]>
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 | ||
8e59f3d8 | 54 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 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 | ||
8e59f3d8 AK |
197 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
198 | ||
199 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
200 | ||
201 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must | |
202 | * be recovered using srcMap */ | |
203 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
204 | * already been migrated and must | |
205 | * be recovered from checkpoint area */ | |
206 | struct migr_record { | |
207 | __u32 rec_status; /* Status used to determine how to restart | |
208 | * migration in case it aborts | |
209 | * in some fashion */ | |
210 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
211 | __u32 family_num; /* Family number of MPB | |
212 | * containing the RaidDev | |
213 | * that is migrating */ | |
214 | __u32 ascending_migr; /* True if migrating in increasing | |
215 | * order of lbas */ | |
216 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
217 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
218 | * member disk | |
219 | * advances per unit-of-operation */ | |
220 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
221 | __u32 dest_1st_member_lba; /* First member lba on first | |
222 | * stripe of destination */ | |
223 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
224 | __u32 post_migr_vol_cap; /* Size of volume after | |
225 | * migration completes */ | |
226 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
227 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
228 | * migration ckpt record was read from | |
229 | * (for recovered migrations) */ | |
230 | } __attribute__ ((__packed__)); | |
231 | ||
1484e727 DW |
232 | static __u8 migr_type(struct imsm_dev *dev) |
233 | { | |
234 | if (dev->vol.migr_type == MIGR_VERIFY && | |
235 | dev->status & DEV_VERIFY_AND_FIX) | |
236 | return MIGR_REPAIR; | |
237 | else | |
238 | return dev->vol.migr_type; | |
239 | } | |
240 | ||
241 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
242 | { | |
243 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
244 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
245 | */ | |
246 | if (migr_type == MIGR_REPAIR) { | |
247 | dev->vol.migr_type = MIGR_VERIFY; | |
248 | dev->status |= DEV_VERIFY_AND_FIX; | |
249 | } else { | |
250 | dev->vol.migr_type = migr_type; | |
251 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
252 | } | |
253 | } | |
254 | ||
87eb16df | 255 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 256 | { |
87eb16df DW |
257 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
258 | } | |
cdddbdbc | 259 | |
87eb16df DW |
260 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
261 | { | |
262 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
263 | } |
264 | ||
ba2de7ba DW |
265 | struct intel_dev { |
266 | struct imsm_dev *dev; | |
267 | struct intel_dev *next; | |
f21e18ca | 268 | unsigned index; |
ba2de7ba DW |
269 | }; |
270 | ||
88654014 LM |
271 | struct intel_hba { |
272 | enum sys_dev_type type; | |
273 | char *path; | |
274 | char *pci_id; | |
275 | struct intel_hba *next; | |
276 | }; | |
277 | ||
1a64be56 LM |
278 | enum action { |
279 | DISK_REMOVE = 1, | |
280 | DISK_ADD | |
281 | }; | |
cdddbdbc DW |
282 | /* internal representation of IMSM metadata */ |
283 | struct intel_super { | |
284 | union { | |
949c47a0 DW |
285 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
286 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 287 | }; |
8e59f3d8 AK |
288 | union { |
289 | void *migr_rec_buf; /* buffer for I/O operations */ | |
290 | struct migr_record *migr_rec; /* migration record */ | |
291 | }; | |
949c47a0 | 292 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
293 | void *next_buf; /* for realloc'ing buf from the manager */ |
294 | size_t next_len; | |
c2c087e6 | 295 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 296 | int current_vol; /* index of raid device undergoing creation */ |
0dcecb2e | 297 | __u32 create_offset; /* common start for 'current_vol' */ |
148acb7b | 298 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 299 | struct intel_dev *devlist; |
cdddbdbc DW |
300 | struct dl { |
301 | struct dl *next; | |
302 | int index; | |
303 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
304 | int major, minor; | |
305 | char *devname; | |
b9f594fe | 306 | struct imsm_disk disk; |
cdddbdbc | 307 | int fd; |
0dcecb2e DW |
308 | int extent_cnt; |
309 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 310 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 311 | enum action action; |
cdddbdbc | 312 | } *disks; |
1a64be56 LM |
313 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
314 | active */ | |
47ee5a45 | 315 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 316 | struct bbm_log *bbm_log; |
88654014 | 317 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 318 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
319 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
320 | }; | |
321 | ||
322 | struct intel_disk { | |
323 | struct imsm_disk disk; | |
324 | #define IMSM_UNKNOWN_OWNER (-1) | |
325 | int owner; | |
326 | struct intel_disk *next; | |
cdddbdbc DW |
327 | }; |
328 | ||
c2c087e6 DW |
329 | struct extent { |
330 | unsigned long long start, size; | |
331 | }; | |
332 | ||
694575e7 KW |
333 | /* definitions of reshape process types */ |
334 | enum imsm_reshape_type { | |
335 | CH_TAKEOVER, | |
b5347799 | 336 | CH_MIGRATION, |
694575e7 KW |
337 | }; |
338 | ||
88758e9d DW |
339 | /* definition of messages passed to imsm_process_update */ |
340 | enum imsm_update_type { | |
341 | update_activate_spare, | |
8273f55e | 342 | update_create_array, |
33414a01 | 343 | update_kill_array, |
aa534678 | 344 | update_rename_array, |
1a64be56 | 345 | update_add_remove_disk, |
78b10e66 | 346 | update_reshape_container_disks, |
48c5303a | 347 | update_reshape_migration, |
bb025c2f | 348 | update_takeover |
88758e9d DW |
349 | }; |
350 | ||
351 | struct imsm_update_activate_spare { | |
352 | enum imsm_update_type type; | |
d23fe947 | 353 | struct dl *dl; |
88758e9d DW |
354 | int slot; |
355 | int array; | |
356 | struct imsm_update_activate_spare *next; | |
357 | }; | |
358 | ||
78b10e66 N |
359 | struct geo_params { |
360 | int dev_id; | |
361 | char *dev_name; | |
362 | long long size; | |
363 | int level; | |
364 | int layout; | |
365 | int chunksize; | |
366 | int raid_disks; | |
367 | }; | |
368 | ||
bb025c2f KW |
369 | enum takeover_direction { |
370 | R10_TO_R0, | |
371 | R0_TO_R10 | |
372 | }; | |
373 | struct imsm_update_takeover { | |
374 | enum imsm_update_type type; | |
375 | int subarray; | |
376 | enum takeover_direction direction; | |
377 | }; | |
78b10e66 N |
378 | |
379 | struct imsm_update_reshape { | |
380 | enum imsm_update_type type; | |
381 | int old_raid_disks; | |
382 | int new_raid_disks; | |
48c5303a PC |
383 | |
384 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
385 | }; | |
386 | ||
387 | struct imsm_update_reshape_migration { | |
388 | enum imsm_update_type type; | |
389 | int old_raid_disks; | |
390 | int new_raid_disks; | |
391 | /* fields for array migration changes | |
392 | */ | |
393 | int subdev; | |
394 | int new_level; | |
395 | int new_layout; | |
4bba0439 | 396 | int new_chunksize; |
48c5303a | 397 | |
d195167d | 398 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
399 | }; |
400 | ||
54c2c1ea DW |
401 | struct disk_info { |
402 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
403 | }; | |
404 | ||
8273f55e DW |
405 | struct imsm_update_create_array { |
406 | enum imsm_update_type type; | |
8273f55e | 407 | int dev_idx; |
6a3e913e | 408 | struct imsm_dev dev; |
8273f55e DW |
409 | }; |
410 | ||
33414a01 DW |
411 | struct imsm_update_kill_array { |
412 | enum imsm_update_type type; | |
413 | int dev_idx; | |
414 | }; | |
415 | ||
aa534678 DW |
416 | struct imsm_update_rename_array { |
417 | enum imsm_update_type type; | |
418 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
419 | int dev_idx; | |
420 | }; | |
421 | ||
1a64be56 | 422 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
423 | enum imsm_update_type type; |
424 | }; | |
425 | ||
88654014 LM |
426 | |
427 | static const char *_sys_dev_type[] = { | |
428 | [SYS_DEV_UNKNOWN] = "Unknown", | |
429 | [SYS_DEV_SAS] = "SAS", | |
430 | [SYS_DEV_SATA] = "SATA" | |
431 | }; | |
432 | ||
433 | const char *get_sys_dev_type(enum sys_dev_type type) | |
434 | { | |
435 | if (type >= SYS_DEV_MAX) | |
436 | type = SYS_DEV_UNKNOWN; | |
437 | ||
438 | return _sys_dev_type[type]; | |
439 | } | |
440 | ||
441 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
442 | { | |
443 | struct intel_hba *result = malloc(sizeof(*result)); | |
444 | if (result) { | |
445 | result->type = device->type; | |
446 | result->path = strdup(device->path); | |
447 | result->next = NULL; | |
448 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
449 | result->pci_id++; | |
450 | } | |
451 | return result; | |
452 | } | |
453 | ||
454 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
455 | { | |
456 | struct intel_hba *result=NULL; | |
457 | for (result = hba; result; result = result->next) { | |
458 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
459 | break; | |
460 | } | |
461 | return result; | |
462 | } | |
463 | ||
b4cf4cba | 464 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
465 | { |
466 | struct intel_hba *hba; | |
467 | ||
468 | /* check if disk attached to Intel HBA */ | |
469 | hba = find_intel_hba(super->hba, device); | |
470 | if (hba != NULL) | |
471 | return 1; | |
472 | /* Check if HBA is already attached to super */ | |
473 | if (super->hba == NULL) { | |
474 | super->hba = alloc_intel_hba(device); | |
475 | return 1; | |
476 | } | |
477 | ||
478 | hba = super->hba; | |
479 | /* Intel metadata allows for all disks attached to the same type HBA. | |
480 | * Do not sypport odf HBA types mixing | |
481 | */ | |
482 | if (device->type != hba->type) | |
483 | return 2; | |
484 | ||
485 | while (hba->next) | |
486 | hba = hba->next; | |
487 | ||
488 | hba->next = alloc_intel_hba(device); | |
489 | return 1; | |
490 | } | |
491 | ||
492 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
493 | { | |
494 | struct sys_dev *list, *elem, *prev; | |
495 | char *disk_path; | |
496 | ||
497 | if ((list = find_intel_devices()) == NULL) | |
498 | return 0; | |
499 | ||
500 | if (fd < 0) | |
501 | disk_path = (char *) devname; | |
502 | else | |
503 | disk_path = diskfd_to_devpath(fd); | |
504 | ||
505 | if (!disk_path) { | |
506 | free_sys_dev(&list); | |
507 | return 0; | |
508 | } | |
509 | ||
510 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
511 | if (path_attached_to_hba(disk_path, elem->path)) { | |
512 | if (prev == NULL) | |
513 | list = list->next; | |
514 | else | |
515 | prev->next = elem->next; | |
516 | elem->next = NULL; | |
517 | if (disk_path != devname) | |
518 | free(disk_path); | |
519 | free_sys_dev(&list); | |
520 | return elem; | |
521 | } | |
522 | } | |
523 | if (disk_path != devname) | |
524 | free(disk_path); | |
525 | free_sys_dev(&list); | |
526 | ||
527 | return NULL; | |
528 | } | |
529 | ||
530 | ||
d424212e N |
531 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
532 | char *devname); | |
f2f5c343 | 533 | |
cdddbdbc DW |
534 | static struct supertype *match_metadata_desc_imsm(char *arg) |
535 | { | |
536 | struct supertype *st; | |
537 | ||
538 | if (strcmp(arg, "imsm") != 0 && | |
539 | strcmp(arg, "default") != 0 | |
540 | ) | |
541 | return NULL; | |
542 | ||
543 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
544 | if (!st) |
545 | return NULL; | |
ef609477 | 546 | memset(st, 0, sizeof(*st)); |
d1d599ea | 547 | st->container_dev = NoMdDev; |
cdddbdbc DW |
548 | st->ss = &super_imsm; |
549 | st->max_devs = IMSM_MAX_DEVICES; | |
550 | st->minor_version = 0; | |
551 | st->sb = NULL; | |
552 | return st; | |
553 | } | |
554 | ||
0e600426 | 555 | #ifndef MDASSEMBLE |
cdddbdbc DW |
556 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
557 | { | |
558 | return &mpb->sig[MPB_SIG_LEN]; | |
559 | } | |
0e600426 | 560 | #endif |
cdddbdbc | 561 | |
949c47a0 DW |
562 | /* retrieve a disk directly from the anchor when the anchor is known to be |
563 | * up-to-date, currently only at load time | |
564 | */ | |
565 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 566 | { |
949c47a0 | 567 | if (index >= mpb->num_disks) |
cdddbdbc DW |
568 | return NULL; |
569 | return &mpb->disk[index]; | |
570 | } | |
571 | ||
95d07a2c LM |
572 | /* retrieve the disk description based on a index of the disk |
573 | * in the sub-array | |
574 | */ | |
575 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 576 | { |
b9f594fe DW |
577 | struct dl *d; |
578 | ||
579 | for (d = super->disks; d; d = d->next) | |
580 | if (d->index == index) | |
95d07a2c LM |
581 | return d; |
582 | ||
583 | return NULL; | |
584 | } | |
585 | /* retrieve a disk from the parsed metadata */ | |
586 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
587 | { | |
588 | struct dl *dl; | |
589 | ||
590 | dl = get_imsm_dl_disk(super, index); | |
591 | if (dl) | |
592 | return &dl->disk; | |
593 | ||
b9f594fe | 594 | return NULL; |
949c47a0 DW |
595 | } |
596 | ||
597 | /* generate a checksum directly from the anchor when the anchor is known to be | |
598 | * up-to-date, currently only at load or write_super after coalescing | |
599 | */ | |
600 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
601 | { |
602 | __u32 end = mpb->mpb_size / sizeof(end); | |
603 | __u32 *p = (__u32 *) mpb; | |
604 | __u32 sum = 0; | |
605 | ||
97f734fd N |
606 | while (end--) { |
607 | sum += __le32_to_cpu(*p); | |
608 | p++; | |
609 | } | |
cdddbdbc DW |
610 | |
611 | return sum - __le32_to_cpu(mpb->check_sum); | |
612 | } | |
613 | ||
a965f303 DW |
614 | static size_t sizeof_imsm_map(struct imsm_map *map) |
615 | { | |
616 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
617 | } | |
618 | ||
619 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 620 | { |
5e7b0330 AK |
621 | /* A device can have 2 maps if it is in the middle of a migration. |
