hba(device); | |
423 | return 1; | |
424 | } | |
425 | ||
426 | hba = super->hba; | |
427 | /* Intel metadata allows for all disks attached to the same type HBA. | |
428 | * Do not sypport odf HBA types mixing | |
429 | */ | |
430 | if (device->type != hba->type) | |
431 | return 2; | |
432 | ||
433 | while (hba->next) | |
434 | hba = hba->next; | |
435 | ||
436 | hba->next = alloc_intel_hba(device); | |
437 | return 1; | |
438 | } | |
439 | ||
440 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
441 | { | |
442 | struct sys_dev *list, *elem, *prev; | |
443 | char *disk_path; | |
444 | ||
445 | if ((list = find_intel_devices()) == NULL) | |
446 | return 0; | |
447 | ||
448 | if (fd < 0) | |
449 | disk_path = (char *) devname; | |
450 | else | |
451 | disk_path = diskfd_to_devpath(fd); | |
452 | ||
453 | if (!disk_path) { | |
454 | free_sys_dev(&list); | |
455 | return 0; | |
456 | } | |
457 | ||
458 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
459 | if (path_attached_to_hba(disk_path, elem->path)) { | |
460 | if (prev == NULL) | |
461 | list = list->next; | |
462 | else | |
463 | prev->next = elem->next; | |
464 | elem->next = NULL; | |
465 | if (disk_path != devname) | |
466 | free(disk_path); | |
467 | free_sys_dev(&list); | |
468 | return elem; | |
469 | } | |
470 | } | |
471 | if (disk_path != devname) | |
472 | free(disk_path); | |
473 | free_sys_dev(&list); | |
474 | ||
475 | return NULL; | |
476 | } | |
71204a50 | 477 | #endif /* MDASSEMBLE */ |
88654014 LM |
478 | |
479 | ||
cdddbdbc DW |
480 | static struct supertype *match_metadata_desc_imsm(char *arg) |
481 | { | |
482 | struct supertype *st; | |
483 | ||
484 | if (strcmp(arg, "imsm") != 0 && | |
485 | strcmp(arg, "default") != 0 | |
486 | ) | |
487 | return NULL; | |
488 | ||
489 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
490 | if (!st) |
491 | return NULL; | |
ef609477 | 492 | memset(st, 0, sizeof(*st)); |
d1d599ea | 493 | st->container_dev = NoMdDev; |
cdddbdbc DW |
494 | st->ss = &super_imsm; |
495 | st->max_devs = IMSM_MAX_DEVICES; | |
496 | st->minor_version = 0; | |
497 | st->sb = NULL; | |
498 | return st; | |
499 | } | |
500 | ||
0e600426 | 501 | #ifndef MDASSEMBLE |
cdddbdbc DW |
502 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
503 | { | |
504 | return &mpb->sig[MPB_SIG_LEN]; | |
505 | } | |
0e600426 | 506 | #endif |
cdddbdbc | 507 | |
949c47a0 DW |
508 | /* retrieve a disk directly from the anchor when the anchor is known to be |
509 | * up-to-date, currently only at load time | |
510 | */ | |
511 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 512 | { |
949c47a0 | 513 | if (index >= mpb->num_disks) |
cdddbdbc DW |
514 | return NULL; |
515 | return &mpb->disk[index]; | |
516 | } | |
517 | ||
95d07a2c LM |
518 | /* retrieve the disk description based on a index of the disk |
519 | * in the sub-array | |
520 | */ | |
521 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 522 | { |
b9f594fe DW |
523 | struct dl *d; |
524 | ||
525 | for (d = super->disks; d; d = d->next) | |
526 | if (d->index == index) | |
95d07a2c LM |
527 | return d; |
528 | ||
529 | return NULL; | |
530 | } | |
531 | /* retrieve a disk from the parsed metadata */ | |
532 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
533 | { | |
534 | struct dl *dl; | |
535 | ||
536 | dl = get_imsm_dl_disk(super, index); | |
537 | if (dl) | |
538 | return &dl->disk; | |
539 | ||
b9f594fe | 540 | return NULL; |
949c47a0 DW |
541 | } |
542 | ||
543 | /* generate a checksum directly from the anchor when the anchor is known to be | |
544 | * up-to-date, currently only at load or write_super after coalescing | |
545 | */ | |
546 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
547 | { |
548 | __u32 end = mpb->mpb_size / sizeof(end); | |
549 | __u32 *p = (__u32 *) mpb; | |
550 | __u32 sum = 0; | |
551 | ||
97f734fd N |
552 | while (end--) { |
553 | sum += __le32_to_cpu(*p); | |
554 | p++; | |
555 | } | |
cdddbdbc DW |
556 | |
557 | return sum - __le32_to_cpu(mpb->check_sum); | |
558 | } | |
559 | ||
a965f303 DW |
560 | static size_t sizeof_imsm_map(struct imsm_map *map) |
561 | { | |
562 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
563 | } | |
564 | ||
565 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 566 | { |
5e7b0330 AK |
567 | /* A device can have 2 maps if it is in the middle of a migration. |
568 | * If second_map is: | |
569 | * 0 - we return the first map | |
570 | * 1 - we return the second map if it exists, else NULL | |
571 | * -1 - we return the second map if it exists, else the first | |
572 | */ | |
a965f303 DW |
573 | struct imsm_map *map = &dev->vol.map[0]; |
574 | ||
5e7b0330 | 575 | if (second_map == 1 && !dev->vol.migr_state) |
a965f303 | 576 | return NULL; |
5e7b0330 AK |
577 | else if (second_map == 1 || |
578 | (second_map < 0 && dev->vol.migr_state)) { | |
a965f303 DW |
579 | void *ptr = map; |
580 | ||
581 | return ptr + sizeof_imsm_map(map); | |
582 | } else | |
583 | return map; | |
5e7b0330 | 584 | |
a965f303 | 585 | } |
cdddbdbc | 586 | |
3393c6af DW |
587 | /* return the size of the device. |
588 | * migr_state increases the returned size if map[0] were to be duplicated | |
589 | */ | |
590 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
591 | { |
592 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
593 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
594 | |
595 | /* migrating means an additional map */ | |
a965f303 DW |
596 | if (dev->vol.migr_state) |
597 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
598 | else if (migr_state) |
599 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
600 | |
601 | return size; | |
602 | } | |
603 | ||
54c2c1ea DW |
604 | #ifndef MDASSEMBLE |
605 | /* retrieve disk serial number list from a metadata update */ | |
606 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
607 | { | |
608 | void *u = update; | |
609 | struct disk_info *inf; | |
610 | ||
611 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
612 | sizeof_imsm_dev(&update->dev, 0); | |
613 | ||
614 | return inf; | |
615 | } | |
616 | #endif | |
617 | ||
949c47a0 | 618 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
619 | { |
620 | int offset; | |
621 | int i; | |
622 | void *_mpb = mpb; | |
623 | ||
949c47a0 | 624 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
625 | return NULL; |
626 | ||
627 | /* devices start after all disks */ | |
628 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
629 | ||
630 | for (i = 0; i <= index; i++) | |
631 | if (i == index) | |
632 | return _mpb + offset; | |
633 | else | |
3393c6af | 634 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
635 | |
636 | return NULL; | |
637 | } | |
638 | ||
949c47a0 DW |
639 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
640 | { | |
ba2de7ba DW |
641 | struct intel_dev *dv; |
642 | ||
949c47a0 DW |
643 | if (index >= super->anchor->num_raid_devs) |
644 | return NULL; | |
ba2de7ba DW |
645 | for (dv = super->devlist; dv; dv = dv->next) |
646 | if (dv->index == index) | |
647 | return dv->dev; | |
648 | return NULL; | |
949c47a0 DW |
649 | } |
650 | ||
98130f40 AK |
651 | /* |
652 | * for second_map: | |
653 | * == 0 get first map | |
654 | * == 1 get second map | |
655 | * == -1 than get map according to the current migr_state | |
656 | */ | |
657 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
658 | int slot, | |
659 | int second_map) | |
7eef0453 DW |
660 | { |
661 | struct imsm_map *map; | |
662 | ||
5e7b0330 | 663 | map = get_imsm_map(dev, second_map); |
7eef0453 | 664 | |
ff077194 DW |
665 | /* top byte identifies disk under rebuild */ |
666 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
667 | } | |
668 | ||
669 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 670 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 671 | { |
98130f40 | 672 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
673 | |
674 | return ord_to_idx(ord); | |
7eef0453 DW |
675 | } |
676 | ||
be73972f DW |
677 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
678 | { | |
679 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
680 | } | |
681 | ||
f21e18ca | 682 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
683 | { |
684 | int slot; | |
685 | __u32 ord; | |
686 | ||
687 | for (slot = 0; slot < map->num_members; slot++) { | |
688 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
689 | if (ord_to_idx(ord) == idx) | |
690 | return slot; | |
691 | } | |
692 | ||
693 | return -1; | |
694 | } | |
695 | ||
cdddbdbc DW |
696 | static int get_imsm_raid_level(struct imsm_map *map) |
697 | { | |
698 | if (map->raid_level == 1) { | |
699 | if (map->num_members == 2) | |
700 | return 1; | |
701 | else | |
702 | return 10; | |
703 | } | |
704 | ||
705 | return map->raid_level; | |
706 | } | |
707 | ||
c2c087e6 DW |
708 | static int cmp_extent(const void *av, const void *bv) |
709 | { | |
710 | const struct extent *a = av; | |
711 | const struct extent *b = bv; | |
712 | if (a->start < b->start) | |
713 | return -1; | |
714 | if (a->start > b->start) | |
715 | return 1; | |
716 | return 0; | |
717 | } | |
718 | ||
0dcecb2e | 719 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 720 | { |
c2c087e6 | 721 | int memberships = 0; |
620b1713 | 722 | int i; |
c2c087e6 | 723 | |
949c47a0 DW |
724 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
725 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 726 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 727 | |
620b1713 DW |
728 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
729 | memberships++; | |
c2c087e6 | 730 | } |
0dcecb2e DW |
731 | |
732 | return memberships; | |
733 | } | |
734 | ||
735 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
736 | { | |
737 | /* find a list of used extents on the given physical device */ | |
738 | struct extent *rv, *e; | |
620b1713 | 739 | int i; |
0dcecb2e DW |
740 | int memberships = count_memberships(dl, super); |
741 | __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
742 | ||
c2c087e6 DW |
743 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
744 | if (!rv) | |
745 | return NULL; | |
746 | e = rv; | |
747 | ||
949c47a0 DW |
748 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
749 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 750 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 751 | |
620b1713 DW |
752 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
753 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
754 | e->size = __le32_to_cpu(map->blocks_per_member); | |
755 | e++; | |
c2c087e6 DW |
756 | } |
757 | } | |
758 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
759 | ||
14e8215b DW |
760 | /* determine the start of the metadata |
761 | * when no raid devices are defined use the default | |
762 | * ...otherwise allow the metadata to truncate the value | |
763 | * as is the case with older versions of imsm | |
764 | */ | |
765 | if (memberships) { | |
766 | struct extent *last = &rv[memberships - 1]; | |
767 | __u32 remainder; | |
768 | ||
769 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
770 | (last->start + last->size); | |
dda5855f DW |
771 | /* round down to 1k block to satisfy precision of the kernel |
772 | * 'size' interface | |
773 | */ | |
774 | remainder &= ~1UL; | |
775 | /* make sure remainder is still sane */ | |
f21e18ca | 776 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 777 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
778 | if (reservation > remainder) |
779 | reservation = remainder; | |
780 | } | |
781 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
782 | e->size = 0; |
783 | return rv; | |
784 | } | |
785 | ||
14e8215b DW |
786 | /* try to determine how much space is reserved for metadata from |
787 | * the last get_extents() entry, otherwise fallback to the | |
788 | * default | |
789 | */ | |
790 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
791 | { | |
792 | struct extent *e; | |
793 | int i; | |
794 | __u32 rv; | |
795 | ||
796 | /* for spares just return a minimal reservation which will grow | |
797 | * once the spare is picked up by an array | |
798 | */ | |
799 | if (dl->index == -1) | |
800 | return MPB_SECTOR_CNT; | |
801 | ||
802 | e = get_extents(super, dl); | |
803 | if (!e) | |
804 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
805 | ||
806 | /* scroll to last entry */ | |
807 | for (i = 0; e[i].size; i++) | |
808 | continue; | |
809 | ||
810 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
811 | ||
812 | free(e); | |
813 | ||
814 | return rv; | |
815 | } | |
816 | ||
25ed7e59 DW |
817 | static int is_spare(struct imsm_disk *disk) |
818 | { | |
819 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
820 | } | |
821 | ||
822 | static int is_configured(struct imsm_disk *disk) | |
823 | { | |
824 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
825 | } | |
826 | ||
827 | static int is_failed(struct imsm_disk *disk) | |
828 | { | |
829 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
830 | } | |
831 | ||
80e7f8c3 AC |
832 | /* Return minimum size of a spare that can be used in this array*/ |
833 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
834 | { | |
835 | struct intel_super *super = st->sb; | |
836 | struct dl *dl; | |
837 | struct extent *e; | |
838 | int i; | |
839 | unsigned long long rv = 0; | |
840 | ||
841 | if (!super) | |
842 | return rv; | |
843 | /* find first active disk in array */ | |
844 | dl = super->disks; | |
845 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
846 | dl = dl->next; | |
847 | if (!dl) | |
848 | return rv; | |
849 | /* find last lba used by subarrays */ | |
850 | e = get_extents(super, dl); | |
851 | if (!e) | |
852 | return rv; | |
853 | for (i = 0; e[i].size; i++) | |
854 | continue; | |
855 | if (i > 0) | |
856 | rv = e[i-1].start + e[i-1].size; | |
857 | free(e); | |
858 | /* add the amount of space needed for metadata */ | |
859 | rv = rv + MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
860 | return rv * 512; | |
861 | } | |
862 | ||
1799c9e8 | 863 | #ifndef MDASSEMBLE |
1e5c6983 DW |
864 | static __u64 blocks_per_migr_unit(struct imsm_dev *dev); |
865 | ||
44470971 | 866 | static void print_imsm_dev(struct imsm_dev *dev, char *uuid, int disk_idx) |
cdddbdbc DW |
867 | { |
868 | __u64 sz; | |
0d80bb2f | 869 | int slot, i; |
a965f303 | 870 | struct imsm_map *map = get_imsm_map(dev, 0); |
dd8bcb3b | 871 | struct imsm_map *map2 = get_imsm_map(dev, 1); |
b10b37b8 | 872 | __u32 ord; |
cdddbdbc DW |
873 | |
874 | printf("\n"); | |
1e7bc0ed | 875 | printf("[%.16s]:\n", dev->volume); |
44470971 | 876 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
877 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
878 | if (map2) | |
879 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
880 | printf("\n"); | |
881 | printf(" Members : %d", map->num_members); | |
882 | if (map2) | |
883 | printf(" <-- %d", map2->num_members); | |
884 | printf("\n"); | |
0d80bb2f DW |
885 | printf(" Slots : ["); |
886 | for (i = 0; i < map->num_members; i++) { | |
dd8bcb3b | 887 | ord = get_imsm_ord_tbl_ent(dev, i, 0); |
0d80bb2f DW |
888 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
889 | } | |
dd8bcb3b AK |
890 | printf("]"); |
891 | if (map2) { | |
892 | printf(" <-- ["); | |
893 | for (i = 0; i < map2->num_members; i++) { | |
894 | ord = get_imsm_ord_tbl_ent(dev, i, 1); | |
895 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); | |
896 | } | |
897 | printf("]"); | |
898 | } | |
899 | printf("\n"); | |
7095bccb AK |
900 | printf(" Failed disk : "); |
901 | if (map->failed_disk_num == 0xff) | |
902 | printf("none"); | |
903 | else | |
904 | printf("%i", map->failed_disk_num); | |
905 | printf("\n"); | |
620b1713 DW |
906 | slot = get_imsm_disk_slot(map, disk_idx); |
907 | if (slot >= 0) { | |
98130f40 | 908 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
b10b37b8 DW |
909 | printf(" This Slot : %d%s\n", slot, |
910 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
911 | } else | |
cdddbdbc DW |
912 | printf(" This Slot : ?\n"); |
913 | sz = __le32_to_cpu(dev->size_high); | |
914 | sz <<= 32; | |
915 | sz += __le32_to_cpu(dev->size_low); | |
916 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
917 | human_size(sz * 512)); | |
918 | sz = __le32_to_cpu(map->blocks_per_member); | |
919 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
920 | human_size(sz * 512)); | |
921 | printf(" Sector Offset : %u\n", | |
922 | __le32_to_cpu(map->pba_of_lba0)); | |
923 | printf(" Num Stripes : %u\n", | |
924 | __le32_to_cpu(map->num_data_stripes)); | |
dd8bcb3b | 925 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 926 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
927 | if (map2) |
928 | printf(" <-- %u KiB", | |
929 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
930 | printf("\n"); | |
cdddbdbc | 931 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 932 | printf(" Migrate State : "); |
1484e727 DW |
933 | if (dev->vol.migr_state) { |
934 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 935 | printf("initialize\n"); |
1484e727 | 936 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 937 | printf("rebuild\n"); |
1484e727 | 938 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 939 | printf("check\n"); |
1484e727 | 940 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 941 | printf("general migration\n"); |
1484e727 | 942 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 943 | printf("state change\n"); |
1484e727 | 944 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 945 | printf("repair\n"); |
1484e727 | 946 | else |
8655a7b1 DW |
947 | printf("<unknown:%d>\n", migr_type(dev)); |
948 | } else | |
949 | printf("idle\n"); | |
3393c6af DW |
950 | printf(" Map State : %s", map_state_str[map->map_state]); |
951 | if (dev->vol.migr_state) { | |
952 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1e5c6983 | 953 | |
b10b37b8 | 954 | printf(" <-- %s", map_state_str[map->map_state]); |
1e5c6983 DW |
955 | printf("\n Checkpoint : %u (%llu)", |
956 | __le32_to_cpu(dev->vol.curr_migr_unit), | |
94fcb80a | 957 | (unsigned long long)blocks_per_migr_unit(dev)); |
3393c6af DW |
958 | } |
959 | printf("\n"); | |
cdddbdbc | 960 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
961 | } |
962 | ||
14e8215b | 963 | static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved) |
cdddbdbc | 964 | { |
949c47a0 | 965 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); |
1f24f035 | 966 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
967 | __u64 sz; |
968 | ||
d362da3d | 969 | if (index < 0 || !disk) |
e9d82038 DW |
970 | return; |
971 | ||
cdddbdbc | 972 | printf("\n"); |
1f24f035 | 973 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
cdddbdbc | 974 | printf(" Disk%02d Serial : %s\n", index, str); |
25ed7e59 DW |
975 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
976 | is_configured(disk) ? " active" : "", | |
977 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 978 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
14e8215b | 979 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
980 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
981 | human_size(sz * 512)); | |
982 | } | |
983 | ||
a5d85af7 | 984 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 985 | |
cdddbdbc DW |
986 | static void examine_super_imsm(struct supertype *st, char *homehost) |
987 | { | |
988 | struct intel_super *super = st->sb; | |
949c47a0 | 989 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
990 | char str[MAX_SIGNATURE_LENGTH]; |
991 | int i; | |
27fd6274 DW |
992 | struct mdinfo info; |
993 | char nbuf[64]; | |
cdddbdbc | 994 | __u32 sum; |
14e8215b | 995 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 996 | struct dl *dl; |
27fd6274 | 997 | |
cdddbdbc DW |
998 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
999 | printf(" Magic : %s\n", str); | |
1000 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1001 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1002 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1003 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1004 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
a5d85af7 | 1005 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1006 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1007 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1008 | sum = __le32_to_cpu(mpb->check_sum); |
1009 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1010 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1011 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1012 | printf(" Disks : %d\n", mpb->num_disks); |
1013 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
14e8215b | 1014 | print_imsm_disk(mpb, super->disks->index, reserved); |
604b746f JD |
1015 | if (super->bbm_log) { |
1016 | struct bbm_log *log = super->bbm_log; | |
1017 | ||
1018 | printf("\n"); | |
1019 | printf("Bad Block Management Log:\n"); | |
1020 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1021 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1022 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1023 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1024 | printf(" First Spare : %llx\n", |
1025 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1026 | } |
44470971 DW |
1027 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1028 | struct mdinfo info; | |
1029 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1030 | ||
1031 | super->current_vol = i; | |
a5d85af7 | 1032 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1033 | fname_from_uuid(st, &info, nbuf, ':'); |
44470971 DW |
1034 | print_imsm_dev(dev, nbuf + 5, super->disks->index); |
1035 | } | |
cdddbdbc DW |
1036 | for (i = 0; i < mpb->num_disks; i++) { |
1037 | if (i == super->disks->index) | |
1038 | continue; | |
14e8215b | 1039 | print_imsm_disk(mpb, i, reserved); |
cdddbdbc | 1040 | } |
94827db3 N |
1041 | for (dl = super->disks ; dl; dl = dl->next) { |
1042 | struct imsm_disk *disk; | |
1043 | char str[MAX_RAID_SERIAL_LEN + 1]; | |
1044 | __u64 sz; | |
1045 | ||
1046 | if (dl->index >= 0) | |
1047 | continue; | |
1048 | ||
1049 | disk = &dl->disk; | |
1050 | printf("\n"); | |
1051 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); | |
1052 | printf(" Disk Serial : %s\n", str); | |
1053 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", | |
1054 | is_configured(disk) ? " active" : "", | |
1055 | is_failed(disk) ? " failed" : ""); | |
1056 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
1057 | sz = __le32_to_cpu(disk->total_blocks) - reserved; | |
1058 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, | |
1059 | human_size(sz * 512)); | |
1060 | } | |
cdddbdbc DW |
1061 | } |
1062 | ||
061f2c6a | 1063 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1064 | { |
27fd6274 | 1065 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1066 | struct mdinfo info; |
1067 | char nbuf[64]; | |
1e7bc0ed | 1068 | struct intel_super *super = st->sb; |
1e7bc0ed | 1069 | |
0d5a423f DW |
1070 | if (!super->anchor->num_raid_devs) { |
1071 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1072 | return; |
0d5a423f | 1073 | } |
ff54de6e | 1074 | |
a5d85af7 | 1075 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1076 | fname_from_uuid(st, &info, nbuf, ':'); |
1077 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1078 | } | |
1079 | ||
1080 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1081 | { | |
1082 | /* We just write a generic IMSM ARRAY entry */ | |
1083 | struct mdinfo info; | |
1084 | char nbuf[64]; | |
1085 | char nbuf1[64]; | |
1086 | struct intel_super *super = st->sb; | |
1087 | int i; | |
1088 | ||
1089 | if (!super->anchor->num_raid_devs) | |
1090 | return; | |
1091 | ||
a5d85af7 | 1092 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1093 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1094 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1095 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1096 | ||
1097 | super->current_vol = i; | |
a5d85af7 | 1098 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1099 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1100 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1101 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1102 | } |
cdddbdbc DW |
1103 | } |
1104 | ||
9d84c8ea DW |
1105 | static void export_examine_super_imsm(struct supertype *st) |
1106 | { | |
1107 | struct intel_super *super = st->sb; | |
1108 | struct imsm_super *mpb = super->anchor; | |
1109 | struct mdinfo info; | |
1110 | char nbuf[64]; | |
1111 | ||
a5d85af7 | 1112 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1113 | fname_from_uuid(st, &info, nbuf, ':'); |
1114 | printf("MD_METADATA=imsm\n"); | |
1115 | printf("MD_LEVEL=container\n"); | |
1116 | printf("MD_UUID=%s\n", nbuf+5); | |
1117 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1118 | } | |
1119 | ||
cdddbdbc DW |
1120 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1121 | { | |
3ebe00a1 DW |
1122 | struct mdinfo info; |
1123 | char nbuf[64]; | |
1124 | ||
a5d85af7 | 1125 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1126 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1127 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1128 | } |
1129 | ||
1130 | static void brief_detail_super_imsm(struct supertype *st) | |
1131 | { | |
ff54de6e N |
1132 | struct mdinfo info; |
1133 | char nbuf[64]; | |
a5d85af7 | 1134 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1135 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1136 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1137 | } |
d665cc31 DW |
1138 | |
1139 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1140 | static void fd2devname(int fd, char *name); | |
1141 | ||
120dc887 | 1142 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1143 | { |
120dc887 LM |
1144 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1145 | * controller, as well as non-connected ports | |
d665cc31 DW |
1146 | */ |
1147 | int hba_len = strlen(hba_path) + 1; | |
1148 | struct dirent *ent; | |
1149 | DIR *dir; | |
1150 | char *path = NULL; | |
1151 | int err = 0; | |
1152 | unsigned long port_mask = (1 << port_count) - 1; | |
1153 | ||
f21e18ca | 1154 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1155 | if (verbose) |
1156 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1157 | return 2; | |
1158 | } | |
1159 | ||
1160 | /* scroll through /sys/dev/block looking for devices attached to | |
1161 | * this hba | |
1162 | */ | |
1163 | dir = opendir("/sys/dev/block"); | |
1164 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1165 | int fd; | |
1166 | char model[64]; | |
1167 | char vendor[64]; | |
1168 | char buf[1024]; | |
1169 | int major, minor; | |
1170 | char *device; | |
1171 | char *c; | |
1172 | int port; | |
1173 | int type; | |
1174 | ||
1175 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1176 | continue; | |
1177 | path = devt_to_devpath(makedev(major, minor)); | |
1178 | if (!path) | |
1179 | continue; | |
1180 | if (!path_attached_to_hba(path, hba_path)) { | |
1181 | free(path); | |
1182 | path = NULL; | |
1183 | continue; | |
1184 | } | |
1185 | ||
1186 | /* retrieve the scsi device type */ | |
1187 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1188 | if (verbose) | |
1189 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1190 | err = 2; | |
1191 | break; | |
1192 | } | |
1193 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1194 | if (load_sys(device, buf) != 0) { | |
1195 | if (verbose) | |
1196 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1197 | path); | |
1198 | err = 2; | |
1199 | free(device); | |
1200 | break; | |
1201 | } | |
1202 | type = strtoul(buf, NULL, 10); | |
1203 | ||
1204 | /* if it's not a disk print the vendor and model */ | |
1205 | if (!(type == 0 || type == 7 || type == 14)) { | |
1206 | vendor[0] = '\0'; | |
1207 | model[0] = '\0'; | |
1208 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1209 | if (load_sys(device, buf) == 0) { | |
1210 | strncpy(vendor, buf, sizeof(vendor)); | |
1211 | vendor[sizeof(vendor) - 1] = '\0'; | |
1212 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1213 | while (isspace(*c) || *c == '\0') | |
1214 | *c-- = '\0'; | |
1215 | ||
1216 | } | |
1217 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1218 | if (load_sys(device, buf) == 0) { | |
1219 | strncpy(model, buf, sizeof(model)); | |
1220 | model[sizeof(model) - 1] = '\0'; | |
1221 | c = (char *) &model[sizeof(model) - 1]; | |
1222 | while (isspace(*c) || *c == '\0') | |
1223 | *c-- = '\0'; | |
1224 | } | |
1225 | ||
1226 | if (vendor[0] && model[0]) | |
1227 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1228 | else | |
1229 | switch (type) { /* numbers from hald/linux/device.c */ | |
1230 | case 1: sprintf(buf, "tape"); break; | |
1231 | case 2: sprintf(buf, "printer"); break; | |
1232 | case 3: sprintf(buf, "processor"); break; | |
1233 | case 4: | |
1234 | case 5: sprintf(buf, "cdrom"); break; | |
1235 | case 6: sprintf(buf, "scanner"); break; | |
1236 | case 8: sprintf(buf, "media_changer"); break; | |
1237 | case 9: sprintf(buf, "comm"); break; | |
1238 | case 12: sprintf(buf, "raid"); break; | |
1239 | default: sprintf(buf, "unknown"); | |
1240 | } | |
1241 | } else | |
1242 | buf[0] = '\0'; | |
1243 | free(device); | |
1244 | ||
1245 | /* chop device path to 'host%d' and calculate the port number */ | |
1246 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1247 | if (!c) { |
1248 | if (verbose) | |
1249 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1250 | err = 2; | |
1251 | break; | |
1252 | } | |
d665cc31 DW |
1253 | *c = '\0'; |
1254 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1255 | port -= host_base; | |
1256 | else { | |
1257 | if (verbose) { | |
1258 | *c = '/'; /* repair the full string */ | |
1259 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1260 | path); | |
1261 | } | |
1262 | err = 2; | |
1263 | break; | |
1264 | } | |
1265 | ||
1266 | /* mark this port as used */ | |
1267 | port_mask &= ~(1 << port); | |
1268 | ||
1269 | /* print out the device information */ | |
1270 | if (buf[0]) { | |
1271 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1272 | continue; | |
1273 | } | |
1274 | ||
1275 | fd = dev_open(ent->d_name, O_RDONLY); | |
1276 | if (fd < 0) | |
1277 | printf(" Port%d : - disk info unavailable -\n", port); | |
1278 | else { | |
1279 | fd2devname(fd, buf); | |
1280 | printf(" Port%d : %s", port, buf); | |
1281 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
1282 | printf(" (%s)\n", buf); | |
1283 | else | |
1284 | printf("()\n"); | |
1285 | } | |
1286 | close(fd); | |
1287 | free(path); | |
1288 | path = NULL; | |
1289 | } | |
1290 | if (path) | |
1291 | free(path); | |
1292 | if (dir) | |