622 | * If second_map is: | |
623 | * 0 - we return the first map | |
624 | * 1 - we return the second map if it exists, else NULL | |
625 | * -1 - we return the second map if it exists, else the first | |
626 | */ | |
a965f303 DW |
627 | struct imsm_map *map = &dev->vol.map[0]; |
628 | ||
5e7b0330 | 629 | if (second_map == 1 && !dev->vol.migr_state) |
a965f303 | 630 | return NULL; |
5e7b0330 AK |
631 | else if (second_map == 1 || |
632 | (second_map < 0 && dev->vol.migr_state)) { | |
a965f303 DW |
633 | void *ptr = map; |
634 | ||
635 | return ptr + sizeof_imsm_map(map); | |
636 | } else | |
637 | return map; | |
5e7b0330 | 638 | |
a965f303 | 639 | } |
cdddbdbc | 640 | |
3393c6af DW |
641 | /* return the size of the device. |
642 | * migr_state increases the returned size if map[0] were to be duplicated | |
643 | */ | |
644 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
645 | { |
646 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
647 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
648 | |
649 | /* migrating means an additional map */ | |
a965f303 DW |
650 | if (dev->vol.migr_state) |
651 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
652 | else if (migr_state) |
653 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
654 | |
655 | return size; | |
656 | } | |
657 | ||
54c2c1ea DW |
658 | #ifndef MDASSEMBLE |
659 | /* retrieve disk serial number list from a metadata update */ | |
660 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
661 | { | |
662 | void *u = update; | |
663 | struct disk_info *inf; | |
664 | ||
665 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
666 | sizeof_imsm_dev(&update->dev, 0); | |
667 | ||
668 | return inf; | |
669 | } | |
670 | #endif | |
671 | ||
949c47a0 | 672 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
673 | { |
674 | int offset; | |
675 | int i; | |
676 | void *_mpb = mpb; | |
677 | ||
949c47a0 | 678 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
679 | return NULL; |
680 | ||
681 | /* devices start after all disks */ | |
682 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
683 | ||
684 | for (i = 0; i <= index; i++) | |
685 | if (i == index) | |
686 | return _mpb + offset; | |
687 | else | |
3393c6af | 688 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
689 | |
690 | return NULL; | |
691 | } | |
692 | ||
949c47a0 DW |
693 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
694 | { | |
ba2de7ba DW |
695 | struct intel_dev *dv; |
696 | ||
949c47a0 DW |
697 | if (index >= super->anchor->num_raid_devs) |
698 | return NULL; | |
ba2de7ba DW |
699 | for (dv = super->devlist; dv; dv = dv->next) |
700 | if (dv->index == index) | |
701 | return dv->dev; | |
702 | return NULL; | |
949c47a0 DW |
703 | } |
704 | ||
98130f40 AK |
705 | /* |
706 | * for second_map: | |
707 | * == 0 get first map | |
708 | * == 1 get second map | |
709 | * == -1 than get map according to the current migr_state | |
710 | */ | |
711 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
712 | int slot, | |
713 | int second_map) | |
7eef0453 DW |
714 | { |
715 | struct imsm_map *map; | |
716 | ||
5e7b0330 | 717 | map = get_imsm_map(dev, second_map); |
7eef0453 | 718 | |
ff077194 DW |
719 | /* top byte identifies disk under rebuild */ |
720 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
721 | } | |
722 | ||
723 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 724 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 725 | { |
98130f40 | 726 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
727 | |
728 | return ord_to_idx(ord); | |
7eef0453 DW |
729 | } |
730 | ||
be73972f DW |
731 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
732 | { | |
733 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
734 | } | |
735 | ||
f21e18ca | 736 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
737 | { |
738 | int slot; | |
739 | __u32 ord; | |
740 | ||
741 | for (slot = 0; slot < map->num_members; slot++) { | |
742 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
743 | if (ord_to_idx(ord) == idx) | |
744 | return slot; | |
745 | } | |
746 | ||
747 | return -1; | |
748 | } | |
749 | ||
cdddbdbc DW |
750 | static int get_imsm_raid_level(struct imsm_map *map) |
751 | { | |
752 | if (map->raid_level == 1) { | |
753 | if (map->num_members == 2) | |
754 | return 1; | |
755 | else | |
756 | return 10; | |
757 | } | |
758 | ||
759 | return map->raid_level; | |
760 | } | |
761 | ||
c2c087e6 DW |
762 | static int cmp_extent(const void *av, const void *bv) |
763 | { | |
764 | const struct extent *a = av; | |
765 | const struct extent *b = bv; | |
766 | if (a->start < b->start) | |
767 | return -1; | |
768 | if (a->start > b->start) | |
769 | return 1; | |
770 | return 0; | |
771 | } | |
772 | ||
0dcecb2e | 773 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 774 | { |
c2c087e6 | 775 | int memberships = 0; |
620b1713 | 776 | int i; |
c2c087e6 | 777 | |
949c47a0 DW |
778 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
779 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 780 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 781 | |
620b1713 DW |
782 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
783 | memberships++; | |
c2c087e6 | 784 | } |
0dcecb2e DW |
785 | |
786 | return memberships; | |
787 | } | |
788 | ||
789 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
790 | { | |
791 | /* find a list of used extents on the given physical device */ | |
792 | struct extent *rv, *e; | |
620b1713 | 793 | int i; |
0dcecb2e DW |
794 | int memberships = count_memberships(dl, super); |
795 | __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
796 | ||
c2c087e6 DW |
797 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
798 | if (!rv) | |
799 | return NULL; | |
800 | e = rv; | |
801 | ||
949c47a0 DW |
802 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
803 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 804 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 805 | |
620b1713 DW |
806 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
807 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
808 | e->size = __le32_to_cpu(map->blocks_per_member); | |
809 | e++; | |
c2c087e6 DW |
810 | } |
811 | } | |
812 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
813 | ||
14e8215b DW |
814 | /* determine the start of the metadata |
815 | * when no raid devices are defined use the default | |
816 | * ...otherwise allow the metadata to truncate the value | |
817 | * as is the case with older versions of imsm | |
818 | */ | |
819 | if (memberships) { | |
820 | struct extent *last = &rv[memberships - 1]; | |
821 | __u32 remainder; | |
822 | ||
823 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
824 | (last->start + last->size); | |
dda5855f DW |
825 | /* round down to 1k block to satisfy precision of the kernel |
826 | * 'size' interface | |
827 | */ | |
828 | remainder &= ~1UL; | |
829 | /* make sure remainder is still sane */ | |
f21e18ca | 830 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 831 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
832 | if (reservation > remainder) |
833 | reservation = remainder; | |
834 | } | |
835 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
836 | e->size = 0; |
837 | return rv; | |
838 | } | |
839 | ||
14e8215b DW |
840 | /* try to determine how much space is reserved for metadata from |
841 | * the last get_extents() entry, otherwise fallback to the | |
842 | * default | |
843 | */ | |
844 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
845 | { | |
846 | struct extent *e; | |
847 | int i; | |
848 | __u32 rv; | |
849 | ||
850 | /* for spares just return a minimal reservation which will grow | |
851 | * once the spare is picked up by an array | |
852 | */ | |
853 | if (dl->index == -1) | |
854 | return MPB_SECTOR_CNT; | |
855 | ||
856 | e = get_extents(super, dl); | |
857 | if (!e) | |
858 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
859 | ||
860 | /* scroll to last entry */ | |
861 | for (i = 0; e[i].size; i++) | |
862 | continue; | |
863 | ||
864 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
865 | ||
866 | free(e); | |
867 | ||
868 | return rv; | |
869 | } | |
870 | ||
25ed7e59 DW |
871 | static int is_spare(struct imsm_disk *disk) |
872 | { | |
873 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
874 | } | |
875 | ||
876 | static int is_configured(struct imsm_disk *disk) | |
877 | { | |
878 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
879 | } | |
880 | ||
881 | static int is_failed(struct imsm_disk *disk) | |
882 | { | |
883 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
884 | } | |
885 | ||
80e7f8c3 AC |
886 | /* Return minimum size of a spare that can be used in this array*/ |
887 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
888 | { | |
889 | struct intel_super *super = st->sb; | |
890 | struct dl *dl; | |
891 | struct extent *e; | |
892 | int i; | |
893 | unsigned long long rv = 0; | |
894 | ||
895 | if (!super) | |
896 | return rv; | |
897 | /* find first active disk in array */ | |
898 | dl = super->disks; | |
899 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
900 | dl = dl->next; | |
901 | if (!dl) | |
902 | return rv; | |
903 | /* find last lba used by subarrays */ | |
904 | e = get_extents(super, dl); | |
905 | if (!e) | |
906 | return rv; | |
907 | for (i = 0; e[i].size; i++) | |
908 | continue; | |
909 | if (i > 0) | |
910 | rv = e[i-1].start + e[i-1].size; | |
911 | free(e); | |
912 | /* add the amount of space needed for metadata */ | |
913 | rv = rv + MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
914 | return rv * 512; | |
915 | } | |
916 | ||
1799c9e8 | 917 | #ifndef MDASSEMBLE |
1e5c6983 DW |
918 | static __u64 blocks_per_migr_unit(struct imsm_dev *dev); |
919 | ||
44470971 | 920 | static void print_imsm_dev(struct imsm_dev *dev, char *uuid, int disk_idx) |
cdddbdbc DW |
921 | { |
922 | __u64 sz; | |
0d80bb2f | 923 | int slot, i; |
a965f303 | 924 | struct imsm_map *map = get_imsm_map(dev, 0); |
dd8bcb3b | 925 | struct imsm_map *map2 = get_imsm_map(dev, 1); |
b10b37b8 | 926 | __u32 ord; |
cdddbdbc DW |
927 | |
928 | printf("\n"); | |
1e7bc0ed | 929 | printf("[%.16s]:\n", dev->volume); |
44470971 | 930 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
931 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
932 | if (map2) | |
933 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
934 | printf("\n"); | |
935 | printf(" Members : %d", map->num_members); | |
936 | if (map2) | |
937 | printf(" <-- %d", map2->num_members); | |
938 | printf("\n"); | |
0d80bb2f DW |
939 | printf(" Slots : ["); |
940 | for (i = 0; i < map->num_members; i++) { | |
dd8bcb3b | 941 | ord = get_imsm_ord_tbl_ent(dev, i, 0); |
0d80bb2f DW |
942 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
943 | } | |
dd8bcb3b AK |
944 | printf("]"); |
945 | if (map2) { | |
946 | printf(" <-- ["); | |
947 | for (i = 0; i < map2->num_members; i++) { | |
948 | ord = get_imsm_ord_tbl_ent(dev, i, 1); | |
949 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); | |
950 | } | |
951 | printf("]"); | |
952 | } | |
953 | printf("\n"); | |
7095bccb AK |
954 | printf(" Failed disk : "); |
955 | if (map->failed_disk_num == 0xff) | |
956 | printf("none"); | |
957 | else | |
958 | printf("%i", map->failed_disk_num); | |
959 | printf("\n"); | |
620b1713 DW |
960 | slot = get_imsm_disk_slot(map, disk_idx); |
961 | if (slot >= 0) { | |
98130f40 | 962 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
b10b37b8 DW |
963 | printf(" This Slot : %d%s\n", slot, |
964 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
965 | } else | |
cdddbdbc DW |
966 | printf(" This Slot : ?\n"); |
967 | sz = __le32_to_cpu(dev->size_high); | |
968 | sz <<= 32; | |
969 | sz += __le32_to_cpu(dev->size_low); | |
970 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
971 | human_size(sz * 512)); | |
972 | sz = __le32_to_cpu(map->blocks_per_member); | |
973 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
974 | human_size(sz * 512)); | |
975 | printf(" Sector Offset : %u\n", | |
976 | __le32_to_cpu(map->pba_of_lba0)); | |
977 | printf(" Num Stripes : %u\n", | |
978 | __le32_to_cpu(map->num_data_stripes)); | |
dd8bcb3b | 979 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 980 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
981 | if (map2) |
982 | printf(" <-- %u KiB", | |
983 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
984 | printf("\n"); | |
cdddbdbc | 985 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 986 | printf(" Migrate State : "); |
1484e727 DW |
987 | if (dev->vol.migr_state) { |
988 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 989 | printf("initialize\n"); |
1484e727 | 990 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 991 | printf("rebuild\n"); |
1484e727 | 992 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 993 | printf("check\n"); |
1484e727 | 994 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 995 | printf("general migration\n"); |
1484e727 | 996 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 997 | printf("state change\n"); |
1484e727 | 998 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 999 | printf("repair\n"); |
1484e727 | 1000 | else |
8655a7b1 DW |
1001 | printf("<unknown:%d>\n", migr_type(dev)); |
1002 | } else | |
1003 | printf("idle\n"); | |
3393c6af DW |
1004 | printf(" Map State : %s", map_state_str[map->map_state]); |
1005 | if (dev->vol.migr_state) { | |
1006 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1e5c6983 | 1007 | |
b10b37b8 | 1008 | printf(" <-- %s", map_state_str[map->map_state]); |
1e5c6983 DW |
1009 | printf("\n Checkpoint : %u (%llu)", |
1010 | __le32_to_cpu(dev->vol.curr_migr_unit), | |
94fcb80a | 1011 | (unsigned long long)blocks_per_migr_unit(dev)); |
3393c6af DW |
1012 | } |
1013 | printf("\n"); | |
cdddbdbc | 1014 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1015 | } |
1016 | ||
14e8215b | 1017 | static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved) |
cdddbdbc | 1018 | { |
949c47a0 | 1019 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); |
1f24f035 | 1020 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1021 | __u64 sz; |
1022 | ||
d362da3d | 1023 | if (index < 0 || !disk) |
e9d82038 DW |
1024 | return; |
1025 | ||
cdddbdbc | 1026 | printf("\n"); |
1f24f035 | 1027 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
cdddbdbc | 1028 | printf(" Disk%02d Serial : %s\n", index, str); |
25ed7e59 DW |
1029 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1030 | is_configured(disk) ? " active" : "", | |
1031 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1032 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
14e8215b | 1033 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
1034 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1035 | human_size(sz * 512)); | |
1036 | } | |
1037 | ||
a5d85af7 | 1038 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1039 | |
cdddbdbc DW |
1040 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1041 | { | |
1042 | struct intel_super *super = st->sb; | |
949c47a0 | 1043 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1044 | char str[MAX_SIGNATURE_LENGTH]; |
1045 | int i; | |
27fd6274 DW |
1046 | struct mdinfo info; |
1047 | char nbuf[64]; | |
cdddbdbc | 1048 | __u32 sum; |
14e8215b | 1049 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1050 | struct dl *dl; |
27fd6274 | 1051 | |
cdddbdbc DW |
1052 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1053 | printf(" Magic : %s\n", str); | |
1054 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1055 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1056 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1057 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1058 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
a5d85af7 | 1059 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1060 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1061 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1062 | sum = __le32_to_cpu(mpb->check_sum); |
1063 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1064 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1065 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1066 | printf(" Disks : %d\n", mpb->num_disks); |
1067 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
14e8215b | 1068 | print_imsm_disk(mpb, super->disks->index, reserved); |
604b746f JD |
1069 | if (super->bbm_log) { |
1070 | struct bbm_log *log = super->bbm_log; | |
1071 | ||
1072 | printf("\n"); | |
1073 | printf("Bad Block Management Log:\n"); | |
1074 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1075 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1076 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1077 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1078 | printf(" First Spare : %llx\n", |
1079 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1080 | } |
44470971 DW |
1081 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1082 | struct mdinfo info; | |
1083 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1084 | ||
1085 | super->current_vol = i; | |
a5d85af7 | 1086 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1087 | fname_from_uuid(st, &info, nbuf, ':'); |
44470971 DW |
1088 | print_imsm_dev(dev, nbuf + 5, super->disks->index); |
1089 | } | |
cdddbdbc DW |
1090 | for (i = 0; i < mpb->num_disks; i++) { |
1091 | if (i == super->disks->index) | |
1092 | continue; | |
14e8215b | 1093 | print_imsm_disk(mpb, i, reserved); |
cdddbdbc | 1094 | } |
94827db3 N |
1095 | for (dl = super->disks ; dl; dl = dl->next) { |
1096 | struct imsm_disk *disk; | |
1097 | char str[MAX_RAID_SERIAL_LEN + 1]; | |
1098 | __u64 sz; | |
1099 | ||
1100 | if (dl->index >= 0) | |
1101 | continue; | |
1102 | ||
1103 | disk = &dl->disk; | |
1104 | printf("\n"); | |
1105 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); | |
1106 | printf(" Disk Serial : %s\n", str); | |
1107 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", | |
1108 | is_configured(disk) ? " active" : "", | |
1109 | is_failed(disk) ? " failed" : ""); | |
1110 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
1111 | sz = __le32_to_cpu(disk->total_blocks) - reserved; | |
1112 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, | |
1113 | human_size(sz * 512)); | |
1114 | } | |
cdddbdbc DW |
1115 | } |
1116 | ||
061f2c6a | 1117 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1118 | { |
27fd6274 | 1119 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1120 | struct mdinfo info; |
1121 | char nbuf[64]; | |
1e7bc0ed | 1122 | struct intel_super *super = st->sb; |
1e7bc0ed | 1123 | |
0d5a423f DW |
1124 | if (!super->anchor->num_raid_devs) { |
1125 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1126 | return; |
0d5a423f | 1127 | } |
ff54de6e | 1128 | |
a5d85af7 | 1129 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1130 | fname_from_uuid(st, &info, nbuf, ':'); |
1131 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1132 | } | |
1133 | ||
1134 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1135 | { | |
1136 | /* We just write a generic IMSM ARRAY entry */ | |
1137 | struct mdinfo info; | |
1138 | char nbuf[64]; | |
1139 | char nbuf1[64]; | |
1140 | struct intel_super *super = st->sb; | |
1141 | int i; | |
1142 | ||
1143 | if (!super->anchor->num_raid_devs) | |
1144 | return; | |
1145 | ||
a5d85af7 | 1146 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1147 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1148 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1149 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1150 | ||
1151 | super->current_vol = i; | |
a5d85af7 | 1152 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1153 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1154 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1155 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1156 | } |
cdddbdbc DW |
1157 | } |
1158 | ||
9d84c8ea DW |
1159 | static void export_examine_super_imsm(struct supertype *st) |
1160 | { | |
1161 | struct intel_super *super = st->sb; | |
1162 | struct imsm_super *mpb = super->anchor; | |
1163 | struct mdinfo info; | |
1164 | char nbuf[64]; | |
1165 | ||
a5d85af7 | 1166 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1167 | fname_from_uuid(st, &info, nbuf, ':'); |
1168 | printf("MD_METADATA=imsm\n"); | |
1169 | printf("MD_LEVEL=container\n"); | |
1170 | printf("MD_UUID=%s\n", nbuf+5); | |
1171 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1172 | } | |
1173 | ||
cdddbdbc DW |
1174 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1175 | { | |
3ebe00a1 DW |
1176 | struct mdinfo info; |
1177 | char nbuf[64]; | |
1178 | ||
a5d85af7 | 1179 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1180 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1181 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1182 | } |
1183 | ||
1184 | static void brief_detail_super_imsm(struct supertype *st) | |
1185 | { | |
ff54de6e N |
1186 | struct mdinfo info; |
1187 | char nbuf[64]; | |
a5d85af7 | 1188 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1189 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1190 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1191 | } |
d665cc31 DW |
1192 | |
1193 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1194 | static void fd2devname(int fd, char *name); | |
1195 | ||
120dc887 | 1196 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1197 | { |
120dc887 LM |
1198 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1199 | * controller, as well as non-connected ports | |
d665cc31 DW |
1200 | */ |
1201 | int hba_len = strlen(hba_path) + 1; | |
1202 | struct dirent *ent; | |
1203 | DIR *dir; | |
1204 | char *path = NULL; | |
1205 | int err = 0; | |
1206 | unsigned long port_mask = (1 << port_count) - 1; | |
1207 | ||
f21e18ca | 1208 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1209 | if (verbose) |
1210 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1211 | return 2; | |
1212 | } | |
1213 | ||
1214 | /* scroll through /sys/dev/block looking for devices attached to | |
1215 | * this hba | |
1216 | */ | |
1217 | dir = opendir("/sys/dev/block"); | |
1218 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1219 | int fd; | |
1220 | char model[64]; | |
1221 | char vendor[64]; | |
1222 | char buf[1024]; | |
1223 | int major, minor; | |
1224 | char *device; | |
1225 | char *c; | |
1226 | int port; | |
1227 | int type; | |
1228 | ||
1229 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1230 | continue; | |
1231 | path = devt_to_devpath(makedev(major, minor)); | |
1232 | if (!path) | |
1233 | continue; | |
1234 | if (!path_attached_to_hba(path, hba_path)) { | |
1235 | free(path); | |
1236 | path = NULL; | |
1237 | continue; | |
1238 | } | |
1239 | ||
1240 | /* retrieve the scsi device type */ | |
1241 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1242 | if (verbose) | |
1243 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1244 | err = 2; | |
1245 | break; | |
1246 | } | |
1247 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1248 | if (load_sys(device, buf) != 0) { | |
1249 | if (verbose) | |
1250 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1251 | path); | |
1252 | err = 2; | |
1253 | free(device); | |
1254 | break; | |
1255 | } | |
1256 | type = strtoul(buf, NULL, 10); | |
1257 | ||
1258 | /* if it's not a disk print the vendor and model */ | |
1259 | if (!(type == 0 || type == 7 || type == 14)) { | |
1260 | vendor[0] = '\0'; | |
1261 | model[0] = '\0'; | |
1262 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1263 | if (load_sys(device, buf) == 0) { | |
1264 | strncpy(vendor, buf, sizeof(vendor)); | |
1265 | vendor[sizeof(vendor) - 1] = '\0'; | |
1266 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1267 | while (isspace(*c) || *c == '\0') | |
1268 | *c-- = '\0'; | |
1269 | ||
1270 | } | |
1271 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1272 | if (load_sys(device, buf) == 0) { | |
1273 | strncpy(model, buf, sizeof(model)); | |
1274 | model[sizeof(model) - 1] = '\0'; | |
1275 | c = (char *) &model[sizeof(model) - 1]; | |
1276 | while (isspace(*c) || *c == '\0') | |
1277 | *c-- = '\0'; | |
1278 | } | |
1279 | ||
1280 | if (vendor[0] && model[0]) | |
1281 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1282 | else | |
1283 | switch (type) { /* numbers from hald/linux/device.c */ | |
1284 | case 1: sprintf(buf, "tape"); break; | |
1285 | case 2: sprintf(buf, "printer"); break; | |
1286 | case 3: sprintf(buf, "processor"); break; | |
1287 | case 4: | |
1288 | case 5: sprintf(buf, "cdrom"); break; | |
1289 | case 6: sprintf(buf, "scanner"); break; | |
1290 | case 8: sprintf(buf, "media_changer"); break; | |
1291 | case 9: sprintf(buf, "comm"); break; | |
1292 | case 12: sprintf(buf, "raid"); break; | |
1293 | default: sprintf(buf, "unknown"); | |
1294 | } | |
1295 | } else | |
1296 | buf[0] = '\0'; | |
1297 | free(device); | |
1298 | ||
1299 | /* chop device path to 'host%d' and calculate the port number */ | |
1300 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1301 | if (!c) { |
1302 | if (verbose) | |
1303 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1304 | err = 2; | |
1305 | break; | |
1306 | } | |
d665cc31 DW |
1307 | *c = '\0'; |
1308 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1309 | port -= host_base; | |
1310 | else { | |
1311 | if (verbose) { | |
1312 | *c = '/'; /* repair the full string */ | |
1313 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1314 | path); | |
1315 | } | |
1316 | err = 2; | |
1317 | break; | |
1318 | } | |
1319 | ||
1320 | /* mark this port as used */ | |
1321 | port_mask &= ~(1 << port); | |
1322 | ||
1323 | /* print out the device information */ | |
1324 | if (buf[0]) { | |
1325 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1326 | continue; | |
1327 | } | |
1328 | ||
1329 | fd = dev_open(ent->d_name, O_RDONLY); | |
1330 | if (fd < 0) | |
1331 | printf(" Port%d : - disk info unavailable -\n", port); | |
1332 | else { | |
1333 | fd2devname(fd, buf); | |
1334 | printf(" Port%d : %s", port, buf); | |
1335 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
1336 | printf(" (%s)\n", buf); | |
1337 | else | |
1338 | printf("()\n"); | |
1339 | } | |
1340 | close(fd); | |
1341 | free(path); | |
1342 | path = NULL; | |
1343 | } | |
1344 | if (path) | |
1345 | free(path); | |
1346 | if (dir) | |
1347 | closedir(dir); | |
1348 | if (err == 0) { | |
1349 | int i; | |
1350 | ||
1351 | for (i = 0; i < port_count; i++) | |
1352 | if (port_mask & (1 << i)) | |
1353 | printf(" Port%d : - no device attached -\n", i); | |
1354 | } | |
1355 | ||
1356 | return err; | |
1357 | } | |
1358 | ||
120dc887 | 1359 | |
155cbb4c | 1360 | |
120dc887 LM |
1361 | static void print_found_intel_controllers(struct sys_dev *elem) |
1362 | { | |
1363 | for (; elem; elem = elem->next) { | |
1364 | fprintf(stderr, Name ": found Intel(R) "); | |
1365 | if (elem->type == SYS_DEV_SATA) | |
1366 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1367 | else if (elem->type == SYS_DEV_SAS) |
1368 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1369 | fprintf(stderr, "RAID controller"); |
1370 | if (elem->pci_id) | |
1371 | fprintf(stderr, " at %s", elem->pci_id); | |
1372 | fprintf(stderr, ".\n"); | |
1373 | } | |
1374 | fflush(stderr); | |
1375 | } | |
1376 | ||
120dc887 LM |
1377 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1378 | { | |
1379 | struct dirent *ent; | |
1380 | DIR *dir; | |
1381 | int host_base = -1; | |
1382 | ||
1383 | *port_count = 0; | |
1384 | if ((dir = opendir(hba_path)) == NULL) | |
1385 | return -1; | |
1386 | ||
1387 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1388 | int host; | |
1389 | ||
1390 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1391 | continue; | |
1392 | if (*port_count == 0) | |
1393 | host_base = host; | |
1394 | else if (host < host_base) | |
1395 | host_base = host; | |
1396 | ||
1397 | if (host + 1 > *port_count + host_base) | |
1398 | *port_count = host + 1 - host_base; | |
1399 | } | |
1400 | closedir(dir); | |
1401 | return host_base; | |
1402 | } | |
1403 | ||
a891a3c2 LM |
1404 | static void print_imsm_capability(const struct imsm_orom *orom) |
1405 | { | |
1406 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1407 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1408 | orom->hotfix_ver, orom->build); | |