1293 | closedir(dir); | |
1294 | if (err == 0) { | |
1295 | int i; | |
1296 | ||
1297 | for (i = 0; i < port_count; i++) | |
1298 | if (port_mask & (1 << i)) | |
1299 | printf(" Port%d : - no device attached -\n", i); | |
1300 | } | |
1301 | ||
1302 | return err; | |
1303 | } | |
1304 | ||
120dc887 | 1305 | |
155cbb4c | 1306 | |
120dc887 LM |
1307 | static void print_found_intel_controllers(struct sys_dev *elem) |
1308 | { | |
1309 | for (; elem; elem = elem->next) { | |
1310 | fprintf(stderr, Name ": found Intel(R) "); | |
1311 | if (elem->type == SYS_DEV_SATA) | |
1312 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1313 | else if (elem->type == SYS_DEV_SAS) |
1314 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1315 | fprintf(stderr, "RAID controller"); |
1316 | if (elem->pci_id) | |
1317 | fprintf(stderr, " at %s", elem->pci_id); | |
1318 | fprintf(stderr, ".\n"); | |
1319 | } | |
1320 | fflush(stderr); | |
1321 | } | |
1322 | ||
120dc887 LM |
1323 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1324 | { | |
1325 | struct dirent *ent; | |
1326 | DIR *dir; | |
1327 | int host_base = -1; | |
1328 | ||
1329 | *port_count = 0; | |
1330 | if ((dir = opendir(hba_path)) == NULL) | |
1331 | return -1; | |
1332 | ||
1333 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1334 | int host; | |
1335 | ||
1336 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1337 | continue; | |
1338 | if (*port_count == 0) | |
1339 | host_base = host; | |
1340 | else if (host < host_base) | |
1341 | host_base = host; | |
1342 | ||
1343 | if (host + 1 > *port_count + host_base) | |
1344 | *port_count = host + 1 - host_base; | |
1345 | } | |
1346 | closedir(dir); | |
1347 | return host_base; | |
1348 | } | |
1349 | ||
5615172f | 1350 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1351 | { |
1352 | /* There are two components to imsm platform support, the ahci SATA | |
1353 | * controller and the option-rom. To find the SATA controller we | |
1354 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1355 | * controller with the Intel vendor id is present. This approach | |
1356 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1357 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1358 | * detect platform raid capabilities. The option-rom resides in a | |
1359 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1360 | * platform capabilities. If raid support is disabled in the BIOS the | |
1361 | * option-rom capability structure will not be available. | |
1362 | */ | |
1363 | const struct imsm_orom *orom; | |
1364 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1365 | int host_base = 0; |
1366 | int port_count = 0; | |
120dc887 | 1367 | int result=0; |
d665cc31 | 1368 | |
5615172f DW |
1369 | if (enumerate_only) { |
1370 | if (check_env("IMSM_NO_PLATFORM") || find_imsm_orom()) | |
1371 | return 0; | |
1372 | return 2; | |
1373 | } | |
1374 | ||
155cbb4c LM |
1375 | list = find_intel_devices(); |
1376 | if (!list) { | |
d665cc31 | 1377 | if (verbose) |
155cbb4c LM |
1378 | fprintf(stderr, Name ": no active Intel(R) RAID " |
1379 | "controller found.\n"); | |
d665cc31 DW |
1380 | free_sys_dev(&list); |
1381 | return 2; | |
1382 | } else if (verbose) | |
155cbb4c | 1383 | print_found_intel_controllers(list); |
d665cc31 DW |
1384 | |
1385 | orom = find_imsm_orom(); | |
1386 | if (!orom) { | |
155cbb4c | 1387 | free_sys_dev(&list); |
d665cc31 DW |
1388 | if (verbose) |
1389 | fprintf(stderr, Name ": imsm option-rom not found\n"); | |
1390 | return 2; | |
1391 | } | |
1392 | ||
1393 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1394 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1395 | orom->hotfix_ver, orom->build); | |
1396 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1397 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1398 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1399 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1400 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1401 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
8be094f0 DW |
1402 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", |
1403 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1404 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1405 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1406 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1407 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1408 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1409 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1410 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1411 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1412 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1413 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1414 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1415 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1416 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1417 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1418 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
d665cc31 DW |
1419 | printf(" Max Disks : %d\n", orom->tds); |
1420 | printf(" Max Volumes : %d\n", orom->vpa); | |
d665cc31 | 1421 | |
120dc887 LM |
1422 | for (hba = list; hba; hba = hba->next) { |
1423 | printf(" I/O Controller : %s (%s)\n", | |
1424 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1425 | |
120dc887 LM |
1426 | if (hba->type == SYS_DEV_SATA) { |
1427 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1428 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1429 | if (verbose) | |
1430 | fprintf(stderr, Name ": failed to enumerate " | |
1431 | "ports on SATA controller at %s.", hba->pci_id); | |
1432 | result |= 2; | |
1433 | } | |
155cbb4c LM |
1434 | } else if (hba->type == SYS_DEV_SAS) { |
1435 | if (verbose) | |
1436 | fprintf(stderr, Name ": failed to enumerate " | |
1437 | "devices on SAS controller at %s.", hba->pci_id); | |
1438 | result |= 2; | |
120dc887 | 1439 | } |
d665cc31 | 1440 | } |
155cbb4c | 1441 | |
120dc887 LM |
1442 | free_sys_dev(&list); |
1443 | return result; | |
d665cc31 | 1444 | } |
cdddbdbc DW |
1445 | #endif |
1446 | ||
1447 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1448 | { | |
5115ca67 DW |
1449 | /* the imsm metadata format does not specify any host |
1450 | * identification information. We return -1 since we can never | |
1451 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1452 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1453 | * exclude member disks that do not belong, and we rely on |
1454 | * mdadm.conf to specify the arrays that should be assembled. | |
1455 | * Auto-assembly may still pick up "foreign" arrays. | |
1456 | */ | |
cdddbdbc | 1457 | |
9362c1c8 | 1458 | return -1; |
cdddbdbc DW |
1459 | } |
1460 | ||
1461 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1462 | { | |
51006d85 N |
1463 | /* The uuid returned here is used for: |
1464 | * uuid to put into bitmap file (Create, Grow) | |
1465 | * uuid for backup header when saving critical section (Grow) | |
1466 | * comparing uuids when re-adding a device into an array | |
1467 | * In these cases the uuid required is that of the data-array, | |
1468 | * not the device-set. | |
1469 | * uuid to recognise same set when adding a missing device back | |
1470 | * to an array. This is a uuid for the device-set. | |
1471 | * | |
1472 | * For each of these we can make do with a truncated | |
1473 | * or hashed uuid rather than the original, as long as | |
1474 | * everyone agrees. | |
1475 | * In each case the uuid required is that of the data-array, | |
1476 | * not the device-set. | |
43dad3d6 | 1477 | */ |
51006d85 N |
1478 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1479 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1480 | * - the orig_family_num of the container |
51006d85 N |
1481 | * - the index number of the volume |
1482 | * - the 'serial' number of the volume. | |
1483 | * Hopefully these are all constant. | |
1484 | */ | |
1485 | struct intel_super *super = st->sb; | |
43dad3d6 | 1486 | |
51006d85 N |
1487 | char buf[20]; |
1488 | struct sha1_ctx ctx; | |
1489 | struct imsm_dev *dev = NULL; | |
148acb7b | 1490 | __u32 family_num; |
51006d85 | 1491 | |
148acb7b DW |
1492 | /* some mdadm versions failed to set ->orig_family_num, in which |
1493 | * case fall back to ->family_num. orig_family_num will be | |
1494 | * fixed up with the first metadata update. | |
1495 | */ | |
1496 | family_num = super->anchor->orig_family_num; | |
1497 | if (family_num == 0) | |
1498 | family_num = super->anchor->family_num; | |
51006d85 | 1499 | sha1_init_ctx(&ctx); |
92bd8f8d | 1500 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1501 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1502 | if (super->current_vol >= 0) |
1503 | dev = get_imsm_dev(super, super->current_vol); | |
1504 | if (dev) { | |
1505 | __u32 vol = super->current_vol; | |
1506 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1507 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1508 | } | |
1509 | sha1_finish_ctx(&ctx, buf); | |
1510 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1511 | } |
1512 | ||
0d481d37 | 1513 | #if 0 |
4f5bc454 DW |
1514 | static void |
1515 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1516 | { |
cdddbdbc DW |
1517 | __u8 *v = get_imsm_version(mpb); |
1518 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1519 | char major[] = { 0, 0, 0 }; | |
1520 | char minor[] = { 0 ,0, 0 }; | |
1521 | char patch[] = { 0, 0, 0 }; | |
1522 | char *ver_parse[] = { major, minor, patch }; | |
1523 | int i, j; | |
1524 | ||
1525 | i = j = 0; | |
1526 | while (*v != '\0' && v < end) { | |
1527 | if (*v != '.' && j < 2) | |
1528 | ver_parse[i][j++] = *v; | |
1529 | else { | |
1530 | i++; | |
1531 | j = 0; | |
1532 | } | |
1533 | v++; | |
1534 | } | |
1535 | ||
4f5bc454 DW |
1536 | *m = strtol(minor, NULL, 0); |
1537 | *p = strtol(patch, NULL, 0); | |
1538 | } | |
0d481d37 | 1539 | #endif |
4f5bc454 | 1540 | |
1e5c6983 DW |
1541 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1542 | { | |
1543 | /* migr_strip_size when repairing or initializing parity */ | |
1544 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1545 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1546 | ||
1547 | switch (get_imsm_raid_level(map)) { | |
1548 | case 5: | |
1549 | case 10: | |
1550 | return chunk; | |
1551 | default: | |
1552 | return 128*1024 >> 9; | |
1553 | } | |
1554 | } | |
1555 | ||
1556 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1557 | { | |
1558 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1559 | * this is different than migr_strip_size_resync(), but it's good | |
1560 | * to be compatible | |
1561 | */ | |
1562 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1563 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1564 | ||
1565 | switch (get_imsm_raid_level(map)) { | |
1566 | case 1: | |
1567 | case 10: | |
1568 | if (map->num_members % map->num_domains == 0) | |
1569 | return 128*1024 >> 9; | |
1570 | else | |
1571 | return chunk; | |
1572 | case 5: | |
1573 | return max((__u32) 64*1024 >> 9, chunk); | |
1574 | default: | |
1575 | return 128*1024 >> 9; | |
1576 | } | |
1577 | } | |
1578 | ||
1579 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
1580 | { | |
1581 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1582 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1583 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); | |
1584 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
1585 | ||
1586 | return max((__u32) 1, hi_chunk / lo_chunk); | |
1587 | } | |
1588 | ||
1589 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
1590 | { | |
1591 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1592 | int level = get_imsm_raid_level(lo); | |
1593 | ||
1594 | if (level == 1 || level == 10) { | |
1595 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1596 | ||
1597 | return hi->num_domains; | |
1598 | } else | |
1599 | return num_stripes_per_unit_resync(dev); | |
1600 | } | |
1601 | ||
98130f40 | 1602 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
1603 | { |
1604 | /* named 'imsm_' because raid0, raid1 and raid10 | |
1605 | * counter-intuitively have the same number of data disks | |
1606 | */ | |
98130f40 | 1607 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
1608 | |
1609 | switch (get_imsm_raid_level(map)) { | |
1610 | case 0: | |
1611 | case 1: | |
1612 | case 10: | |
1613 | return map->num_members; | |
1614 | case 5: | |
1615 | return map->num_members - 1; | |
1616 | default: | |
1617 | dprintf("%s: unsupported raid level\n", __func__); | |
1618 | return 0; | |
1619 | } | |
1620 | } | |
1621 | ||
1622 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
1623 | { | |
1624 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1625 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1626 | ||
1627 | switch(get_imsm_raid_level(map)) { | |
1628 | case 1: | |
1629 | case 10: | |
1630 | return chunk * map->num_domains; | |
1631 | case 5: | |
1632 | return chunk * map->num_members; | |
1633 | default: | |
1634 | return chunk; | |
1635 | } | |
1636 | } | |
1637 | ||
1638 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
1639 | { | |
1640 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1641 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1642 | __u32 strip = block / chunk; | |
1643 | ||
1644 | switch (get_imsm_raid_level(map)) { | |
1645 | case 1: | |
1646 | case 10: { | |
1647 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
1648 | __u32 vol_stripe = vol_strip / map->num_members; | |
1649 | ||
1650 | return vol_stripe * chunk + block % chunk; | |
1651 | } case 5: { | |
1652 | __u32 stripe = strip / (map->num_members - 1); | |
1653 | ||
1654 | return stripe * chunk + block % chunk; | |
1655 | } | |
1656 | default: | |
1657 | return 0; | |
1658 | } | |
1659 | } | |
1660 | ||
1661 | static __u64 blocks_per_migr_unit(struct imsm_dev *dev) | |
1662 | { | |
1663 | /* calculate the conversion factor between per member 'blocks' | |
1664 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
1665 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
1666 | */ | |
1667 | if (!dev->vol.migr_state) | |
1668 | return 0; | |
1669 | ||
1670 | switch (migr_type(dev)) { | |
6345120e | 1671 | case MIGR_GEN_MIGR: |
1e5c6983 DW |
1672 | case MIGR_VERIFY: |
1673 | case MIGR_REPAIR: | |
1674 | case MIGR_INIT: { | |
1675 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1676 | __u32 stripes_per_unit; | |
1677 | __u32 blocks_per_unit; | |
1678 | __u32 parity_depth; | |
1679 | __u32 migr_chunk; | |
1680 | __u32 block_map; | |
1681 | __u32 block_rel; | |
1682 | __u32 segment; | |
1683 | __u32 stripe; | |
1684 | __u8 disks; | |
1685 | ||
1686 | /* yes, this is really the translation of migr_units to | |
1687 | * per-member blocks in the 'resync' case | |
1688 | */ | |
1689 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
1690 | migr_chunk = migr_strip_blocks_resync(dev); | |
98130f40 | 1691 | disks = imsm_num_data_members(dev, 0); |
1e5c6983 DW |
1692 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
1693 | stripe = __le32_to_cpu(map->blocks_per_strip) * disks; | |
1694 | segment = blocks_per_unit / stripe; | |
1695 | block_rel = blocks_per_unit - segment * stripe; | |
1696 | parity_depth = parity_segment_depth(dev); | |
1697 | block_map = map_migr_block(dev, block_rel); | |
1698 | return block_map + parity_depth * segment; | |
1699 | } | |
1700 | case MIGR_REBUILD: { | |
1701 | __u32 stripes_per_unit; | |
1702 | __u32 migr_chunk; | |
1703 | ||
1704 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
1705 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
1706 | return migr_chunk * stripes_per_unit; | |
1707 | } | |
1e5c6983 DW |
1708 | case MIGR_STATE_CHANGE: |
1709 | default: | |
1710 | return 0; | |
1711 | } | |
1712 | } | |
1713 | ||
c2c087e6 DW |
1714 | static int imsm_level_to_layout(int level) |
1715 | { | |
1716 | switch (level) { | |
1717 | case 0: | |
1718 | case 1: | |
1719 | return 0; | |
1720 | case 5: | |
1721 | case 6: | |
a380c027 | 1722 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 1723 | case 10: |
c92a2527 | 1724 | return 0x102; |
c2c087e6 | 1725 | } |
a18a888e | 1726 | return UnSet; |
c2c087e6 DW |
1727 | } |
1728 | ||
a5d85af7 | 1729 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
1730 | { |
1731 | struct intel_super *super = st->sb; | |
949c47a0 | 1732 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 1733 | struct imsm_map *map = get_imsm_map(dev, 0); |
81ac8b4d | 1734 | struct imsm_map *prev_map = get_imsm_map(dev, 1); |
b335e593 | 1735 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 1736 | struct dl *dl; |
e207da2f | 1737 | char *devname; |
a5d85af7 | 1738 | int map_disks = info->array.raid_disks; |
bf5a934a | 1739 | |
b335e593 AK |
1740 | if (prev_map) |
1741 | map_to_analyse = prev_map; | |
1742 | ||
efb30e7f DW |
1743 | for (dl = super->disks; dl; dl = dl->next) |
1744 | if (dl->raiddisk == info->disk.raid_disk) | |
1745 | break; | |
bf5a934a | 1746 | info->container_member = super->current_vol; |
b335e593 AK |
1747 | info->array.raid_disks = map_to_analyse->num_members; |
1748 | info->array.level = get_imsm_raid_level(map_to_analyse); | |
bf5a934a DW |
1749 | info->array.layout = imsm_level_to_layout(info->array.level); |
1750 | info->array.md_minor = -1; | |
1751 | info->array.ctime = 0; | |
1752 | info->array.utime = 0; | |
b335e593 AK |
1753 | info->array.chunk_size = |
1754 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 1755 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
1756 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
1757 | info->custom_array_size <<= 32; | |
1758 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
b335e593 AK |
1759 | if (prev_map) { |
1760 | info->new_level = get_imsm_raid_level(map); | |
1761 | info->new_layout = imsm_level_to_layout(info->new_level); | |
1762 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
1763 | } else { | |
1764 | info->new_level = UnSet; | |
1765 | info->new_layout = UnSet; | |
1766 | info->new_chunk = info->array.chunk_size; | |
1767 | } | |
301406c9 DW |
1768 | info->disk.major = 0; |
1769 | info->disk.minor = 0; | |
efb30e7f DW |
1770 | if (dl) { |
1771 | info->disk.major = dl->major; | |
1772 | info->disk.minor = dl->minor; | |
1773 | } | |
bf5a934a | 1774 | |
b335e593 AK |
1775 | info->data_offset = __le32_to_cpu(map_to_analyse->pba_of_lba0); |
1776 | info->component_size = | |
1777 | __le32_to_cpu(map_to_analyse->blocks_per_member); | |
301406c9 | 1778 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 1779 | info->recovery_start = MaxSector; |
21f3b9cc CA |
1780 | info->reshape_active = (prev_map != NULL) && |
1781 | (map->map_state == prev_map->map_state); | |
81ac8b4d AK |
1782 | if (info->reshape_active) |
1783 | info->delta_disks = map->num_members - prev_map->num_members; | |
1784 | else | |
1785 | info->delta_disks = 0; | |
bf5a934a | 1786 | |
d2e6d5d6 | 1787 | info->reshape_progress = 0; |
b335e593 AK |
1788 | if (map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
1789 | dev->vol.dirty) { | |
301406c9 | 1790 | info->resync_start = 0; |
1e5c6983 DW |
1791 | } else if (dev->vol.migr_state) { |
1792 | switch (migr_type(dev)) { | |
1793 | case MIGR_REPAIR: | |
1794 | case MIGR_INIT: { | |
1795 | __u64 blocks_per_unit = blocks_per_migr_unit(dev); | |
1796 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); | |
1797 | ||
1798 | info->resync_start = blocks_per_unit * units; | |
1799 | break; | |
1800 | } | |
d2e6d5d6 AK |
1801 | case MIGR_GEN_MIGR: { |
1802 | __u64 blocks_per_unit = blocks_per_migr_unit(dev); | |
1803 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); | |
1804 | ||
1805 | info->reshape_progress = blocks_per_unit * units; | |
1806 | dprintf("IMSM: General Migration checkpoint : %llu " | |
1807 | "(%llu) -> read reshape progress : %llu\n", | |
1808 | units, blocks_per_unit, info->reshape_progress); | |
75156c46 AK |
1809 | unsigned long long array_blocks; |
1810 | int used_disks; | |
1811 | ||
1812 | used_disks = imsm_num_data_members(dev, 1); | |
1813 | if (used_disks > 0) { | |
1814 | array_blocks = map->blocks_per_member * | |
1815 | used_disks; | |
1816 | /* round array size down to closest MB | |
1817 | */ | |
1818 | info->custom_array_size = (array_blocks | |
1819 | >> SECT_PER_MB_SHIFT) | |
1820 | << SECT_PER_MB_SHIFT; | |
1821 | } | |
d2e6d5d6 | 1822 | } |
1e5c6983 DW |
1823 | case MIGR_VERIFY: |
1824 | /* we could emulate the checkpointing of | |
1825 | * 'sync_action=check' migrations, but for now | |
1826 | * we just immediately complete them | |
1827 | */ | |
1828 | case MIGR_REBUILD: | |
1829 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
1830 | case MIGR_STATE_CHANGE: |
1831 | /* FIXME handle other migrations */ | |
1832 | default: | |
1833 | /* we are not dirty, so... */ | |
1834 | info->resync_start = MaxSector; | |
1835 | } | |
1836 | } else | |
b7528a20 | 1837 | info->resync_start = MaxSector; |
301406c9 DW |
1838 | |
1839 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
1840 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 1841 | |
f35f2525 N |
1842 | info->array.major_version = -1; |
1843 | info->array.minor_version = -2; | |
e207da2f AW |
1844 | devname = devnum2devname(st->container_dev); |
1845 | *info->text_version = '\0'; | |
1846 | if (devname) | |
1847 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
1848 | free(devname); | |
a67dd8cc | 1849 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 1850 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
1851 | |
1852 | if (dmap) { | |
1853 | int i, j; | |
1854 | for (i=0; i<map_disks; i++) { | |
1855 | dmap[i] = 0; | |
1856 | if (i < info->array.raid_disks) { | |
1857 | struct imsm_disk *dsk; | |
98130f40 | 1858 | j = get_imsm_disk_idx(dev, i, -1); |
a5d85af7 N |
1859 | dsk = get_imsm_disk(super, j); |
1860 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
1861 | dmap[i] = 1; | |
1862 | } | |
1863 | } | |
1864 | } | |
81ac8b4d | 1865 | } |
bf5a934a | 1866 | |
97b4d0e9 DW |
1867 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed); |
1868 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev); | |
1869 | ||
1870 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
1871 | { | |
1872 | struct dl *d; | |
1873 | ||
1874 | for (d = super->missing; d; d = d->next) | |
1875 | if (d->index == index) | |
1876 | return &d->disk; | |
1877 | return NULL; | |
1878 | } | |
1879 | ||
a5d85af7 | 1880 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
1881 | { |
1882 | struct intel_super *super = st->sb; | |
4f5bc454 | 1883 | struct imsm_disk *disk; |
a5d85af7 | 1884 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
1885 | int max_enough = -1; |
1886 | int i; | |
1887 | struct imsm_super *mpb; | |
4f5bc454 | 1888 | |
bf5a934a | 1889 | if (super->current_vol >= 0) { |
a5d85af7 | 1890 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
1891 | return; |
1892 | } | |
d23fe947 DW |
1893 | |
1894 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
1895 | * spares | |
1896 | */ | |
1897 | info->array.raid_disks = 0; | |
cdddbdbc DW |
1898 | info->array.level = LEVEL_CONTAINER; |
1899 | info->array.layout = 0; | |
1900 | info->array.md_minor = -1; | |
c2c087e6 | 1901 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
1902 | info->array.utime = 0; |
1903 | info->array.chunk_size = 0; | |
1904 | ||
1905 | info->disk.major = 0; | |
1906 | info->disk.minor = 0; | |
cdddbdbc | 1907 | info->disk.raid_disk = -1; |
c2c087e6 | 1908 | info->reshape_active = 0; |
f35f2525 N |
1909 | info->array.major_version = -1; |
1910 | info->array.minor_version = -2; | |
c2c087e6 | 1911 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 1912 | info->safe_mode_delay = 0; |
c2c087e6 DW |
1913 | info->disk.number = -1; |
1914 | info->disk.state = 0; | |
c5afc314 | 1915 | info->name[0] = 0; |
921d9e16 | 1916 | info->recovery_start = MaxSector; |
c2c087e6 | 1917 | |
97b4d0e9 | 1918 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 1919 | mpb = super->anchor; |
97b4d0e9 | 1920 | |
ab3cb6b3 N |
1921 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1922 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1923 | int failed, enough, j, missing = 0; | |
1924 | struct imsm_map *map; | |
1925 | __u8 state; | |
97b4d0e9 | 1926 | |
ab3cb6b3 N |
1927 | failed = imsm_count_failed(super, dev); |
1928 | state = imsm_check_degraded(super, dev, failed); | |
1929 | map = get_imsm_map(dev, dev->vol.migr_state); | |
1930 | ||
1931 | /* any newly missing disks? | |
1932 | * (catches single-degraded vs double-degraded) | |
1933 | */ | |
1934 | for (j = 0; j < map->num_members; j++) { | |
98130f40 | 1935 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
ab3cb6b3 N |
1936 | __u32 idx = ord_to_idx(ord); |
1937 | ||
1938 | if (!(ord & IMSM_ORD_REBUILD) && | |
1939 | get_imsm_missing(super, idx)) { | |
1940 | missing = 1; | |
1941 | break; | |
1942 | } | |
97b4d0e9 | 1943 | } |
ab3cb6b3 N |
1944 | |
1945 | if (state == IMSM_T_STATE_FAILED) | |
1946 | enough = -1; | |
1947 | else if (state == IMSM_T_STATE_DEGRADED && | |
1948 | (state != map->map_state || missing)) | |
1949 | enough = 0; | |
1950 | else /* we're normal, or already degraded */ | |
1951 | enough = 1; | |
1952 | ||
1953 | /* in the missing/failed disk case check to see | |
1954 | * if at least one array is runnable | |
1955 | */ | |
1956 | max_enough = max(max_enough, enough); | |
1957 | } | |
1958 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
1959 | info->container_enough = max_enough; | |
97b4d0e9 | 1960 | |
4a04ec6c | 1961 | if (super->disks) { |
14e8215b DW |
1962 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
1963 | ||
b9f594fe | 1964 | disk = &super->disks->disk; |
14e8215b DW |
1965 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
1966 | info->component_size = reserved; | |
25ed7e59 | 1967 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
1968 | /* we don't change info->disk.raid_disk here because |
1969 | * this state will be finalized in mdmon after we have | |
1970 | * found the 'most fresh' version of the metadata | |
1971 | */ | |
25ed7e59 DW |
1972 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
1973 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 1974 | } |
a575e2a7 DW |
1975 | |
1976 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
1977 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
1978 | */ | |
1979 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 1980 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
1981 | else |
1982 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
1983 | |
1984 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
1985 | if (map) { | |
1986 | int i; | |
1987 | for (i = 0; i < map_disks; i++) | |
1988 | map[i] = 1; | |
1989 | } | |
1990 | ||
cdddbdbc DW |
1991 | } |
1992 | ||
5c4cd5da AC |
1993 | /* allocates memory and fills disk in mdinfo structure |
1994 | * for each disk in array */ | |
1995 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
1996 | { | |
1997 | struct mdinfo *mddev = NULL; | |
1998 | struct intel_super *super = st->sb; | |
1999 | struct imsm_disk *disk; | |
2000 | int count = 0; | |
2001 | struct dl *dl; | |
2002 | if (!super || !super->disks) | |
2003 | return NULL; | |
2004 | dl = super->disks; | |
2005 | mddev = malloc(sizeof(*mddev)); | |
2006 | if (!mddev) { | |
2007 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2008 | return NULL; | |
2009 | } | |
2010 | memset(mddev, 0, sizeof(*mddev)); | |
2011 | while (dl) { | |
2012 | struct mdinfo *tmp; | |
2013 | disk = &dl->disk; | |
2014 | tmp = malloc(sizeof(*tmp)); | |
2015 | if (!tmp) { | |
2016 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2017 | if (mddev) | |
2018 | sysfs_free(mddev); | |
2019 | return NULL; | |
2020 | } | |
2021 | memset(tmp, 0, sizeof(*tmp)); | |
2022 | if (mddev->devs) | |
2023 | tmp->next = mddev->devs; | |
2024 | mddev->devs = tmp; | |
2025 | tmp->disk.number = count++; | |
2026 | tmp->disk.major = dl->major; | |
2027 | tmp->disk.minor = dl->minor; | |
2028 | tmp->disk.state = is_configured(disk) ? | |
2029 | (1 << MD_DISK_ACTIVE) : 0; | |
2030 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2031 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2032 | tmp->disk.raid_disk = -1; | |
2033 | dl = dl->next; | |
2034 | } | |
2035 | return mddev; | |
2036 | } | |
2037 | ||
cdddbdbc DW |
2038 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2039 | char *update, char *devname, int verbose, | |
2040 | int uuid_set, char *homehost) | |
2041 | { | |
f352c545 DW |
2042 | /* For 'assemble' and 'force' we need to return non-zero if any |
2043 | * change was made. For others, the return value is ignored. | |
2044 | * Update options are: | |
2045 | * force-one : This device looks a bit old but needs to be included, | |
2046 | * update age info appropriately. | |
2047 | * assemble: clear any 'faulty' flag to allow this device to | |
2048 | * be assembled. | |
2049 | * force-array: Array is degraded but being forced, mark it clean | |
2050 | * if that will be needed to assemble it. | |
2051 | * | |
2052 | * newdev: not used ???? | |
2053 | * grow: Array has gained a new device - this is currently for | |
2054 | * linear only | |
2055 | * resync: mark as dirty so a resync will happen. | |
2056 | * name: update the name - preserving the homehost | |
6e46bf34 | 2057 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2058 | * |
2059 | * Following are not relevant for this imsm: | |
2060 | * sparc2.2 : update from old dodgey metadata | |
2061 | * super-minor: change the preferred_minor number | |
2062 | * summaries: update redundant counters. | |
f352c545 DW |
2063 | * homehost: update the recorded homehost |
2064 | * _reshape_progress: record new reshape_progress position. | |
2065 | */ | |
6e46bf34 DW |
2066 | int rv = 1; |
2067 | struct intel_super *super = st->sb; | |
2068 | struct imsm_super *mpb; | |
f352c545 | 2069 | |
6e46bf34 DW |
2070 | /* we can only update container info */ |
2071 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2072 | return 1; | |
2073 | ||
2074 | mpb = super->anchor; | |
2075 | ||
2076 | if (strcmp(update, "uuid") == 0 && uuid_set && !info->update_private) | |
1e2b2765 | 2077 | rv = -1; |
6e46bf34 DW |
2078 | else if (strcmp(update, "uuid") == 0 && uuid_set && info->update_private) { |
2079 | mpb->orig_family_num = *((__u32 *) info->update_private); | |
2080 | rv = 0; | |
2081 | } else if (strcmp(update, "uuid") == 0) { | |
2082 | __u32 *new_family = malloc(sizeof(*new_family)); | |
2083 | ||
2084 | /* update orig_family_number with the incoming random | |
2085 | * data, report the new effective uuid, and store the | |