1409 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1410 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1411 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1412 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1413 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1414 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1415 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1416 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1417 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1418 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1419 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1420 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1421 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1422 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1423 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1424 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1425 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1426 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1427 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1428 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1429 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1430 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1431 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
1432 | printf(" Max Disks : %d\n", orom->tds); | |
1433 | printf(" Max Volumes : %d\n", orom->vpa); | |
1434 | return; | |
1435 | } | |
1436 | ||
5615172f | 1437 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1438 | { |
1439 | /* There are two components to imsm platform support, the ahci SATA | |
1440 | * controller and the option-rom. To find the SATA controller we | |
1441 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1442 | * controller with the Intel vendor id is present. This approach | |
1443 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1444 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1445 | * detect platform raid capabilities. The option-rom resides in a | |
1446 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1447 | * platform capabilities. If raid support is disabled in the BIOS the | |
1448 | * option-rom capability structure will not be available. | |
1449 | */ | |
1450 | const struct imsm_orom *orom; | |
1451 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1452 | int host_base = 0; |
1453 | int port_count = 0; | |
120dc887 | 1454 | int result=0; |
d665cc31 | 1455 | |
5615172f | 1456 | if (enumerate_only) { |
a891a3c2 | 1457 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1458 | return 0; |
a891a3c2 LM |
1459 | list = find_intel_devices(); |
1460 | if (!list) | |
1461 | return 2; | |
1462 | for (hba = list; hba; hba = hba->next) { | |
1463 | orom = find_imsm_capability(hba->type); | |
1464 | if (!orom) { | |
1465 | result = 2; | |
1466 | break; | |
1467 | } | |
1468 | } | |
1469 | free_sys_dev(&list); | |
1470 | return result; | |
5615172f DW |
1471 | } |
1472 | ||
155cbb4c LM |
1473 | list = find_intel_devices(); |
1474 | if (!list) { | |
d665cc31 | 1475 | if (verbose) |
155cbb4c LM |
1476 | fprintf(stderr, Name ": no active Intel(R) RAID " |
1477 | "controller found.\n"); | |
d665cc31 DW |
1478 | free_sys_dev(&list); |
1479 | return 2; | |
1480 | } else if (verbose) | |
155cbb4c | 1481 | print_found_intel_controllers(list); |
d665cc31 | 1482 | |
a891a3c2 LM |
1483 | for (hba = list; hba; hba = hba->next) { |
1484 | orom = find_imsm_capability(hba->type); | |
1485 | if (!orom) | |
1486 | fprintf(stderr, Name ": imsm capabilities not found for controller: %s (type %s)\n", | |
1487 | hba->path, get_sys_dev_type(hba->type)); | |
1488 | else | |
1489 | print_imsm_capability(orom); | |
d665cc31 DW |
1490 | } |
1491 | ||
120dc887 LM |
1492 | for (hba = list; hba; hba = hba->next) { |
1493 | printf(" I/O Controller : %s (%s)\n", | |
1494 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1495 | |
120dc887 LM |
1496 | if (hba->type == SYS_DEV_SATA) { |
1497 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1498 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1499 | if (verbose) | |
1500 | fprintf(stderr, Name ": failed to enumerate " | |
1501 | "ports on SATA controller at %s.", hba->pci_id); | |
1502 | result |= 2; | |
1503 | } | |
1504 | } | |
d665cc31 | 1505 | } |
155cbb4c | 1506 | |
120dc887 LM |
1507 | free_sys_dev(&list); |
1508 | return result; | |
d665cc31 | 1509 | } |
cdddbdbc DW |
1510 | #endif |
1511 | ||
1512 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1513 | { | |
5115ca67 DW |
1514 | /* the imsm metadata format does not specify any host |
1515 | * identification information. We return -1 since we can never | |
1516 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1517 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1518 | * exclude member disks that do not belong, and we rely on |
1519 | * mdadm.conf to specify the arrays that should be assembled. | |
1520 | * Auto-assembly may still pick up "foreign" arrays. | |
1521 | */ | |
cdddbdbc | 1522 | |
9362c1c8 | 1523 | return -1; |
cdddbdbc DW |
1524 | } |
1525 | ||
1526 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1527 | { | |
51006d85 N |
1528 | /* The uuid returned here is used for: |
1529 | * uuid to put into bitmap file (Create, Grow) | |
1530 | * uuid for backup header when saving critical section (Grow) | |
1531 | * comparing uuids when re-adding a device into an array | |
1532 | * In these cases the uuid required is that of the data-array, | |
1533 | * not the device-set. | |
1534 | * uuid to recognise same set when adding a missing device back | |
1535 | * to an array. This is a uuid for the device-set. | |
1536 | * | |
1537 | * For each of these we can make do with a truncated | |
1538 | * or hashed uuid rather than the original, as long as | |
1539 | * everyone agrees. | |
1540 | * In each case the uuid required is that of the data-array, | |
1541 | * not the device-set. | |
43dad3d6 | 1542 | */ |
51006d85 N |
1543 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1544 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1545 | * - the orig_family_num of the container |
51006d85 N |
1546 | * - the index number of the volume |
1547 | * - the 'serial' number of the volume. | |
1548 | * Hopefully these are all constant. | |
1549 | */ | |
1550 | struct intel_super *super = st->sb; | |
43dad3d6 | 1551 | |
51006d85 N |
1552 | char buf[20]; |
1553 | struct sha1_ctx ctx; | |
1554 | struct imsm_dev *dev = NULL; | |
148acb7b | 1555 | __u32 family_num; |
51006d85 | 1556 | |
148acb7b DW |
1557 | /* some mdadm versions failed to set ->orig_family_num, in which |
1558 | * case fall back to ->family_num. orig_family_num will be | |
1559 | * fixed up with the first metadata update. | |
1560 | */ | |
1561 | family_num = super->anchor->orig_family_num; | |
1562 | if (family_num == 0) | |
1563 | family_num = super->anchor->family_num; | |
51006d85 | 1564 | sha1_init_ctx(&ctx); |
92bd8f8d | 1565 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1566 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1567 | if (super->current_vol >= 0) |
1568 | dev = get_imsm_dev(super, super->current_vol); | |
1569 | if (dev) { | |
1570 | __u32 vol = super->current_vol; | |
1571 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1572 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1573 | } | |
1574 | sha1_finish_ctx(&ctx, buf); | |
1575 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1576 | } |
1577 | ||
0d481d37 | 1578 | #if 0 |
4f5bc454 DW |
1579 | static void |
1580 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1581 | { |
cdddbdbc DW |
1582 | __u8 *v = get_imsm_version(mpb); |
1583 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1584 | char major[] = { 0, 0, 0 }; | |
1585 | char minor[] = { 0 ,0, 0 }; | |
1586 | char patch[] = { 0, 0, 0 }; | |
1587 | char *ver_parse[] = { major, minor, patch }; | |
1588 | int i, j; | |
1589 | ||
1590 | i = j = 0; | |
1591 | while (*v != '\0' && v < end) { | |
1592 | if (*v != '.' && j < 2) | |
1593 | ver_parse[i][j++] = *v; | |
1594 | else { | |
1595 | i++; | |
1596 | j = 0; | |
1597 | } | |
1598 | v++; | |
1599 | } | |
1600 | ||
4f5bc454 DW |
1601 | *m = strtol(minor, NULL, 0); |
1602 | *p = strtol(patch, NULL, 0); | |
1603 | } | |
0d481d37 | 1604 | #endif |
4f5bc454 | 1605 | |
1e5c6983 DW |
1606 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1607 | { | |
1608 | /* migr_strip_size when repairing or initializing parity */ | |
1609 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1610 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1611 | ||
1612 | switch (get_imsm_raid_level(map)) { | |
1613 | case 5: | |
1614 | case 10: | |
1615 | return chunk; | |
1616 | default: | |
1617 | return 128*1024 >> 9; | |
1618 | } | |
1619 | } | |
1620 | ||
1621 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1622 | { | |
1623 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1624 | * this is different than migr_strip_size_resync(), but it's good | |
1625 | * to be compatible | |
1626 | */ | |
1627 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1628 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1629 | ||
1630 | switch (get_imsm_raid_level(map)) { | |
1631 | case 1: | |
1632 | case 10: | |
1633 | if (map->num_members % map->num_domains == 0) | |
1634 | return 128*1024 >> 9; | |
1635 | else | |
1636 | return chunk; | |
1637 | case 5: | |
1638 | return max((__u32) 64*1024 >> 9, chunk); | |
1639 | default: | |
1640 | return 128*1024 >> 9; | |
1641 | } | |
1642 | } | |
1643 | ||
1644 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
1645 | { | |
1646 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1647 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1648 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); | |
1649 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
1650 | ||
1651 | return max((__u32) 1, hi_chunk / lo_chunk); | |
1652 | } | |
1653 | ||
1654 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
1655 | { | |
1656 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1657 | int level = get_imsm_raid_level(lo); | |
1658 | ||
1659 | if (level == 1 || level == 10) { | |
1660 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1661 | ||
1662 | return hi->num_domains; | |
1663 | } else | |
1664 | return num_stripes_per_unit_resync(dev); | |
1665 | } | |
1666 | ||
98130f40 | 1667 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
1668 | { |
1669 | /* named 'imsm_' because raid0, raid1 and raid10 | |
1670 | * counter-intuitively have the same number of data disks | |
1671 | */ | |
98130f40 | 1672 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
1673 | |
1674 | switch (get_imsm_raid_level(map)) { | |
1675 | case 0: | |
1676 | case 1: | |
1677 | case 10: | |
1678 | return map->num_members; | |
1679 | case 5: | |
1680 | return map->num_members - 1; | |
1681 | default: | |
1682 | dprintf("%s: unsupported raid level\n", __func__); | |
1683 | return 0; | |
1684 | } | |
1685 | } | |
1686 | ||
1687 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
1688 | { | |
1689 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1690 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1691 | ||
1692 | switch(get_imsm_raid_level(map)) { | |
1693 | case 1: | |
1694 | case 10: | |
1695 | return chunk * map->num_domains; | |
1696 | case 5: | |
1697 | return chunk * map->num_members; | |
1698 | default: | |
1699 | return chunk; | |
1700 | } | |
1701 | } | |
1702 | ||
1703 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
1704 | { | |
1705 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1706 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1707 | __u32 strip = block / chunk; | |
1708 | ||
1709 | switch (get_imsm_raid_level(map)) { | |
1710 | case 1: | |
1711 | case 10: { | |
1712 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
1713 | __u32 vol_stripe = vol_strip / map->num_members; | |
1714 | ||
1715 | return vol_stripe * chunk + block % chunk; | |
1716 | } case 5: { | |
1717 | __u32 stripe = strip / (map->num_members - 1); | |
1718 | ||
1719 | return stripe * chunk + block % chunk; | |
1720 | } | |
1721 | default: | |
1722 | return 0; | |
1723 | } | |
1724 | } | |
1725 | ||
1726 | static __u64 blocks_per_migr_unit(struct imsm_dev *dev) | |
1727 | { | |
1728 | /* calculate the conversion factor between per member 'blocks' | |
1729 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
1730 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
1731 | */ | |
1732 | if (!dev->vol.migr_state) | |
1733 | return 0; | |
1734 | ||
1735 | switch (migr_type(dev)) { | |
6345120e | 1736 | case MIGR_GEN_MIGR: |
1e5c6983 DW |
1737 | case MIGR_VERIFY: |
1738 | case MIGR_REPAIR: | |
1739 | case MIGR_INIT: { | |
1740 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1741 | __u32 stripes_per_unit; | |
1742 | __u32 blocks_per_unit; | |
1743 | __u32 parity_depth; | |
1744 | __u32 migr_chunk; | |
1745 | __u32 block_map; | |
1746 | __u32 block_rel; | |
1747 | __u32 segment; | |
1748 | __u32 stripe; | |
1749 | __u8 disks; | |
1750 | ||
1751 | /* yes, this is really the translation of migr_units to | |
1752 | * per-member blocks in the 'resync' case | |
1753 | */ | |
1754 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
1755 | migr_chunk = migr_strip_blocks_resync(dev); | |
98130f40 | 1756 | disks = imsm_num_data_members(dev, 0); |
1e5c6983 DW |
1757 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
1758 | stripe = __le32_to_cpu(map->blocks_per_strip) * disks; | |
1759 | segment = blocks_per_unit / stripe; | |
1760 | block_rel = blocks_per_unit - segment * stripe; | |
1761 | parity_depth = parity_segment_depth(dev); | |
1762 | block_map = map_migr_block(dev, block_rel); | |
1763 | return block_map + parity_depth * segment; | |
1764 | } | |
1765 | case MIGR_REBUILD: { | |
1766 | __u32 stripes_per_unit; | |
1767 | __u32 migr_chunk; | |
1768 | ||
1769 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
1770 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
1771 | return migr_chunk * stripes_per_unit; | |
1772 | } | |
1e5c6983 DW |
1773 | case MIGR_STATE_CHANGE: |
1774 | default: | |
1775 | return 0; | |
1776 | } | |
1777 | } | |
1778 | ||
c2c087e6 DW |