2086 | * new orig_family_num for future updates. | |
2087 | */ | |
2088 | if (new_family) { | |
2089 | memcpy(&mpb->orig_family_num, info->uuid, sizeof(__u32)); | |
2090 | uuid_from_super_imsm(st, info->uuid); | |
2091 | *new_family = mpb->orig_family_num; | |
2092 | info->update_private = new_family; | |
2093 | rv = 0; | |
2094 | } | |
2095 | } else if (strcmp(update, "assemble") == 0) | |
2096 | rv = 0; | |
2097 | else | |
1e2b2765 | 2098 | rv = -1; |
f352c545 | 2099 | |
6e46bf34 DW |
2100 | /* successful update? recompute checksum */ |
2101 | if (rv == 0) | |
2102 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2103 | |
2104 | return rv; | |
cdddbdbc DW |
2105 | } |
2106 | ||
c2c087e6 | 2107 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2108 | { |
c2c087e6 | 2109 | size_t size; |
cdddbdbc | 2110 | |
c2c087e6 DW |
2111 | size = sizeof(struct imsm_super); |
2112 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2113 | size += 2 * sizeof(struct imsm_dev); | |
2114 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2115 | size += (4 - 2) * sizeof(struct imsm_map); | |
2116 | /* 4 possible disk_ord_tbl's */ | |
2117 | size += 4 * (disks - 1) * sizeof(__u32); | |
2118 | ||
2119 | return size; | |
2120 | } | |
2121 | ||
2122 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
2123 | { | |
2124 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
2125 | return 0; | |
2126 | ||
2127 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
2128 | } |
2129 | ||
ba2de7ba DW |
2130 | static void free_devlist(struct intel_super *super) |
2131 | { | |
2132 | struct intel_dev *dv; | |
2133 | ||
2134 | while (super->devlist) { | |
2135 | dv = super->devlist->next; | |
2136 | free(super->devlist->dev); | |
2137 | free(super->devlist); | |
2138 | super->devlist = dv; | |
2139 | } | |
2140 | } | |
2141 | ||
2142 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
2143 | { | |
2144 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
2145 | } | |
2146 | ||
cdddbdbc DW |
2147 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
2148 | { | |
2149 | /* | |
2150 | * return: | |
2151 | * 0 same, or first was empty, and second was copied | |
2152 | * 1 second had wrong number | |
2153 | * 2 wrong uuid | |
2154 | * 3 wrong other info | |
2155 | */ | |
2156 | struct intel_super *first = st->sb; | |
2157 | struct intel_super *sec = tst->sb; | |
2158 | ||
2159 | if (!first) { | |
2160 | st->sb = tst->sb; | |
2161 | tst->sb = NULL; | |
2162 | return 0; | |
2163 | } | |
2164 | ||
d23fe947 DW |
2165 | /* if an anchor does not have num_raid_devs set then it is a free |
2166 | * floating spare | |
2167 | */ | |
2168 | if (first->anchor->num_raid_devs > 0 && | |
2169 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
2170 | /* Determine if these disks might ever have been |
2171 | * related. Further disambiguation can only take place | |
2172 | * in load_super_imsm_all | |
2173 | */ | |
2174 | __u32 first_family = first->anchor->orig_family_num; | |
2175 | __u32 sec_family = sec->anchor->orig_family_num; | |
2176 | ||
f796af5d DW |
2177 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
2178 | MAX_SIGNATURE_LENGTH) != 0) | |
2179 | return 3; | |
2180 | ||
a2b97981 DW |
2181 | if (first_family == 0) |
2182 | first_family = first->anchor->family_num; | |
2183 | if (sec_family == 0) | |
2184 | sec_family = sec->anchor->family_num; | |
2185 | ||
2186 | if (first_family != sec_family) | |
d23fe947 | 2187 | return 3; |
f796af5d | 2188 | |
d23fe947 | 2189 | } |
cdddbdbc | 2190 | |
f796af5d | 2191 | |
3e372e5a DW |
2192 | /* if 'first' is a spare promote it to a populated mpb with sec's |
2193 | * family number | |
2194 | */ | |
2195 | if (first->anchor->num_raid_devs == 0 && | |
2196 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 2197 | int i; |
ba2de7ba DW |
2198 | struct intel_dev *dv; |
2199 | struct imsm_dev *dev; | |
78d30f94 DW |
2200 | |
2201 | /* we need to copy raid device info from sec if an allocation | |
2202 | * fails here we don't associate the spare | |
2203 | */ | |
2204 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
2205 | dv = malloc(sizeof(*dv)); |
2206 | if (!dv) | |
2207 | break; | |
2208 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
2209 | if (!dev) { | |
2210 | free(dv); | |
2211 | break; | |
78d30f94 | 2212 | } |
ba2de7ba DW |
2213 | dv->dev = dev; |
2214 | dv->index = i; | |
2215 | dv->next = first->devlist; | |
2216 | first->devlist = dv; | |
78d30f94 | 2217 | } |
709743c5 | 2218 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
2219 | /* allocation failure */ |
2220 | free_devlist(first); | |
2221 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
2222 | return 3; | |
78d30f94 | 2223 | } |
3e372e5a | 2224 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 2225 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 2226 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 2227 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
2228 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
2229 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
2230 | } |
2231 | ||
cdddbdbc DW |
2232 | return 0; |
2233 | } | |
2234 | ||
0030e8d6 DW |
2235 | static void fd2devname(int fd, char *name) |
2236 | { | |
2237 | struct stat st; | |
2238 | char path[256]; | |
33a6535d | 2239 | char dname[PATH_MAX]; |
0030e8d6 DW |
2240 | char *nm; |
2241 | int rv; | |
2242 | ||
2243 | name[0] = '\0'; | |
2244 | if (fstat(fd, &st) != 0) | |
2245 | return; | |
2246 | sprintf(path, "/sys/dev/block/%d:%d", | |
2247 | major(st.st_rdev), minor(st.st_rdev)); | |
2248 | ||
2249 | rv = readlink(path, dname, sizeof(dname)); | |
2250 | if (rv <= 0) | |
2251 | return; | |
2252 | ||
2253 | dname[rv] = '\0'; | |
2254 | nm = strrchr(dname, '/'); | |
2255 | nm++; | |
2256 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
2257 | } | |
2258 | ||
cdddbdbc DW |
2259 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
2260 | ||
2261 | static int imsm_read_serial(int fd, char *devname, | |
2262 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
2263 | { | |
2264 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
2265 | int rv; |
2266 | int rsp_len; | |
1f24f035 | 2267 | int len; |
316e2bf4 DW |
2268 | char *dest; |
2269 | char *src; | |
2270 | char *rsp_buf; | |
2271 | int i; | |
cdddbdbc DW |
2272 | |
2273 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 2274 | |
f9ba0ff1 DW |
2275 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
2276 | ||
40ebbb9c | 2277 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
2278 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
2279 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
2280 | return 0; |
2281 | } | |
2282 | ||
cdddbdbc DW |
2283 | if (rv != 0) { |
2284 | if (devname) | |
2285 | fprintf(stderr, | |
2286 | Name ": Failed to retrieve serial for %s\n", | |
2287 | devname); | |
2288 | return rv; | |
2289 | } | |
2290 | ||
2291 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
2292 | if (!rsp_len) { |
2293 | if (devname) | |
2294 | fprintf(stderr, | |
2295 | Name ": Failed to retrieve serial for %s\n", | |
2296 | devname); | |
2297 | return 2; | |
2298 | } | |
1f24f035 | 2299 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 2300 | |
316e2bf4 DW |
2301 | /* trim all whitespace and non-printable characters and convert |
2302 | * ':' to ';' | |
2303 | */ | |
2304 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
2305 | src = &rsp_buf[i]; | |
2306 | if (*src > 0x20) { | |
2307 | /* ':' is reserved for use in placeholder serial | |
2308 | * numbers for missing disks | |
2309 | */ | |
2310 | if (*src == ':') | |
2311 | *dest++ = ';'; | |
2312 | else | |
2313 | *dest++ = *src; | |
2314 | } | |
2315 | } | |
2316 | len = dest - rsp_buf; | |
2317 | dest = rsp_buf; | |
2318 | ||
2319 | /* truncate leading characters */ | |
2320 | if (len > MAX_RAID_SERIAL_LEN) { | |
2321 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 2322 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 2323 | } |
5c3db629 | 2324 | |
5c3db629 | 2325 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 2326 | memcpy(serial, dest, len); |
cdddbdbc DW |
2327 | |
2328 | return 0; | |
2329 | } | |
2330 | ||
1f24f035 DW |
2331 | static int serialcmp(__u8 *s1, __u8 *s2) |
2332 | { | |
2333 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
2334 | } | |
2335 | ||
2336 | static void serialcpy(__u8 *dest, __u8 *src) | |
2337 | { | |
2338 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
2339 | } | |
2340 | ||
1799c9e8 | 2341 | #ifndef MDASSEMBLE |
54c2c1ea DW |
2342 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
2343 | { | |
2344 | struct dl *dl; | |
2345 | ||
2346 | for (dl = super->disks; dl; dl = dl->next) | |
2347 | if (serialcmp(dl->serial, serial) == 0) | |
2348 | break; | |
2349 | ||
2350 | return dl; | |
2351 | } | |
1799c9e8 | 2352 | #endif |
54c2c1ea | 2353 | |
a2b97981 DW |
2354 | static struct imsm_disk * |
2355 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
2356 | { | |
2357 | int i; | |
2358 | ||
2359 | for (i = 0; i < mpb->num_disks; i++) { | |
2360 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
2361 | ||
2362 | if (serialcmp(disk->serial, serial) == 0) { | |
2363 | if (idx) | |
2364 | *idx = i; | |
2365 | return disk; | |
2366 | } | |
2367 | } | |
2368 | ||
2369 | return NULL; | |
2370 | } | |
2371 | ||
cdddbdbc DW |
2372 | static int |
2373 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
2374 | { | |
a2b97981 | 2375 | struct imsm_disk *disk; |
cdddbdbc DW |
2376 | struct dl *dl; |
2377 | struct stat stb; | |
cdddbdbc | 2378 | int rv; |
a2b97981 | 2379 | char name[40]; |
d23fe947 DW |
2380 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
2381 | ||
2382 | rv = imsm_read_serial(fd, devname, serial); | |
2383 | ||
2384 | if (rv != 0) | |
2385 | return 2; | |
2386 | ||
a2b97981 | 2387 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 2388 | if (!dl) { |
cdddbdbc DW |
2389 | if (devname) |
2390 | fprintf(stderr, | |
2391 | Name ": failed to allocate disk buffer for %s\n", | |
2392 | devname); | |
2393 | return 2; | |
2394 | } | |
cdddbdbc | 2395 | |
a2b97981 DW |
2396 | fstat(fd, &stb); |
2397 | dl->major = major(stb.st_rdev); | |
2398 | dl->minor = minor(stb.st_rdev); | |
2399 | dl->next = super->disks; | |
2400 | dl->fd = keep_fd ? fd : -1; | |
2401 | assert(super->disks == NULL); | |
2402 | super->disks = dl; | |
2403 | serialcpy(dl->serial, serial); | |
2404 | dl->index = -2; | |
2405 | dl->e = NULL; | |
2406 | fd2devname(fd, name); | |
2407 | if (devname) | |
2408 | dl->devname = strdup(devname); | |
2409 | else | |
2410 | dl->devname = strdup(name); | |
cdddbdbc | 2411 | |
d23fe947 | 2412 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
2413 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
2414 | if (disk) { | |
2415 | dl->disk = *disk; | |
2416 | /* only set index on disks that are a member of a | |
2417 | * populated contianer, i.e. one with raid_devs | |
2418 | */ | |
2419 | if (is_failed(&dl->disk)) | |
3f6efecc | 2420 | dl->index = -2; |
a2b97981 DW |
2421 | else if (is_spare(&dl->disk)) |
2422 | dl->index = -1; | |
3f6efecc DW |
2423 | } |
2424 | ||
949c47a0 DW |
2425 | return 0; |
2426 | } | |
2427 | ||
0e600426 | 2428 | #ifndef MDASSEMBLE |
0c046afd DW |
2429 | /* When migrating map0 contains the 'destination' state while map1 |
2430 | * contains the current state. When not migrating map0 contains the | |
2431 | * current state. This routine assumes that map[0].map_state is set to | |
2432 | * the current array state before being called. | |
2433 | * | |
2434 | * Migration is indicated by one of the following states | |
2435 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 2436 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 2437 | * map1state=unitialized) |
1484e727 | 2438 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 2439 | * map1state=normal) |
e3bba0e0 | 2440 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd DW |
2441 | * map1state=degraded) |
2442 | */ | |
0556e1a2 | 2443 | static void migrate(struct imsm_dev *dev, __u8 to_state, int migr_type) |
3393c6af | 2444 | { |
0c046afd | 2445 | struct imsm_map *dest; |
3393c6af DW |
2446 | struct imsm_map *src = get_imsm_map(dev, 0); |
2447 | ||
0c046afd | 2448 | dev->vol.migr_state = 1; |
1484e727 | 2449 | set_migr_type(dev, migr_type); |
f8f603f1 | 2450 | dev->vol.curr_migr_unit = 0; |
0c046afd DW |
2451 | dest = get_imsm_map(dev, 1); |
2452 | ||
0556e1a2 | 2453 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 2454 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
2455 | if ((migr_type == MIGR_REBUILD) || |
2456 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
2457 | __u32 ord; |
2458 | int i; | |
2459 | ||
2460 | for (i = 0; i < src->num_members; i++) { | |
2461 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
2462 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
2463 | } | |
2464 | } | |
2465 | ||
0c046afd | 2466 | src->map_state = to_state; |
949c47a0 | 2467 | } |
f8f603f1 DW |
2468 | |
2469 | static void end_migration(struct imsm_dev *dev, __u8 map_state) | |
2470 | { | |
2471 | struct imsm_map *map = get_imsm_map(dev, 0); | |
0556e1a2 | 2472 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
28bce06f | 2473 | int i, j; |
0556e1a2 DW |
2474 | |
2475 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
2476 | * completed in the last migration. | |
2477 | * | |
28bce06f | 2478 | * FIXME add support for raid-level-migration |
0556e1a2 DW |
2479 | */ |
2480 | for (i = 0; i < prev->num_members; i++) | |
28bce06f AK |
2481 | for (j = 0; j < map->num_members; j++) |
2482 | /* during online capacity expansion | |
2483 | * disks position can be changed if takeover is used | |
2484 | */ | |
2485 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
2486 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
2487 | map->disk_ord_tbl[j] |= prev->disk_ord_tbl[i]; | |
2488 | break; | |
2489 | } | |
f8f603f1 DW |
2490 | |
2491 | dev->vol.migr_state = 0; | |
28bce06f | 2492 | dev->vol.migr_type = 0; |
f8f603f1 DW |
2493 | dev->vol.curr_migr_unit = 0; |
2494 | map->map_state = map_state; | |
2495 | } | |
0e600426 | 2496 | #endif |
949c47a0 DW |
2497 | |
2498 | static int parse_raid_devices(struct intel_super *super) | |
2499 | { | |
2500 | int i; | |
2501 | struct imsm_dev *dev_new; | |
4d7b1503 | 2502 | size_t len, len_migr; |
401d313b | 2503 | size_t max_len = 0; |
4d7b1503 DW |
2504 | size_t space_needed = 0; |
2505 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
2506 | |
2507 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
2508 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 2509 | struct intel_dev *dv; |
949c47a0 | 2510 | |
4d7b1503 DW |
2511 | len = sizeof_imsm_dev(dev_iter, 0); |
2512 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
2513 | if (len_migr > len) | |
2514 | space_needed += len_migr - len; | |
2515 | ||
ba2de7ba DW |
2516 | dv = malloc(sizeof(*dv)); |
2517 | if (!dv) | |
2518 | return 1; | |
401d313b AK |
2519 | if (max_len < len_migr) |
2520 | max_len = len_migr; | |
2521 | if (max_len > len_migr) | |
2522 | space_needed += max_len - len_migr; | |
2523 | dev_new = malloc(max_len); | |
ba2de7ba DW |
2524 | if (!dev_new) { |
2525 | free(dv); | |
949c47a0 | 2526 | return 1; |
ba2de7ba | 2527 | } |
949c47a0 | 2528 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
2529 | dv->dev = dev_new; |
2530 | dv->index = i; | |
2531 | dv->next = super->devlist; | |
2532 | super->devlist = dv; | |
949c47a0 | 2533 | } |
cdddbdbc | 2534 | |
4d7b1503 DW |
2535 | /* ensure that super->buf is large enough when all raid devices |
2536 | * are migrating | |
2537 | */ | |
2538 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
2539 | void *buf; | |
2540 | ||
2541 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
2542 | if (posix_memalign(&buf, 512, len) != 0) | |
2543 | return 1; | |
2544 | ||
1f45a8ad DW |
2545 | memcpy(buf, super->buf, super->len); |
2546 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
2547 | free(super->buf); |
2548 | super->buf = buf; | |
2549 | super->len = len; | |
2550 | } | |
2551 | ||
cdddbdbc DW |
2552 | return 0; |
2553 | } | |
2554 | ||
604b746f JD |
2555 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
2556 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
2557 | { | |
2558 | void *ptr = NULL; | |
2559 | ||
2560 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
2561 | ptr = mpb; | |
2562 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
2563 | } | |
2564 | ||
2565 | return ptr; | |
2566 | } | |
2567 | ||
d23fe947 | 2568 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 2569 | |
cdddbdbc DW |
2570 | /* load_imsm_mpb - read matrix metadata |
2571 | * allocates super->mpb to be freed by free_super | |
2572 | */ | |
2573 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
2574 | { | |
2575 | unsigned long long dsize; | |
cdddbdbc DW |
2576 | unsigned long long sectors; |
2577 | struct stat; | |
6416d527 | 2578 | struct imsm_super *anchor; |
cdddbdbc DW |
2579 | __u32 check_sum; |
2580 | ||
cdddbdbc | 2581 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
2582 | if (dsize < 1024) { |
2583 | if (devname) | |
2584 | fprintf(stderr, | |
2585 | Name ": %s: device to small for imsm\n", | |
2586 | devname); | |
2587 | return 1; | |
2588 | } | |
cdddbdbc DW |
2589 | |
2590 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
2591 | if (devname) | |
2592 | fprintf(stderr, | |
2593 | Name ": Cannot seek to anchor block on %s: %s\n", | |
2594 | devname, strerror(errno)); | |
2595 | return 1; | |
2596 | } | |
2597 | ||
949c47a0 | 2598 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
2599 | if (devname) |
2600 | fprintf(stderr, | |
2601 | Name ": Failed to allocate imsm anchor buffer" | |
2602 | " on %s\n", devname); | |
2603 | return 1; | |
2604 | } | |
949c47a0 | 2605 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
2606 | if (devname) |
2607 | fprintf(stderr, | |
2608 | Name ": Cannot read anchor block on %s: %s\n", | |
2609 | devname, strerror(errno)); | |
6416d527 | 2610 | free(anchor); |
cdddbdbc DW |
2611 | return 1; |
2612 | } | |
2613 | ||
6416d527 | 2614 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
2615 | if (devname) |
2616 | fprintf(stderr, | |
2617 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 2618 | free(anchor); |
cdddbdbc DW |
2619 | return 2; |
2620 | } | |
2621 | ||
d23fe947 | 2622 | __free_imsm(super, 0); |
949c47a0 DW |
2623 | super->len = ROUND_UP(anchor->mpb_size, 512); |
2624 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
2625 | if (devname) |
2626 | fprintf(stderr, | |
2627 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 2628 | super->len); |
6416d527 | 2629 | free(anchor); |
cdddbdbc DW |
2630 | return 2; |
2631 | } | |
949c47a0 | 2632 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 2633 | |
6416d527 NB |
2634 | sectors = mpb_sectors(anchor) - 1; |
2635 | free(anchor); | |
949c47a0 | 2636 | if (!sectors) { |
ecf45690 DW |
2637 | check_sum = __gen_imsm_checksum(super->anchor); |
2638 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
2639 | if (devname) | |
2640 | fprintf(stderr, | |
2641 | Name ": IMSM checksum %x != %x on %s\n", | |
2642 | check_sum, | |
2643 | __le32_to_cpu(super->anchor->check_sum), | |
2644 | devname); | |
2645 | return 2; | |
2646 | } | |
2647 | ||
a2b97981 | 2648 | return 0; |
949c47a0 | 2649 | } |
cdddbdbc DW |
2650 | |
2651 | /* read the extended mpb */ | |
2652 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
2653 | if (devname) | |
2654 | fprintf(stderr, | |
2655 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
2656 | devname, strerror(errno)); | |
2657 | return 1; | |
2658 | } | |
2659 | ||
f21e18ca | 2660 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
2661 | if (devname) |
2662 | fprintf(stderr, | |
2663 | Name ": Cannot read extended mpb on %s: %s\n", | |
2664 | devname, strerror(errno)); | |
2665 | return 2; | |
2666 | } | |
2667 | ||
949c47a0 DW |
2668 | check_sum = __gen_imsm_checksum(super->anchor); |
2669 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
2670 | if (devname) |
2671 | fprintf(stderr, | |
2672 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 2673 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 2674 | devname); |
db575f3b | 2675 | return 3; |
cdddbdbc DW |
2676 | } |
2677 | ||
604b746f JD |
2678 | /* FIXME the BBM log is disk specific so we cannot use this global |
2679 | * buffer for all disks. Ok for now since we only look at the global | |
2680 | * bbm_log_size parameter to gate assembly | |
2681 | */ | |
2682 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
2683 | ||
a2b97981 DW |
2684 | return 0; |
2685 | } | |
2686 | ||
2687 | static int | |
2688 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
2689 | { | |
2690 | int err; | |
2691 | ||
2692 | err = load_imsm_mpb(fd, super, devname); | |
2693 | if (err) | |
2694 | return err; | |
2695 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
2696 | if (err) | |
2697 | return err; | |
2698 | err = parse_raid_devices(super); | |
4d7b1503 | 2699 | |
a2b97981 | 2700 | return err; |
cdddbdbc DW |
2701 | } |
2702 | ||
ae6aad82 DW |
2703 | static void __free_imsm_disk(struct dl *d) |
2704 | { | |
2705 | if (d->fd >= 0) | |
2706 | close(d->fd); | |
2707 | if (d->devname) | |
2708 | free(d->devname); | |
0dcecb2e DW |
2709 | if (d->e) |
2710 | free(d->e); | |
ae6aad82 DW |
2711 | free(d); |
2712 | ||
2713 | } | |
1a64be56 | 2714 | |
cdddbdbc DW |
2715 | static void free_imsm_disks(struct intel_super *super) |
2716 | { | |
47ee5a45 | 2717 | struct dl *d; |
cdddbdbc | 2718 | |
47ee5a45 DW |
2719 | while (super->disks) { |
2720 | d = super->disks; | |
cdddbdbc | 2721 | super->disks = d->next; |
ae6aad82 | 2722 | __free_imsm_disk(d); |
cdddbdbc | 2723 | } |
cb82edca AK |
2724 | while (super->disk_mgmt_list) { |
2725 | d = super->disk_mgmt_list; | |
2726 | super->disk_mgmt_list = d->next; | |
2727 | __free_imsm_disk(d); | |
2728 | } | |
47ee5a45 DW |
2729 | while (super->missing) { |
2730 | d = super->missing; | |
2731 | super->missing = d->next; | |
2732 | __free_imsm_disk(d); | |
2733 | } | |
2734 | ||
cdddbdbc DW |
2735 | } |
2736 | ||
9ca2c81c | 2737 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 2738 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 2739 | { |
88654014 LM |
2740 | struct intel_hba *elem, *next; |
2741 | ||
9ca2c81c | 2742 | if (super->buf) { |
949c47a0 | 2743 | free(super->buf); |
9ca2c81c DW |
2744 | super->buf = NULL; |
2745 | } | |
d23fe947 DW |
2746 | if (free_disks) |
2747 | free_imsm_disks(super); | |
ba2de7ba | 2748 | free_devlist(super); |
88654014 LM |
2749 | elem = super->hba; |
2750 | while (elem) { | |
2751 | if (elem->path) | |
2752 | free((void *)elem->path); | |
2753 | next = elem->next; | |
2754 | free(elem); | |
2755 | elem = next; | |
88c32bb1 | 2756 | } |
88654014 | 2757 | super->hba = NULL; |
cdddbdbc DW |
2758 | } |
2759 | ||
9ca2c81c DW |
2760 | static void free_imsm(struct intel_super *super) |
2761 | { | |
d23fe947 | 2762 | __free_imsm(super, 1); |
9ca2c81c DW |
2763 | free(super); |
2764 | } | |
cdddbdbc DW |
2765 | |
2766 | static void free_super_imsm(struct supertype *st) | |
2767 | { | |
2768 | struct intel_super *super = st->sb; | |
2769 | ||
2770 | if (!super) | |
2771 | return; | |
2772 | ||
2773 | free_imsm(super); | |
2774 | st->sb = NULL; | |
2775 | } | |
2776 | ||
49133e57 | 2777 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
2778 | { |
2779 | struct intel_super *super = malloc(sizeof(*super)); | |
2780 | ||
2781 | if (super) { | |
2782 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 2783 | super->current_vol = -1; |
0dcecb2e | 2784 | super->create_offset = ~((__u32 ) 0); |
88c32bb1 DW |
2785 | if (!check_env("IMSM_NO_PLATFORM")) |
2786 | super->orom = find_imsm_orom(); | |
c2c087e6 DW |
2787 | } |
2788 | ||
2789 | return super; | |
2790 | } | |
2791 | ||
cdddbdbc | 2792 | #ifndef MDASSEMBLE |
47ee5a45 DW |
2793 | /* find_missing - helper routine for load_super_imsm_all that identifies |
2794 | * disks that have disappeared from the system. This routine relies on | |
2795 | * the mpb being uptodate, which it is at load time. | |
2796 | */ | |
2797 | static int find_missing(struct intel_super *super) | |
2798 | { | |
2799 | int i; | |
2800 | struct imsm_super *mpb = super->anchor; | |
2801 | struct dl *dl; | |
2802 | struct imsm_disk *disk; | |
47ee5a45 DW |
2803 | |
2804 | for (i = 0; i < mpb->num_disks; i++) { | |
2805 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 2806 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
2807 | if (dl) |
2808 | continue; | |
47ee5a45 DW |
2809 | |
2810 | dl = malloc(sizeof(*dl)); | |
2811 | if (!dl) | |
2812 | return 1; | |
2813 | dl->major = 0; | |
2814 | dl->minor = 0; | |
2815 | dl->fd = -1; | |
2816 | dl->devname = strdup("missing"); | |
2817 | dl->index = i; | |
2818 | serialcpy(dl->serial, disk->serial); | |
2819 | dl->disk = *disk; | |
689c9bf3 | 2820 | dl->e = NULL; |
47ee5a45 DW |
2821 | dl->next = super->missing; |
2822 | super->missing = dl; | |
2823 | } | |
2824 | ||
2825 | return 0; | |
2826 | } | |
2827 | ||
a2b97981 DW |
2828 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
2829 | { | |
2830 | struct intel_disk *idisk = disk_list; | |
2831 | ||
2832 | while (idisk) { | |
2833 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
2834 | break; | |
2835 | idisk = idisk->next; | |
2836 | } | |
2837 | ||
2838 | return idisk; | |
2839 | } | |
2840 | ||
2841 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
2842 | struct intel_super *super, | |
2843 | struct intel_disk **disk_list) | |
2844 | { | |
2845 | struct imsm_disk *d = &super->disks->disk; | |
2846 | struct imsm_super *mpb = super->anchor; | |
2847 | int i, j; | |
2848 | ||
2849 | for (i = 0; i < tbl_size; i++) { | |
2850 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
2851 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
2852 | ||
2853 | if (tbl_mpb->family_num == mpb->family_num) { | |
2854 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
2855 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
2856 | __func__, super->disks->major, | |
2857 | super->disks->minor, | |
2858 | table[i]->disks->major, | |
2859 | table[i]->disks->minor); | |
2860 | break; | |
2861 | } | |
2862 | ||
2863 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
2864 | is_configured(d) == is_configured(tbl_d)) && | |
2865 | tbl_mpb->generation_num < mpb->generation_num) { | |
2866 | /* current version of the mpb is a | |
2867 | * better candidate than the one in | |
2868 | * super_table, but copy over "cross | |
2869 | * generational" status | |
2870 | */ | |
2871 | struct intel_disk *idisk; | |
2872 | ||
2873 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
2874 | __func__, super->disks->major, | |
2875 | super->disks->minor, | |
2876 | table[i]->disks->major, | |
2877 | table[i]->disks->minor); | |
2878 | ||
2879 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
2880 | if (idisk && is_failed(&idisk->disk)) | |
2881 | tbl_d->status |= FAILED_DISK; | |
2882 | break; | |
2883 | } else { | |
2884 | struct intel_disk *idisk; | |
2885 | struct imsm_disk *disk; | |
2886 | ||
2887 | /* tbl_mpb is more up to date, but copy | |
2888 | * over cross generational status before | |
2889 | * returning | |
2890 | */ | |
2891 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
2892 | if (disk && is_failed(disk)) | |
2893 | d->status |= FAILED_DISK; | |
2894 | ||
2895 | idisk = disk_list_get(d->serial, *disk_list); | |
2896 | if (idisk) { | |
2897 | idisk->owner = i; | |
2898 | if (disk && is_configured(disk)) | |
2899 | idisk->disk.status |= CONFIGURED_DISK; | |
2900 | } | |
2901 | ||
2902 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
2903 | __func__, super->disks->major, | |
2904 | super->disks->minor, | |
2905 | table[i]->disks->major, | |
2906 | table[i]->disks->minor); | |
2907 | ||
2908 | return tbl_size; | |
2909 | } | |
2910 | } | |
2911 | } | |
2912 | ||
2913 | if (i >= tbl_size) | |
2914 | table[tbl_size++] = super; | |
2915 | else | |
2916 | table[i] = super; | |
2917 | ||
2918 | /* update/extend the merged list of imsm_disk records */ | |
2919 | for (j = 0; j < mpb->num_disks; j++) { | |
2920 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
2921 | struct intel_disk *idisk; | |
2922 | ||
2923 | idisk = disk_list_get(disk->serial, *disk_list); | |
2924 | if (idisk) { | |
2925 | idisk->disk.status |= disk->status; | |
2926 | if (is_configured(&idisk->disk) || | |
2927 | is_failed(&idisk->disk)) | |
2928 | idisk->disk.status &= ~(SPARE_DISK); | |
2929 | } else { | |
2930 | idisk = calloc(1, sizeof(*idisk)); | |
2931 | if (!idisk) | |
2932 | return -1; | |
2933 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
2934 | idisk->disk = *disk; | |
2935 | idisk->next = *disk_list; | |
2936 | *disk_list = idisk; | |
2937 | } | |
2938 | ||
2939 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
2940 | idisk->owner = i; | |
2941 | } | |
2942 | ||
2943 | return tbl_size; | |
2944 | } | |
2945 | ||
2946 | static struct intel_super * | |
2947 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
2948 | const int owner) | |
2949 | { | |
2950 | struct imsm_super *mpb = super->anchor; | |
2951 | int ok_count = 0; | |
2952 | int i; | |
2953 | ||
2954 | for (i = 0; i < mpb->num_disks; i++) { | |
2955 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
2956 | struct intel_disk *idisk; | |
2957 | ||
2958 | idisk = disk_list_get(disk->serial, disk_list); | |
2959 | if (idisk) { | |
2960 | if (idisk->owner == owner || | |
2961 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
2962 | ok_count++; | |
2963 | else | |
2964 | dprintf("%s: '%.16s' owner %d != %d\n", | |
2965 | __func__, disk->serial, idisk->owner, | |
2966 | owner); | |
2967 | } else { | |
2968 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
2969 | __func__, __le32_to_cpu(mpb->family_num), i, | |
2970 | disk->serial); | |
2971 | break; | |
2972 | } | |
2973 | } | |
2974 | ||
2975 | if (ok_count == mpb->num_disks) | |
2976 | return super; | |
2977 | return NULL; | |
2978 | } | |
2979 | ||
2980 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
2981 | { | |
2982 | struct intel_super *s; | |
2983 | ||
2984 | for (s = super_list; s; s = s->next) { | |
2985 | if (family_num != s->anchor->family_num) | |
2986 | continue; | |
2987 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
2988 | __le32_to_cpu(family_num), s->disks->devname); | |
2989 | } | |
2990 | } | |
2991 | ||
2992 | static struct intel_super * | |
2993 | imsm_thunderdome(struct intel_super **super_list, int len) | |
2994 | { | |
2995 | struct intel_super *super_table[len]; | |
2996 | struct intel_disk *disk_list = NULL; | |
2997 | struct intel_super *champion, *spare; | |
2998 | struct intel_super *s, **del; | |
2999 | int tbl_size = 0; | |
3000 | int conflict; | |
3001 | int i; | |
3002 | ||
3003 | memset(super_table, 0, sizeof(super_table)); | |
3004 | for (s = *super_list; s; s = s->next) | |
3005 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
3006 | ||
3007 | for (i = 0; i < tbl_size; i++) { | |
3008 | struct imsm_disk *d; | |
3009 | struct intel_disk *idisk; | |
3010 | struct imsm_super *mpb = super_table[i]->anchor; | |
3011 | ||
3012 | s = super_table[i]; | |
3013 | d = &s->disks->disk; | |
3014 | ||
3015 | /* 'd' must appear in merged disk list for its | |
3016 | * configuration to be valid | |
3017 | */ | |
3018 | idisk = disk_list_get(d->serial, disk_list); | |
3019 | if (idisk && idisk->owner == i) | |
3020 | s = validate_members(s, disk_list, i); | |
3021 | else | |
3022 | s = NULL; | |
3023 | ||
3024 | if (!s) | |
3025 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
3026 | __func__, mpb->family_num, | |
3027 | super_table[i]->disks->major, | |
3028 | super_table[i]->disks->minor); | |
3029 | super_table[i] = s; | |
3030 | } | |
3031 | ||
3032 | /* This is where the mdadm implementation differs from the Windows | |
3033 | * driver which has no strict concept of a container. We can only | |
3034 | * assemble one family from a container, so when returning a prodigal | |
3035 | * array member to this system the code will not be able to disambiguate | |
3036 | * the container contents that should be assembled ("foreign" versus | |
3037 | * "local"). It requires user intervention to set the orig_family_num | |
3038 | * to a new value to establish a new container. The Windows driver in | |
3039 | * this situation fixes up the volume name in place and manages the | |
3040 | * foreign array as an independent entity. | |
3041 | */ | |
3042 | s = NULL; | |
3043 | spare = NULL; | |
3044 | conflict = 0; | |
3045 | for (i = 0; i < tbl_size; i++) { | |
3046 | struct intel_super *tbl_ent = super_table[i]; | |
3047 | int is_spare = 0; | |
3048 | ||
3049 | if (!tbl_ent) | |
3050 | continue; | |
3051 | ||
3052 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
3053 | spare = tbl_ent; | |
3054 | is_spare = 1; | |
3055 | } | |
3056 | ||
3057 | if (s && !is_spare) { | |
3058 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
3059 | conflict++; | |
3060 | } else if (!s && !is_spare) | |
3061 | s = tbl_ent; | |
3062 | } | |
3063 | ||
3064 | if (!s) | |
3065 | s = spare; | |
3066 | if (!s) { | |
3067 | champion = NULL; | |
3068 | goto out; | |
3069 | } | |
3070 | champion = s; | |
3071 | ||
3072 | if (conflict) | |
3073 | fprintf(stderr, "Chose family %#x on '%s', " | |
3074 | "assemble conflicts to new container with '--update=uuid'\n", | |
3075 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
3076 | ||
3077 | /* collect all dl's onto 'champion', and update them to | |
3078 | * champion's version of the status | |
3079 | */ | |
3080 | for (s = *super_list; s; s = s->next) { | |
3081 | struct imsm_super *mpb = champion->anchor; | |
3082 | struct dl *dl = s->disks; | |
3083 | ||
3084 | if (s == champion) | |
3085 | continue; | |
3086 | ||
3087 | for (i = 0; i < mpb->num_disks; i++) { | |
3088 | struct imsm_disk *disk; | |
3089 | ||
3090 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
3091 | if (disk) { | |
3092 | dl->disk = *disk; | |
3093 | /* only set index on disks that are a member of | |
3094 | * a populated contianer, i.e. one with | |
3095 | * raid_devs | |
3096 | */ | |
3097 | if (is_failed(&dl->disk)) | |
3098 | dl->index = -2; | |
3099 | else if (is_spare(&dl->disk)) | |
3100 | dl->index = -1; | |
3101 | break; | |
3102 | } | |
3103 | } | |
3104 | ||
3105 | if (i >= mpb->num_disks) { | |
3106 | struct intel_disk *idisk; | |
3107 | ||
3108 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 3109 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
3110 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
3111 | dl->index = -1; | |
3112 | else { | |
3113 | dl->index = -2; | |
3114 | continue; | |
3115 | } | |
3116 | } | |
3117 | ||
3118 | dl->next = champion->disks; | |
3119 | champion->disks = dl; | |
3120 | s->disks = NULL; | |
3121 | } | |
3122 | ||
3123 | /* delete 'champion' from super_list */ | |
3124 | for (del = super_list; *del; ) { | |
3125 | if (*del == champion) { | |
3126 | *del = (*del)->next; | |
3127 | break; | |
3128 | } else | |
3129 | del = &(*del)->next; | |
3130 | } | |
3131 | champion->next = NULL; | |
3132 | ||
3133 | out: | |
3134 | while (disk_list) { | |
3135 | struct intel_disk *idisk = disk_list; | |
3136 | ||
3137 | disk_list = disk_list->next; | |
3138 | free(idisk); | |
3139 | } | |
3140 | ||
3141 | return champion; | |
3142 | } | |
3143 | ||
cdddbdbc | 3144 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
e1902a7b | 3145 | char *devname) |
cdddbdbc DW |
3146 | { |
3147 | struct mdinfo *sra; | |
a2b97981 DW |
3148 | struct intel_super *super_list = NULL; |
3149 | struct intel_super *super = NULL; | |
db575f3b | 3150 | int devnum = fd2devnum(fd); |
a2b97981 | 3151 | struct mdinfo *sd; |
db575f3b | 3152 | int retry; |
a2b97981 DW |
3153 | int err = 0; |
3154 | int i; | |
dab4a513 DW |
3155 | |
3156 | /* check if 'fd' an opened container */ | |
b526e52d | 3157 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
cdddbdbc DW |
3158 | if (!sra) |
3159 | return 1; | |
3160 | ||
3161 | if (sra->array.major_version != -1 || | |
3162 | sra->array.minor_version != -2 || | |
1602d52c AW |
3163 | strcmp(sra->text_version, "imsm") != 0) { |
3164 | err = 1; | |
3165 | goto error; | |
3166 | } | |
a2b97981 DW |
3167 | /* load all mpbs */ |
3168 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
49133e57 | 3169 | struct intel_super *s = alloc_super(); |
7a6ecd55 | 3170 | char nm[32]; |
a2b97981 DW |
3171 | int dfd; |
3172 | ||
3173 | err = 1; | |
3174 | if (!s) | |
3175 | goto error; | |
3176 | s->next = super_list; | |
3177 | super_list = s; | |
cdddbdbc | 3178 | |
a2b97981 | 3179 | err = 2; |
cdddbdbc | 3180 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3181 | dfd = dev_open(nm, O_RDWR); |
a2b97981 DW |
3182 | if (dfd < 0) |
3183 | goto error; | |
3184 | ||
e1902a7b | 3185 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
db575f3b DW |
3186 | |
3187 | /* retry the load if we might have raced against mdmon */ | |
a2b97981 | 3188 | if (err == 3 && mdmon_running(devnum)) |
db575f3b DW |
3189 | for (retry = 0; retry < 3; retry++) { |
3190 | usleep(3000); | |
e1902a7b | 3191 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
a2b97981 | 3192 | if (err != 3) |
db575f3b DW |
3193 | break; |
3194 | } | |
a2b97981 DW |
3195 | if (err) |
3196 | goto error; | |
cdddbdbc DW |
3197 | } |
3198 | ||
a2b97981 DW |
3199 | /* all mpbs enter, maybe one leaves */ |
3200 | super = imsm_thunderdome(&super_list, i); | |
3201 | if (!super) { | |
3202 | err = 1; | |
3203 | goto error; | |
cdddbdbc DW |
3204 | } |
3205 | ||
47ee5a45 DW |
3206 | if (find_missing(super) != 0) { |
3207 | free_imsm(super); | |
a2b97981 DW |
3208 | err = 2; |
3209 | goto error; | |
47ee5a45 | 3210 | } |
a2b97981 DW |
3211 | err = 0; |
3212 | ||
3213 | error: | |
3214 | while (super_list) { | |
3215 | struct intel_super *s = super_list; | |
3216 | ||
3217 | super_list = super_list->next; | |
3218 | free_imsm(s); | |
3219 | } | |
1602d52c | 3220 | sysfs_free(sra); |
a2b97981 DW |
3221 | |
3222 | if (err) | |
3223 | return err; | |
f7e7067b | 3224 | |
cdddbdbc | 3225 | *sbp = super; |
db575f3b | 3226 | st->container_dev = devnum; |
a2b97981 | 3227 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 3228 | st->ss = &super_imsm; |
cdddbdbc DW |
3229 | st->minor_version = 0; |
3230 | st->max_devs = IMSM_MAX_DEVICES; | |
3231 | } | |
cdddbdbc DW |
3232 | return 0; |
3233 | } | |
2b959fbf N |
3234 | |
3235 | static int load_container_imsm(struct supertype *st, int fd, char *devname) | |
3236 | { | |
3237 | return load_super_imsm_all(st, fd, &st->sb, devname); | |
3238 | } | |
cdddbdbc DW |
3239 | #endif |
3240 | ||
3241 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
3242 | { | |
3243 | struct intel_super *super; | |
3244 | int rv; | |
3245 | ||
691c6ee1 N |
3246 | if (test_partition(fd)) |
3247 | /* IMSM not allowed on partitions */ | |
3248 | return 1; | |
3249 | ||
37424f13 DW |
3250 | free_super_imsm(st); |
3251 | ||
49133e57 | 3252 | super = alloc_super(); |
cdddbdbc DW |
3253 | if (!super) { |
3254 | fprintf(stderr, | |
3255 | Name ": malloc of %zu failed.\n", | |
3256 | sizeof(*super)); | |
3257 | return 1; | |
3258 | } | |
3259 | ||
a2b97981 | 3260 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
3261 | |
3262 | if (rv) { | |
3263 | if (devname) | |
3264 | fprintf(stderr, | |
3265 | Name ": Failed to load all information " | |
3266 | "sections on %s\n", devname); | |
3267 | free_imsm(super); | |
3268 | return rv; | |
3269 | } | |
3270 | ||
3271 | st->sb = super; | |
3272 | if (st->ss == NULL) { | |
3273 | st->ss = &super_imsm; | |
3274 | st->minor_version = 0; | |
3275 | st->max_devs = IMSM_MAX_DEVICES; | |
3276 | } | |
cdddbdbc DW |
3277 | return 0; |
3278 | } | |
3279 | ||
ef6ffade DW |
3280 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
3281 | { | |
3282 | if (info->level == 1) | |
3283 | return 128; | |
3284 | return info->chunk_size >> 9; | |
3285 | } | |
3286 | ||
ff596308 | 3287 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains) |
ef6ffade DW |
3288 | { |
3289 | __u32 num_stripes; | |
3290 | ||
3291 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
ff596308 | 3292 | num_stripes /= num_domains; |
ef6ffade DW |
3293 | |
3294 | return num_stripes; | |
3295 | } | |
3296 | ||
fcfd9599 DW |
3297 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
3298 | { | |
4025c288 DW |
3299 | if (info->level == 1) |
3300 | return info->size * 2; | |
3301 | else | |
3302 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
fcfd9599 DW |
3303 | } |
3304 | ||
4d1313e9 DW |
3305 | static void imsm_update_version_info(struct intel_super *super) |
3306 | { | |
3307 | /* update the version and attributes */ | |
3308 | struct imsm_super *mpb = super->anchor; | |
3309 | char *version; | |
3310 | struct imsm_dev *dev; | |
3311 | struct imsm_map *map; | |
3312 | int i; | |
3313 | ||
3314 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
3315 | dev = get_imsm_dev(super, i); | |
3316 | map = get_imsm_map(dev, 0); | |
3317 | if (__le32_to_cpu(dev->size_high) > 0) | |
3318 | mpb->attributes |= MPB_ATTRIB_2TB; | |
3319 | ||
3320 | /* FIXME detect when an array spans a port multiplier */ | |
3321 | #if 0 | |
3322 | mpb->attributes |= MPB_ATTRIB_PM; | |
3323 | #endif | |
3324 | ||
3325 | if (mpb->num_raid_devs > 1 || | |
3326 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
3327 | version = MPB_VERSION_ATTRIBS; | |
3328 | switch (get_imsm_raid_level(map)) { | |
3329 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
3330 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
3331 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
3332 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
3333 | } | |
3334 | } else { | |
3335 | if (map->num_members >= 5) | |
3336 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
3337 | else if (dev->status == DEV_CLONE_N_GO) | |
3338 | version = MPB_VERSION_CNG; | |
3339 | else if (get_imsm_raid_level(map) == 5) | |
3340 | version = MPB_VERSION_RAID5; | |
3341 | else if (map->num_members >= 3) | |
3342 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
3343 | else if (get_imsm_raid_level(map) == 1) | |
3344 | version = MPB_VERSION_RAID1; | |
3345 | else | |
3346 | version = MPB_VERSION_RAID0; | |
3347 | } | |
3348 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
3349 | } | |
3350 | } | |
3351 | ||
aa534678 DW |
3352 | static int check_name(struct intel_super *super, char *name, int quiet) |
3353 | { | |
3354 | struct imsm_super *mpb = super->anchor; | |
3355 | char *reason = NULL; | |
3356 | int i; | |
3357 | ||
3358 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
3359 | reason = "must be 16 characters or less"; | |
3360 | ||
3361 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
3362 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3363 | ||
3364 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
3365 | reason = "already exists"; | |
3366 | break; | |
3367 | } | |
3368 | } | |
3369 | ||
3370 | if (reason && !quiet) | |
3371 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
3372 | ||
3373 | return !reason; | |
3374 | } | |
3375 | ||
8b353278 DW |
3376 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
3377 | unsigned long long size, char *name, | |
3378 | char *homehost, int *uuid) | |
cdddbdbc | 3379 | { |
c2c087e6 DW |
3380 | /* We are creating a volume inside a pre-existing container. |
3381 | * so st->sb is already set. | |
3382 | */ | |
3383 | struct intel_super *super = st->sb; | |
949c47a0 | 3384 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 3385 | struct intel_dev *dv; |
c2c087e6 DW |
3386 | struct imsm_dev *dev; |
3387 | struct imsm_vol *vol; | |
3388 | struct imsm_map *map; | |
3389 | int idx = mpb->num_raid_devs; | |
3390 | int i; | |
3391 | unsigned long long array_blocks; | |
2c092cad | 3392 | size_t size_old, size_new; |
ff596308 | 3393 | __u32 num_data_stripes; |
cdddbdbc | 3394 | |
88c32bb1 | 3395 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 3396 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 3397 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
3398 | return 0; |
3399 | } | |
3400 | ||
2c092cad DW |
3401 | /* ensure the mpb is large enough for the new data */ |
3402 | size_old = __le32_to_cpu(mpb->mpb_size); | |
3403 | size_new = disks_to_mpb_size(info->nr_disks); | |
3404 | if (size_new > size_old) { | |
3405 | void *mpb_new; | |
3406 | size_t size_round = ROUND_UP(size_new, 512); | |
3407 | ||
3408 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
3409 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
3410 | return 0; | |
3411 | } | |
3412 | memcpy(mpb_new, mpb, size_old); | |
3413 | free(mpb); | |
3414 | mpb = mpb_new; | |
949c47a0 | 3415 | super->anchor = mpb_new; |
2c092cad DW |
3416 | mpb->mpb_size = __cpu_to_le32(size_new); |
3417 | memset(mpb_new + size_old, 0, size_round - size_old); | |
3418 | } | |
bf5a934a | 3419 | super->current_vol = idx; |
d23fe947 DW |
3420 | /* when creating the first raid device in this container set num_disks |
3421 | * to zero, i.e. delete this spare and add raid member devices in | |
3422 | * add_to_super_imsm_volume() | |
3423 | */ | |
3424 | if (super->current_vol == 0) | |
3425 | mpb->num_disks = 0; | |
5a038140 | 3426 | |
aa534678 DW |
3427 | if (!check_name(super, name, 0)) |
3428 | return 0; | |
ba2de7ba DW |
3429 | dv = malloc(sizeof(*dv)); |
3430 | if (!dv) { | |
3431 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
3432 | return 0; | |
3433 | } | |
1a2487c2 | 3434 | dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
949c47a0 | 3435 | if (!dev) { |
ba2de7ba | 3436 | free(dv); |
949c47a0 DW |
3437 | fprintf(stderr, Name": could not allocate raid device\n"); |
3438 | return 0; | |
3439 | } | |
1a2487c2 | 3440 | |
c2c087e6 | 3441 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
03bcbc65 DW |
3442 | if (info->level == 1) |
3443 | array_blocks = info_to_blocks_per_member(info); | |
3444 | else | |
3445 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
3446 | info->layout, info->chunk_size, | |
3447 | info->size*2); | |
979d38be DW |
3448 | /* round array size down to closest MB */ |
3449 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
3450 | ||
c2c087e6 DW |
3451 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
3452 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 3453 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
3454 | vol = &dev->vol; |
3455 | vol->migr_state = 0; | |
1484e727 | 3456 | set_migr_type(dev, MIGR_INIT); |
c2c087e6 | 3457 | vol->dirty = 0; |
f8f603f1 | 3458 | vol->curr_migr_unit = 0; |
a965f303 | 3459 | map = get_imsm_map(dev, 0); |
0dcecb2e | 3460 | map->pba_of_lba0 = __cpu_to_le32(super->create_offset); |
fcfd9599 | 3461 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade | 3462 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 3463 | map->failed_disk_num = ~0; |
c2c087e6 DW |
3464 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
3465 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 3466 | map->ddf = 1; |
ef6ffade DW |
3467 | |
3468 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
3469 | free(dev); |
3470 | free(dv); | |
ef6ffade DW |
3471 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
3472 | "in a raid1 volume\n"); | |
3473 | return 0; | |
3474 | } | |
81062a36 DW |
3475 | |
3476 | map->raid_level = info->level; | |
4d1313e9 | 3477 | if (info->level == 10) { |
c2c087e6 | 3478 | map->raid_level = 1; |
4d1313e9 | 3479 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
3480 | } else if (info->level == 1) |
3481 | map->num_domains = info->raid_disks; | |
3482 | else | |
ff596308 | 3483 | map->num_domains = 1; |
81062a36 | 3484 | |
ff596308 DW |
3485 | num_data_stripes = info_to_num_data_stripes(info, map->num_domains); |
3486 | map->num_data_stripes = __cpu_to_le32(num_data_stripes); | |
ef6ffade | 3487 | |
c2c087e6 DW |
3488 | map->num_members = info->raid_disks; |
3489 | for (i = 0; i < map->num_members; i++) { | |
3490 | /* initialized in add_to_super */ | |
4eb26970 | 3491 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 3492 | } |
949c47a0 | 3493 | mpb->num_raid_devs++; |
ba2de7ba DW |
3494 | |
3495 | dv->dev = dev; | |
3496 | dv->index = super->current_vol; | |
3497 | dv->next = super->devlist; | |
3498 | super->devlist = dv; | |
c2c087e6 | 3499 | |
4d1313e9 DW |
3500 | imsm_update_version_info(super); |
3501 | ||
c2c087e6 | 3502 | return 1; |
cdddbdbc DW |
3503 | } |
3504 | ||
bf5a934a DW |
3505 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
3506 | unsigned long long size, char *name, | |
3507 | char *homehost, int *uuid) | |
3508 | { | |
3509 | /* This is primarily called by Create when creating a new array. | |
3510 | * We will then get add_to_super called for each component, and then | |
3511 | * write_init_super called to write it out to each device. | |
3512 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
3513 | * array. | |
3514 | * To create on a pre-existing array a different method will be called. | |
3515 | * This one is just for fresh drives. | |
3516 | */ | |
3517 | struct intel_super *super; | |
3518 | struct imsm_super *mpb; | |
3519 | size_t mpb_size; | |
4d1313e9 | 3520 | char *version; |
bf5a934a | 3521 | |
bf5a934a | 3522 | if (st->sb) |
e683ca88 DW |
3523 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
3524 | ||
3525 | if (info) | |
3526 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
3527 | else | |
3528 | mpb_size = 512; | |
bf5a934a | 3529 | |
49133e57 | 3530 | super = alloc_super(); |
e683ca88 | 3531 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 3532 | free(super); |
e683ca88 DW |
3533 | super = NULL; |
3534 | } | |
3535 | if (!super) { | |
3536 | fprintf(stderr, Name | |
3537 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
3538 | return 0; |
3539 | } | |
e683ca88 | 3540 | memset(super->buf, 0, mpb_size); |
ef649044 | 3541 | mpb = super->buf; |
e683ca88 DW |
3542 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
3543 | st->sb = super; | |
3544 | ||
3545 | if (info == NULL) { | |
3546 | /* zeroing superblock */ | |
3547 | return 0; | |
3548 | } | |
bf5a934a | 3549 | |
4d1313e9 DW |
3550 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
3551 | ||
3552 | version = (char *) mpb->sig; | |
3553 | strcpy(version, MPB_SIGNATURE); | |
3554 | version += strlen(MPB_SIGNATURE); | |
3555 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 3556 | |
bf5a934a DW |
3557 | return 1; |
3558 | } | |
3559 | ||
0e600426 | 3560 | #ifndef MDASSEMBLE |
f20c3968 | 3561 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
3562 | int fd, char *devname) |
3563 | { | |
3564 | struct intel_super *super = st->sb; | |
d23fe947 | 3565 | struct imsm_super *mpb = super->anchor; |
bf5a934a DW |
3566 | struct dl *dl; |
3567 | struct imsm_dev *dev; | |
3568 | struct imsm_map *map; | |
4eb26970 | 3569 | int slot; |
bf5a934a | 3570 | |
949c47a0 | 3571 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 3572 | map = get_imsm_map(dev, 0); |
bf5a934a | 3573 | |
208933a7 N |
3574 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
3575 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
3576 | devname); | |
3577 | return 1; | |
3578 | } | |
3579 | ||
efb30e7f DW |
3580 | if (fd == -1) { |
3581 | /* we're doing autolayout so grab the pre-marked (in | |
3582 | * validate_geometry) raid_disk | |
3583 | */ | |
3584 | for (dl = super->disks; dl; dl = dl->next) | |
3585 | if (dl->raiddisk == dk->raid_disk) | |
3586 | break; | |
3587 | } else { | |
3588 | for (dl = super->disks; dl ; dl = dl->next) | |
3589 | if (dl->major == dk->major && | |
3590 | dl->minor == dk->minor) | |
3591 | break; | |
3592 | } | |
d23fe947 | 3593 | |
208933a7 N |
3594 | if (!dl) { |
3595 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 3596 | return 1; |
208933a7 | 3597 | } |
bf5a934a | 3598 | |
d23fe947 DW |
3599 | /* add a pristine spare to the metadata */ |
3600 | if (dl->index < 0) { | |
3601 | dl->index = super->anchor->num_disks; | |
3602 | super->anchor->num_disks++; | |
3603 | } | |
4eb26970 DW |
3604 | /* Check the device has not already been added */ |
3605 | slot = get_imsm_disk_slot(map, dl->index); | |
3606 | if (slot >= 0 && | |
98130f40 | 3607 | (get_imsm_ord_tbl_ent(dev, slot, -1) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
3608 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
3609 | devname); | |
3610 | return 1; | |
3611 | } | |
be73972f | 3612 | set_imsm_ord_tbl_ent(map, dk->number, dl->index); |
ee5aad5a | 3613 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 DW |
3614 | |
3615 | /* if we are creating the first raid device update the family number */ | |
3616 | if (super->current_vol == 0) { | |
3617 | __u32 sum; | |
3618 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
3619 | struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index); | |
3620 | ||
791b666a AW |
3621 | if (!_dev || !_disk) { |
3622 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
3623 | return 1; | |
3624 | } | |
d23fe947 DW |
3625 | *_dev = *dev; |
3626 | *_disk = dl->disk; | |
148acb7b DW |
3627 | sum = random32(); |
3628 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 3629 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 3630 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 3631 | } |
f20c3968 DW |
3632 | |
3633 | return 0; | |
bf5a934a DW |
3634 | } |
3635 | ||
88654014 | 3636 | |
f20c3968 | 3637 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 3638 | int fd, char *devname) |
cdddbdbc | 3639 | { |
c2c087e6 | 3640 | struct intel_super *super = st->sb; |
c2c087e6 DW |
3641 | struct dl *dd; |
3642 | unsigned long long size; | |
f2f27e63 | 3643 | __u32 id; |
c2c087e6 DW |
3644 | int rv; |
3645 | struct stat stb; | |
3646 | ||
88654014 LM |
3647 | /* If we are on an RAID enabled platform check that the disk is |
3648 | * attached to the raid controller. | |
3649 | * We do not need to test disks attachment for container based additions, | |
3650 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 3651 | */ |
88654014 LM |
3652 | if ((fd != -1) && !check_env("IMSM_NO_PLATFORM")) { |
3653 | struct sys_dev *hba_name; | |
3654 | struct intel_hba *hba; | |
3655 | ||
3656 | hba_name = find_disk_attached_hba(fd, NULL); | |
3657 | if (!hba_name) { | |
3658 | fprintf(stderr, | |
3659 | Name ": %s is not attached to Intel(R) RAID controller.\n", | |
3660 | devname ? : "disk"); | |
3661 | return 1; | |
3662 | } | |
3663 | rv = attach_hba_to_super(super, hba_name, devname); | |
3664 | switch (rv) { | |
3665 | case 2: | |
3666 | fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID " | |
3667 | "controller (%s),\n but the container is assigned to Intel(R) " | |
3668 | "%s RAID controller (", | |
3669 | devname, | |
3670 | get_sys_dev_type(hba_name->type), | |
3671 | hba_name->pci_id ? : "Err!", | |
3672 | get_sys_dev_type(hba_name->type)); | |
3673 | ||
3674 | hba = super->hba; | |
3675 | while (hba) { | |
3676 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3677 | if (hba->next) | |
3678 | fprintf(stderr, ", "); | |
3679 | hba = hba->next; | |
3680 | } | |
3681 | ||
3682 | fprintf(stderr, ").\n" | |
3683 | " Mixing devices attached to different controllers " | |
3684 | "is not allowed.\n"); | |
3685 | free_sys_dev(&hba_name); | |
3686 | return 1; | |
3687 | } | |
3688 | free_sys_dev(&hba_name); | |
88c32bb1 DW |
3689 | } |
3690 | ||
f20c3968 DW |
3691 | if (super->current_vol >= 0) |
3692 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 3693 | |
c2c087e6 DW |
3694 | fstat(fd, &stb); |
3695 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 3696 | if (!dd) { |
c2c087e6 DW |
3697 | fprintf(stderr, |
3698 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 3699 | return 1; |
c2c087e6 DW |
3700 | } |
3701 | memset(dd, 0, sizeof(*dd)); | |
3702 | dd->major = major(stb.st_rdev); | |
3703 | dd->minor = minor(stb.st_rdev); | |
b9f594fe | 3704 | dd->index = -1; |
c2c087e6 | 3705 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 3706 | dd->fd = fd; |
689c9bf3 | 3707 | dd->e = NULL; |
1a64be56 | 3708 | dd->action = DISK_ADD; |
c2c087e6 | 3709 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 3710 | if (rv) { |
c2c087e6 | 3711 | fprintf(stderr, |
0030e8d6 | 3712 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 3713 | free(dd); |
0030e8d6 | 3714 | abort(); |
c2c087e6 DW |
3715 | } |
3716 | ||
c2c087e6 DW |
3717 | get_dev_size(fd, NULL, &size); |
3718 | size /= 512; | |
1f24f035 | 3719 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 3720 | dd->disk.total_blocks = __cpu_to_le32(size); |
ee5aad5a | 3721 | dd->disk.status = SPARE_DISK; |
c2c087e6 | 3722 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 3723 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 3724 | else |
b9f594fe | 3725 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
3726 | |
3727 | if (st->update_tail) { | |
1a64be56 LM |
3728 | dd->next = super->disk_mgmt_list; |
3729 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
3730 | } else { |
3731 | dd->next = super->disks; | |
3732 | super->disks = dd; | |
3733 | } | |
f20c3968 DW |
3734 | |
3735 | return 0; | |
cdddbdbc DW |
3736 | } |
3737 | ||
1a64be56 LM |
3738 | |
3739 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
3740 | { | |
3741 | struct intel_super *super = st->sb; | |
3742 | struct dl *dd; | |
3743 | ||
3744 | /* remove from super works only in mdmon - for communication | |
3745 | * manager - monitor. Check if communication memory buffer | |
3746 | * is prepared. | |
3747 | */ | |
3748 | if (!st->update_tail) { | |
3749 | fprintf(stderr, | |
3750 | Name ": %s shall be used in mdmon context only" | |
3751 | "(line %d).\n", __func__, __LINE__); | |
3752 | return 1; | |
3753 | } | |
3754 | dd = malloc(sizeof(*dd)); | |
3755 | if (!dd) { | |
3756 | fprintf(stderr, | |
3757 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
3758 | return 1; | |
3759 | } | |
3760 | memset(dd, 0, sizeof(*dd)); | |
3761 | dd->major = dk->major; | |
3762 | dd->minor = dk->minor; | |
3763 | dd->index = -1; | |
3764 | dd->fd = -1; | |
3765 | dd->disk.status = SPARE_DISK; | |
3766 | dd->action = DISK_REMOVE; | |
3767 | ||
3768 | dd->next = super->disk_mgmt_list; | |
3769 | super->disk_mgmt_list = dd; | |
3770 | ||
3771 | ||
3772 | return 0; | |
3773 | } | |
3774 | ||
f796af5d DW |
3775 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
3776 | ||
3777 | static union { | |
3778 | char buf[512]; | |
3779 | struct imsm_super anchor; | |
3780 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 3781 | |
d23fe947 DW |
3782 | /* spare records have their own family number and do not have any defined raid |
3783 | * devices | |
3784 | */ | |
3785 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
3786 | { | |
d23fe947 | 3787 | struct imsm_super *mpb = super->anchor; |
f796af5d | 3788 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
3789 | __u32 sum; |
3790 | struct dl *d; | |
3791 | ||
f796af5d DW |
3792 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
3793 | spare->generation_num = __cpu_to_le32(1UL), | |
3794 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
3795 | spare->num_disks = 1, | |
3796 | spare->num_raid_devs = 0, | |
3797 | spare->cache_size = mpb->cache_size, | |
3798 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
3799 | ||
3800 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
3801 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
3802 | |
3803 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 3804 | if (d->index != -1) |
d23fe947 DW |
3805 | continue; |
3806 | ||
f796af5d DW |
3807 | spare->disk[0] = d->disk; |
3808 | sum = __gen_imsm_checksum(spare); | |
3809 | spare->family_num = __cpu_to_le32(sum); | |
3810 | spare->orig_family_num = 0; | |
3811 | sum = __gen_imsm_checksum(spare); | |
3812 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 3813 | |
f796af5d | 3814 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
3815 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
3816 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 3817 | return 1; |
d23fe947 DW |
3818 | } |
3819 | if (doclose) { | |
3820 | close(d->fd); | |
3821 | d->fd = -1; | |
3822 | } | |
3823 | } | |
3824 | ||
e74255d9 | 3825 | return 0; |
d23fe947 DW |
3826 | } |
3827 | ||
36988a3d | 3828 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 3829 | { |
36988a3d | 3830 | struct intel_super *super = st->sb; |
949c47a0 | 3831 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
3832 | struct dl *d; |
3833 | __u32 generation; | |
3834 | __u32 sum; | |
d23fe947 | 3835 | int spares = 0; |
949c47a0 | 3836 | int i; |
a48ac0a8 | 3837 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 3838 | int num_disks = 0; |
cdddbdbc | 3839 | |
c2c087e6 DW |
3840 | /* 'generation' is incremented everytime the metadata is written */ |
3841 | generation = __le32_to_cpu(mpb->generation_num); | |
3842 | generation++; | |
3843 | mpb->generation_num = __cpu_to_le32(generation); | |
3844 | ||
148acb7b DW |
3845 | /* fix up cases where previous mdadm releases failed to set |
3846 | * orig_family_num | |
3847 | */ | |
3848 | if (mpb->orig_family_num == 0) | |
3849 | mpb->orig_family_num = mpb->family_num; | |
3850 | ||
d23fe947 | 3851 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 3852 | if (d->index == -1) |
d23fe947 | 3853 | spares++; |
36988a3d | 3854 | else { |
d23fe947 | 3855 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