1779 | static int imsm_level_to_layout(int level) |
1780 | { | |
1781 | switch (level) { | |
1782 | case 0: | |
1783 | case 1: | |
1784 | return 0; | |
1785 | case 5: | |
1786 | case 6: | |
a380c027 | 1787 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 1788 | case 10: |
c92a2527 | 1789 | return 0x102; |
c2c087e6 | 1790 | } |
a18a888e | 1791 | return UnSet; |
c2c087e6 DW |
1792 | } |
1793 | ||
8e59f3d8 AK |
1794 | /******************************************************************************* |
1795 | * Function: read_imsm_migr_rec | |
1796 | * Description: Function reads imsm migration record from last sector of disk | |
1797 | * Parameters: | |
1798 | * fd : disk descriptor | |
1799 | * super : metadata info | |
1800 | * Returns: | |
1801 | * 0 : success, | |
1802 | * -1 : fail | |
1803 | ******************************************************************************/ | |
1804 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
1805 | { | |
1806 | int ret_val = -1; | |
1807 | unsigned long long dsize; | |
1808 | ||
1809 | get_dev_size(fd, NULL, &dsize); | |
1810 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
1811 | fprintf(stderr, | |
1812 | Name ": Cannot seek to anchor block: %s\n", | |
1813 | strerror(errno)); | |
1814 | goto out; | |
1815 | } | |
1816 | if (read(fd, super->migr_rec_buf, 512) != 512) { | |
1817 | fprintf(stderr, | |
1818 | Name ": Cannot read migr record block: %s\n", | |
1819 | strerror(errno)); | |
1820 | goto out; | |
1821 | } | |
1822 | ret_val = 0; | |
1823 | ||
1824 | out: | |
1825 | return ret_val; | |
1826 | } | |
1827 | ||
1828 | /******************************************************************************* | |
1829 | * Function: load_imsm_migr_rec | |
1830 | * Description: Function reads imsm migration record (it is stored at the last | |
1831 | * sector of disk) | |
1832 | * Parameters: | |
1833 | * super : imsm internal array info | |
1834 | * info : general array info | |
1835 | * Returns: | |
1836 | * 0 : success | |
1837 | * -1 : fail | |
1838 | ******************************************************************************/ | |
1839 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
1840 | { | |
1841 | struct mdinfo *sd; | |
1842 | struct dl *dl = NULL; | |
1843 | char nm[30]; | |
1844 | int retval = -1; | |
1845 | int fd = -1; | |
1846 | ||
1847 | if (info) { | |
1848 | for (sd = info->devs ; sd ; sd = sd->next) { | |
1849 | /* read only from one of the first two slots */ | |
1850 | if ((sd->disk.raid_disk > 1) || | |
1851 | (sd->disk.raid_disk < 0)) | |
1852 | continue; | |
1853 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1854 | fd = dev_open(nm, O_RDONLY); | |
1855 | if (fd >= 0) | |
1856 | break; | |
1857 | } | |
1858 | } | |
1859 | if (fd < 0) { | |
1860 | for (dl = super->disks; dl; dl = dl->next) { | |
1861 | /* read only from one of the first two slots */ | |
1862 | if (dl->index > 1) | |
1863 | continue; | |
1864 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
1865 | fd = dev_open(nm, O_RDONLY); | |
1866 | if (fd >= 0) | |
1867 | break; | |
1868 | } | |
1869 | } | |
1870 | if (fd < 0) | |
1871 | goto out; | |
1872 | retval = read_imsm_migr_rec(fd, super); | |
1873 | ||
1874 | out: | |
1875 | if (fd >= 0) | |
1876 | close(fd); | |
1877 | return retval; | |
1878 | } | |
1879 | ||
687629c2 AK |
1880 | /******************************************************************************* |
1881 | * Function: write_imsm_migr_rec | |
1882 | * Description: Function writes imsm migration record | |
1883 | * (at the last sector of disk) | |
1884 | * Parameters: | |
1885 | * super : imsm internal array info | |
1886 | * Returns: | |
1887 | * 0 : success | |
1888 | * -1 : if fail | |
1889 | ******************************************************************************/ | |
1890 | static int write_imsm_migr_rec(struct supertype *st) | |
1891 | { | |
1892 | struct intel_super *super = st->sb; | |
1893 | unsigned long long dsize; | |
1894 | char nm[30]; | |
1895 | int fd = -1; | |
1896 | int retval = -1; | |
1897 | struct dl *sd; | |
1898 | ||
1899 | for (sd = super->disks ; sd ; sd = sd->next) { | |
1900 | /* write to 2 first slots only */ | |
1901 | if ((sd->index < 0) || (sd->index > 1)) | |
1902 | continue; | |
1903 | sprintf(nm, "%d:%d", sd->major, sd->minor); | |
1904 | fd = dev_open(nm, O_RDWR); | |
1905 | if (fd < 0) | |
1906 | continue; | |
1907 | get_dev_size(fd, NULL, &dsize); | |
1908 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
1909 | fprintf(stderr, | |
1910 | Name ": Cannot seek to anchor block: %s\n", | |
1911 | strerror(errno)); | |
1912 | goto out; | |
1913 | } | |
1914 | if (write(fd, super->migr_rec_buf, 512) != 512) { | |
1915 | fprintf(stderr, | |
1916 | Name ": Cannot write migr record block: %s\n", | |
1917 | strerror(errno)); | |
1918 | goto out; | |
1919 | } | |
1920 | close(fd); | |
1921 | fd = -1; | |
1922 | } | |
1923 | ||
1924 | retval = 0; | |
1925 | out: | |
1926 | if (fd >= 0) | |
1927 | close(fd); | |
1928 | return retval; | |
1929 | } | |
1930 | ||
a5d85af7 | 1931 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
1932 | { |
1933 | struct intel_super *super = st->sb; | |
949c47a0 | 1934 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 1935 | struct imsm_map *map = get_imsm_map(dev, 0); |
81ac8b4d | 1936 | struct imsm_map *prev_map = get_imsm_map(dev, 1); |
b335e593 | 1937 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 1938 | struct dl *dl; |
e207da2f | 1939 | char *devname; |
139dae11 | 1940 | unsigned int component_size_alligment; |
a5d85af7 | 1941 | int map_disks = info->array.raid_disks; |
bf5a934a | 1942 | |
95eeceeb | 1943 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
1944 | if (prev_map) |
1945 | map_to_analyse = prev_map; | |
1946 | ||
efb30e7f DW |
1947 | for (dl = super->disks; dl; dl = dl->next) |
1948 | if (dl->raiddisk == info->disk.raid_disk) | |
1949 | break; | |
bf5a934a | 1950 | info->container_member = super->current_vol; |
cd0430a1 | 1951 | info->array.raid_disks = map->num_members; |
b335e593 | 1952 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
1953 | info->array.layout = imsm_level_to_layout(info->array.level); |
1954 | info->array.md_minor = -1; | |
1955 | info->array.ctime = 0; | |
1956 | info->array.utime = 0; | |
b335e593 AK |
1957 | info->array.chunk_size = |
1958 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 1959 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
1960 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
1961 | info->custom_array_size <<= 32; | |
1962 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3f83228a N |
1963 | if (prev_map && map->map_state == prev_map->map_state) { |
1964 | info->reshape_active = 1; | |
b335e593 AK |
1965 | info->new_level = get_imsm_raid_level(map); |
1966 | info->new_layout = imsm_level_to_layout(info->new_level); | |
1967 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 1968 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
1969 | if (info->delta_disks) { |
1970 | /* this needs to be applied to every array | |
1971 | * in the container. | |
1972 | */ | |
1973 | info->reshape_active = 2; | |
1974 | } | |
3f83228a N |
1975 | /* We shape information that we give to md might have to be |
1976 | * modify to cope with md's requirement for reshaping arrays. | |
1977 | * For example, when reshaping a RAID0, md requires it to be | |
1978 | * presented as a degraded RAID4. | |
1979 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
1980 | * the array as already being RAID5, but the 'before' layout | |
1981 | * is a RAID4-like layout. | |
1982 | */ | |
1983 | switch (info->array.level) { | |
1984 | case 0: | |
1985 | switch(info->new_level) { | |
1986 | case 0: | |
1987 | /* conversion is happening as RAID4 */ | |
1988 | info->array.level = 4; | |
1989 | info->array.raid_disks += 1; | |
1990 | break; | |
1991 | case 5: | |
1992 | /* conversion is happening as RAID5 */ | |
1993 | info->array.level = 5; | |
1994 | info->array.layout = ALGORITHM_PARITY_N; | |
1995 | info->array.raid_disks += 1; | |
1996 | info->delta_disks -= 1; | |
1997 | break; | |
1998 | default: | |
1999 | /* FIXME error message */ | |
2000 | info->array.level = UnSet; | |
2001 | break; | |
2002 | } | |
2003 | break; | |
2004 | } | |
b335e593 AK |
2005 | } else { |
2006 | info->new_level = UnSet; | |
2007 | info->new_layout = UnSet; | |
2008 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2009 | info->delta_disks = 0; |
b335e593 | 2010 | } |
301406c9 DW |
2011 | info->disk.major = 0; |
2012 | info->disk.minor = 0; | |
efb30e7f DW |
2013 | if (dl) { |
2014 | info->disk.major = dl->major; | |
2015 | info->disk.minor = dl->minor; | |
2016 | } | |
bf5a934a | 2017 | |
b335e593 AK |
2018 | info->data_offset = __le32_to_cpu(map_to_analyse->pba_of_lba0); |
2019 | info->component_size = | |
2020 | __le32_to_cpu(map_to_analyse->blocks_per_member); | |
139dae11 AK |
2021 | |
2022 | /* check component size aligment | |
2023 | */ | |
2024 | component_size_alligment = | |
2025 | info->component_size % (info->array.chunk_size/512); | |
2026 | ||
2027 | if (component_size_alligment && | |
2028 | (info->array.level != 1) && (info->array.level != UnSet)) { | |
2029 | dprintf("imsm: reported component size alligned from %llu ", | |
2030 | info->component_size); | |
2031 | info->component_size -= component_size_alligment; | |
2032 | dprintf("to %llu (%i).\n", | |
2033 | info->component_size, component_size_alligment); | |
2034 | } | |
2035 | ||
301406c9 | 2036 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2037 | info->recovery_start = MaxSector; |
bf5a934a | 2038 | |
d2e6d5d6 | 2039 | info->reshape_progress = 0; |
b6796ce1 | 2040 | info->resync_start = MaxSector; |
b335e593 AK |
2041 | if (map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2042 | dev->vol.dirty) { | |
301406c9 | 2043 | info->resync_start = 0; |
b6796ce1 AK |
2044 | } |
2045 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2046 | switch (migr_type(dev)) { |
2047 | case MIGR_REPAIR: | |
2048 | case MIGR_INIT: { | |
2049 | __u64 blocks_per_unit = blocks_per_migr_unit(dev); | |
2050 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); | |
2051 | ||
2052 | info->resync_start = blocks_per_unit * units; | |
2053 | break; | |
2054 | } | |
d2e6d5d6 AK |
2055 | case MIGR_GEN_MIGR: { |
2056 | __u64 blocks_per_unit = blocks_per_migr_unit(dev); | |
2057 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); | |
04fa9523 AK |
2058 | unsigned long long array_blocks; |
2059 | int used_disks; | |
d2e6d5d6 AK |
2060 | |
2061 | info->reshape_progress = blocks_per_unit * units; | |
6289d1e0 | 2062 | |
d2e6d5d6 AK |
2063 | dprintf("IMSM: General Migration checkpoint : %llu " |
2064 | "(%llu) -> read reshape progress : %llu\n", | |
2065 | units, blocks_per_unit, info->reshape_progress); | |
75156c46 AK |
2066 | |
2067 | used_disks = imsm_num_data_members(dev, 1); | |
2068 | if (used_disks > 0) { | |
2069 | array_blocks = map->blocks_per_member * | |
2070 | used_disks; | |
2071 | /* round array size down to closest MB | |
2072 | */ | |
2073 | info->custom_array_size = (array_blocks | |
2074 | >> SECT_PER_MB_SHIFT) | |
2075 | << SECT_PER_MB_SHIFT; | |
2076 | } | |
d2e6d5d6 | 2077 | } |
1e5c6983 DW |
2078 | case MIGR_VERIFY: |
2079 | /* we could emulate the checkpointing of | |
2080 | * 'sync_action=check' migrations, but for now | |
2081 | * we just immediately complete them | |
2082 | */ | |
2083 | case MIGR_REBUILD: | |
2084 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2085 | case MIGR_STATE_CHANGE: |
2086 | /* FIXME handle other migrations */ | |
2087 | default: | |
2088 | /* we are not dirty, so... */ | |
2089 | info->resync_start = MaxSector; | |
2090 | } | |
b6796ce1 | 2091 | } |
301406c9 DW |
2092 | |
2093 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2094 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2095 | |
f35f2525 N |
2096 | info->array.major_version = -1; |
2097 | info->array.minor_version = -2; | |
e207da2f AW |
2098 | devname = devnum2devname(st->container_dev); |
2099 | *info->text_version = '\0'; | |
2100 | if (devname) | |
2101 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2102 | free(devname); | |
a67dd8cc | 2103 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2104 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2105 | |
2106 | if (dmap) { | |
2107 | int i, j; | |
2108 | for (i=0; i<map_disks; i++) { | |
2109 | dmap[i] = 0; | |
2110 | if (i < info->array.raid_disks) { | |
2111 | struct imsm_disk *dsk; | |
98130f40 | 2112 | j = get_imsm_disk_idx(dev, i, -1); |
a5d85af7 N |
2113 | dsk = get_imsm_disk(super, j); |
2114 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2115 | dmap[i] = 1; | |
2116 | } | |
2117 | } | |
2118 | } | |
81ac8b4d | 2119 | } |
bf5a934a | 2120 | |
97b4d0e9 DW |
2121 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed); |
2122 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev); | |
2123 | ||
2124 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2125 | { | |
2126 | struct dl *d; | |
2127 | ||
2128 | for (d = super->missing; d; d = d->next) | |
2129 | if (d->index == index) | |
2130 | return &d->disk; | |
2131 | return NULL; | |
2132 | } | |
2133 | ||
a5d85af7 | 2134 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2135 | { |
2136 | struct intel_super *super = st->sb; | |
4f5bc454 | 2137 | struct imsm_disk *disk; |
a5d85af7 | 2138 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2139 | int max_enough = -1; |
2140 | int i; | |
2141 | struct imsm_super *mpb; | |
4f5bc454 | 2142 | |
bf5a934a | 2143 | if (super->current_vol >= 0) { |
a5d85af7 | 2144 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2145 | return; |
2146 | } | |
95eeceeb | 2147 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2148 | |
2149 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2150 | * spares | |
2151 | */ | |
2152 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2153 | info->array.level = LEVEL_CONTAINER; |
2154 | info->array.layout = 0; | |
2155 | info->array.md_minor = -1; | |
c2c087e6 | 2156 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2157 | info->array.utime = 0; |
2158 | info->array.chunk_size = 0; | |
2159 | ||
2160 | info->disk.major = 0; | |
2161 | info->disk.minor = 0; | |