3856 | num_disks++; |
3857 | } | |
d23fe947 | 3858 | } |
36988a3d | 3859 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 3860 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
3861 | num_disks++; |
3862 | } | |
3863 | mpb->num_disks = num_disks; | |
3864 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 3865 | |
949c47a0 DW |
3866 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3867 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
3868 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
3869 | if (dev && dev2) { | |
3870 | imsm_copy_dev(dev, dev2); | |
3871 | mpb_size += sizeof_imsm_dev(dev, 0); | |
3872 | } | |
949c47a0 | 3873 | } |
a48ac0a8 DW |
3874 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
3875 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 3876 | |
c2c087e6 | 3877 | /* recalculate checksum */ |
949c47a0 | 3878 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
3879 | mpb->check_sum = __cpu_to_le32(sum); |
3880 | ||
d23fe947 | 3881 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 3882 | for (d = super->disks; d ; d = d->next) { |
d23fe947 DW |
3883 | if (d->index < 0) |
3884 | continue; | |
f796af5d | 3885 | if (store_imsm_mpb(d->fd, mpb)) |
c2c087e6 DW |
3886 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
3887 | __func__, d->major, d->minor, strerror(errno)); | |
c2c087e6 DW |
3888 | if (doclose) { |
3889 | close(d->fd); | |
3890 | d->fd = -1; | |
3891 | } | |
3892 | } | |
3893 | ||
d23fe947 DW |
3894 | if (spares) |
3895 | return write_super_imsm_spares(super, doclose); | |
3896 | ||
e74255d9 | 3897 | return 0; |
c2c087e6 DW |
3898 | } |
3899 | ||
0e600426 | 3900 | |
9b1fb677 | 3901 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
3902 | { |
3903 | size_t len; | |
3904 | struct imsm_update_create_array *u; | |
3905 | struct intel_super *super = st->sb; | |
9b1fb677 | 3906 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
54c2c1ea DW |
3907 | struct imsm_map *map = get_imsm_map(dev, 0); |
3908 | struct disk_info *inf; | |
3909 | struct imsm_disk *disk; | |
3910 | int i; | |
43dad3d6 | 3911 | |
54c2c1ea DW |
3912 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
3913 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
3914 | u = malloc(len); |
3915 | if (!u) { | |
3916 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
3917 | __func__); | |
3918 | return 1; | |
3919 | } | |
3920 | ||
3921 | u->type = update_create_array; | |
9b1fb677 | 3922 | u->dev_idx = dev_idx; |
43dad3d6 | 3923 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
3924 | inf = get_disk_info(u); |
3925 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 3926 | int idx = get_imsm_disk_idx(dev, i, -1); |
9b1fb677 | 3927 | |
54c2c1ea DW |
3928 | disk = get_imsm_disk(super, idx); |
3929 | serialcpy(inf[i].serial, disk->serial); | |
3930 | } | |
43dad3d6 DW |
3931 | append_metadata_update(st, u, len); |
3932 | ||
3933 | return 0; | |
3934 | } | |
3935 | ||
1a64be56 | 3936 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
3937 | { |
3938 | struct intel_super *super = st->sb; | |
3939 | size_t len; | |
1a64be56 | 3940 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 3941 | |
1a64be56 | 3942 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
3943 | return 0; |
3944 | ||
3945 | len = sizeof(*u); | |
3946 | u = malloc(len); | |
3947 | if (!u) { | |
3948 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
3949 | __func__); | |
3950 | return 1; | |
3951 | } | |
3952 | ||
1a64be56 | 3953 | u->type = update_add_remove_disk; |
43dad3d6 DW |
3954 | append_metadata_update(st, u, len); |
3955 | ||
3956 | return 0; | |
3957 | } | |
3958 | ||
c2c087e6 DW |
3959 | static int write_init_super_imsm(struct supertype *st) |
3960 | { | |
9b1fb677 DW |
3961 | struct intel_super *super = st->sb; |
3962 | int current_vol = super->current_vol; | |
3963 | ||
3964 | /* we are done with current_vol reset it to point st at the container */ | |
3965 | super->current_vol = -1; | |
3966 | ||
8273f55e | 3967 | if (st->update_tail) { |
43dad3d6 DW |
3968 | /* queue the recently created array / added disk |
3969 | * as a metadata update */ | |
43dad3d6 | 3970 | int rv; |
8273f55e | 3971 | |
43dad3d6 | 3972 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 3973 | if (current_vol < 0) { |
1a64be56 LM |
3974 | /* in the mgmt (add/remove) disk case we are running |
3975 | * in mdmon context, so don't close fd's | |
43dad3d6 | 3976 | */ |
1a64be56 | 3977 | return mgmt_disk(st); |
43dad3d6 | 3978 | } else |
9b1fb677 | 3979 | rv = create_array(st, current_vol); |
8273f55e | 3980 | |
43dad3d6 | 3981 | return rv; |
d682f344 N |
3982 | } else { |
3983 | struct dl *d; | |
3984 | for (d = super->disks; d; d = d->next) | |
3985 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 3986 | return write_super_imsm(st, 1); |
d682f344 | 3987 | } |
cdddbdbc | 3988 | } |
0e600426 | 3989 | #endif |
cdddbdbc | 3990 | |
e683ca88 | 3991 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 3992 | { |
e683ca88 DW |
3993 | struct intel_super *super = st->sb; |
3994 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 3995 | |
e683ca88 | 3996 | if (!mpb) |
ad97895e DW |
3997 | return 1; |
3998 | ||
1799c9e8 | 3999 | #ifndef MDASSEMBLE |
e683ca88 | 4000 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
4001 | #else |
4002 | return 1; | |
4003 | #endif | |
cdddbdbc DW |
4004 | } |
4005 | ||
0e600426 N |
4006 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
4007 | { | |
4008 | return __le32_to_cpu(mpb->bbm_log_size); | |
4009 | } | |
4010 | ||
4011 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
4012 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
4013 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 4014 | unsigned long long size, char *dev, |
2c514b71 NB |
4015 | unsigned long long *freesize, |
4016 | int verbose) | |
cdddbdbc | 4017 | { |
c2c087e6 DW |
4018 | int fd; |
4019 | unsigned long long ldsize; | |
88c32bb1 | 4020 | const struct imsm_orom *orom; |
cdddbdbc | 4021 | |
c2c087e6 DW |
4022 | if (level != LEVEL_CONTAINER) |
4023 | return 0; | |
4024 | if (!dev) | |
4025 | return 1; | |
4026 | ||
88c32bb1 DW |
4027 | if (check_env("IMSM_NO_PLATFORM")) |
4028 | orom = NULL; | |
4029 | else | |
4030 | orom = find_imsm_orom(); | |
4031 | if (orom && raiddisks > orom->tds) { | |
4032 | if (verbose) | |
4033 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
4034 | " platform supported disks: %d\n", | |
4035 | raiddisks, orom->tds); | |
4036 | return 0; | |
4037 | } | |
4038 | ||
c2c087e6 DW |
4039 | fd = open(dev, O_RDONLY|O_EXCL, 0); |
4040 | if (fd < 0) { | |
2c514b71 NB |
4041 | if (verbose) |
4042 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
4043 | dev, strerror(errno)); | |
c2c087e6 DW |
4044 | return 0; |
4045 | } | |
4046 | if (!get_dev_size(fd, dev, &ldsize)) { | |
4047 | close(fd); | |
4048 | return 0; | |
4049 | } | |
4050 | close(fd); | |
4051 | ||
4052 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
4053 | ||
4054 | return 1; | |
cdddbdbc DW |
4055 | } |
4056 | ||
0dcecb2e DW |
4057 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
4058 | { | |
4059 | const unsigned long long base_start = e[*idx].start; | |
4060 | unsigned long long end = base_start + e[*idx].size; | |
4061 | int i; | |
4062 | ||
4063 | if (base_start == end) | |
4064 | return 0; | |
4065 | ||
4066 | *idx = *idx + 1; | |
4067 | for (i = *idx; i < num_extents; i++) { | |
4068 | /* extend overlapping extents */ | |
4069 | if (e[i].start >= base_start && | |
4070 | e[i].start <= end) { | |
4071 | if (e[i].size == 0) | |
4072 | return 0; | |
4073 | if (e[i].start + e[i].size > end) | |
4074 | end = e[i].start + e[i].size; | |
4075 | } else if (e[i].start > end) { | |
4076 | *idx = i; | |
4077 | break; | |
4078 | } | |
4079 | } | |
4080 | ||
4081 | return end - base_start; | |
4082 | } | |
4083 | ||
4084 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
4085 | { | |
4086 | /* build a composite disk with all known extents and generate a new | |
4087 | * 'maxsize' given the "all disks in an array must share a common start | |
4088 | * offset" constraint | |
4089 | */ | |
4090 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
4091 | struct dl *dl; | |
4092 | int i, j; | |
4093 | int start_extent; | |
4094 | unsigned long long pos; | |
b9d77223 | 4095 | unsigned long long start = 0; |
0dcecb2e DW |
4096 | unsigned long long maxsize; |
4097 | unsigned long reserve; | |
4098 | ||
4099 | if (!e) | |
a7dd165b | 4100 | return 0; |
0dcecb2e DW |
4101 | |
4102 | /* coalesce and sort all extents. also, check to see if we need to | |
4103 | * reserve space between member arrays | |
4104 | */ | |
4105 | j = 0; | |
4106 | for (dl = super->disks; dl; dl = dl->next) { | |
4107 | if (!dl->e) | |
4108 | continue; | |
4109 | for (i = 0; i < dl->extent_cnt; i++) | |
4110 | e[j++] = dl->e[i]; | |
4111 | } | |
4112 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
4113 | ||
4114 | /* merge extents */ | |
4115 | i = 0; | |
4116 | j = 0; | |
4117 | while (i < sum_extents) { | |
4118 | e[j].start = e[i].start; | |
4119 | e[j].size = find_size(e, &i, sum_extents); | |
4120 | j++; | |
4121 | if (e[j-1].size == 0) | |
4122 | break; | |
4123 | } | |
4124 | ||
4125 | pos = 0; | |
4126 | maxsize = 0; | |
4127 | start_extent = 0; | |
4128 | i = 0; | |
4129 | do { | |
4130 | unsigned long long esize; | |
4131 | ||
4132 | esize = e[i].start - pos; | |
4133 | if (esize >= maxsize) { | |
4134 | maxsize = esize; | |
4135 | start = pos; | |
4136 | start_extent = i; | |
4137 | } | |
4138 | pos = e[i].start + e[i].size; | |
4139 | i++; | |
4140 | } while (e[i-1].size); | |
4141 | free(e); | |
4142 | ||
a7dd165b DW |
4143 | if (maxsize == 0) |
4144 | return 0; | |
4145 | ||
4146 | /* FIXME assumes volume at offset 0 is the first volume in a | |
4147 | * container | |
4148 | */ | |
0dcecb2e DW |
4149 | if (start_extent > 0) |
4150 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
4151 | else | |
4152 | reserve = 0; | |
4153 | ||
4154 | if (maxsize < reserve) | |
a7dd165b | 4155 | return 0; |
0dcecb2e DW |
4156 | |
4157 | super->create_offset = ~((__u32) 0); | |
4158 | if (start + reserve > super->create_offset) | |
a7dd165b | 4159 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
4160 | super->create_offset = start + reserve; |
4161 | ||
4162 | return maxsize - reserve; | |
4163 | } | |
4164 | ||
88c32bb1 DW |
4165 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
4166 | { | |
4167 | if (level < 0 || level == 6 || level == 4) | |
4168 | return 0; | |
4169 | ||
4170 | /* if we have an orom prevent invalid raid levels */ | |
4171 | if (orom) | |
4172 | switch (level) { | |
4173 | case 0: return imsm_orom_has_raid0(orom); | |
4174 | case 1: | |
4175 | if (raiddisks > 2) | |
4176 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
4177 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
4178 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
4179 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
4180 | } |
4181 | else | |
4182 | return 1; /* not on an Intel RAID platform so anything goes */ | |
4183 | ||
4184 | return 0; | |
4185 | } | |
4186 | ||
35f81cbb | 4187 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) |
6592ce37 DW |
4188 | static int |
4189 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
c21e737b | 4190 | int raiddisks, int *chunk, int verbose) |
6592ce37 DW |
4191 | { |
4192 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { | |
4193 | pr_vrb(": platform does not support raid%d with %d disk%s\n", | |
4194 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
4195 | return 0; | |
4196 | } | |
c21e737b CA |
4197 | if (super->orom && level != 1) { |
4198 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
4199 | *chunk = imsm_orom_default_chunk(super->orom); | |
4200 | else if (chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
4201 | pr_vrb(": platform does not support a chunk size of: " | |
4202 | "%d\n", *chunk); | |
4203 | return 0; | |
4204 | } | |
6592ce37 DW |
4205 | } |
4206 | if (layout != imsm_level_to_layout(level)) { | |
4207 | if (level == 5) | |
4208 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
4209 | else if (level == 10) | |
4210 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
4211 | else | |
4212 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
4213 | layout, level); | |
4214 | return 0; | |
4215 | } | |
4216 | ||
4217 | return 1; | |
4218 | } | |
4219 | ||
c2c087e6 DW |
4220 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
4221 | * FIX ME add ahci details | |
4222 | */ | |
8b353278 | 4223 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 4224 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 4225 | unsigned long long size, char *dev, |
2c514b71 NB |
4226 | unsigned long long *freesize, |
4227 | int verbose) | |
cdddbdbc | 4228 | { |
c2c087e6 DW |
4229 | struct stat stb; |
4230 | struct intel_super *super = st->sb; | |
a20d2ba5 | 4231 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4232 | struct dl *dl; |
4233 | unsigned long long pos = 0; | |
4234 | unsigned long long maxsize; | |
4235 | struct extent *e; | |
4236 | int i; | |
cdddbdbc | 4237 | |
88c32bb1 DW |
4238 | /* We must have the container info already read in. */ |
4239 | if (!super) | |
c2c087e6 DW |
4240 | return 0; |
4241 | ||
6592ce37 | 4242 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, verbose)) |
c2c087e6 | 4243 | return 0; |
c2c087e6 DW |
4244 | |
4245 | if (!dev) { | |
4246 | /* General test: make sure there is space for | |
2da8544a DW |
4247 | * 'raiddisks' device extents of size 'size' at a given |
4248 | * offset | |
c2c087e6 | 4249 | */ |
e46273eb | 4250 | unsigned long long minsize = size; |
b7528a20 | 4251 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
4252 | int dcnt = 0; |
4253 | if (minsize == 0) | |
4254 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
4255 | for (dl = super->disks; dl ; dl = dl->next) { | |
4256 | int found = 0; | |
4257 | ||
bf5a934a | 4258 | pos = 0; |
c2c087e6 DW |
4259 | i = 0; |
4260 | e = get_extents(super, dl); | |
4261 | if (!e) continue; | |
4262 | do { | |
4263 | unsigned long long esize; | |
4264 | esize = e[i].start - pos; | |
4265 | if (esize >= minsize) | |
4266 | found = 1; | |
b7528a20 | 4267 | if (found && start_offset == MaxSector) { |
2da8544a DW |
4268 | start_offset = pos; |
4269 | break; | |
4270 | } else if (found && pos != start_offset) { | |
4271 | found = 0; | |
4272 | break; | |
4273 | } | |
c2c087e6 DW |
4274 | pos = e[i].start + e[i].size; |
4275 | i++; | |
4276 | } while (e[i-1].size); | |
4277 | if (found) | |
4278 | dcnt++; | |
4279 | free(e); | |
4280 | } | |
4281 | if (dcnt < raiddisks) { | |
2c514b71 NB |
4282 | if (verbose) |
4283 | fprintf(stderr, Name ": imsm: Not enough " | |
4284 | "devices with space for this array " | |
4285 | "(%d < %d)\n", | |
4286 | dcnt, raiddisks); | |
c2c087e6 DW |
4287 | return 0; |
4288 | } | |
4289 | return 1; | |
4290 | } | |
0dcecb2e | 4291 | |
c2c087e6 DW |
4292 | /* This device must be a member of the set */ |
4293 | if (stat(dev, &stb) < 0) | |
4294 | return 0; | |
4295 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
4296 | return 0; | |
4297 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
4298 | if (dl->major == (int)major(stb.st_rdev) && |
4299 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
4300 | break; |
4301 | } | |
4302 | if (!dl) { | |
2c514b71 NB |
4303 | if (verbose) |
4304 | fprintf(stderr, Name ": %s is not in the " | |
4305 | "same imsm set\n", dev); | |
c2c087e6 | 4306 | return 0; |
a20d2ba5 DW |
4307 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
4308 | /* If a volume is present then the current creation attempt | |
4309 | * cannot incorporate new spares because the orom may not | |
4310 | * understand this configuration (all member disks must be | |
4311 | * members of each array in the container). | |
4312 | */ | |
4313 | fprintf(stderr, Name ": %s is a spare and a volume" | |
4314 | " is already defined for this container\n", dev); | |
4315 | fprintf(stderr, Name ": The option-rom requires all member" | |
4316 | " disks to be a member of all volumes\n"); | |
4317 | return 0; | |
c2c087e6 | 4318 | } |
0dcecb2e DW |
4319 | |
4320 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
4321 | e = get_extents(super, dl); |
4322 | maxsize = 0; | |
4323 | i = 0; | |
0dcecb2e DW |
4324 | if (e) { |
4325 | do { | |
4326 | unsigned long long esize; | |
4327 | ||
4328 | esize = e[i].start - pos; | |
4329 | if (esize >= maxsize) | |
4330 | maxsize = esize; | |
4331 | pos = e[i].start + e[i].size; | |
4332 | i++; | |
4333 | } while (e[i-1].size); | |
4334 | dl->e = e; | |
4335 | dl->extent_cnt = i; | |
4336 | } else { | |
4337 | if (verbose) | |
4338 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
4339 | dev); | |
4340 | return 0; | |
4341 | } | |
4342 | if (maxsize < size) { | |
4343 | if (verbose) | |
4344 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
4345 | dev, maxsize, size); | |
4346 | return 0; | |
4347 | } | |
4348 | ||
4349 | /* count total number of extents for merge */ | |
4350 | i = 0; | |
4351 | for (dl = super->disks; dl; dl = dl->next) | |
4352 | if (dl->e) | |
4353 | i += dl->extent_cnt; | |
4354 | ||
4355 | maxsize = merge_extents(super, i); | |
a7dd165b | 4356 | if (maxsize < size || maxsize == 0) { |
0dcecb2e DW |
4357 | if (verbose) |
4358 | fprintf(stderr, Name ": not enough space after merge (%llu < %llu)\n", | |
4359 | maxsize, size); | |
4360 | return 0; | |
0dcecb2e DW |
4361 | } |
4362 | ||
c2c087e6 DW |
4363 | *freesize = maxsize; |
4364 | ||
4365 | return 1; | |
cdddbdbc DW |
4366 | } |
4367 | ||
efb30e7f DW |
4368 | static int reserve_space(struct supertype *st, int raiddisks, |
4369 | unsigned long long size, int chunk, | |
4370 | unsigned long long *freesize) | |
4371 | { | |
4372 | struct intel_super *super = st->sb; | |
4373 | struct imsm_super *mpb = super->anchor; | |
4374 | struct dl *dl; | |
4375 | int i; | |
4376 | int extent_cnt; | |
4377 | struct extent *e; | |
4378 | unsigned long long maxsize; | |
4379 | unsigned long long minsize; | |
4380 | int cnt; | |
4381 | int used; | |
4382 | ||
4383 | /* find the largest common start free region of the possible disks */ | |
4384 | used = 0; | |
4385 | extent_cnt = 0; | |
4386 | cnt = 0; | |
4387 | for (dl = super->disks; dl; dl = dl->next) { | |
4388 | dl->raiddisk = -1; | |
4389 | ||
4390 | if (dl->index >= 0) | |
4391 | used++; | |
4392 | ||
4393 | /* don't activate new spares if we are orom constrained | |
4394 | * and there is already a volume active in the container | |
4395 | */ | |
4396 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
4397 | continue; | |
4398 | ||
4399 | e = get_extents(super, dl); | |
4400 | if (!e) | |
4401 | continue; | |
4402 | for (i = 1; e[i-1].size; i++) | |
4403 | ; | |
4404 | dl->e = e; | |
4405 | dl->extent_cnt = i; | |
4406 | extent_cnt += i; | |
4407 | cnt++; | |
4408 | } | |
4409 | ||
4410 | maxsize = merge_extents(super, extent_cnt); | |
4411 | minsize = size; | |
4412 | if (size == 0) | |
612e59d8 CA |
4413 | /* chunk is in K */ |
4414 | minsize = chunk * 2; | |
efb30e7f DW |
4415 | |
4416 | if (cnt < raiddisks || | |
4417 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
4418 | maxsize < minsize || |
4419 | maxsize == 0) { | |
efb30e7f DW |
4420 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
4421 | return 0; /* No enough free spaces large enough */ | |
4422 | } | |
4423 | ||
4424 | if (size == 0) { | |
4425 | size = maxsize; | |
4426 | if (chunk) { | |
612e59d8 CA |
4427 | size /= 2 * chunk; |
4428 | size *= 2 * chunk; | |
efb30e7f DW |
4429 | } |
4430 | } | |
4431 | ||
4432 | cnt = 0; | |
4433 | for (dl = super->disks; dl; dl = dl->next) | |
4434 | if (dl->e) | |
4435 | dl->raiddisk = cnt++; | |
4436 | ||
4437 | *freesize = size; | |
4438 | ||
4439 | return 1; | |
4440 | } | |
4441 | ||
bf5a934a | 4442 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 4443 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
4444 | char *dev, unsigned long long *freesize, |
4445 | int verbose) | |
4446 | { | |
4447 | int fd, cfd; | |
4448 | struct mdinfo *sra; | |
20cbe8d2 | 4449 | int is_member = 0; |
bf5a934a DW |
4450 | |
4451 | /* if given unused devices create a container | |
4452 | * if given given devices in a container create a member volume | |
4453 | */ | |
4454 | if (level == LEVEL_CONTAINER) { | |
4455 | /* Must be a fresh device to add to a container */ | |
4456 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
4457 | raiddisks, |
4458 | chunk?*chunk:0, size, | |
bf5a934a DW |
4459 | dev, freesize, |
4460 | verbose); | |
4461 | } | |
4462 | ||
8592f29d N |
4463 | if (!dev) { |
4464 | if (st->sb && freesize) { | |
efb30e7f DW |
4465 | /* we are being asked to automatically layout a |
4466 | * new volume based on the current contents of | |
4467 | * the container. If the the parameters can be | |
4468 | * satisfied reserve_space will record the disks, | |
4469 | * start offset, and size of the volume to be | |
4470 | * created. add_to_super and getinfo_super | |
4471 | * detect when autolayout is in progress. | |
4472 | */ | |
6592ce37 DW |
4473 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
4474 | raiddisks, chunk, | |
4475 | verbose)) | |
4476 | return 0; | |
c21e737b CA |
4477 | return reserve_space(st, raiddisks, size, |
4478 | chunk?*chunk:0, freesize); | |
8592f29d N |
4479 | } |
4480 | return 1; | |
4481 | } | |
bf5a934a DW |
4482 | if (st->sb) { |
4483 | /* creating in a given container */ | |
4484 | return validate_geometry_imsm_volume(st, level, layout, | |
4485 | raiddisks, chunk, size, | |
4486 | dev, freesize, verbose); | |
4487 | } | |
4488 | ||
bf5a934a DW |
4489 | /* This device needs to be a device in an 'imsm' container */ |
4490 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4491 | if (fd >= 0) { | |
4492 | if (verbose) | |
4493 | fprintf(stderr, | |
4494 | Name ": Cannot create this array on device %s\n", | |
4495 | dev); | |
4496 | close(fd); | |
4497 | return 0; | |
4498 | } | |
4499 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
4500 | if (verbose) | |
4501 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
4502 | dev, strerror(errno)); | |
4503 | return 0; | |
4504 | } | |
4505 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
4506 | cfd = open_container(fd); | |
20cbe8d2 | 4507 | close(fd); |
bf5a934a | 4508 | if (cfd < 0) { |
bf5a934a DW |
4509 | if (verbose) |
4510 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
4511 | dev); | |
4512 | return 0; | |
4513 | } | |
4514 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 4515 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
4516 | strcmp(sra->text_version, "imsm") == 0) |
4517 | is_member = 1; | |
4518 | sysfs_free(sra); | |
4519 | if (is_member) { | |
bf5a934a DW |
4520 | /* This is a member of a imsm container. Load the container |
4521 | * and try to create a volume | |
4522 | */ | |
4523 | struct intel_super *super; | |
4524 | ||
e1902a7b | 4525 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL) == 0) { |
bf5a934a DW |
4526 | st->sb = super; |
4527 | st->container_dev = fd2devnum(cfd); | |
4528 | close(cfd); | |
4529 | return validate_geometry_imsm_volume(st, level, layout, | |
4530 | raiddisks, chunk, | |
4531 | size, dev, | |
4532 | freesize, verbose); | |
4533 | } | |
20cbe8d2 | 4534 | } |
bf5a934a | 4535 | |
20cbe8d2 AW |
4536 | if (verbose) |
4537 | fprintf(stderr, Name ": failed container membership check\n"); | |
4538 | ||
4539 | close(cfd); | |
4540 | return 0; | |
bf5a934a | 4541 | } |
0bd16cf2 | 4542 | |
30f58b22 | 4543 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
4544 | { |
4545 | struct intel_super *super = st->sb; | |
4546 | ||
30f58b22 DW |
4547 | if (level && *level == UnSet) |
4548 | *level = LEVEL_CONTAINER; | |
4549 | ||
4550 | if (level && layout && *layout == UnSet) | |
4551 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 4552 | |
1d54f286 N |
4553 | if (chunk && (*chunk == UnSet || *chunk == 0) && |
4554 | super && super->orom) | |
30f58b22 | 4555 | *chunk = imsm_orom_default_chunk(super->orom); |
0bd16cf2 DJ |
4556 | } |
4557 | ||
33414a01 DW |
4558 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
4559 | ||
4560 | static int kill_subarray_imsm(struct supertype *st) | |
4561 | { | |
4562 | /* remove the subarray currently referenced by ->current_vol */ | |
4563 | __u8 i; | |
4564 | struct intel_dev **dp; | |
4565 | struct intel_super *super = st->sb; | |
4566 | __u8 current_vol = super->current_vol; | |
4567 | struct imsm_super *mpb = super->anchor; | |
4568 | ||
4569 | if (super->current_vol < 0) | |
4570 | return 2; | |
4571 | super->current_vol = -1; /* invalidate subarray cursor */ | |
4572 | ||
4573 | /* block deletions that would change the uuid of active subarrays | |
4574 | * | |
4575 | * FIXME when immutable ids are available, but note that we'll | |
4576 | * also need to fixup the invalidated/active subarray indexes in | |
4577 | * mdstat | |
4578 | */ | |
4579 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4580 | char subarray[4]; | |
4581 | ||
4582 | if (i < current_vol) | |
4583 | continue; | |
4584 | sprintf(subarray, "%u", i); | |
4585 | if (is_subarray_active(subarray, st->devname)) { | |
4586 | fprintf(stderr, | |
4587 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
4588 | current_vol, i); | |
4589 | ||
4590 | return 2; | |
4591 | } | |
4592 | } | |
4593 | ||
4594 | if (st->update_tail) { | |
4595 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
4596 | ||
4597 | if (!u) | |
4598 | return 2; | |
4599 | u->type = update_kill_array; | |
4600 | u->dev_idx = current_vol; | |
4601 | append_metadata_update(st, u, sizeof(*u)); | |
4602 | ||
4603 | return 0; | |
4604 | } | |
4605 | ||
4606 | for (dp = &super->devlist; *dp;) | |
4607 | if ((*dp)->index == current_vol) { | |
4608 | *dp = (*dp)->next; | |
4609 | } else { | |
4610 | handle_missing(super, (*dp)->dev); | |
4611 | if ((*dp)->index > current_vol) | |
4612 | (*dp)->index--; | |
4613 | dp = &(*dp)->next; | |
4614 | } | |
4615 | ||
4616 | /* no more raid devices, all active components are now spares, | |
4617 | * but of course failed are still failed | |
4618 | */ | |
4619 | if (--mpb->num_raid_devs == 0) { | |
4620 | struct dl *d; | |
4621 | ||
4622 | for (d = super->disks; d; d = d->next) | |
4623 | if (d->index > -2) { | |
4624 | d->index = -1; | |
4625 | d->disk.status = SPARE_DISK; | |
4626 | } | |
4627 | } | |
4628 | ||
4629 | super->updates_pending++; | |
4630 | ||
4631 | return 0; | |
4632 | } | |
aa534678 | 4633 | |
a951a4f7 | 4634 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 4635 | char *update, struct mddev_ident *ident) |
aa534678 DW |
4636 | { |
4637 | /* update the subarray currently referenced by ->current_vol */ | |
4638 | struct intel_super *super = st->sb; | |
4639 | struct imsm_super *mpb = super->anchor; | |
4640 | ||
aa534678 DW |
4641 | if (strcmp(update, "name") == 0) { |
4642 | char *name = ident->name; | |
a951a4f7 N |
4643 | char *ep; |
4644 | int vol; | |
aa534678 | 4645 | |
a951a4f7 | 4646 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
4647 | fprintf(stderr, |
4648 | Name ": Unable to update name of active subarray\n"); | |
4649 | return 2; | |
4650 | } | |
4651 | ||
4652 | if (!check_name(super, name, 0)) | |
4653 | return 2; | |
4654 | ||
a951a4f7 N |
4655 | vol = strtoul(subarray, &ep, 10); |
4656 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
4657 | return 2; | |
4658 | ||
aa534678 DW |
4659 | if (st->update_tail) { |
4660 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
4661 | ||
4662 | if (!u) | |
4663 | return 2; | |
4664 | u->type = update_rename_array; | |
a951a4f7 | 4665 | u->dev_idx = vol; |
aa534678 DW |
4666 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
4667 | append_metadata_update(st, u, sizeof(*u)); | |
4668 | } else { | |
4669 | struct imsm_dev *dev; | |
4670 | int i; | |
4671 | ||
a951a4f7 | 4672 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
4673 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
4674 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4675 | dev = get_imsm_dev(super, i); | |
4676 | handle_missing(super, dev); | |
4677 | } | |
4678 | super->updates_pending++; | |
4679 | } | |
4680 | } else | |
4681 | return 2; | |
4682 | ||
4683 | return 0; | |
4684 | } | |
bf5a934a | 4685 | |
28bce06f AK |
4686 | static int is_gen_migration(struct imsm_dev *dev) |
4687 | { | |
4688 | if (!dev->vol.migr_state) | |
4689 | return 0; | |
4690 | ||
4691 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
4692 | return 1; | |
4693 | ||
4694 | return 0; | |
4695 | } | |
71204a50 | 4696 | #endif /* MDASSEMBLE */ |
28bce06f | 4697 | |
1e5c6983 DW |
4698 | static int is_rebuilding(struct imsm_dev *dev) |
4699 | { | |
4700 | struct imsm_map *migr_map; | |
4701 | ||
4702 | if (!dev->vol.migr_state) | |
4703 | return 0; | |
4704 | ||
4705 | if (migr_type(dev) != MIGR_REBUILD) | |
4706 | return 0; | |
4707 | ||
4708 | migr_map = get_imsm_map(dev, 1); | |
4709 | ||
4710 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
4711 | return 1; | |
4712 | else | |
4713 | return 0; | |
4714 | } | |
4715 | ||
4716 | static void update_recovery_start(struct imsm_dev *dev, struct mdinfo *array) | |
4717 | { | |
4718 | struct mdinfo *rebuild = NULL; | |
4719 | struct mdinfo *d; | |
4720 | __u32 units; | |
4721 | ||
4722 | if (!is_rebuilding(dev)) | |
4723 | return; | |
4724 | ||
4725 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
4726 | for (d = array->devs; d; d = d->next) | |
4727 | if (d->recovery_start == 0) { | |
4728 | if (rebuild) | |
4729 | return; | |
4730 | rebuild = d; | |
4731 | } | |
4732 | ||
4363fd80 DW |
4733 | if (!rebuild) { |
4734 | /* (?) none of the disks are marked with | |
4735 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
4736 | * disk_ord_tbl was not correctly updated | |
4737 | */ | |
4738 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
4739 | return; | |
4740 | } | |
4741 | ||
1e5c6983 DW |
4742 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
4743 | rebuild->recovery_start = units * blocks_per_migr_unit(dev); | |
4744 | } | |
4745 | ||
4746 | ||
00bbdbda | 4747 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 4748 | { |
4f5bc454 DW |
4749 | /* Given a container loaded by load_super_imsm_all, |
4750 | * extract information about all the arrays into | |
4751 | * an mdinfo tree. | |
00bbdbda | 4752 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
4753 | * |
4754 | * For each imsm_dev create an mdinfo, fill it in, | |
4755 | * then look for matching devices in super->disks | |
4756 | * and create appropriate device mdinfo. | |
4757 | */ | |
4758 | struct intel_super *super = st->sb; | |