cdddbdbc | 2162 | info->disk.raid_disk = -1; |
c2c087e6 | 2163 | info->reshape_active = 0; |
f35f2525 N |
2164 | info->array.major_version = -1; |
2165 | info->array.minor_version = -2; | |
c2c087e6 | 2166 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2167 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2168 | info->disk.number = -1; |
2169 | info->disk.state = 0; | |
c5afc314 | 2170 | info->name[0] = 0; |
921d9e16 | 2171 | info->recovery_start = MaxSector; |
c2c087e6 | 2172 | |
97b4d0e9 | 2173 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2174 | mpb = super->anchor; |
97b4d0e9 | 2175 | |
ab3cb6b3 N |
2176 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2177 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2178 | int failed, enough, j, missing = 0; | |
2179 | struct imsm_map *map; | |
2180 | __u8 state; | |
97b4d0e9 | 2181 | |
ab3cb6b3 N |
2182 | failed = imsm_count_failed(super, dev); |
2183 | state = imsm_check_degraded(super, dev, failed); | |
2184 | map = get_imsm_map(dev, dev->vol.migr_state); | |
2185 | ||
2186 | /* any newly missing disks? | |
2187 | * (catches single-degraded vs double-degraded) | |
2188 | */ | |
2189 | for (j = 0; j < map->num_members; j++) { | |
98130f40 | 2190 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
ab3cb6b3 N |
2191 | __u32 idx = ord_to_idx(ord); |
2192 | ||
2193 | if (!(ord & IMSM_ORD_REBUILD) && | |
2194 | get_imsm_missing(super, idx)) { | |
2195 | missing = 1; | |
2196 | break; | |
2197 | } | |
97b4d0e9 | 2198 | } |
ab3cb6b3 N |
2199 | |
2200 | if (state == IMSM_T_STATE_FAILED) | |
2201 | enough = -1; | |
2202 | else if (state == IMSM_T_STATE_DEGRADED && | |
2203 | (state != map->map_state || missing)) | |
2204 | enough = 0; | |
2205 | else /* we're normal, or already degraded */ | |
2206 | enough = 1; | |
2207 | ||
2208 | /* in the missing/failed disk case check to see | |
2209 | * if at least one array is runnable | |
2210 | */ | |
2211 | max_enough = max(max_enough, enough); | |
2212 | } | |
2213 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2214 | info->container_enough = max_enough; | |
97b4d0e9 | 2215 | |
4a04ec6c | 2216 | if (super->disks) { |
14e8215b DW |
2217 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2218 | ||
b9f594fe | 2219 | disk = &super->disks->disk; |
14e8215b DW |
2220 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
2221 | info->component_size = reserved; | |
25ed7e59 | 2222 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2223 | /* we don't change info->disk.raid_disk here because |
2224 | * this state will be finalized in mdmon after we have | |
2225 | * found the 'most fresh' version of the metadata | |
2226 | */ | |
25ed7e59 DW |
2227 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2228 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2229 | } |
a575e2a7 DW |
2230 | |
2231 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2232 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2233 | */ | |
2234 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2235 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2236 | else |
2237 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2238 | |
2239 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2240 | if (map) { | |
2241 | int i; | |
2242 | for (i = 0; i < map_disks; i++) | |
2243 | map[i] = 1; | |
2244 | } | |
2245 | ||
cdddbdbc DW |
2246 | } |
2247 | ||
5c4cd5da AC |
2248 | /* allocates memory and fills disk in mdinfo structure |
2249 | * for each disk in array */ | |
2250 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2251 | { | |
2252 | struct mdinfo *mddev = NULL; | |
2253 | struct intel_super *super = st->sb; | |
2254 | struct imsm_disk *disk; | |
2255 | int count = 0; | |
2256 | struct dl *dl; | |
2257 | if (!super || !super->disks) | |
2258 | return NULL; | |
2259 | dl = super->disks; | |
2260 | mddev = malloc(sizeof(*mddev)); | |
2261 | if (!mddev) { | |
2262 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2263 | return NULL; | |
2264 | } | |
2265 | memset(mddev, 0, sizeof(*mddev)); | |
2266 | while (dl) { | |
2267 | struct mdinfo *tmp; | |
2268 | disk = &dl->disk; | |
2269 | tmp = malloc(sizeof(*tmp)); | |
2270 | if (!tmp) { | |
2271 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2272 | if (mddev) | |
2273 | sysfs_free(mddev); | |
2274 | return NULL; | |
2275 | } | |
2276 | memset(tmp, 0, sizeof(*tmp)); | |
2277 | if (mddev->devs) | |
2278 | tmp->next = mddev->devs; | |
2279 | mddev->devs = tmp; | |
2280 | tmp->disk.number = count++; | |
2281 | tmp->disk.major = dl->major; | |
2282 | tmp->disk.minor = dl->minor; | |
2283 | tmp->disk.state = is_configured(disk) ? | |
2284 | (1 << MD_DISK_ACTIVE) : 0; | |
2285 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2286 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2287 | tmp->disk.raid_disk = -1; | |
2288 | dl = dl->next; | |
2289 | } | |
2290 | return mddev; | |
2291 | } | |
2292 | ||
cdddbdbc DW |
2293 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2294 | char *update, char *devname, int verbose, | |
2295 | int uuid_set, char *homehost) | |
2296 | { | |
f352c545 DW |
2297 | /* For 'assemble' and 'force' we need to return non-zero if any |
2298 | * change was made. For others, the return value is ignored. | |
2299 | * Update options are: | |
2300 | * force-one : This device looks a bit old but needs to be included, | |
2301 | * update age info appropriately. | |
2302 | * assemble: clear any 'faulty' flag to allow this device to | |
2303 | * be assembled. | |
2304 | * force-array: Array is degraded but being forced, mark it clean | |
2305 | * if that will be needed to assemble it. | |
2306 | * | |
2307 | * newdev: not used ???? | |
2308 | * grow: Array has gained a new device - this is currently for | |
2309 | * linear only | |
2310 | * resync: mark as dirty so a resync will happen. | |
2311 | * name: update the name - preserving the homehost | |
6e46bf34 | 2312 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2313 | * |
2314 | * Following are not relevant for this imsm: | |
2315 | * sparc2.2 : update from old dodgey metadata | |
2316 | * super-minor: change the preferred_minor number | |
2317 | * summaries: update redundant counters. | |
f352c545 DW |
2318 | * homehost: update the recorded homehost |
2319 | * _reshape_progress: record new reshape_progress position. | |
2320 | */ | |
6e46bf34 DW |
2321 | int rv = 1; |
2322 | struct intel_super *super = st->sb; | |
2323 | struct imsm_super *mpb; | |
f352c545 | 2324 | |
6e46bf34 DW |
2325 | /* we can only update container info */ |
2326 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2327 | return 1; | |
2328 | ||
2329 | mpb = super->anchor; | |
2330 | ||
2331 | if (strcmp(update, "uuid") == 0 && uuid_set && !info->update_private) | |
1e2b2765 | 2332 | rv = -1; |
6e46bf34 DW |
2333 | else if (strcmp(update, "uuid") == 0 && uuid_set && info->update_private) { |
2334 | mpb->orig_family_num = *((__u32 *) info->update_private); | |
2335 | rv = 0; | |
2336 | } else if (strcmp(update, "uuid") == 0) { | |
2337 | __u32 *new_family = malloc(sizeof(*new_family)); | |
2338 | ||
2339 | /* update orig_family_number with the incoming random | |
2340 | * data, report the new effective uuid, and store the | |
2341 | * new orig_family_num for future updates. | |
2342 | */ | |
2343 | if (new_family) { | |
2344 | memcpy(&mpb->orig_family_num, info->uuid, sizeof(__u32)); | |
2345 | uuid_from_super_imsm(st, info->uuid); | |
2346 | *new_family = mpb->orig_family_num; | |
2347 | info->update_private = new_family; | |
2348 | rv = 0; | |
2349 | } | |
2350 | } else if (strcmp(update, "assemble") == 0) | |
2351 | rv = 0; | |
2352 | else | |
1e2b2765 | 2353 | rv = -1; |
f352c545 | 2354 | |
6e46bf34 DW |
2355 | /* successful update? recompute checksum */ |
2356 | if (rv == 0) | |
2357 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2358 | |
2359 | return rv; | |
cdddbdbc DW |
2360 | } |
2361 | ||
c2c087e6 | 2362 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2363 | { |
c2c087e6 | 2364 | size_t size; |
cdddbdbc | 2365 | |
c2c087e6 DW |
2366 | size = sizeof(struct imsm_super); |
2367 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2368 | size += 2 * sizeof(struct imsm_dev); | |
2369 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2370 | size += (4 - 2) * sizeof(struct imsm_map); | |
2371 | /* 4 possible disk_ord_tbl's */ | |
2372 | size += 4 * (disks - 1) * sizeof(__u32); | |
2373 | ||
2374 | return size; | |
2375 | } | |
2376 | ||
2377 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
2378 | { | |
2379 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
2380 | return 0; | |
2381 | ||
2382 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
2383 | } |
2384 | ||
ba2de7ba DW |
2385 | static void free_devlist(struct intel_super *super) |
2386 | { | |
2387 | struct intel_dev *dv; | |
2388 | ||
2389 | while (super->devlist) { | |
2390 | dv = super->devlist->next; | |
2391 | free(super->devlist->dev); | |
2392 | free(super->devlist); | |
2393 | super->devlist = dv; | |
2394 | } | |
2395 | } | |
2396 | ||
2397 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
2398 | { | |
2399 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
2400 | } | |
2401 | ||
cdddbdbc DW |
2402 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
2403 | { | |
2404 | /* | |
2405 | * return: | |
2406 | * 0 same, or first was empty, and second was copied | |
2407 | * 1 second had wrong number | |
2408 | * 2 wrong uuid | |
2409 | * 3 wrong other info | |
2410 | */ | |
2411 | struct intel_super *first = st->sb; | |
2412 | struct intel_super *sec = tst->sb; | |
2413 | ||
2414 | if (!first) { | |
2415 | st->sb = tst->sb; | |
2416 | tst->sb = NULL; | |
2417 | return 0; | |
2418 | } | |
8603ea6f LM |
2419 | /* in platform dependent environment test if the disks |
2420 | * use the same Intel hba | |
2421 | */ | |
2422 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
2423 | if (!first->hba || !sec->hba || |
2424 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
2425 | fprintf(stderr, |
2426 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
2427 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
2428 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
2429 | return 3; |
2430 | } | |
2431 | } | |
cdddbdbc | 2432 | |
d23fe947 DW |
2433 | /* if an anchor does not have num_raid_devs set then it is a free |
2434 | * floating spare | |
2435 | */ | |
2436 | if (first->anchor->num_raid_devs > 0 && | |
2437 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
2438 | /* Determine if these disks might ever have been |
2439 | * related. Further disambiguation can only take place | |
2440 | * in load_super_imsm_all | |
2441 | */ | |
2442 | __u32 first_family = first->anchor->orig_family_num; | |
2443 | __u32 sec_family = sec->anchor->orig_family_num; | |
2444 | ||
f796af5d DW |
2445 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
2446 | MAX_SIGNATURE_LENGTH) != 0) | |
2447 | return 3; | |
2448 | ||
a2b97981 DW |
2449 | if (first_family == 0) |
2450 | first_family = first->anchor->family_num; | |
2451 | if (sec_family == 0) | |
2452 | sec_family = sec->anchor->family_num; | |
2453 | ||
2454 | if (first_family != sec_family) | |
d23fe947 | 2455 | return 3; |
f796af5d | 2456 | |
d23fe947 | 2457 | } |
cdddbdbc | 2458 | |
f796af5d | 2459 | |
3e372e5a DW |
2460 | /* if 'first' is a spare promote it to a populated mpb with sec's |
2461 | * family number | |
2462 | */ | |
2463 | if (first->anchor->num_raid_devs == 0 && | |
2464 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 2465 | int i; |
ba2de7ba DW |
2466 | struct intel_dev *dv; |
2467 | struct imsm_dev *dev; | |
78d30f94 DW |
2468 | |
2469 | /* we need to copy raid device info from sec if an allocation | |
2470 | * fails here we don't associate the spare | |
2471 | */ | |
2472 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
2473 | dv = malloc(sizeof(*dv)); |
2474 | if (!dv) | |
2475 | break; | |
2476 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
2477 | if (!dev) { | |
2478 | free(dv); | |
2479 | break; | |
78d30f94 | 2480 | } |
ba2de7ba DW |
2481 | dv->dev = dev; |
2482 | dv->index = i; | |
2483 | dv->next = first->devlist; | |
2484 | first->devlist = dv; | |
78d30f94 | 2485 | } |
709743c5 | 2486 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
2487 | /* allocation failure */ |
2488 | free_devlist(first); | |
2489 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
2490 | return 3; | |
78d30f94 | 2491 | } |
3e372e5a | 2492 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 2493 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 2494 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 2495 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
2496 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
2497 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
2498 | } |
2499 | ||
cdddbdbc DW |
2500 | return 0; |
2501 | } | |
2502 | ||
0030e8d6 DW |
2503 | static void fd2devname(int fd, char *name) |
2504 | { | |
2505 | struct stat st; | |
2506 | char path[256]; | |
33a6535d | 2507 | char dname[PATH_MAX]; |
0030e8d6 DW |
2508 | char *nm; |
2509 | int rv; | |
2510 | ||
2511 | name[0] = '\0'; | |
2512 | if (fstat(fd, &st) != 0) | |
2513 | return; | |
2514 | sprintf(path, "/sys/dev/block/%d:%d", | |
2515 | major(st.st_rdev), minor(st.st_rdev)); | |
2516 | ||
2517 | rv = readlink(path, dname, sizeof(dname)); | |
2518 | if (rv <= 0) | |
2519 | return; | |
2520 | ||
2521 | dname[rv] = '\0'; | |
2522 | nm = strrchr(dname, '/'); | |
2523 | nm++; | |
2524 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
2525 | } | |
2526 | ||
cdddbdbc DW |
2527 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
2528 | ||
2529 | static int imsm_read_serial(int fd, char *devname, | |
2530 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
2531 | { | |
2532 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
2533 | int rv; |
2534 | int rsp_len; | |
1f24f035 | 2535 | int len; |
316e2bf4 DW |
2536 | char *dest; |
2537 | char *src; | |
2538 | char *rsp_buf; | |
2539 | int i; | |
cdddbdbc DW |
2540 | |
2541 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 2542 | |
f9ba0ff1 DW |
2543 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
2544 | ||
40ebbb9c | 2545 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
2546 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
2547 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
2548 | return 0; |
2549 | } | |
2550 | ||
cdddbdbc DW |
2551 | if (rv != 0) { |
2552 | if (devname) | |
2553 | fprintf(stderr, | |
2554 | Name ": Failed to retrieve serial for %s\n", | |
2555 | devname); | |
2556 | return rv; | |
2557 | } | |
2558 | ||
2559 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
2560 | if (!rsp_len) { |
2561 | if (devname) | |
2562 | fprintf(stderr, | |
2563 | Name ": Failed to retrieve serial for %s\n", | |
2564 | devname); | |
2565 | return 2; | |
2566 | } | |
1f24f035 | 2567 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 2568 | |
316e2bf4 DW |
2569 | /* trim all whitespace and non-printable characters and convert |
2570 | * ':' to ';' | |
2571 | */ | |
2572 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
2573 | src = &rsp_buf[i]; | |
2574 | if (*src > 0x20) { | |
2575 | /* ':' is reserved for use in placeholder serial | |
2576 | * numbers for missing disks | |
2577 | */ | |
2578 | if (*src == ':') | |
2579 | *dest++ = ';'; | |
2580 | else | |
2581 | *dest++ = *src; | |
2582 | } | |
2583 | } | |
2584 | len = dest - rsp_buf; | |
2585 | dest = rsp_buf; | |
2586 | ||
2587 | /* truncate leading characters */ | |
2588 | if (len > MAX_RAID_SERIAL_LEN) { | |
2589 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 2590 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 2591 | } |
5c3db629 | 2592 | |
5c3db629 | 2593 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 2594 | memcpy(serial, dest, len); |
cdddbdbc DW |
2595 | |
2596 | return 0; | |
2597 | } | |
2598 | ||
1f24f035 DW |
2599 | static int serialcmp(__u8 *s1, __u8 *s2) |
2600 | { | |
2601 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
2602 | } | |
2603 | ||
2604 | static void serialcpy(__u8 *dest, __u8 *src) | |
2605 | { | |
2606 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
2607 | } | |
2608 | ||
1799c9e8 | 2609 | #ifndef MDASSEMBLE |
54c2c1ea DW |
2610 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
2611 | { | |
2612 | struct dl *dl; | |
2613 | ||
2614 | for (dl = super->disks; dl; dl = dl->next) | |
2615 | if (serialcmp(dl->serial, serial) == 0) | |
2616 | break; | |
2617 | ||
2618 | return dl; | |
2619 | } | |
1799c9e8 | 2620 | #endif |
54c2c1ea | 2621 | |
a2b97981 DW |
2622 | static struct imsm_disk * |
2623 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
2624 | { | |
2625 | int i; | |
2626 | ||
2627 | for (i = 0; i < mpb->num_disks; i++) { | |
2628 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
2629 | ||
2630 | if (serialcmp(disk->serial, serial) == 0) { | |
2631 | if (idx) | |
2632 | *idx = i; | |
2633 | return disk; | |
2634 | } | |
2635 | } | |
2636 | ||
2637 | return NULL; | |
2638 | } | |
2639 | ||
cdddbdbc DW |
2640 | static int |
2641 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
2642 | { | |
a2b97981 | 2643 | struct imsm_disk *disk; |
cdddbdbc DW |
2644 | struct dl *dl; |
2645 | struct stat stb; | |
cdddbdbc | 2646 | int rv; |
a2b97981 | 2647 | char name[40]; |
d23fe947 DW |
2648 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
2649 | ||
2650 | rv = imsm_read_serial(fd, devname, serial); | |
2651 | ||
2652 | if (rv != 0) | |
2653 | return 2; | |
2654 | ||
a2b97981 | 2655 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 2656 | if (!dl) { |
cdddbdbc DW |
2657 | if (devname) |
2658 | fprintf(stderr, | |
2659 | Name ": failed to allocate disk buffer for %s\n", | |
2660 | devname); | |
2661 | return 2; | |
2662 | } | |
cdddbdbc | 2663 | |
a2b97981 DW |
2664 | fstat(fd, &stb); |
2665 | dl->major = major(stb.st_rdev); | |
2666 | dl->minor = minor(stb.st_rdev); | |
2667 | dl->next = super->disks; | |
2668 | dl->fd = keep_fd ? fd : -1; | |
2669 | assert(super->disks == NULL); | |
2670 | super->disks = dl; | |
2671 | serialcpy(dl->serial, serial); | |
2672 | dl->index = -2; | |
2673 | dl->e = NULL; | |
2674 | fd2devname(fd, name); | |
2675 | if (devname) | |
2676 | dl->devname = strdup(devname); | |
2677 | else | |
2678 | dl->devname = strdup(name); | |
cdddbdbc | 2679 | |
d23fe947 | 2680 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
2681 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
2682 | if (disk) { | |
2683 | dl->disk = *disk; | |
2684 | /* only set index on disks that are a member of a | |
2685 | * populated contianer, i.e. one with raid_devs | |
2686 | */ | |
2687 | if (is_failed(&dl->disk)) | |
3f6efecc | 2688 | dl->index = -2; |
a2b97981 DW |
2689 | else if (is_spare(&dl->disk)) |
2690 | dl->index = -1; | |
3f6efecc DW |
2691 | } |
2692 | ||
949c47a0 DW |
2693 | return 0; |
2694 | } | |
2695 | ||
0e600426 | 2696 | #ifndef MDASSEMBLE |
0c046afd DW |
2697 | /* When migrating map0 contains the 'destination' state while map1 |
2698 | * contains the current state. When not migrating map0 contains the | |
2699 | * current state. This routine assumes that map[0].map_state is set to | |
2700 | * the current array state before being called. | |
2701 | * | |
2702 | * Migration is indicated by one of the following states | |
2703 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 2704 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 2705 | * map1state=unitialized) |
1484e727 | 2706 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 2707 | * map1state=normal) |
e3bba0e0 | 2708 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 2709 | * map1state=degraded) |
8e59f3d8 AK |
2710 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
2711 | * map1state=normal) | |
0c046afd | 2712 | */ |
8e59f3d8 AK |
2713 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
2714 | __u8 to_state, int migr_type) | |
3393c6af | 2715 | { |
0c046afd | 2716 | struct imsm_map *dest; |
3393c6af DW |
2717 | struct imsm_map *src = get_imsm_map(dev, 0); |
2718 | ||
0c046afd | 2719 | dev->vol.migr_state = 1; |
1484e727 | 2720 | set_migr_type(dev, migr_type); |
f8f603f1 | 2721 | dev->vol.curr_migr_unit = 0; |
0c046afd DW |
2722 | dest = get_imsm_map(dev, 1); |
2723 | ||
0556e1a2 | 2724 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 2725 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
2726 | if ((migr_type == MIGR_REBUILD) || |
2727 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
2728 | __u32 ord; |
2729 | int i; | |
2730 | ||
2731 | for (i = 0; i < src->num_members; i++) { | |
2732 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
2733 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
2734 | } | |
2735 | } | |
2736 | ||
8e59f3d8 AK |
2737 | if (migr_type == MIGR_GEN_MIGR) |
2738 | /* Clear migration record */ | |
2739 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2740 | ||
0c046afd | 2741 | src->map_state = to_state; |
949c47a0 | 2742 | } |
f8f603f1 DW |
2743 | |
2744 | static void end_migration(struct imsm_dev *dev, __u8 map_state) | |
2745 | { | |
2746 | struct imsm_map *map = get_imsm_map(dev, 0); | |
0556e1a2 | 2747 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
28bce06f | 2748 | int i, j; |
0556e1a2 DW |
2749 | |
2750 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
2751 | * completed in the last migration. | |
2752 | * | |
28bce06f | 2753 | * FIXME add support for raid-level-migration |
0556e1a2 DW |
2754 | */ |
2755 | for (i = 0; i < prev->num_members; i++) | |
28bce06f AK |
2756 | for (j = 0; j < map->num_members; j++) |
2757 | /* during online capacity expansion | |
2758 | * disks position can be changed if takeover is used | |
2759 | */ | |
2760 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
2761 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
2762 | map->disk_ord_tbl[j] |= prev->disk_ord_tbl[i]; | |
2763 | break; | |
2764 | } | |
f8f603f1 DW |
2765 | |
2766 | dev->vol.migr_state = 0; | |
28bce06f | 2767 | dev->vol.migr_type = 0; |
f8f603f1 DW |
2768 | dev->vol.curr_migr_unit = 0; |
2769 | map->map_state = map_state; | |
2770 | } | |
0e600426 | 2771 | #endif |
949c47a0 DW |
2772 | |
2773 | static int parse_raid_devices(struct intel_super *super) | |
2774 | { | |
2775 | int i; | |
2776 | struct imsm_dev *dev_new; | |
4d7b1503 | 2777 | size_t len, len_migr; |
401d313b | 2778 | size_t max_len = 0; |
4d7b1503 DW |
2779 | size_t space_needed = 0; |
2780 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
2781 | |
2782 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
2783 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 2784 | struct intel_dev *dv; |
949c47a0 | 2785 | |
4d7b1503 DW |
2786 | len = sizeof_imsm_dev(dev_iter, 0); |
2787 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
2788 | if (len_migr > len) | |
2789 | space_needed += len_migr - len; | |
2790 | ||
ba2de7ba DW |
2791 | dv = malloc(sizeof(*dv)); |
2792 | if (!dv) | |
2793 | return 1; | |
401d313b AK |
2794 | if (max_len < len_migr) |
2795 | max_len = len_migr; | |
2796 | if (max_len > len_migr) | |
2797 | space_needed += max_len - len_migr; | |
2798 | dev_new = malloc(max_len); | |
ba2de7ba DW |
2799 | if (!dev_new) { |
2800 | free(dv); | |
949c47a0 | 2801 | return 1; |
ba2de7ba | 2802 | } |
949c47a0 | 2803 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
2804 | dv->dev = dev_new; |
2805 | dv->index = i; | |
2806 | dv->next = super->devlist; | |
2807 | super->devlist = dv; | |
949c47a0 | 2808 | } |
cdddbdbc | 2809 | |
4d7b1503 DW |
2810 | /* ensure that super->buf is large enough when all raid devices |
2811 | * are migrating | |
2812 | */ | |
2813 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
2814 | void *buf; | |
2815 | ||
2816 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
2817 | if (posix_memalign(&buf, 512, len) != 0) | |
2818 | return 1; | |
2819 | ||
1f45a8ad DW |
2820 | memcpy(buf, super->buf, super->len); |
2821 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
2822 | free(super->buf); |
2823 | super->buf = buf; | |
2824 | super->len = len; | |
2825 | } | |
2826 | ||
cdddbdbc DW |
2827 | return 0; |
2828 | } | |
2829 | ||
604b746f JD |
2830 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
2831 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
2832 | { | |
2833 | void *ptr = NULL; | |
2834 | ||
2835 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
2836 | ptr = mpb; | |
2837 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
2838 | } | |
2839 | ||
2840 | return ptr; | |
2841 | } | |
2842 | ||
e2f41b2c AK |
2843 | /******************************************************************************* |
2844 | * Function: check_mpb_migr_compatibility | |
2845 | * Description: Function checks for unsupported migration features: | |
2846 | * - migration optimization area (pba_of_lba0) | |
2847 | * - descending reshape (ascending_migr) | |
2848 | * Parameters: | |
2849 | * super : imsm metadata information | |
2850 | * Returns: | |
2851 | * 0 : migration is compatible | |
2852 | * -1 : migration is not compatible | |
2853 | ******************************************************************************/ | |
2854 | int check_mpb_migr_compatibility(struct intel_super *super) | |
2855 | { | |
2856 | struct imsm_map *map0, *map1; | |
2857 | struct migr_record *migr_rec = super->migr_rec; | |
2858 | int i; | |
2859 | ||
2860 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
2861 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
2862 | ||
2863 | if (dev_iter && | |
2864 | dev_iter->vol.migr_state == 1 && | |
2865 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
2866 | /* This device is migrating */ | |
2867 | map0 = get_imsm_map(dev_iter, 0); | |
2868 | map1 = get_imsm_map(dev_iter, 1); | |
2869 | if (map0->pba_of_lba0 != map1->pba_of_lba0) | |
2870 | /* migration optimization area was used */ | |
2871 | return -1; | |
2872 | if (migr_rec->ascending_migr == 0 | |
2873 | && migr_rec->dest_depth_per_unit > 0) | |
2874 | /* descending reshape not supported yet */ | |
2875 | return -1; | |
2876 | } | |
2877 | } | |
2878 | return 0; | |
2879 | } | |
2880 | ||
d23fe947 | 2881 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 2882 | |
cdddbdbc | 2883 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 2884 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
2885 | */ |
2886 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
2887 | { | |
2888 | unsigned long long dsize; | |
cdddbdbc DW |
2889 | unsigned long long sectors; |
2890 | struct stat; | |
6416d527 | 2891 | struct imsm_super *anchor; |
cdddbdbc DW |
2892 | __u32 check_sum; |
2893 | ||
cdddbdbc | 2894 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
2895 | if (dsize < 1024) { |
2896 | if (devname) | |
2897 | fprintf(stderr, | |
2898 | Name ": %s: device to small for imsm\n", | |
2899 | devname); | |
2900 | return 1; | |
2901 | } | |
cdddbdbc DW |
2902 | |
2903 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
2904 | if (devname) | |
2905 | fprintf(stderr, | |
2906 | Name ": Cannot seek to anchor block on %s: %s\n", | |
2907 | devname, strerror(errno)); | |
2908 | return 1; | |
2909 | } | |
2910 | ||
949c47a0 | 2911 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
2912 | if (devname) |
2913 | fprintf(stderr, | |
2914 | Name ": Failed to allocate imsm anchor buffer" | |
2915 | " on %s\n", devname); | |
2916 | return 1; | |
2917 | } | |
949c47a0 | 2918 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
2919 | if (devname) |
2920 | fprintf(stderr, | |
2921 | Name ": Cannot read anchor block on %s: %s\n", | |
2922 | devname, strerror(errno)); | |
6416d527 | 2923 | free(anchor); |
cdddbdbc DW |
2924 | return 1; |
2925 | } | |
2926 | ||
6416d527 | 2927 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
2928 | if (devname) |
2929 | fprintf(stderr, | |
2930 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 2931 | free(anchor); |
cdddbdbc DW |
2932 | return 2; |
2933 | } | |
2934 | ||
d23fe947 | 2935 | __free_imsm(super, 0); |
f2f5c343 LM |
2936 | /* reload capability and hba */ |
2937 | ||
2938 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 2939 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
2940 | super->len = ROUND_UP(anchor->mpb_size, 512); |
2941 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
2942 | if (devname) |
2943 | fprintf(stderr, | |