949c47a0 | 4759 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 4760 | struct mdinfo *rest = NULL; |
00bbdbda | 4761 | unsigned int i; |
a06d022d | 4762 | int bbm_errors = 0; |
abef11a3 AK |
4763 | struct dl *d; |
4764 | int spare_disks = 0; | |
cdddbdbc | 4765 | |
a06d022d KW |
4766 | /* check for bad blocks */ |
4767 | if (imsm_bbm_log_size(super->anchor)) | |
4768 | bbm_errors = 1; | |
604b746f | 4769 | |
abef11a3 AK |
4770 | /* count spare devices, not used in maps |
4771 | */ | |
4772 | for (d = super->disks; d; d = d->next) | |
4773 | if (d->index == -1) | |
4774 | spare_disks++; | |
4775 | ||
4f5bc454 | 4776 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
4777 | struct imsm_dev *dev; |
4778 | struct imsm_map *map; | |
86e3692b | 4779 | struct imsm_map *map2; |
4f5bc454 | 4780 | struct mdinfo *this; |
4f5bc454 | 4781 | int slot; |
00bbdbda N |
4782 | char *ep; |
4783 | ||
4784 | if (subarray && | |
4785 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
4786 | continue; | |
4787 | ||
4788 | dev = get_imsm_dev(super, i); | |
4789 | map = get_imsm_map(dev, 0); | |
86e3692b | 4790 | map2 = get_imsm_map(dev, 1); |
4f5bc454 | 4791 | |
1ce0101c DW |
4792 | /* do not publish arrays that are in the middle of an |
4793 | * unsupported migration | |
4794 | */ | |
4795 | if (dev->vol.migr_state && | |
28bce06f | 4796 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
4797 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
4798 | " unsupported migration in progress\n", | |
4799 | dev->volume); | |
4800 | continue; | |
4801 | } | |
4802 | ||
4f5bc454 | 4803 | this = malloc(sizeof(*this)); |
0fbd635c | 4804 | if (!this) { |
cf1be220 | 4805 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
4806 | sizeof(*this)); |
4807 | break; | |
4808 | } | |
4f5bc454 DW |
4809 | memset(this, 0, sizeof(*this)); |
4810 | this->next = rest; | |
4f5bc454 | 4811 | |
301406c9 | 4812 | super->current_vol = i; |
a5d85af7 | 4813 | getinfo_super_imsm_volume(st, this, NULL); |
4f5bc454 | 4814 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 4815 | unsigned long long recovery_start; |
4f5bc454 DW |
4816 | struct mdinfo *info_d; |
4817 | struct dl *d; | |
4818 | int idx; | |
9a1608e5 | 4819 | int skip; |
7eef0453 | 4820 | __u32 ord; |
4f5bc454 | 4821 | |
9a1608e5 | 4822 | skip = 0; |
98130f40 | 4823 | idx = get_imsm_disk_idx(dev, slot, 0); |
196b0d44 | 4824 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
4f5bc454 DW |
4825 | for (d = super->disks; d ; d = d->next) |
4826 | if (d->index == idx) | |
0fbd635c | 4827 | break; |
4f5bc454 | 4828 | |
1e5c6983 | 4829 | recovery_start = MaxSector; |
4f5bc454 | 4830 | if (d == NULL) |
9a1608e5 | 4831 | skip = 1; |
25ed7e59 | 4832 | if (d && is_failed(&d->disk)) |
9a1608e5 | 4833 | skip = 1; |
7eef0453 | 4834 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 4835 | recovery_start = 0; |
9a1608e5 DW |
4836 | |
4837 | /* | |
4838 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
4839 | * reset resync start to avoid a dirty-degraded |
4840 | * situation when performing the intial sync | |
9a1608e5 DW |
4841 | * |
4842 | * FIXME handle dirty degraded | |
4843 | */ | |
1e5c6983 | 4844 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 4845 | this->resync_start = MaxSector; |
9a1608e5 DW |
4846 | if (skip) |
4847 | continue; | |
4f5bc454 | 4848 | |
1e5c6983 | 4849 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
4850 | if (!info_d) { |
4851 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 4852 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
4853 | info_d = this->devs; |
4854 | while (info_d) { | |
4855 | struct mdinfo *d = info_d->next; | |
4856 | ||
4857 | free(info_d); | |
4858 | info_d = d; | |
4859 | } | |
9a1608e5 DW |
4860 | free(this); |
4861 | this = rest; | |
4862 | break; | |
4863 | } | |
4f5bc454 DW |
4864 | info_d->next = this->devs; |
4865 | this->devs = info_d; | |
4866 | ||
4f5bc454 DW |
4867 | info_d->disk.number = d->index; |
4868 | info_d->disk.major = d->major; | |
4869 | info_d->disk.minor = d->minor; | |
4870 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 4871 | info_d->recovery_start = recovery_start; |
86e3692b AK |
4872 | if (map2) { |
4873 | if (slot < map2->num_members) | |
4874 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
4875 | else |
4876 | this->array.spare_disks++; | |
86e3692b AK |
4877 | } else { |
4878 | if (slot < map->num_members) | |
4879 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
4880 | else |
4881 | this->array.spare_disks++; | |
86e3692b | 4882 | } |
1e5c6983 DW |
4883 | if (info_d->recovery_start == MaxSector) |
4884 | this->array.working_disks++; | |
4f5bc454 DW |
4885 | |
4886 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
4887 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
4888 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
4f5bc454 | 4889 | } |
1e5c6983 DW |
4890 | /* now that the disk list is up-to-date fixup recovery_start */ |
4891 | update_recovery_start(dev, this); | |
abef11a3 | 4892 | this->array.spare_disks += spare_disks; |
9a1608e5 | 4893 | rest = this; |
4f5bc454 DW |
4894 | } |
4895 | ||
a06d022d KW |
4896 | /* if array has bad blocks, set suitable bit in array status */ |
4897 | if (bbm_errors) | |
4898 | rest->array.state |= (1<<MD_SB_BBM_ERRORS); | |
4899 | ||
4f5bc454 | 4900 | return rest; |
cdddbdbc DW |
4901 | } |
4902 | ||
845dea95 | 4903 | |
fb49eef2 | 4904 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed) |
c2a1e7da | 4905 | { |
a965f303 | 4906 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
4907 | |
4908 | if (!failed) | |
3393c6af DW |
4909 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
4910 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
4911 | |
4912 | switch (get_imsm_raid_level(map)) { | |
4913 | case 0: | |
4914 | return IMSM_T_STATE_FAILED; | |
4915 | break; | |
4916 | case 1: | |
4917 | if (failed < map->num_members) | |
4918 | return IMSM_T_STATE_DEGRADED; | |
4919 | else | |
4920 | return IMSM_T_STATE_FAILED; | |
4921 | break; | |
4922 | case 10: | |
4923 | { | |
4924 | /** | |
c92a2527 DW |
4925 | * check to see if any mirrors have failed, otherwise we |
4926 | * are degraded. Even numbered slots are mirrored on | |
4927 | * slot+1 | |
c2a1e7da | 4928 | */ |
c2a1e7da | 4929 | int i; |
d9b420a5 N |
4930 | /* gcc -Os complains that this is unused */ |
4931 | int insync = insync; | |
c2a1e7da DW |
4932 | |
4933 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 4934 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
c92a2527 DW |
4935 | int idx = ord_to_idx(ord); |
4936 | struct imsm_disk *disk; | |
c2a1e7da | 4937 | |
c92a2527 DW |
4938 | /* reset the potential in-sync count on even-numbered |
4939 | * slots. num_copies is always 2 for imsm raid10 | |
4940 | */ | |
4941 | if ((i & 1) == 0) | |
4942 | insync = 2; | |
c2a1e7da | 4943 | |
c92a2527 | 4944 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 4945 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 4946 | insync--; |
c2a1e7da | 4947 | |
c92a2527 DW |
4948 | /* no in-sync disks left in this mirror the |
4949 | * array has failed | |
4950 | */ | |
4951 | if (insync == 0) | |
4952 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
4953 | } |
4954 | ||
4955 | return IMSM_T_STATE_DEGRADED; | |
4956 | } | |
4957 | case 5: | |
4958 | if (failed < 2) | |
4959 | return IMSM_T_STATE_DEGRADED; | |
4960 | else | |
4961 | return IMSM_T_STATE_FAILED; | |
4962 | break; | |
4963 | default: | |
4964 | break; | |
4965 | } | |
4966 | ||
4967 | return map->map_state; | |
4968 | } | |
4969 | ||
ff077194 | 4970 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev) |
c2a1e7da DW |
4971 | { |
4972 | int i; | |
4973 | int failed = 0; | |
4974 | struct imsm_disk *disk; | |
ff077194 | 4975 | struct imsm_map *map = get_imsm_map(dev, 0); |
0556e1a2 DW |
4976 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
4977 | __u32 ord; | |
4978 | int idx; | |
c2a1e7da | 4979 | |
0556e1a2 DW |
4980 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
4981 | * disks that are being rebuilt. New failures are recorded to | |
4982 | * map[0]. So we look through all the disks we started with and | |
4983 | * see if any failures are still present, or if any new ones | |
4984 | * have arrived | |
4985 | * | |
4986 | * FIXME add support for online capacity expansion and | |
4987 | * raid-level-migration | |
4988 | */ | |
4989 | for (i = 0; i < prev->num_members; i++) { | |
4990 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); | |
4991 | ord |= __le32_to_cpu(map->disk_ord_tbl[i]); | |
4992 | idx = ord_to_idx(ord); | |
c2a1e7da | 4993 | |
949c47a0 | 4994 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 4995 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
fcb84475 | 4996 | failed++; |
c2a1e7da DW |
4997 | } |
4998 | ||
4999 | return failed; | |
845dea95 NB |
5000 | } |
5001 | ||
97b4d0e9 DW |
5002 | #ifndef MDASSEMBLE |
5003 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
5004 | char *inst) | |
5005 | { | |
5006 | struct intel_super *super = c->sb; | |
5007 | struct imsm_super *mpb = super->anchor; | |
5008 | ||
5009 | if (atoi(inst) >= mpb->num_raid_devs) { | |
5010 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
5011 | __func__, atoi(inst)); | |
5012 | return -ENODEV; | |
5013 | } | |
5014 | ||
5015 | dprintf("imsm: open_new %s\n", inst); | |
5016 | a->info.container_member = atoi(inst); | |
5017 | return 0; | |
5018 | } | |
5019 | ||
0c046afd DW |
5020 | static int is_resyncing(struct imsm_dev *dev) |
5021 | { | |
5022 | struct imsm_map *migr_map; | |
5023 | ||
5024 | if (!dev->vol.migr_state) | |
5025 | return 0; | |
5026 | ||
1484e727 DW |
5027 | if (migr_type(dev) == MIGR_INIT || |
5028 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
5029 | return 1; |
5030 | ||
4c9bc37b AK |
5031 | if (migr_type(dev) == MIGR_GEN_MIGR) |
5032 | return 0; | |
5033 | ||
0c046afd DW |
5034 | migr_map = get_imsm_map(dev, 1); |
5035 | ||
4c9bc37b AK |
5036 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
5037 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
5038 | return 1; |
5039 | else | |
5040 | return 0; | |
5041 | } | |
5042 | ||
0556e1a2 DW |
5043 | /* return true if we recorded new information */ |
5044 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 5045 | { |
0556e1a2 DW |
5046 | __u32 ord; |
5047 | int slot; | |
5048 | struct imsm_map *map; | |
1a2487c2 KW |
5049 | char buf[MAX_RAID_SERIAL_LEN+3]; |
5050 | unsigned int len, shift = 0; | |
0556e1a2 DW |
5051 | |
5052 | /* new failures are always set in map[0] */ | |
5053 | map = get_imsm_map(dev, 0); | |
5054 | ||
5055 | slot = get_imsm_disk_slot(map, idx); | |
5056 | if (slot < 0) | |
5057 | return 0; | |
5058 | ||
5059 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 5060 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
5061 | return 0; |
5062 | ||
1a2487c2 KW |
5063 | sprintf(buf, "%s:0", disk->serial); |
5064 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) | |
5065 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
5066 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
5067 | ||
f2f27e63 | 5068 | disk->status |= FAILED_DISK; |
0556e1a2 | 5069 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
f21e18ca | 5070 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
5071 | map->failed_disk_num = slot; |
5072 | return 1; | |
5073 | } | |
5074 | ||
5075 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
5076 | { | |
5077 | mark_failure(dev, disk, idx); | |
5078 | ||
5079 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
5080 | return; | |
5081 | ||
47ee5a45 DW |
5082 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
5083 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
5084 | } | |
5085 | ||
33414a01 DW |
5086 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
5087 | { | |
5088 | __u8 map_state; | |
5089 | struct dl *dl; | |
5090 | int failed; | |
5091 | ||
5092 | if (!super->missing) | |
5093 | return; | |
5094 | failed = imsm_count_failed(super, dev); | |
5095 | map_state = imsm_check_degraded(super, dev, failed); | |
5096 | ||
5097 | dprintf("imsm: mark missing\n"); | |
5098 | end_migration(dev, map_state); | |
5099 | for (dl = super->missing; dl; dl = dl->next) | |
5100 | mark_missing(dev, &dl->disk, dl->index); | |
5101 | super->updates_pending++; | |
5102 | } | |
5103 | ||
70bdf0dc AK |
5104 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev) |
5105 | { | |
5106 | int used_disks = imsm_num_data_members(dev, 0); | |
5107 | unsigned long long array_blocks; | |
5108 | struct imsm_map *map; | |
5109 | ||
5110 | if (used_disks == 0) { | |
5111 | /* when problems occures | |
5112 | * return current array_blocks value | |
5113 | */ | |
5114 | array_blocks = __le32_to_cpu(dev->size_high); | |
5115 | array_blocks = array_blocks << 32; | |
5116 | array_blocks += __le32_to_cpu(dev->size_low); | |
5117 | ||
5118 | return array_blocks; | |
5119 | } | |
5120 | ||
5121 | /* set array size in metadata | |
5122 | */ | |
5123 | map = get_imsm_map(dev, 0); | |
5124 | array_blocks = map->blocks_per_member * used_disks; | |
5125 | ||
5126 | /* round array size down to closest MB | |
5127 | */ | |
5128 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
5129 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
5130 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
5131 | ||
5132 | return array_blocks; | |
5133 | } | |
5134 | ||
28bce06f AK |
5135 | static void imsm_set_disk(struct active_array *a, int n, int state); |
5136 | ||
0e2d1a4e AK |
5137 | static void imsm_progress_container_reshape(struct intel_super *super) |
5138 | { | |
5139 | /* if no device has a migr_state, but some device has a | |
5140 | * different number of members than the previous device, start | |
5141 | * changing the number of devices in this device to match | |
5142 | * previous. | |
5143 | */ | |
5144 | struct imsm_super *mpb = super->anchor; | |
5145 | int prev_disks = -1; | |
5146 | int i; | |
1dfaa380 | 5147 | int copy_map_size; |
0e2d1a4e AK |
5148 | |
5149 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5150 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5151 | struct imsm_map *map = get_imsm_map(dev, 0); | |
5152 | struct imsm_map *map2; | |
5153 | int prev_num_members; | |
0e2d1a4e AK |
5154 | |
5155 | if (dev->vol.migr_state) | |
5156 | return; | |
5157 | ||
5158 | if (prev_disks == -1) | |
5159 | prev_disks = map->num_members; | |
5160 | if (prev_disks == map->num_members) | |
5161 | continue; | |
5162 | ||
5163 | /* OK, this array needs to enter reshape mode. | |
5164 | * i.e it needs a migr_state | |
5165 | */ | |
5166 | ||
1dfaa380 | 5167 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
5168 | prev_num_members = map->num_members; |
5169 | map->num_members = prev_disks; | |
5170 | dev->vol.migr_state = 1; | |
5171 | dev->vol.curr_migr_unit = 0; | |
5172 | dev->vol.migr_type = MIGR_GEN_MIGR; | |
5173 | for (i = prev_num_members; | |
5174 | i < map->num_members; i++) | |
5175 | set_imsm_ord_tbl_ent(map, i, i); | |
5176 | map2 = get_imsm_map(dev, 1); | |
5177 | /* Copy the current map */ | |
1dfaa380 | 5178 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
5179 | map2->num_members = prev_num_members; |
5180 | ||
70bdf0dc | 5181 | imsm_set_array_size(dev); |
0e2d1a4e AK |
5182 | super->updates_pending++; |
5183 | } | |
5184 | } | |
5185 | ||
aad6f216 | 5186 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
5187 | * states are handled in imsm_set_disk() with one exception, when a |
5188 | * resync is stopped due to a new failure this routine will set the | |
5189 | * 'degraded' state for the array. | |
5190 | */ | |
01f157d7 | 5191 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
5192 | { |
5193 | int inst = a->info.container_member; | |
5194 | struct intel_super *super = a->container->sb; | |
949c47a0 | 5195 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5196 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd DW |
5197 | int failed = imsm_count_failed(super, dev); |
5198 | __u8 map_state = imsm_check_degraded(super, dev, failed); | |
1e5c6983 | 5199 | __u32 blocks_per_unit; |
a862209d | 5200 | |
1af97990 AK |
5201 | if (dev->vol.migr_state && |
5202 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
5203 | /* array state change is blocked due to reshape action | |
aad6f216 N |
5204 | * We might need to |
5205 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
5206 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
5207 | * - update curr_migr_unit | |
1af97990 | 5208 | */ |
aad6f216 N |
5209 | if (a->curr_action == reshape) { |
5210 | /* still reshaping, maybe update curr_migr_unit */ | |
5211 | long long blocks_per_unit = blocks_per_migr_unit(dev); | |
5212 | long long unit = a->last_checkpoint; | |
6345120e AK |
5213 | if (blocks_per_unit) { |
5214 | unit /= blocks_per_unit; | |
5215 | if (unit > | |
5216 | __le32_to_cpu(dev->vol.curr_migr_unit)) { | |
5217 | dev->vol.curr_migr_unit = | |
5218 | __cpu_to_le32(unit); | |
5219 | super->updates_pending++; | |
5220 | } | |
aad6f216 | 5221 | } |
a4546b61 | 5222 | return 0; |
aad6f216 N |
5223 | } else { |
5224 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
5225 | /* for some reason we aborted the reshape. | |
5226 | * Better clean up | |
5227 | */ | |
5228 | struct imsm_map *map2 = get_imsm_map(dev, 1); | |
5229 | dev->vol.migr_state = 0; | |
5230 | dev->vol.migr_type = 0; | |
5231 | dev->vol.curr_migr_unit = 0; | |
5232 | memcpy(map, map2, sizeof_imsm_map(map2)); | |
5233 | super->updates_pending++; | |
5234 | } | |
5235 | if (a->last_checkpoint >= a->info.component_size) { | |
5236 | unsigned long long array_blocks; | |
5237 | int used_disks; | |
e154ced3 | 5238 | struct mdinfo *mdi; |
aad6f216 | 5239 | |
9653001d | 5240 | used_disks = imsm_num_data_members(dev, 0); |
d55adef9 AK |
5241 | if (used_disks > 0) { |
5242 | array_blocks = | |
5243 | map->blocks_per_member * | |
5244 | used_disks; | |
5245 | /* round array size down to closest MB | |
5246 | */ | |
5247 | array_blocks = (array_blocks | |
5248 | >> SECT_PER_MB_SHIFT) | |
5249 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
5250 | a->info.custom_array_size = array_blocks; |
5251 | /* encourage manager to update array | |
5252 | * size | |
5253 | */ | |
e154ced3 | 5254 | |
d55adef9 | 5255 | a->check_reshape = 1; |
f49ee92d | 5256 | } |
e154ced3 AK |
5257 | /* finalize online capacity expansion/reshape */ |
5258 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
5259 | imsm_set_disk(a, | |
5260 | mdi->disk.raid_disk, | |
5261 | mdi->curr_state); | |
5262 | ||
0e2d1a4e | 5263 | imsm_progress_container_reshape(super); |
e154ced3 | 5264 | } |
aad6f216 | 5265 | } |
1af97990 AK |
5266 | } |
5267 | ||
47ee5a45 | 5268 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
5269 | if (consistent == 2) |
5270 | handle_missing(super, dev); | |
1e5c6983 | 5271 | |
0c046afd | 5272 | if (consistent == 2 && |
b7941fd6 | 5273 | (!is_resync_complete(&a->info) || |
0c046afd DW |
5274 | map_state != IMSM_T_STATE_NORMAL || |
5275 | dev->vol.migr_state)) | |
01f157d7 | 5276 | consistent = 0; |
272906ef | 5277 | |
b7941fd6 | 5278 | if (is_resync_complete(&a->info)) { |
0c046afd | 5279 | /* complete intialization / resync, |
0556e1a2 DW |
5280 | * recovery and interrupted recovery is completed in |
5281 | * ->set_disk | |
0c046afd DW |
5282 | */ |
5283 | if (is_resyncing(dev)) { | |
5284 | dprintf("imsm: mark resync done\n"); | |
f8f603f1 | 5285 | end_migration(dev, map_state); |
115c3803 | 5286 | super->updates_pending++; |
484240d8 | 5287 | a->last_checkpoint = 0; |
115c3803 | 5288 | } |
0c046afd DW |
5289 | } else if (!is_resyncing(dev) && !failed) { |
5290 | /* mark the start of the init process if nothing is failed */ | |
b7941fd6 | 5291 | dprintf("imsm: mark resync start\n"); |
1484e727 | 5292 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
e3bba0e0 | 5293 | migrate(dev, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 DW |
5294 | else |
5295 | migrate(dev, IMSM_T_STATE_NORMAL, MIGR_REPAIR); | |
3393c6af | 5296 | super->updates_pending++; |
115c3803 | 5297 | } |
a862209d | 5298 | |
1e5c6983 DW |
5299 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
5300 | blocks_per_unit = blocks_per_migr_unit(dev); | |
4f0a7acc | 5301 | if (blocks_per_unit) { |
1e5c6983 DW |
5302 | __u32 units32; |
5303 | __u64 units; | |
5304 | ||
4f0a7acc | 5305 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
5306 | units32 = units; |
5307 | ||
5308 | /* check that we did not overflow 32-bits, and that | |
5309 | * curr_migr_unit needs updating | |
5310 | */ | |
5311 | if (units32 == units && | |
5312 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { | |
5313 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
5314 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
5315 | super->updates_pending++; | |
5316 | } | |
5317 | } | |
f8f603f1 | 5318 | |
3393c6af | 5319 | /* mark dirty / clean */ |
0c046afd | 5320 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 5321 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
5322 | if (consistent) |
5323 | dev->vol.dirty = 0; | |
5324 | else | |
5325 | dev->vol.dirty = 1; | |
a862209d DW |
5326 | super->updates_pending++; |
5327 | } | |
28bce06f | 5328 | |
01f157d7 | 5329 | return consistent; |
a862209d DW |
5330 | } |
5331 | ||
8d45d196 | 5332 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 5333 | { |
8d45d196 DW |
5334 | int inst = a->info.container_member; |
5335 | struct intel_super *super = a->container->sb; | |
949c47a0 | 5336 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5337 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 | 5338 | struct imsm_disk *disk; |
0c046afd | 5339 | int failed; |
b10b37b8 | 5340 | __u32 ord; |
0c046afd | 5341 | __u8 map_state; |
8d45d196 DW |
5342 | |
5343 | if (n > map->num_members) | |
5344 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
5345 | n, map->num_members - 1); | |
5346 | ||
5347 | if (n < 0) | |
5348 | return; | |
5349 | ||
4e6e574a | 5350 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 5351 | |
98130f40 | 5352 | ord = get_imsm_ord_tbl_ent(dev, n, -1); |
b10b37b8 | 5353 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 5354 | |
5802a811 | 5355 | /* check for new failures */ |
0556e1a2 DW |
5356 | if (state & DS_FAULTY) { |
5357 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
5358 | super->updates_pending++; | |
8d45d196 | 5359 | } |
47ee5a45 | 5360 | |
19859edc | 5361 | /* check if in_sync */ |
0556e1a2 | 5362 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
b10b37b8 DW |
5363 | struct imsm_map *migr_map = get_imsm_map(dev, 1); |
5364 | ||
5365 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
5366 | super->updates_pending++; |
5367 | } | |
8d45d196 | 5368 | |
0c046afd DW |
5369 | failed = imsm_count_failed(super, dev); |
5370 | map_state = imsm_check_degraded(super, dev, failed); | |
5802a811 | 5371 | |
0c046afd DW |
5372 | /* check if recovery complete, newly degraded, or failed */ |
5373 | if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) { | |
f8f603f1 | 5374 | end_migration(dev, map_state); |
0556e1a2 DW |
5375 | map = get_imsm_map(dev, 0); |
5376 | map->failed_disk_num = ~0; | |
0c046afd | 5377 | super->updates_pending++; |
484240d8 | 5378 | a->last_checkpoint = 0; |
0c046afd DW |
5379 | } else if (map_state == IMSM_T_STATE_DEGRADED && |
5380 | map->map_state != map_state && | |
5381 | !dev->vol.migr_state) { | |
5382 | dprintf("imsm: mark degraded\n"); | |
5383 | map->map_state = map_state; | |
5384 | super->updates_pending++; | |
484240d8 | 5385 | a->last_checkpoint = 0; |
0c046afd DW |
5386 | } else if (map_state == IMSM_T_STATE_FAILED && |
5387 | map->map_state != map_state) { | |
5388 | dprintf("imsm: mark failed\n"); | |
f8f603f1 | 5389 | end_migration(dev, map_state); |
0c046afd | 5390 | super->updates_pending++; |
484240d8 | 5391 | a->last_checkpoint = 0; |
28bce06f AK |
5392 | } else if (is_gen_migration(dev)) { |
5393 | dprintf("imsm: Detected General Migration in state: "); | |
5394 | if (map_state == IMSM_T_STATE_NORMAL) { | |
5395 | end_migration(dev, map_state); | |
5396 | map = get_imsm_map(dev, 0); | |
5397 | map->failed_disk_num = ~0; | |
5398 | dprintf("normal\n"); | |
5399 | } else { | |
5400 | if (map_state == IMSM_T_STATE_DEGRADED) { | |
5401 | printf("degraded\n"); | |
5402 | end_migration(dev, map_state); | |
5403 | } else { | |
5404 | dprintf("failed\n"); | |
5405 | } | |
5406 | map->map_state = map_state; | |
5407 | } | |
5408 | super->updates_pending++; | |
5802a811 | 5409 | } |
845dea95 NB |
5410 | } |
5411 | ||
f796af5d | 5412 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 5413 | { |
f796af5d | 5414 | void *buf = mpb; |
c2a1e7da DW |
5415 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
5416 | unsigned long long dsize; | |
5417 | unsigned long long sectors; | |
5418 | ||
5419 | get_dev_size(fd, NULL, &dsize); | |
5420 | ||
272f648f DW |
5421 | if (mpb_size > 512) { |
5422 | /* -1 to account for anchor */ | |
5423 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 5424 | |
272f648f DW |
5425 | /* write the extended mpb to the sectors preceeding the anchor */ |
5426 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
5427 | return 1; | |
c2a1e7da | 5428 | |
f21e18ca N |
5429 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
5430 | != 512 * sectors) | |
272f648f DW |
5431 | return 1; |
5432 | } | |
c2a1e7da | 5433 | |
272f648f DW |
5434 | /* first block is stored on second to last sector of the disk */ |
5435 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
5436 | return 1; |
5437 | ||
f796af5d | 5438 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
5439 | return 1; |
5440 | ||
c2a1e7da DW |
5441 | return 0; |
5442 | } | |
5443 | ||
2e735d19 | 5444 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 5445 | { |
2e735d19 | 5446 | struct intel_super *super = container->sb; |
c2a1e7da | 5447 | |
1a64be56 | 5448 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
5449 | if (!super->updates_pending) |
5450 | return; | |
5451 | ||
36988a3d | 5452 | write_super_imsm(container, 0); |
c2a1e7da DW |
5453 | |
5454 | super->updates_pending = 0; | |
845dea95 NB |
5455 | } |
5456 | ||
272906ef DW |
5457 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
5458 | { | |
5459 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 5460 | int i = get_imsm_disk_idx(dev, idx, -1); |
272906ef DW |
5461 | struct dl *dl; |
5462 | ||
5463 | for (dl = super->disks; dl; dl = dl->next) | |
5464 | if (dl->index == i) | |
5465 | break; | |
5466 | ||
25ed7e59 | 5467 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
5468 | dl = NULL; |
5469 | ||
5470 | if (dl) | |
5471 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
5472 | ||
5473 | return dl; | |
5474 | } | |
5475 | ||
a20d2ba5 | 5476 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
5477 | struct active_array *a, int activate_new, |
5478 | struct mdinfo *additional_test_list) | |
272906ef DW |
5479 | { |
5480 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 5481 | int idx = get_imsm_disk_idx(dev, slot, -1); |
a20d2ba5 DW |
5482 | struct imsm_super *mpb = super->anchor; |
5483 | struct imsm_map *map; | |
272906ef DW |
5484 | unsigned long long pos; |
5485 | struct mdinfo *d; | |
5486 | struct extent *ex; | |
a20d2ba5 | 5487 | int i, j; |
272906ef | 5488 | int found; |
569cc43f DW |
5489 | __u32 array_start = 0; |
5490 | __u32 array_end = 0; | |
272906ef | 5491 | struct dl *dl; |
6c932028 | 5492 | struct mdinfo *test_list; |
272906ef DW |
5493 | |
5494 | for (dl = super->disks; dl; dl = dl->next) { | |
5495 | /* If in this array, skip */ | |
5496 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
5497 | if (d->state_fd >= 0 && |
5498 | d->disk.major == dl->major && | |
272906ef | 5499 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
5500 | dprintf("%x:%x already in array\n", |
5501 | dl->major, dl->minor); | |
272906ef DW |
5502 | break; |
5503 | } | |
5504 | if (d) | |
5505 | continue; | |
6c932028 AK |
5506 | test_list = additional_test_list; |
5507 | while (test_list) { | |
5508 | if (test_list->disk.major == dl->major && | |
5509 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
5510 | dprintf("%x:%x already in additional test list\n", |
5511 | dl->major, dl->minor); | |
5512 | break; | |
5513 | } | |
6c932028 | 5514 | test_list = test_list->next; |
8ba77d32 | 5515 | } |
6c932028 | 5516 | if (test_list) |
8ba77d32 | 5517 | continue; |
272906ef | 5518 | |
e553d2a4 | 5519 | /* skip in use or failed drives */ |
25ed7e59 | 5520 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
5521 | dl->index == -2) { |
5522 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 5523 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
5524 | continue; |
5525 | } | |
5526 | ||
a20d2ba5 DW |
5527 | /* skip pure spares when we are looking for partially |
5528 | * assimilated drives | |
5529 | */ | |
5530 | if (dl->index == -1 && !activate_new) | |
5531 | continue; | |
5532 | ||
272906ef | 5533 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 5534 | * It needs to be able to cover all member volumes |
272906ef DW |
5535 | */ |
5536 | ex = get_extents(super, dl); | |
5537 | if (!ex) { | |
5538 | dprintf("cannot get extents\n"); | |
5539 | continue; | |
5540 | } | |
a20d2ba5 DW |
5541 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5542 | dev = get_imsm_dev(super, i); | |
5543 | map = get_imsm_map(dev, 0); | |
272906ef | 5544 | |
a20d2ba5 DW |
5545 | /* check if this disk is already a member of |
5546 | * this array | |
272906ef | 5547 | */ |
620b1713 | 5548 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
5549 | continue; |
5550 | ||
5551 | found = 0; | |
5552 | j = 0; | |
5553 | pos = 0; | |
5554 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
329c8278 DW |
5555 | array_end = array_start + |
5556 | __le32_to_cpu(map->blocks_per_member) - 1; | |
a20d2ba5 DW |
5557 | |
5558 | do { | |
5559 | /* check that we can start at pba_of_lba0 with | |
5560 | * blocks_per_member of space | |
5561 | */ | |
329c8278 | 5562 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
5563 | found = 1; |
5564 | break; | |
5565 | } | |
5566 | pos = ex[j].start + ex[j].size; | |
5567 | j++; | |
5568 | } while (ex[j-1].size); | |
5569 | ||
5570 | if (!found) | |
272906ef | 5571 | break; |
a20d2ba5 | 5572 | } |
272906ef DW |
5573 | |
5574 | free(ex); | |
a20d2ba5 | 5575 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
5576 | dprintf("%x:%x does not have %u to %u available\n", |
5577 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
5578 | /* No room */ |
5579 | continue; | |
a20d2ba5 DW |
5580 | } |
5581 | return dl; | |
272906ef DW |
5582 | } |
5583 | ||
5584 | return dl; | |
5585 | } | |
5586 | ||
95d07a2c LM |
5587 | |
5588 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
5589 | { | |
5590 | struct imsm_dev *dev2; | |
5591 | struct imsm_map *map; | |
5592 | struct dl *idisk; | |
5593 | int slot; | |
5594 | int idx; | |
5595 | __u8 state; | |
5596 | ||
5597 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
5598 | if (dev2) { | |
5599 | state = imsm_check_degraded(cont->sb, dev2, failed); | |
5600 | if (state == IMSM_T_STATE_FAILED) { | |
5601 | map = get_imsm_map(dev2, 0); | |
5602 | if (!map) | |
5603 | return 1; | |
5604 | for (slot = 0; slot < map->num_members; slot++) { | |
5605 | /* | |
5606 | * Check if failed disks are deleted from intel | |
5607 | * disk list or are marked to be deleted | |
5608 | */ | |
98130f40 | 5609 | idx = get_imsm_disk_idx(dev2, slot, -1); |
95d07a2c LM |
5610 | idisk = get_imsm_dl_disk(cont->sb, idx); |
5611 | /* | |
5612 | * Do not rebuild the array if failed disks | |
5613 | * from failed sub-array are not removed from | |
5614 | * container. | |
5615 | */ | |
5616 | if (idisk && | |
5617 | is_failed(&idisk->disk) && | |
5618 | (idisk->action != DISK_REMOVE)) | |
5619 | return 0; | |
5620 | } | |
5621 | } | |
5622 | } | |
5623 | return 1; | |
5624 | } | |
5625 | ||
88758e9d DW |
5626 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
5627 | struct metadata_update **updates) | |
5628 | { | |
5629 | /** | |
d23fe947 DW |
5630 | * Find a device with unused free space and use it to replace a |
5631 | * failed/vacant region in an array. We replace failed regions one a | |
5632 | * array at a time. The result is that a new spare disk will be added | |
5633 | * to the first failed array and after the monitor has finished | |
5634 | * propagating failures the remainder will be consumed. | |
88758e9d | 5635 | * |
d23fe947 DW |
5636 | * FIXME add a capability for mdmon to request spares from another |
5637 | * container. | |
88758e9d DW |
5638 | */ |
5639 | ||
5640 | struct intel_super *super = a->container->sb; | |
88758e9d | 5641 | int inst = a->info.container_member; |
949c47a0 | 5642 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5643 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
5644 | int failed = a->info.array.raid_disks; |
5645 | struct mdinfo *rv = NULL; | |
5646 | struct mdinfo *d; | |
5647 | struct mdinfo *di; | |
5648 | struct metadata_update *mu; | |
5649 | struct dl *dl; | |
5650 | struct imsm_update_activate_spare *u; | |
5651 | int num_spares = 0; | |
5652 | int i; | |
95d07a2c | 5653 | int allowed; |
88758e9d DW |
5654 | |
5655 | for (d = a->info.devs ; d ; d = d->next) { | |
5656 | if ((d->curr_state & DS_FAULTY) && | |
5657 | d->state_fd >= 0) | |
5658 | /* wait for Removal to happen */ | |
5659 | return NULL; | |
5660 | if (d->state_fd >= 0) | |
5661 | failed--; | |
5662 | } | |
5663 | ||
5664 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
5665 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 AK |
5666 | |
5667 | if (dev->vol.migr_state && | |
5668 | dev->vol.migr_type == MIGR_GEN_MIGR) | |
5669 | /* No repair during migration */ | |
5670 | return NULL; | |
5671 | ||
89c67882 AK |
5672 | if (a->info.array.level == 4) |
5673 | /* No repair for takeovered array | |
5674 | * imsm doesn't support raid4 | |
5675 | */ | |
5676 | return NULL; | |
5677 | ||
fb49eef2 | 5678 | if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
5679 | return NULL; |
5680 | ||
95d07a2c LM |
5681 | /* |
5682 | * If there are any failed disks check state of the other volume. | |
5683 | * Block rebuild if the another one is failed until failed disks | |
5684 | * are removed from container. | |
5685 | */ | |
5686 | if (failed) { | |
5687 | dprintf("found failed disks in %s, check if there another" | |
5688 | "failed sub-array.\n", | |
5689 | dev->volume); | |
5690 | /* check if states of the other volumes allow for rebuild */ | |
5691 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
5692 | if (i != inst) { | |
5693 | allowed = imsm_rebuild_allowed(a->container, | |
5694 | i, failed); | |
5695 | if (!allowed) | |
5696 | return NULL; | |
5697 | } | |
5698 | } | |
5699 | } | |
5700 | ||
88758e9d | 5701 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
5702 | for (i = 0; i < a->info.array.raid_disks; i++) { |
5703 | for (d = a->info.devs ; d ; d = d->next) | |
5704 | if (d->disk.raid_disk == i) | |
5705 | break; | |
5706 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
5707 | if (d && (d->state_fd >= 0)) | |
5708 | continue; | |
5709 | ||
272906ef | 5710 | /* |
a20d2ba5 DW |
5711 | * OK, this device needs recovery. Try to re-add the |
5712 | * previous occupant of this slot, if this fails see if | |
5713 | * we can continue the assimilation of a spare that was | |
5714 | * partially assimilated, finally try to activate a new | |
5715 | * spare. | |
272906ef DW |
5716 | */ |
5717 | dl = imsm_readd(super, i, a); | |
5718 | if (!dl) | |
8ba77d32 | 5719 | dl = imsm_add_spare(super, i, a, 0, NULL); |
a20d2ba5 | 5720 | if (!dl) |
8ba77d32 | 5721 | dl = imsm_add_spare(super, i, a, 1, NULL); |
272906ef DW |
5722 | if (!dl) |
5723 | continue; | |
5724 | ||
5725 | /* found a usable disk with enough space */ | |
5726 | di = malloc(sizeof(*di)); | |
79244939 DW |
5727 | if (!di) |
5728 | continue; | |
272906ef DW |
5729 | memset(di, 0, sizeof(*di)); |
5730 | ||
5731 | /* dl->index will be -1 in the case we are activating a | |
5732 | * pristine spare. imsm_process_update() will create a | |
5733 | * new index in this case. Once a disk is found to be | |
5734 | * failed in all member arrays it is kicked from the | |
5735 | * metadata | |
5736 | */ | |
5737 | di->disk.number = dl->index; | |
d23fe947 | 5738 | |
272906ef DW |
5739 | /* (ab)use di->devs to store a pointer to the device |
5740 | * we chose | |
5741 | */ | |
5742 | di->devs = (struct mdinfo *) dl; | |
5743 | ||
5744 | di->disk.raid_disk = i; | |
5745 | di->disk.major = dl->major; | |
5746 | di->disk.minor = dl->minor; | |
5747 | di->disk.state = 0; | |
d23534e4 | 5748 | di->recovery_start = 0; |
272906ef DW |
5749 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); |
5750 | di->component_size = a->info.component_size; | |
5751 | di->container_member = inst; | |
148acb7b | 5752 | super->random = random32(); |
272906ef DW |
5753 | di->next = rv; |
5754 | rv = di; | |
5755 | num_spares++; | |
5756 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
5757 | i, di->data_offset); | |
88758e9d | 5758 | |
272906ef | 5759 | break; |
88758e9d DW |
5760 | } |
5761 | ||
5762 | if (!rv) | |
5763 | /* No spares found */ | |
5764 | return rv; | |
5765 | /* Now 'rv' has a list of devices to return. | |
5766 | * Create a metadata_update record to update the | |
5767 | * disk_ord_tbl for the array | |
5768 | */ | |
5769 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
5770 | if (mu) { |
5771 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
5772 | if (mu->buf == NULL) { | |
5773 | free(mu); | |
5774 | mu = NULL; | |
5775 | } | |
5776 | } | |
5777 | if (!mu) { | |
5778 | while (rv) { | |
5779 | struct mdinfo *n = rv->next; | |
5780 | ||
5781 | free(rv); | |
5782 | rv = n; | |
5783 | } | |
5784 | return NULL; | |
5785 | } | |
5786 | ||
88758e9d | 5787 | mu->space = NULL; |
cb23f1f4 | 5788 | mu->space_list = NULL; |
88758e9d DW |
5789 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
5790 | mu->next = *updates; | |
5791 | u = (struct imsm_update_activate_spare *) mu->buf; | |
5792 | ||
5793 | for (di = rv ; di ; di = di->next) { | |
5794 | u->type = update_activate_spare; | |
d23fe947 DW |
5795 | u->dl = (struct dl *) di->devs; |
5796 | di->devs = NULL; | |
88758e9d DW |
5797 | u->slot = di->disk.raid_disk; |
5798 | u->array = inst; | |
5799 | u->next = u + 1; | |
5800 | u++; | |
5801 | } | |
5802 | (u-1)->next = NULL; | |
5803 | *updates = mu; | |
5804 | ||
5805 | return rv; | |
5806 | } | |
5807 | ||
54c2c1ea | 5808 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 5809 | { |
54c2c1ea DW |
5810 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
5811 | struct imsm_map *map = get_imsm_map(dev, 0); | |
5812 | struct imsm_map *new_map = get_imsm_map(&u->dev, 0); | |
5813 | struct disk_info *inf = get_disk_info(u); | |
5814 | struct imsm_disk *disk; | |
8273f55e DW |
5815 | int i; |
5816 | int j; | |
8273f55e | 5817 | |
54c2c1ea | 5818 | for (i = 0; i < map->num_members; i++) { |
98130f40 | 5819 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, -1)); |
54c2c1ea DW |
5820 | for (j = 0; j < new_map->num_members; j++) |
5821 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
5822 | return 1; |
5823 | } | |
5824 | ||
5825 | return 0; | |
5826 | } | |
5827 | ||
1a64be56 LM |
5828 | |
5829 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
5830 | { | |
5831 | struct dl *dl = NULL; | |
5832 | for (dl = super->disks; dl; dl = dl->next) | |
5833 | if ((dl->major == major) && (dl->minor == minor)) | |
5834 | return dl; | |
5835 | return NULL; | |
5836 | } | |
5837 | ||
5838 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
5839 | { | |
5840 | struct dl *prev = NULL; | |
5841 | struct dl *dl; | |
5842 | ||
5843 | prev = NULL; | |
5844 | for (dl = super->disks; dl; dl = dl->next) { | |
5845 | if ((dl->major == major) && (dl->minor == minor)) { | |
5846 | /* remove */ | |
5847 | if (prev) | |
5848 | prev->next = dl->next; | |
5849 | else | |
5850 | super->disks = dl->next; | |
5851 | dl->next = NULL; | |
5852 | __free_imsm_disk(dl); | |
5853 | dprintf("%s: removed %x:%x\n", | |
5854 | __func__, major, minor); | |
5855 | break; | |
5856 | } | |
5857 | prev = dl; | |
5858 | } | |
5859 | return 0; | |
5860 | } | |
5861 | ||
f21e18ca | 5862 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 5863 | |
1a64be56 LM |
5864 | static int add_remove_disk_update(struct intel_super *super) |
5865 | { | |
5866 | int check_degraded = 0; | |
5867 | struct dl *disk = NULL; | |
5868 | /* add/remove some spares to/from the metadata/contrainer */ | |
5869 | while (super->disk_mgmt_list) { | |
5870 | struct dl *disk_cfg; | |
5871 | ||
5872 | disk_cfg = super->disk_mgmt_list; | |
5873 | super->disk_mgmt_list = disk_cfg->next; | |
5874 | disk_cfg->next = NULL; | |
5875 | ||
5876 | if (disk_cfg->action == DISK_ADD) { | |
5877 | disk_cfg->next = super->disks; | |
5878 | super->disks = disk_cfg; | |
5879 | check_degraded = 1; | |
5880 | dprintf("%s: added %x:%x\n", | |
5881 | __func__, disk_cfg->major, | |
5882 | disk_cfg->minor); | |
5883 | } else if (disk_cfg->action == DISK_REMOVE) { | |
5884 | dprintf("Disk remove action processed: %x.%x\n", | |
5885 | disk_cfg->major, disk_cfg->minor); | |
5886 | disk = get_disk_super(super, | |
5887 | disk_cfg->major, | |
5888 | disk_cfg->minor); | |
5889 | if (disk) { | |
5890 | /* store action status */ | |
5891 | disk->action = DISK_REMOVE; | |
5892 | /* remove spare disks only */ | |
5893 | if (disk->index == -1) { | |
5894 | remove_disk_super(super, | |
5895 | disk_cfg->major, | |
5896 | disk_cfg->minor); | |
5897 | } | |
5898 | } | |
5899 | /* release allocate disk structure */ | |
5900 | __free_imsm_disk(disk_cfg); | |
5901 | } | |
5902 | } | |
5903 | return check_degraded; | |
5904 | } | |
5905 | ||
2e5dc010 N |
5906 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
5907 | struct intel_super *super, | |
5908 | void ***space_list) | |
5909 | { | |
5910 | struct dl *new_disk; | |
5911 | struct intel_dev *id; | |
5912 | int i; | |
5913 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 5914 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
5915 | void **tofree = NULL; |
5916 | int devices_to_reshape = 1; | |
5917 | struct imsm_super *mpb = super->anchor; | |
5918 | int ret_val = 0; | |
d098291a | 5919 | unsigned int dev_id; |
2e5dc010 | 5920 | |
ed7333bd | 5921 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
5922 | |
5923 | /* enable spares to use in array */ | |
5924 | for (i = 0; i < delta_disks; i++) { | |
5925 | new_disk = get_disk_super(super, | |
5926 | major(u->new_disks[i]), | |
5927 | minor(u->new_disks[i])); | |
ed7333bd AK |
5928 | dprintf("imsm: new disk for reshape is: %i:%i " |
5929 | "(%p, index = %i)\n", | |
2e5dc010 N |
5930 | major(u->new_disks[i]), minor(u->new_disks[i]), |
5931 | new_disk, new_disk->index); | |
5932 | if ((new_disk == NULL) || | |
5933 | ((new_disk->index >= 0) && | |
5934 | (new_disk->index < u->old_raid_disks))) | |
5935 | goto update_reshape_exit; | |
ee4beede | 5936 | new_disk->index = disk_count++; |
2e5dc010 N |
5937 | /* slot to fill in autolayout |
5938 | */ | |
5939 | new_disk->raiddisk = new_disk->index; | |
5940 | new_disk->disk.status |= | |
5941 | CONFIGURED_DISK; | |
5942 | new_disk->disk.status &= ~SPARE_DISK; | |
5943 | } | |
5944 | ||
ed7333bd AK |
5945 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
5946 | mpb->num_raid_devs); | |
2e5dc010 N |
5947 | /* manage changes in volume |
5948 | */ | |
d098291a | 5949 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
5950 | void **sp = *space_list; |
5951 | struct imsm_dev *newdev; | |
5952 | struct imsm_map *newmap, *oldmap; | |
5953 | ||
d098291a AK |
5954 | for (id = super->devlist ; id; id = id->next) { |
5955 | if (id->index == dev_id) | |
5956 | break; | |
5957 | } | |
5958 | if (id == NULL) | |
5959 | break; | |
2e5dc010 N |
5960 | if (!sp) |
5961 | continue; | |
5962 | *space_list = *sp; | |
5963 | newdev = (void*)sp; | |
5964 | /* Copy the dev, but not (all of) the map */ | |
5965 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
5966 | oldmap = get_imsm_map(id->dev, 0); | |
5967 | newmap = get_imsm_map(newdev, 0); | |
5968 | /* Copy the current map */ | |
5969 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
5970 | /* update one device only | |
5971 | */ | |
5972 | if (devices_to_reshape) { | |
ed7333bd AK |
5973 | dprintf("imsm: modifying subdev: %i\n", |
5974 | id->index); | |
2e5dc010 N |
5975 | devices_to_reshape--; |
5976 | newdev->vol.migr_state = 1; | |
5977 | newdev->vol.curr_migr_unit = 0; | |
5978 | newdev->vol.migr_type = MIGR_GEN_MIGR; | |
5979 | newmap->num_members = u->new_raid_disks; | |
5980 | for (i = 0; i < delta_disks; i++) { | |
5981 | set_imsm_ord_tbl_ent(newmap, | |
5982 | u->old_raid_disks + i, | |
5983 | u->old_raid_disks + i); | |
5984 | } | |
5985 | /* New map is correct, now need to save old map | |
5986 | */ | |
5987 | newmap = get_imsm_map(newdev, 1); | |
5988 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
5989 | ||
70bdf0dc | 5990 | imsm_set_array_size(newdev); |
2e5dc010 N |
5991 | } |
5992 | ||
5993 | sp = (void **)id->dev; | |
5994 | id->dev = newdev; | |
5995 | *sp = tofree; | |
5996 | tofree = sp; | |
5997 | } | |
819bc634 AK |
5998 | if (tofree) |
5999 | *space_list = tofree; | |
2e5dc010 N |
6000 | ret_val = 1; |
6001 | ||
6002 | update_reshape_exit: | |
6003 | ||
6004 | return ret_val; | |
6005 | } | |
6006 | ||
bb025c2f | 6007 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
6008 | struct intel_super *super, |
6009 | void ***space_list) | |
bb025c2f KW |
6010 | { |
6011 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
6012 | struct intel_dev *dv; |
6013 | struct imsm_dev *dev_new; | |
bb025c2f KW |
6014 | struct imsm_map *map; |
6015 | struct dl *dm, *du; | |
8ca6df95 | 6016 | int i; |
bb025c2f KW |
6017 | |
6018 | for (dv = super->devlist; dv; dv = dv->next) | |
6019 | if (dv->index == (unsigned int)u->subarray) { | |
6020 | dev = dv->dev; | |
6021 | break; | |
6022 | } | |
6023 | ||
6024 | if (dev == NULL) | |
6025 | return 0; | |
6026 | ||
6027 | map = get_imsm_map(dev, 0); | |
6028 | ||
6029 | if (u->direction == R10_TO_R0) { | |
43d5ec18 KW |
6030 | /* Number of failed disks must be half of initial disk number */ |
6031 | if (imsm_count_failed(super, dev) != (map->num_members / 2)) | |
6032 | return 0; | |
6033 | ||
bb025c2f KW |
6034 | /* iterate through devices to mark removed disks as spare */ |
6035 | for (dm = super->disks; dm; dm = dm->next) { | |
6036 | if (dm->disk.status & FAILED_DISK) { | |
6037 | int idx = dm->index; | |
6038 | /* update indexes on the disk list */ | |
6039 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
6040 | the index values will end up being correct.... NB */ | |
6041 | for (du = super->disks; du; du = du->next) | |
6042 | if (du->index > idx) | |
6043 | du->index--; | |
6044 | /* mark as spare disk */ | |
6045 | dm->disk.status = SPARE_DISK; | |
6046 | dm->index = -1; | |
6047 | } | |
6048 | } | |
bb025c2f KW |
6049 | /* update map */ |
6050 | map->num_members = map->num_members / 2; | |
6051 | map->map_state = IMSM_T_STATE_NORMAL; | |
6052 | map->num_domains = 1; | |
6053 | map->raid_level = 0; | |
6054 | map->failed_disk_num = -1; | |
6055 | } | |
6056 | ||
8ca6df95 KW |
6057 | if (u->direction == R0_TO_R10) { |
6058 | void **space; | |
6059 | /* update slots in current disk list */ | |
6060 | for (dm = super->disks; dm; dm = dm->next) { | |
6061 | if (dm->index >= 0) | |
6062 | dm->index *= 2; | |
6063 | } | |
6064 | /* create new *missing* disks */ | |
6065 | for (i = 0; i < map->num_members; i++) { | |
6066 | space = *space_list; | |
6067 | if (!space) | |
6068 | continue; | |
6069 | *space_list = *space; | |
6070 | du = (void *)space; | |
6071 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
6072 | du->fd = -1; |
6073 | du->minor = 0; | |
6074 | du->major = 0; | |
6075 | du->index = (i * 2) + 1; | |
6076 | sprintf((char *)du->disk.serial, | |
6077 | " MISSING_%d", du->index); | |
6078 | sprintf((char *)du->serial, | |
6079 | "MISSING_%d", du->index); | |
6080 | du->next = super->missing; | |
6081 | super->missing = du; | |
6082 | } | |
6083 | /* create new dev and map */ | |
6084 | space = *space_list; | |
6085 | if (!space) | |
6086 | return 0; | |
6087 | *space_list = *space; | |
6088 | dev_new = (void *)space; | |
6089 | memcpy(dev_new, dev, sizeof(*dev)); | |
6090 | /* update new map */ | |
6091 | map = get_imsm_map(dev_new, 0); | |
8ca6df95 | 6092 | map->num_members = map->num_members * 2; |
1a2487c2 | 6093 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
6094 | map->num_domains = 2; |
6095 | map->raid_level = 1; | |
6096 | /* replace dev<->dev_new */ | |
6097 | dv->dev = dev_new; | |
6098 | } | |
bb025c2f KW |
6099 | /* update disk order table */ |
6100 | for (du = super->disks; du; du = du->next) | |
6101 | if (du->index >= 0) | |
6102 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 6103 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
6104 | if (du->index >= 0) { |
6105 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
6106 | mark_missing(dev_new, &du->disk, du->index); | |
6107 | } | |
bb025c2f KW |
6108 | |
6109 | return 1; | |
6110 | } | |
6111 | ||
e8319a19 DW |
6112 | static void imsm_process_update(struct supertype *st, |
6113 | struct metadata_update *update) | |
6114 | { | |
6115 | /** | |
6116 | * crack open the metadata_update envelope to find the update record | |
6117 | * update can be one of: | |
d195167d AK |
6118 | * update_reshape_container_disks - all the arrays in the container |
6119 | * are being reshaped to have more devices. We need to mark | |
6120 | * the arrays for general migration and convert selected spares | |
6121 | * into active devices. | |
6122 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
6123 | * device in an array, update the disk_ord_tbl. If this disk is |
6124 | * present in all member arrays then also clear the SPARE_DISK | |
6125 | * flag | |
d195167d AK |
6126 | * update_create_array |
6127 | * update_kill_array | |
6128 | * update_rename_array | |
6129 | * update_add_remove_disk | |
e8319a19 DW |
6130 | */ |
6131 | struct intel_super *super = st->sb; | |
4d7b1503 | 6132 | struct imsm_super *mpb; |
e8319a19 DW |
6133 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
6134 | ||
4d7b1503 DW |
6135 | /* update requires a larger buf but the allocation failed */ |
6136 | if (super->next_len && !super->next_buf) { | |
6137 | super->next_len = 0; | |
6138 | return; | |
6139 | } | |
6140 | ||
6141 | if (super->next_buf) { | |
6142 | memcpy(super->next_buf, super->buf, super->len); | |
6143 | free(super->buf); | |
6144 | super->len = super->next_len; | |
6145 | super->buf = super->next_buf; | |
6146 | ||
6147 | super->next_len = 0; | |
6148 | super->next_buf = NULL; | |
6149 | } | |
6150 | ||
6151 | mpb = super->anchor; | |
6152 | ||
e8319a19 | 6153 | switch (type) { |
bb025c2f KW |
6154 | case update_takeover: { |
6155 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
6156 | if (apply_takeover_update(u, super, &update->space_list)) { |
6157 | imsm_update_version_info(super); | |
bb025c2f | 6158 | super->updates_pending++; |
1a2487c2 | 6159 | } |
bb025c2f KW |
6160 | break; |
6161 | } | |
6162 | ||
78b10e66 | 6163 | case update_reshape_container_disks: { |
d195167d | 6164 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
6165 | if (apply_reshape_container_disks_update( |
6166 | u, super, &update->space_list)) | |
6167 | super->updates_pending++; | |
78b10e66 N |
6168 | break; |
6169 | } | |
e8319a19 DW |
6170 | case update_activate_spare: { |
6171 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 6172 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 6173 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd | 6174 | struct imsm_map *migr_map; |
e8319a19 DW |
6175 | struct active_array *a; |
6176 | struct imsm_disk *disk; | |
0c046afd | 6177 | __u8 to_state; |
e8319a19 | 6178 | struct dl *dl; |
e8319a19 | 6179 | unsigned int found; |
0c046afd | 6180 | int failed; |
98130f40 | 6181 | int victim = get_imsm_disk_idx(dev, u->slot, -1); |
e8319a19 DW |
6182 | int i; |
6183 | ||
6184 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 6185 | if (dl == u->dl) |
e8319a19 DW |
6186 | break; |
6187 | ||
6188 | if (!dl) { | |
6189 | fprintf(stderr, "error: imsm_activate_spare passed " | |
1f24f035 DW |
6190 | "an unknown disk (index: %d)\n", |
6191 | u->dl->index); | |
e8319a19 DW |
6192 | return; |
6193 | } | |
6194 | ||
6195 | super->updates_pending++; | |
6196 | ||
0c046afd DW |
6197 | /* count failures (excluding rebuilds and the victim) |
6198 | * to determine map[0] state | |
6199 | */ | |
6200 | failed = 0; | |
6201 | for (i = 0; i < map->num_members; i++) { | |
6202 | if (i == u->slot) | |
6203 | continue; | |
98130f40 AK |
6204 | disk = get_imsm_disk(super, |
6205 | get_imsm_disk_idx(dev, i, -1)); | |
25ed7e59 | 6206 | if (!disk || is_failed(disk)) |
0c046afd DW |
6207 | failed++; |
6208 | } | |
6209 | ||
d23fe947 DW |
6210 | /* adding a pristine spare, assign a new index */ |
6211 | if (dl->index < 0) { | |
6212 | dl->index = super->anchor->num_disks; | |
6213 | super->anchor->num_disks++; | |
6214 | } | |
d23fe947 | 6215 | disk = &dl->disk; |
f2f27e63 DW |
6216 | disk->status |= CONFIGURED_DISK; |
6217 | disk->status &= ~SPARE_DISK; | |
e8319a19 | 6218 | |
0c046afd DW |
6219 | /* mark rebuild */ |
6220 | to_state = imsm_check_degraded(super, dev, failed); | |
6221 | map->map_state = IMSM_T_STATE_DEGRADED; | |
e3bba0e0 | 6222 | migrate(dev, to_state, MIGR_REBUILD); |
0c046afd DW |
6223 | migr_map = get_imsm_map(dev, 1); |
6224 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); | |
6225 | set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD); | |
6226 | ||
148acb7b DW |
6227 | /* update the family_num to mark a new container |
6228 | * generation, being careful to record the existing | |
6229 | * family_num in orig_family_num to clean up after | |
6230 | * earlier mdadm versions that neglected to set it. | |
6231 | */ | |
6232 | if (mpb->orig_family_num == 0) | |
6233 | mpb->orig_family_num = mpb->family_num; | |
6234 | mpb->family_num += super->random; | |
6235 | ||
e8319a19 DW |
6236 | /* count arrays using the victim in the metadata */ |
6237 | found = 0; | |
6238 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 6239 | dev = get_imsm_dev(super, a->info.container_member); |
620b1713 DW |
6240 | map = get_imsm_map(dev, 0); |
6241 | ||
6242 | if (get_imsm_disk_slot(map, victim) >= 0) | |
6243 | found++; | |
e8319a19 DW |
6244 | } |
6245 | ||
24565c9a | 6246 | /* delete the victim if it is no longer being |
e8319a19 DW |
6247 | * utilized anywhere |
6248 | */ | |
e8319a19 | 6249 | if (!found) { |
ae6aad82 | 6250 | struct dl **dlp; |
24565c9a | 6251 | |
47ee5a45 DW |
6252 | /* We know that 'manager' isn't touching anything, |
6253 | * so it is safe to delete | |
6254 | */ | |
24565c9a | 6255 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) |
ae6aad82 DW |
6256 | if ((*dlp)->index == victim) |
6257 | break; | |
47ee5a45 DW |
6258 | |
6259 | /* victim may be on the missing list */ | |
6260 | if (!*dlp) | |
6261 | for (dlp = &super->missing; *dlp; dlp = &(*dlp)->next) | |
6262 | if ((*dlp)->index == victim) | |
6263 | break; | |
24565c9a | 6264 | imsm_delete(super, dlp, victim); |
e8319a19 | 6265 | } |
8273f55e DW |
6266 | break; |
6267 | } | |
6268 | case update_create_array: { | |
6269 | /* someone wants to create a new array, we need to be aware of | |
6270 | * a few races/collisions: | |
6271 | * 1/ 'Create' called by two separate instances of mdadm | |
6272 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
6273 | * devices that have since been assimilated via | |
6274 | * activate_spare. | |
6275 | * In the event this update can not be carried out mdadm will | |
6276 | * (FIX ME) notice that its update did not take hold. | |
6277 | */ | |
6278 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 6279 | struct intel_dev *dv; |
8273f55e DW |
6280 | struct imsm_dev *dev; |
6281 | struct imsm_map *map, *new_map; | |
6282 | unsigned long long start, end; | |
6283 | unsigned long long new_start, new_end; | |
6284 | int i; | |
54c2c1ea DW |
6285 | struct disk_info *inf; |
6286 | struct dl *dl; | |
8273f55e DW |
6287 | |
6288 | /* handle racing creates: first come first serve */ | |
6289 | if (u->dev_idx < mpb->num_raid_devs) { | |
6290 | dprintf("%s: subarray %d already defined\n", | |
6291 | __func__, u->dev_idx); | |
ba2de7ba | 6292 | goto create_error; |
8273f55e DW |
6293 | } |
6294 | ||
6295 | /* check update is next in sequence */ | |
6296 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
6297 | dprintf("%s: can not create array %d expected index %d\n", |
6298 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 6299 | goto create_error; |
8273f55e DW |
6300 | } |
6301 | ||
a965f303 | 6302 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
6303 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
6304 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
54c2c1ea | 6305 | inf = get_disk_info(u); |
8273f55e DW |
6306 | |
6307 | /* handle activate_spare versus create race: | |
6308 | * check to make sure that overlapping arrays do not include | |
6309 | * overalpping disks | |
6310 | */ | |
6311 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 6312 | dev = get_imsm_dev(super, i); |
a965f303 | 6313 | map = get_imsm_map(dev, 0); |
8273f55e DW |
6314 | start = __le32_to_cpu(map->pba_of_lba0); |
6315 | end = start + __le32_to_cpu(map->blocks_per_member); | |
6316 | if ((new_start >= start && new_start <= end) || | |
6317 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
6318 | /* overlap */; |
6319 | else | |
6320 | continue; | |
6321 | ||
6322 | if (disks_overlap(super, i, u)) { | |
8273f55e | 6323 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 6324 | goto create_error; |
8273f55e DW |
6325 | } |
6326 | } | |
8273f55e | 6327 | |
949c47a0 DW |
6328 | /* check that prepare update was successful */ |
6329 | if (!update->space) { | |
6330 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 6331 | goto create_error; |
949c47a0 DW |
6332 | } |
6333 | ||
54c2c1ea DW |
6334 | /* check that all disks are still active before committing |
6335 | * changes. FIXME: could we instead handle this by creating a | |
6336 | * degraded array? That's probably not what the user expects, | |
6337 | * so better to drop this update on the floor. | |
6338 | */ | |
6339 | for (i = 0; i < new_map->num_members; i++) { | |
6340 | dl = serial_to_dl(inf[i].serial, super); | |
6341 | if (!dl) { | |
6342 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 6343 | goto create_error; |
54c2c1ea | 6344 | } |
949c47a0 DW |
6345 | } |
6346 | ||
8273f55e | 6347 | super->updates_pending++; |
54c2c1ea DW |
6348 | |
6349 | /* convert spares to members and fixup ord_tbl */ | |
6350 | for (i = 0; i < new_map->num_members; i++) { | |
6351 | dl = serial_to_dl(inf[i].serial, super); | |
6352 | if (dl->index == -1) { | |
6353 | dl->index = mpb->num_disks; | |
6354 | mpb->num_disks++; | |
6355 | dl->disk.status |= CONFIGURED_DISK; | |
6356 | dl->disk.status &= ~SPARE_DISK; | |
6357 | } | |
6358 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
6359 | } | |
6360 | ||
ba2de7ba DW |
6361 | dv = update->space; |
6362 | dev = dv->dev; | |
949c47a0 DW |
6363 | update->space = NULL; |
6364 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
6365 | dv->index = u->dev_idx; |
6366 | dv->next = super->devlist; | |
6367 | super->devlist = dv; | |
8273f55e | 6368 | mpb->num_raid_devs++; |
8273f55e | 6369 | |
4d1313e9 | 6370 | imsm_update_version_info(super); |
8273f55e | 6371 | break; |
ba2de7ba DW |
6372 | create_error: |
6373 | /* mdmon knows how to release update->space, but not | |
6374 | * ((struct intel_dev *) update->space)->dev | |
6375 | */ | |
6376 | if (update->space) { | |
6377 | dv = update->space; | |
6378 | free(dv->dev); | |
6379 | } | |
8273f55e | 6380 | break; |
e8319a19 | 6381 | } |
33414a01 DW |
6382 | case update_kill_array: { |
6383 | struct imsm_update_kill_array *u = (void *) update->buf; | |
6384 | int victim = u->dev_idx; | |
6385 | struct active_array *a; | |
6386 | struct intel_dev **dp; | |
6387 | struct imsm_dev *dev; | |
6388 | ||
6389 | /* sanity check that we are not affecting the uuid of | |
6390 | * active arrays, or deleting an active array | |
6391 | * | |
6392 | * FIXME when immutable ids are available, but note that | |
6393 | * we'll also need to fixup the invalidated/active | |
6394 | * subarray indexes in mdstat | |
6395 | */ | |
6396 | for (a = st->arrays; a; a = a->next) | |
6397 | if (a->info.container_member >= victim) | |
6398 | break; | |
6399 | /* by definition if mdmon is running at least one array | |
6400 | * is active in the container, so checking | |
6401 | * mpb->num_raid_devs is just extra paranoia | |
6402 | */ | |
6403 | dev = get_imsm_dev(super, victim); | |
6404 | if (a || !dev || mpb->num_raid_devs == 1) { | |
6405 | dprintf("failed to delete subarray-%d\n", victim); | |
6406 | break; | |
6407 | } | |
6408 | ||
6409 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 6410 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
6411 | *dp = (*dp)->next; |
6412 | } else { | |
f21e18ca | 6413 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
6414 | (*dp)->index--; |
6415 | dp = &(*dp)->next; | |
6416 | } | |
6417 | mpb->num_raid_devs--; | |
6418 | super->updates_pending++; | |
6419 | break; | |
6420 | } | |
aa534678 DW |
6421 | case update_rename_array: { |
6422 | struct imsm_update_rename_array *u = (void *) update->buf; | |
6423 | char name[MAX_RAID_SERIAL_LEN+1]; | |
6424 | int target = u->dev_idx; | |
6425 | struct active_array *a; | |
6426 | struct imsm_dev *dev; | |
6427 | ||
6428 | /* sanity check that we are not affecting the uuid of | |
6429 | * an active array | |
6430 | */ | |
6431 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
6432 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
6433 | for (a = st->arrays; a; a = a->next) | |
6434 | if (a->info.container_member == target) | |
6435 | break; | |
6436 | dev = get_imsm_dev(super, u->dev_idx); | |
6437 | if (a || !dev || !check_name(super, name, 1)) { | |
6438 | dprintf("failed to rename subarray-%d\n", target); | |
6439 | break; | |
6440 | } | |
6441 | ||
cdbe98cd | 6442 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
6443 | super->updates_pending++; |
6444 | break; | |
6445 | } | |
1a64be56 | 6446 | case update_add_remove_disk: { |
43dad3d6 | 6447 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
6448 | * being added, check teh status of add_remove_disk |
6449 | * if discs has been added. | |
6450 | */ | |
6451 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 6452 | struct active_array *a; |
072b727f DW |
6453 | |
6454 | super->updates_pending++; | |
1a64be56 | 6455 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
6456 | a->check_degraded = 1; |
6457 | } | |
43dad3d6 | 6458 | break; |
e8319a19 | 6459 | } |
1a64be56 LM |
6460 | default: |
6461 | fprintf(stderr, "error: unsuported process update type:" | |
6462 | "(type: %d)\n", type); | |
6463 | } | |
e8319a19 | 6464 | } |
88758e9d | 6465 | |
8273f55e DW |
6466 | static void imsm_prepare_update(struct supertype *st, |
6467 | struct metadata_update *update) | |
6468 | { | |
949c47a0 | 6469 | /** |
4d7b1503 DW |
6470 | * Allocate space to hold new disk entries, raid-device entries or a new |
6471 | * mpb if necessary. The manager synchronously waits for updates to | |
6472 | * complete in the monitor, so new mpb buffers allocated here can be | |
6473 | * integrated by the monitor thread without worrying about live pointers | |
6474 | * in the manager thread. | |
8273f55e | 6475 | */ |
949c47a0 | 6476 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
6477 | struct intel_super *super = st->sb; |
6478 | struct imsm_super *mpb = super->anchor; | |
6479 | size_t buf_len; | |
6480 | size_t len = 0; | |
949c47a0 DW |
6481 | |
6482 | switch (type) { | |
abedf5fc KW |
6483 | case update_takeover: { |
6484 | struct imsm_update_takeover *u = (void *)update->buf; | |
6485 | if (u->direction == R0_TO_R10) { | |
6486 | void **tail = (void **)&update->space_list; | |
6487 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
6488 | struct imsm_map *map = get_imsm_map(dev, 0); | |
6489 | int num_members = map->num_members; | |
6490 | void *space; | |
6491 | int size, i; | |
6492 | int err = 0; | |
6493 | /* allocate memory for added disks */ | |
6494 | for (i = 0; i < num_members; i++) { | |
6495 | size = sizeof(struct dl); | |
6496 | space = malloc(size); | |
6497 | if (!space) { | |
6498 | err++; | |
6499 | break; | |
6500 | } | |
6501 | *tail = space; | |
6502 | tail = space; | |
6503 | *tail = NULL; | |
6504 | } | |
6505 | /* allocate memory for new device */ | |
6506 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
6507 | (num_members * sizeof(__u32)); | |
6508 | space = malloc(size); | |
6509 | if (!space) | |
6510 | err++; | |
6511 | else { | |
6512 | *tail = space; | |
6513 | tail = space; | |
6514 | *tail = NULL; | |
6515 | } | |
6516 | if (!err) { | |
6517 | len = disks_to_mpb_size(num_members * 2); | |
6518 | } else { | |
6519 | /* if allocation didn't success, free buffer */ | |
6520 | while (update->space_list) { | |
6521 | void **sp = update->space_list; | |
6522 | update->space_list = *sp; | |
6523 | free(sp); | |
6524 | } | |
6525 | } | |
6526 | } | |
6527 | ||
6528 | break; | |
6529 | } | |
78b10e66 | 6530 | case update_reshape_container_disks: { |
d195167d AK |
6531 | /* Every raid device in the container is about to |
6532 | * gain some more devices, and we will enter a | |
6533 | * reconfiguration. | |
6534 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
6535 | * will now hold 2 of them. | |
6536 | * Thus we need new 'struct imsm_dev' allocations sized | |
6537 | * as sizeof_imsm_dev but with more devices in both maps. | |
6538 | */ | |
6539 | struct imsm_update_reshape *u = (void *)update->buf; | |
6540 | struct intel_dev *dl; | |
6541 | void **space_tail = (void**)&update->space_list; | |
6542 | ||
6543 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
6544 | ||
6545 | for (dl = super->devlist; dl; dl = dl->next) { | |
6546 | int size = sizeof_imsm_dev(dl->dev, 1); | |
6547 | void *s; | |
d677e0b8 AK |
6548 | if (u->new_raid_disks > u->old_raid_disks) |
6549 | size += sizeof(__u32)*2* | |
6550 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
6551 | s = malloc(size); |
6552 | if (!s) | |
6553 | break; | |
6554 | *space_tail = s; | |
6555 | space_tail = s; | |
6556 | *space_tail = NULL; | |
6557 | } | |
6558 | ||
6559 | len = disks_to_mpb_size(u->new_raid_disks); | |
6560 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
6561 | break; |
6562 | } | |
949c47a0 DW |
6563 | case update_create_array: { |
6564 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 6565 | struct intel_dev *dv; |
54c2c1ea DW |
6566 | struct imsm_dev *dev = &u->dev; |
6567 | struct imsm_map *map = get_imsm_map(dev, 0); | |
6568 | struct dl *dl; | |
6569 | struct disk_info *inf; | |
6570 | int i; | |
6571 | int activate = 0; | |
949c47a0 | 6572 | |
54c2c1ea DW |
6573 | inf = get_disk_info(u); |
6574 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
6575 | /* allocate a new super->devlist entry */ |
6576 | dv = malloc(sizeof(*dv)); | |
6577 | if (dv) { | |
6578 | dv->dev = malloc(len); | |
6579 | if (dv->dev) | |
6580 | update->space = dv; | |
6581 | else { | |
6582 | free(dv); | |
6583 | update->space = NULL; | |
6584 | } | |
6585 | } | |
949c47a0 | 6586 | |
54c2c1ea DW |
6587 | /* count how many spares will be converted to members */ |
6588 | for (i = 0; i < map->num_members; i++) { | |
6589 | dl = serial_to_dl(inf[i].serial, super); | |
6590 | if (!dl) { | |
6591 | /* hmm maybe it failed?, nothing we can do about | |
6592 | * it here | |
6593 | */ | |
6594 | continue; | |
6595 | } | |
6596 | if (count_memberships(dl, super) == 0) | |
6597 | activate++; | |
6598 | } | |
6599 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
6600 | break; |
6601 | default: | |
6602 | break; | |
6603 | } | |
6604 | } | |
8273f55e | 6605 | |
4d7b1503 DW |
6606 | /* check if we need a larger metadata buffer */ |
6607 | if (super->next_buf) | |
6608 | buf_len = super->next_len; | |
6609 | else | |
6610 | buf_len = super->len; | |
6611 | ||
6612 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
6613 | /* ok we need a larger buf than what is currently allocated | |
6614 | * if this allocation fails process_update will notice that | |
6615 | * ->next_len is set and ->next_buf is NULL | |
6616 | */ | |
6617 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
6618 | if (super->next_buf) | |
6619 | free(super->next_buf); | |
6620 | ||
6621 | super->next_len = buf_len; | |
1f45a8ad DW |
6622 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
6623 | memset(super->next_buf, 0, buf_len); | |
6624 | else | |
4d7b1503 DW |
6625 | super->next_buf = NULL; |
6626 | } | |
8273f55e DW |
6627 | } |
6628 | ||
ae6aad82 | 6629 | /* must be called while manager is quiesced */ |
f21e18ca | 6630 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
6631 | { |
6632 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
6633 | struct dl *iter; |
6634 | struct imsm_dev *dev; | |
6635 | struct imsm_map *map; | |
24565c9a DW |
6636 | int i, j, num_members; |
6637 | __u32 ord; | |
ae6aad82 | 6638 | |
24565c9a DW |
6639 | dprintf("%s: deleting device[%d] from imsm_super\n", |
6640 | __func__, index); | |
ae6aad82 DW |
6641 | |
6642 | /* shift all indexes down one */ | |
6643 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 6644 | if (iter->index > (int)index) |
ae6aad82 | 6645 | iter->index--; |
47ee5a45 | 6646 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 6647 | if (iter->index > (int)index) |
47ee5a45 | 6648 | iter->index--; |
ae6aad82 DW |
6649 | |
6650 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6651 | dev = get_imsm_dev(super, i); | |
6652 | map = get_imsm_map(dev, 0); | |
24565c9a DW |
6653 | num_members = map->num_members; |
6654 | for (j = 0; j < num_members; j++) { | |
6655 | /* update ord entries being careful not to propagate | |
6656 | * ord-flags to the first map | |
6657 | */ | |
98130f40 | 6658 | ord = get_imsm_ord_tbl_ent(dev, j, -1); |
ae6aad82 | 6659 | |
24565c9a DW |
6660 | if (ord_to_idx(ord) <= index) |
6661 | continue; | |
ae6aad82 | 6662 | |
24565c9a DW |
6663 | map = get_imsm_map(dev, 0); |
6664 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); | |
6665 | map = get_imsm_map(dev, 1); | |
6666 | if (map) | |
6667 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
6668 | } |
6669 | } | |
6670 | ||
6671 | mpb->num_disks--; | |
6672 | super->updates_pending++; | |
24565c9a DW |
6673 | if (*dlp) { |
6674 | struct dl *dl = *dlp; | |
6675 | ||
6676 | *dlp = (*dlp)->next; | |
6677 | __free_imsm_disk(dl); | |
6678 | } | |
ae6aad82 DW |
6679 | } |
6680 | ||
2cda7640 ML |
6681 | static char disk_by_path[] = "/dev/disk/by-path/"; |
6682 | ||
6683 | static const char *imsm_get_disk_controller_domain(const char *path) | |
6684 | { | |
2cda7640 | 6685 | char disk_path[PATH_MAX]; |
96234762 LM |
6686 | char *drv=NULL; |
6687 | struct stat st; | |
2cda7640 | 6688 | |
96234762 LM |
6689 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
6690 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
6691 | if (stat(disk_path, &st) == 0) { | |
6692 | struct sys_dev* hba; | |
6693 | char *path=NULL; | |
6694 | ||
6695 | path = devt_to_devpath(st.st_rdev); | |
6696 | if (path == NULL) | |
6697 | return "unknown"; | |
6698 | hba = find_disk_attached_hba(-1, path); | |
6699 | if (hba && hba->type == SYS_DEV_SAS) | |
6700 | drv = "isci"; | |
6701 | else if (hba && hba->type == SYS_DEV_SATA) | |
6702 | drv = "ahci"; | |
6703 | else | |
6704 | drv = "unknown"; | |
6705 | dprintf("path: %s hba: %s attached: %s\n", | |
6706 | path, (hba) ? hba->path : "NULL", drv); | |
6707 | free(path); | |
6708 | if (hba) | |
6709 | free_sys_dev(&hba); | |
2cda7640 | 6710 | } |
96234762 | 6711 | return drv; |
2cda7640 ML |
6712 | } |
6713 | ||
78b10e66 N |
6714 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
6715 | { | |
6716 | char subdev_name[20]; | |
6717 | struct mdstat_ent *mdstat; | |
6718 | ||
6719 | sprintf(subdev_name, "%d", subdev); | |
6720 | mdstat = mdstat_by_subdev(subdev_name, container); | |
6721 | if (!mdstat) | |
6722 | return -1; | |
6723 | ||
6724 | *minor = mdstat->devnum; | |
6725 | free_mdstat(mdstat); | |
6726 | return 0; | |
6727 | } | |
6728 | ||
6729 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
6730 | struct geo_params *geo, | |
6731 | int *old_raid_disks) | |
6732 | { | |
694575e7 KW |
6733 | /* currently we only support increasing the number of devices |
6734 | * for a container. This increases the number of device for each | |
6735 | * member array. They must all be RAID0 or RAID5. | |
6736 | */ | |
78b10e66 N |
6737 | int ret_val = 0; |
6738 | struct mdinfo *info, *member; | |
6739 | int devices_that_can_grow = 0; | |
6740 | ||
6741 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
6742 | "st->devnum = (%i)\n", | |
6743 | st->devnum); | |
6744 | ||
6745 | if (geo->size != -1 || | |
6746 | geo->level != UnSet || | |
6747 | geo->layout != UnSet || | |
6748 | geo->chunksize != 0 || | |
6749 | geo->raid_disks == UnSet) { | |
6750 | dprintf("imsm: Container operation is allowed for " | |
6751 | "raid disks number change only.\n"); | |
6752 | return ret_val; | |
6753 | } | |
6754 | ||
6755 | info = container_content_imsm(st, NULL); | |
6756 | for (member = info; member; member = member->next) { | |
6757 | int result; | |
6758 | int minor; | |
6759 | ||
6760 | dprintf("imsm: checking device_num: %i\n", | |
6761 | member->container_member); | |
6762 | ||
d7d205bd | 6763 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
6764 | /* we work on container for Online Capacity Expansion |
6765 | * only so raid_disks has to grow | |
6766 | */ | |
6767 | dprintf("imsm: for container operation raid disks " | |
6768 | "increase is required\n"); | |
6769 | break; | |
6770 | } | |
6771 | ||
6772 | if ((info->array.level != 0) && | |
6773 | (info->array.level != 5)) { | |
6774 | /* we cannot use this container with other raid level | |
6775 | */ | |
690aae1a | 6776 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
6777 | " raid level (%i) detected\n", |
6778 | info->array.level); | |
6779 | break; | |
6780 | } else { | |
6781 | /* check for platform support | |
6782 | * for this raid level configuration | |
6783 | */ | |
6784 | struct intel_super *super = st->sb; | |
6785 | if (!is_raid_level_supported(super->orom, | |
6786 | member->array.level, | |
6787 | geo->raid_disks)) { | |
690aae1a | 6788 | dprintf("platform does not support raid%d with" |
78b10e66 N |
6789 | " %d disk%s\n", |
6790 | info->array.level, | |
6791 | geo->raid_disks, | |
6792 | geo->raid_disks > 1 ? "s" : ""); | |
6793 | break; | |
6794 | } | |
6795 | } | |
6796 | ||
6797 | if (*old_raid_disks && | |
6798 | info->array.raid_disks != *old_raid_disks) | |
6799 | break; | |
6800 | *old_raid_disks = info->array.raid_disks; | |
6801 | ||
6802 | /* All raid5 and raid0 volumes in container | |
6803 | * have to be ready for Online Capacity Expansion | |
6804 | * so they need to be assembled. We have already | |
6805 | * checked that no recovery etc is happening. | |
6806 | */ | |
6807 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
6808 | st->container_dev, | |
6809 | &minor); | |
6810 | if (result < 0) { | |
6811 | dprintf("imsm: cannot find array\n"); | |
6812 | break; | |
6813 | } | |
6814 | devices_that_can_grow++; | |
6815 | } | |
6816 | sysfs_free(info); | |
6817 | if (!member && devices_that_can_grow) | |
6818 | ret_val = 1; | |
6819 | ||
6820 | if (ret_val) | |
6821 | dprintf("\tContainer operation allowed\n"); | |
6822 | else | |
6823 | dprintf("\tError: %i\n", ret_val); | |
6824 | ||
6825 | return ret_val; | |
6826 | } | |
6827 | ||
6828 | /* Function: get_spares_for_grow | |
6829 | * Description: Allocates memory and creates list of spare devices | |
6830 | * avaliable in container. Checks if spare drive size is acceptable. | |
6831 | * Parameters: Pointer to the supertype structure | |
6832 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
6833 | * NULL if fail | |
6834 | */ | |
6835 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
6836 | { | |
78b10e66 | 6837 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 6838 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
6839 | } |
6840 | ||
6841 | /****************************************************************************** | |
6842 | * function: imsm_create_metadata_update_for_reshape | |
6843 | * Function creates update for whole IMSM container. | |
6844 | * | |
6845 | ******************************************************************************/ | |
6846 | static int imsm_create_metadata_update_for_reshape( | |
6847 | struct supertype *st, | |
6848 | struct geo_params *geo, | |
6849 | int old_raid_disks, | |
6850 | struct imsm_update_reshape **updatep) | |
6851 | { | |
6852 | struct intel_super *super = st->sb; | |
6853 | struct imsm_super *mpb = super->anchor; | |
6854 | int update_memory_size = 0; | |
6855 | struct imsm_update_reshape *u = NULL; | |
6856 | struct mdinfo *spares = NULL; | |
6857 | int i; | |
6858 | int delta_disks = 0; | |
bbd24d86 | 6859 | struct mdinfo *dev; |
78b10e66 N |
6860 | |
6861 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
6862 | geo->raid_disks); | |
6863 | ||
6864 | delta_disks = geo->raid_disks - old_raid_disks; | |
6865 | ||
6866 | /* size of all update data without anchor */ | |
6867 | update_memory_size = sizeof(struct imsm_update_reshape); | |
6868 | ||
6869 | /* now add space for spare disks that we need to add. */ | |
6870 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
6871 | ||
6872 | u = calloc(1, update_memory_size); | |
6873 | if (u == NULL) { | |
6874 | dprintf("error: " | |
6875 | "cannot get memory for imsm_update_reshape update\n"); | |
6876 | return 0; | |
6877 | } | |
6878 | u->type = update_reshape_container_disks; | |
6879 | u->old_raid_disks = old_raid_disks; | |
6880 | u->new_raid_disks = geo->raid_disks; | |
6881 | ||
6882 | /* now get spare disks list | |
6883 | */ | |
6884 | spares = get_spares_for_grow(st); | |
6885 | ||
6886 | if (spares == NULL | |
6887 | || delta_disks > spares->array.spare_disks) { | |
e14e5960 KW |
6888 | fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices " |
6889 | "for %s.\n", geo->dev_name); | |
78b10e66 N |
6890 | goto abort; |
6891 | } | |
6892 | ||
6893 | /* we have got spares | |
6894 | * update disk list in imsm_disk list table in anchor | |
6895 | */ | |
6896 | dprintf("imsm: %i spares are available.\n\n", | |
6897 | spares->array.spare_disks); | |
6898 | ||
bbd24d86 | 6899 | dev = spares->devs; |
78b10e66 | 6900 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
6901 | struct dl *dl; |
6902 | ||
bbd24d86 AK |
6903 | if (dev == NULL) |
6904 | break; | |
78b10e66 N |
6905 | u->new_disks[i] = makedev(dev->disk.major, |
6906 | dev->disk.minor); | |
6907 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
6908 | dl->index = mpb->num_disks; |
6909 | mpb->num_disks++; | |
bbd24d86 | 6910 | dev = dev->next; |
78b10e66 | 6911 | } |
78b10e66 N |
6912 | |
6913 | abort: | |
6914 | /* free spares | |
6915 | */ | |
6916 | sysfs_free(spares); | |
6917 | ||
d677e0b8 | 6918 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 6919 | if (i == delta_disks) { |
d677e0b8 | 6920 | dprintf(" OK\n"); |
78b10e66 N |
6921 | *updatep = u; |
6922 | return update_memory_size; | |
6923 | } | |
6924 | free(u); | |
d677e0b8 | 6925 | dprintf(" Error\n"); |
78b10e66 N |
6926 | |
6927 | return 0; | |
6928 | } | |
6929 | ||
8dd70bce AK |
6930 | static void imsm_update_metadata_locally(struct supertype *st, |
6931 | void *buf, int len) | |
6932 | { | |
6933 | struct metadata_update mu; | |
6934 | ||
6935 | mu.buf = buf; | |
6936 | mu.len = len; | |
6937 | mu.space = NULL; | |
6938 | mu.space_list = NULL; | |
6939 | mu.next = NULL; | |
6940 | imsm_prepare_update(st, &mu); | |
6941 | imsm_process_update(st, &mu); | |
6942 | ||
6943 | while (mu.space_list) { | |
6944 | void **space = mu.space_list; | |
6945 | mu.space_list = *space; | |
6946 | free(space); | |
6947 | } | |
6948 | } | |
78b10e66 | 6949 | |
471bceb6 | 6950 | /*************************************************************************** |
694575e7 | 6951 | * Function: imsm_analyze_change |
471bceb6 KW |
6952 | * Description: Function analyze change for single volume |
6953 | * and validate if transition is supported | |
694575e7 KW |
6954 | * Parameters: Geometry parameters, supertype structure |
6955 | * Returns: Operation type code on success, -1 if fail | |
471bceb6 KW |
6956 | ****************************************************************************/ |
6957 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
6958 | struct geo_params *geo) | |
694575e7 | 6959 | { |
471bceb6 KW |
6960 | struct mdinfo info; |
6961 | int change = -1; | |
6962 | int check_devs = 0; | |
c21e737b | 6963 | int chunk; |
471bceb6 KW |
6964 | |
6965 | getinfo_super_imsm_volume(st, &info, NULL); | |
6966 | ||
6967 | if ((geo->level != info.array.level) && | |
6968 | (geo->level >= 0) && | |
6969 | (geo->level != UnSet)) { | |
6970 | switch (info.array.level) { | |
6971 | case 0: | |
6972 | if (geo->level == 5) { | |
b5347799 | 6973 | change = CH_MIGRATION; |
471bceb6 KW |
6974 | check_devs = 1; |
6975 | } | |
6976 | if (geo->level == 10) { | |
6977 | change = CH_TAKEOVER; | |
6978 | check_devs = 1; | |
6979 | } | |
dfe77a9e KW |
6980 | break; |
6981 | case 1: | |
6982 | if (geo->level == 0) { | |
6983 | change = CH_TAKEOVER; | |
6984 | check_devs = 1; | |
6985 | } | |
471bceb6 KW |
6986 | break; |
6987 | case 5: | |
41bf155e | 6988 | if (geo->level == 0) |
b5347799 | 6989 | change = CH_MIGRATION; |
471bceb6 KW |
6990 | break; |
6991 | case 10: | |
6992 | if (geo->level == 0) { | |
6993 | change = CH_TAKEOVER; | |
6994 | check_devs = 1; | |
6995 | } | |
6996 | break; | |
6997 | } | |
6998 | if (change == -1) { | |
6999 | fprintf(stderr, | |
7000 | Name " Error. Level Migration from %d to %d " | |
7001 | "not supported!\n", | |
7002 | info.array.level, geo->level); | |
7003 | goto analyse_change_exit; | |
7004 | } | |
7005 | } else | |
7006 | geo->level = info.array.level; | |
7007 | ||
7008 | if ((geo->layout != info.array.layout) | |
7009 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 7010 | change = CH_MIGRATION; |
471bceb6 KW |
7011 | if ((info.array.layout == 0) |
7012 | && (info.array.level == 5) | |
7013 | && (geo->layout == 5)) { | |
7014 | /* reshape 5 -> 4 */ | |
7015 | } else if ((info.array.layout == 5) | |
7016 | && (info.array.level == 5) | |
7017 | && (geo->layout == 0)) { | |
7018 | /* reshape 4 -> 5 */ | |
7019 | geo->layout = 0; | |
7020 | geo->level = 5; | |
7021 | } else { | |
7022 | fprintf(stderr, | |
7023 | Name " Error. Layout Migration from %d to %d " | |
7024 | "not supported!\n", | |
7025 | info.array.layout, geo->layout); | |
7026 | change = -1; | |
7027 | goto analyse_change_exit; | |
7028 | } | |
7029 | } else | |
7030 | geo->layout = info.array.layout; | |
7031 | ||
7032 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
7033 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 7034 | change = CH_MIGRATION; |
471bceb6 KW |
7035 | else |
7036 | geo->chunksize = info.array.chunk_size; | |
7037 | ||
c21e737b | 7038 | chunk = geo->chunksize / 1024; |
471bceb6 KW |
7039 | if (!validate_geometry_imsm(st, |
7040 | geo->level, | |
7041 | geo->layout, | |
7042 | geo->raid_disks, | |
c21e737b | 7043 | &chunk, |
471bceb6 KW |
7044 | geo->size, |
7045 | 0, 0, 1)) | |
7046 | change = -1; | |
7047 | ||
7048 | if (check_devs) { | |
7049 | struct intel_super *super = st->sb; | |
7050 | struct imsm_super *mpb = super->anchor; | |
7051 | ||
7052 | if (mpb->num_raid_devs > 1) { | |
7053 | fprintf(stderr, | |
7054 | Name " Error. Cannot perform operation on %s" | |
7055 | "- for this operation it MUST be single " | |
7056 | "array in container\n", | |
7057 | geo->dev_name); | |
7058 | change = -1; | |
7059 | } | |
7060 | } | |
7061 | ||
7062 | analyse_change_exit: | |
7063 | ||
7064 | return change; | |
694575e7 KW |
7065 | } |
7066 | ||
bb025c2f KW |
7067 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
7068 | { | |
7069 | struct intel_super *super = st->sb; | |
7070 | struct imsm_update_takeover *u; | |
7071 | ||
7072 | u = malloc(sizeof(struct imsm_update_takeover)); | |
7073 | if (u == NULL) | |
7074 | return 1; | |
7075 | ||
7076 | u->type = update_takeover; | |
7077 | u->subarray = super->current_vol; | |
7078 | ||
7079 | /* 10->0 transition */ | |
7080 | if (geo->level == 0) | |
7081 | u->direction = R10_TO_R0; | |
7082 | ||
0529c688 KW |
7083 | /* 0->10 transition */ |
7084 | if (geo->level == 10) | |
7085 | u->direction = R0_TO_R10; | |
7086 | ||
bb025c2f KW |
7087 | /* update metadata locally */ |
7088 | imsm_update_metadata_locally(st, u, | |
7089 | sizeof(struct imsm_update_takeover)); | |
7090 | /* and possibly remotely */ | |
7091 | if (st->update_tail) | |
7092 | append_metadata_update(st, u, | |
7093 | sizeof(struct imsm_update_takeover)); | |
7094 | else | |
7095 | free(u); | |
7096 | ||
7097 | return 0; | |
7098 | } | |
7099 | ||
78b10e66 N |
7100 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
7101 | int layout, int chunksize, int raid_disks, | |
41784c88 AK |
7102 | int delta_disks, char *backup, char *dev, |
7103 | int verbose) | |
78b10e66 | 7104 | { |
78b10e66 N |
7105 | int ret_val = 1; |
7106 | struct geo_params geo; | |
7107 | ||
7108 | dprintf("imsm: reshape_super called.\n"); | |
7109 | ||
71204a50 | 7110 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
7111 | |
7112 | geo.dev_name = dev; | |
694575e7 | 7113 | geo.dev_id = st->devnum; |
78b10e66 N |
7114 | geo.size = size; |
7115 | geo.level = level; | |
7116 | geo.layout = layout; | |
7117 | geo.chunksize = chunksize; | |
7118 | geo.raid_disks = raid_disks; | |
41784c88 AK |
7119 | if (delta_disks != UnSet) |
7120 | geo.raid_disks += delta_disks; | |
78b10e66 N |
7121 | |
7122 | dprintf("\tfor level : %i\n", geo.level); | |
7123 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
7124 | ||
7125 | if (experimental() == 0) | |
7126 | return ret_val; | |
7127 | ||
78b10e66 | 7128 | if (st->container_dev == st->devnum) { |
694575e7 KW |
7129 | /* On container level we can only increase number of devices. */ |
7130 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 N |
7131 | int old_raid_disks = 0; |
7132 | if (imsm_reshape_is_allowed_on_container( | |
7133 | st, &geo, &old_raid_disks)) { | |
7134 | struct imsm_update_reshape *u = NULL; | |
7135 | int len; | |
7136 | ||
7137 | len = imsm_create_metadata_update_for_reshape( | |
7138 | st, &geo, old_raid_disks, &u); | |
7139 | ||
ed08d51c AK |
7140 | if (len <= 0) { |
7141 | dprintf("imsm: Cannot prepare update\n"); | |
7142 | goto exit_imsm_reshape_super; | |
7143 | } | |
7144 | ||
8dd70bce AK |
7145 | ret_val = 0; |
7146 | /* update metadata locally */ | |
7147 | imsm_update_metadata_locally(st, u, len); | |
7148 | /* and possibly remotely */ | |
7149 | if (st->update_tail) | |
7150 | append_metadata_update(st, u, len); | |
7151 | else | |
ed08d51c | 7152 | free(u); |
8dd70bce | 7153 | |
694575e7 | 7154 | } else { |
e7ff7e40 AK |
7155 | fprintf(stderr, Name ": (imsm) Operation " |
7156 | "is not allowed on this container\n"); | |
694575e7 KW |
7157 | } |
7158 | } else { | |
7159 | /* On volume level we support following operations | |
471bceb6 KW |
7160 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
7161 | * - chunk size migration | |
7162 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
7163 | */ | |
7164 | struct intel_super *super = st->sb; | |
7165 | struct intel_dev *dev = super->devlist; | |
7166 | int change, devnum; | |
694575e7 | 7167 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
7168 | /* find requested device */ |
7169 | while (dev) { | |
7170 | imsm_find_array_minor_by_subdev(dev->index, st->container_dev, &devnum); | |
7171 | if (devnum == geo.dev_id) | |
7172 | break; | |
7173 | dev = dev->next; | |
7174 | } | |
7175 | if (dev == NULL) { | |
7176 | fprintf(stderr, Name " Cannot find %s (%i) subarray\n", | |
7177 | geo.dev_name, geo.dev_id); | |
7178 | goto exit_imsm_reshape_super; | |
7179 | } | |
7180 | super->current_vol = dev->index; | |
694575e7 KW |
7181 | change = imsm_analyze_change(st, &geo); |
7182 | switch (change) { | |
471bceb6 | 7183 | case CH_TAKEOVER: |
bb025c2f | 7184 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 7185 | break; |
b5347799 | 7186 | case CH_MIGRATION: |
471bceb6 | 7187 | ret_val = 0; |
694575e7 | 7188 | break; |
471bceb6 KW |
7189 | default: |
7190 | ret_val = 1; | |
694575e7 | 7191 | } |
694575e7 | 7192 | } |
78b10e66 | 7193 | |
ed08d51c | 7194 | exit_imsm_reshape_super: |
78b10e66 N |
7195 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
7196 | return ret_val; | |
7197 | } | |
2cda7640 | 7198 | |
999b4972 N |
7199 | static int imsm_manage_reshape( |
7200 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
7201 | struct supertype *st, unsigned long stripes, | |
7202 | int *fds, unsigned long long *offsets, | |
7203 | int dests, int *destfd, unsigned long long *destoffsets) | |
7204 | { | |
7205 | /* Just use child_monitor for now */ | |
7206 | return child_monitor( | |
7207 | afd, sra, reshape, st, stripes, | |
7208 | fds, offsets, dests, destfd, destoffsets); | |
7209 | } | |
71204a50 | 7210 | #endif /* MDASSEMBLE */ |
999b4972 | 7211 | |
cdddbdbc DW |
7212 | struct superswitch super_imsm = { |
7213 | #ifndef MDASSEMBLE | |
7214 | .examine_super = examine_super_imsm, | |
7215 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 7216 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 7217 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
7218 | .detail_super = detail_super_imsm, |
7219 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 7220 | .write_init_super = write_init_super_imsm, |
0e600426 N |
7221 | .validate_geometry = validate_geometry_imsm, |
7222 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 7223 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 7224 | .detail_platform = detail_platform_imsm, |
33414a01 | 7225 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 7226 | .update_subarray = update_subarray_imsm, |
2b959fbf | 7227 | .load_container = load_container_imsm, |
71204a50 N |
7228 | .default_geometry = default_geometry_imsm, |
7229 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
7230 | .reshape_super = imsm_reshape_super, | |
7231 | .manage_reshape = imsm_manage_reshape, | |
cdddbdbc DW |
7232 | #endif |
7233 | .match_home = match_home_imsm, | |
7234 | .uuid_from_super= uuid_from_super_imsm, | |
7235 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 7236 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
7237 | .update_super = update_super_imsm, |
7238 | ||
7239 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 7240 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
7241 | |
7242 | .compare_super = compare_super_imsm, | |
7243 | ||
7244 | .load_super = load_super_imsm, | |
bf5a934a | 7245 | .init_super = init_super_imsm, |
e683ca88 | 7246 | .store_super = store_super_imsm, |
cdddbdbc DW |
7247 | .free_super = free_super_imsm, |
7248 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 7249 | .container_content = container_content_imsm, |
cdddbdbc | 7250 | |
cdddbdbc | 7251 | .external = 1, |
4cce4069 | 7252 | .name = "imsm", |
845dea95 | 7253 | |
0e600426 | 7254 | #ifndef MDASSEMBLE |
845dea95 NB |
7255 | /* for mdmon */ |
7256 | .open_new = imsm_open_new, | |
ed9d66aa | 7257 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
7258 | .set_disk = imsm_set_disk, |
7259 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 7260 | .activate_spare = imsm_activate_spare, |
e8319a19 | 7261 | .process_update = imsm_process_update, |
8273f55e | 7262 | .prepare_update = imsm_prepare_update, |
0e600426 | 7263 | #endif /* MDASSEMBLE */ |
cdddbdbc | 7264 | }; |