2944 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 2945 | super->len); |
6416d527 | 2946 | free(anchor); |
cdddbdbc DW |
2947 | return 2; |
2948 | } | |
949c47a0 | 2949 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 2950 | |
6416d527 NB |
2951 | sectors = mpb_sectors(anchor) - 1; |
2952 | free(anchor); | |
8e59f3d8 AK |
2953 | |
2954 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { | |
2955 | fprintf(stderr, Name | |
2956 | ": %s could not allocate migr_rec buffer\n", __func__); | |
2957 | free(super->buf); | |
2958 | return 2; | |
2959 | } | |
2960 | ||
949c47a0 | 2961 | if (!sectors) { |
ecf45690 DW |
2962 | check_sum = __gen_imsm_checksum(super->anchor); |
2963 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
2964 | if (devname) | |
2965 | fprintf(stderr, | |
2966 | Name ": IMSM checksum %x != %x on %s\n", | |
2967 | check_sum, | |
2968 | __le32_to_cpu(super->anchor->check_sum), | |
2969 | devname); | |
2970 | return 2; | |
2971 | } | |
2972 | ||
a2b97981 | 2973 | return 0; |
949c47a0 | 2974 | } |
cdddbdbc DW |
2975 | |
2976 | /* read the extended mpb */ | |
2977 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
2978 | if (devname) | |
2979 | fprintf(stderr, | |
2980 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
2981 | devname, strerror(errno)); | |
2982 | return 1; | |
2983 | } | |
2984 | ||
f21e18ca | 2985 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
2986 | if (devname) |
2987 | fprintf(stderr, | |
2988 | Name ": Cannot read extended mpb on %s: %s\n", | |
2989 | devname, strerror(errno)); | |
2990 | return 2; | |
2991 | } | |
2992 | ||
949c47a0 DW |
2993 | check_sum = __gen_imsm_checksum(super->anchor); |
2994 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
2995 | if (devname) |
2996 | fprintf(stderr, | |
2997 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 2998 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 2999 | devname); |
db575f3b | 3000 | return 3; |
cdddbdbc DW |
3001 | } |
3002 | ||
604b746f JD |
3003 | /* FIXME the BBM log is disk specific so we cannot use this global |
3004 | * buffer for all disks. Ok for now since we only look at the global | |
3005 | * bbm_log_size parameter to gate assembly | |
3006 | */ | |
3007 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3008 | ||
a2b97981 DW |
3009 | return 0; |
3010 | } | |
3011 | ||
8e59f3d8 AK |
3012 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3013 | ||
a2b97981 DW |
3014 | static int |
3015 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3016 | { | |
3017 | int err; | |
3018 | ||
3019 | err = load_imsm_mpb(fd, super, devname); | |
3020 | if (err) | |
3021 | return err; | |
3022 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3023 | if (err) | |
3024 | return err; | |
3025 | err = parse_raid_devices(super); | |
4d7b1503 | 3026 | |
a2b97981 | 3027 | return err; |
cdddbdbc DW |
3028 | } |
3029 | ||
ae6aad82 DW |
3030 | static void __free_imsm_disk(struct dl *d) |
3031 | { | |
3032 | if (d->fd >= 0) | |
3033 | close(d->fd); | |
3034 | if (d->devname) | |
3035 | free(d->devname); | |
0dcecb2e DW |
3036 | if (d->e) |
3037 | free(d->e); | |
ae6aad82 DW |
3038 | free(d); |
3039 | ||
3040 | } | |
1a64be56 | 3041 | |
cdddbdbc DW |
3042 | static void free_imsm_disks(struct intel_super *super) |
3043 | { | |
47ee5a45 | 3044 | struct dl *d; |
cdddbdbc | 3045 | |
47ee5a45 DW |
3046 | while (super->disks) { |
3047 | d = super->disks; | |
cdddbdbc | 3048 | super->disks = d->next; |
ae6aad82 | 3049 | __free_imsm_disk(d); |
cdddbdbc | 3050 | } |
cb82edca AK |
3051 | while (super->disk_mgmt_list) { |
3052 | d = super->disk_mgmt_list; | |
3053 | super->disk_mgmt_list = d->next; | |
3054 | __free_imsm_disk(d); | |
3055 | } | |
47ee5a45 DW |
3056 | while (super->missing) { |
3057 | d = super->missing; | |
3058 | super->missing = d->next; | |
3059 | __free_imsm_disk(d); | |
3060 | } | |
3061 | ||
cdddbdbc DW |
3062 | } |
3063 | ||
9ca2c81c | 3064 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3065 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3066 | { |
88654014 LM |
3067 | struct intel_hba *elem, *next; |
3068 | ||
9ca2c81c | 3069 | if (super->buf) { |
949c47a0 | 3070 | free(super->buf); |
9ca2c81c DW |
3071 | super->buf = NULL; |
3072 | } | |
f2f5c343 LM |
3073 | /* unlink capability description */ |
3074 | super->orom = NULL; | |
8e59f3d8 AK |
3075 | if (super->migr_rec_buf) { |
3076 | free(super->migr_rec_buf); | |
3077 | super->migr_rec_buf = NULL; | |
3078 | } | |
d23fe947 DW |
3079 | if (free_disks) |
3080 | free_imsm_disks(super); | |
ba2de7ba | 3081 | free_devlist(super); |
88654014 LM |
3082 | elem = super->hba; |
3083 | while (elem) { | |
3084 | if (elem->path) | |
3085 | free((void *)elem->path); | |
3086 | next = elem->next; | |
3087 | free(elem); | |
3088 | elem = next; | |
88c32bb1 | 3089 | } |
88654014 | 3090 | super->hba = NULL; |
cdddbdbc DW |
3091 | } |
3092 | ||
9ca2c81c DW |
3093 | static void free_imsm(struct intel_super *super) |
3094 | { | |
d23fe947 | 3095 | __free_imsm(super, 1); |
9ca2c81c DW |
3096 | free(super); |
3097 | } | |
cdddbdbc DW |
3098 | |
3099 | static void free_super_imsm(struct supertype *st) | |
3100 | { | |
3101 | struct intel_super *super = st->sb; | |
3102 | ||
3103 | if (!super) | |
3104 | return; | |
3105 | ||
3106 | free_imsm(super); | |
3107 | st->sb = NULL; | |
3108 | } | |
3109 | ||
49133e57 | 3110 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
3111 | { |
3112 | struct intel_super *super = malloc(sizeof(*super)); | |
3113 | ||
3114 | if (super) { | |
3115 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 3116 | super->current_vol = -1; |
0dcecb2e | 3117 | super->create_offset = ~((__u32 ) 0); |
c2c087e6 | 3118 | } |
c2c087e6 DW |
3119 | return super; |
3120 | } | |
3121 | ||
f0f5a016 LM |
3122 | /* |
3123 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3124 | */ | |
d424212e | 3125 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3126 | { |
3127 | struct sys_dev *hba_name; | |
3128 | int rv = 0; | |
3129 | ||
3130 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3131 | super->orom = NULL; |
f0f5a016 LM |
3132 | super->hba = NULL; |
3133 | return 0; | |
3134 | } | |
3135 | hba_name = find_disk_attached_hba(fd, NULL); | |
3136 | if (!hba_name) { | |
d424212e | 3137 | if (devname) |
f0f5a016 LM |
3138 | fprintf(stderr, |
3139 | Name ": %s is not attached to Intel(R) RAID controller.\n", | |
d424212e | 3140 | devname); |
f0f5a016 LM |
3141 | return 1; |
3142 | } | |
3143 | rv = attach_hba_to_super(super, hba_name); | |
3144 | if (rv == 2) { | |
d424212e N |
3145 | if (devname) { |
3146 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3147 | |
f0f5a016 LM |
3148 | fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID " |
3149 | "controller (%s),\n" | |
3150 | " but the container is assigned to Intel(R) " | |
3151 | "%s RAID controller (", | |
d424212e | 3152 | devname, |
f0f5a016 LM |
3153 | hba_name->path, |
3154 | hba_name->pci_id ? : "Err!", | |
3155 | get_sys_dev_type(hba_name->type)); | |
3156 | ||
f0f5a016 LM |
3157 | while (hba) { |
3158 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3159 | if (hba->next) | |
3160 | fprintf(stderr, ", "); | |
3161 | hba = hba->next; | |
3162 | } | |
3163 | ||
3164 | fprintf(stderr, ").\n" | |
3165 | " Mixing devices attached to different controllers " | |
3166 | "is not allowed.\n"); | |
3167 | } | |
3168 | free_sys_dev(&hba_name); | |
3169 | return 2; | |
3170 | } | |
f2f5c343 | 3171 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3172 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3173 | if (!super->orom) |
3174 | return 3; | |
f0f5a016 LM |
3175 | return 0; |
3176 | } | |
3177 | ||
cdddbdbc | 3178 | #ifndef MDASSEMBLE |
47ee5a45 DW |
3179 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3180 | * disks that have disappeared from the system. This routine relies on | |
3181 | * the mpb being uptodate, which it is at load time. | |
3182 | */ | |
3183 | static int find_missing(struct intel_super *super) | |
3184 | { | |
3185 | int i; | |
3186 | struct imsm_super *mpb = super->anchor; | |
3187 | struct dl *dl; | |
3188 | struct imsm_disk *disk; | |
47ee5a45 DW |
3189 | |
3190 | for (i = 0; i < mpb->num_disks; i++) { | |
3191 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3192 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3193 | if (dl) |
3194 | continue; | |
47ee5a45 DW |
3195 | |
3196 | dl = malloc(sizeof(*dl)); | |
3197 | if (!dl) | |
3198 | return 1; | |
3199 | dl->major = 0; | |
3200 | dl->minor = 0; | |
3201 | dl->fd = -1; | |
3202 | dl->devname = strdup("missing"); | |
3203 | dl->index = i; | |
3204 | serialcpy(dl->serial, disk->serial); | |
3205 | dl->disk = *disk; | |
689c9bf3 | 3206 | dl->e = NULL; |
47ee5a45 DW |
3207 | dl->next = super->missing; |
3208 | super->missing = dl; | |
3209 | } | |
3210 | ||
3211 | return 0; | |
3212 | } | |
3213 | ||
a2b97981 DW |
3214 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3215 | { | |
3216 | struct intel_disk *idisk = disk_list; | |
3217 | ||
3218 | while (idisk) { | |
3219 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3220 | break; | |
3221 | idisk = idisk->next; | |
3222 | } | |
3223 | ||
3224 | return idisk; | |
3225 | } | |
3226 | ||
3227 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3228 | struct intel_super *super, | |
3229 | struct intel_disk **disk_list) | |
3230 | { | |
3231 | struct imsm_disk *d = &super->disks->disk; | |
3232 | struct imsm_super *mpb = super->anchor; | |
3233 | int i, j; | |
3234 | ||
3235 | for (i = 0; i < tbl_size; i++) { | |
3236 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3237 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3238 | ||
3239 | if (tbl_mpb->family_num == mpb->family_num) { | |
3240 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3241 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3242 | __func__, super->disks->major, | |
3243 | super->disks->minor, | |
3244 | table[i]->disks->major, | |
3245 | table[i]->disks->minor); | |
3246 | break; | |
3247 | } | |
3248 | ||
3249 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3250 | is_configured(d) == is_configured(tbl_d)) && | |
3251 | tbl_mpb->generation_num < mpb->generation_num) { | |
3252 | /* current version of the mpb is a | |
3253 | * better candidate than the one in | |
3254 | * super_table, but copy over "cross | |
3255 | * generational" status | |
3256 | */ | |
3257 | struct intel_disk *idisk; | |
3258 | ||
3259 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3260 | __func__, super->disks->major, | |
3261 | super->disks->minor, | |
3262 | table[i]->disks->major, | |
3263 | table[i]->disks->minor); | |
3264 | ||
3265 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3266 | if (idisk && is_failed(&idisk->disk)) | |
3267 | tbl_d->status |= FAILED_DISK; | |
3268 | break; | |
3269 | } else { | |
3270 | struct intel_disk *idisk; | |
3271 | struct imsm_disk *disk; | |
3272 | ||
3273 | /* tbl_mpb is more up to date, but copy | |
3274 | * over cross generational status before | |
3275 | * returning | |
3276 | */ | |
3277 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3278 | if (disk && is_failed(disk)) | |
3279 | d->status |= FAILED_DISK; | |
3280 | ||
3281 | idisk = disk_list_get(d->serial, *disk_list); | |
3282 | if (idisk) { | |
3283 | idisk->owner = i; | |
3284 | if (disk && is_configured(disk)) | |
3285 | idisk->disk.status |= CONFIGURED_DISK; | |
3286 | } | |
3287 | ||
3288 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3289 | __func__, super->disks->major, | |
3290 | super->disks->minor, | |
3291 | table[i]->disks->major, | |
3292 | table[i]->disks->minor); | |
3293 | ||
3294 | return tbl_size; | |
3295 | } | |
3296 | } | |
3297 | } | |
3298 | ||
3299 | if (i >= tbl_size) | |
3300 | table[tbl_size++] = super; | |
3301 | else | |
3302 | table[i] = super; | |
3303 | ||
3304 | /* update/extend the merged list of imsm_disk records */ | |
3305 | for (j = 0; j < mpb->num_disks; j++) { | |
3306 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3307 | struct intel_disk *idisk; | |
3308 | ||
3309 | idisk = disk_list_get(disk->serial, *disk_list); | |
3310 | if (idisk) { | |
3311 | idisk->disk.status |= disk->status; | |
3312 | if (is_configured(&idisk->disk) || | |
3313 | is_failed(&idisk->disk)) | |
3314 | idisk->disk.status &= ~(SPARE_DISK); | |
3315 | } else { | |
3316 | idisk = calloc(1, sizeof(*idisk)); | |
3317 | if (!idisk) | |
3318 | return -1; | |
3319 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
3320 | idisk->disk = *disk; | |
3321 | idisk->next = *disk_list; | |
3322 | *disk_list = idisk; | |
3323 | } | |
3324 | ||
3325 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3326 | idisk->owner = i; | |
3327 | } | |
3328 | ||
3329 | return tbl_size; | |
3330 | } | |
3331 | ||
3332 | static struct intel_super * | |
3333 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3334 | const int owner) | |
3335 | { | |
3336 | struct imsm_super *mpb = super->anchor; | |
3337 | int ok_count = 0; | |
3338 | int i; | |
3339 | ||
3340 | for (i = 0; i < mpb->num_disks; i++) { | |
3341 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3342 | struct intel_disk *idisk; | |
3343 | ||
3344 | idisk = disk_list_get(disk->serial, disk_list); | |
3345 | if (idisk) { | |
3346 | if (idisk->owner == owner || | |
3347 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3348 | ok_count++; | |
3349 | else | |
3350 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3351 | __func__, disk->serial, idisk->owner, | |
3352 | owner); | |
3353 | } else { | |
3354 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
3355 | __func__, __le32_to_cpu(mpb->family_num), i, | |
3356 | disk->serial); | |
3357 | break; | |
3358 | } | |
3359 | } | |
3360 | ||
3361 | if (ok_count == mpb->num_disks) | |
3362 | return super; | |
3363 | return NULL; | |
3364 | } | |
3365 | ||
3366 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
3367 | { | |
3368 | struct intel_super *s; | |
3369 | ||
3370 | for (s = super_list; s; s = s->next) { | |
3371 | if (family_num != s->anchor->family_num) | |
3372 | continue; | |
3373 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
3374 | __le32_to_cpu(family_num), s->disks->devname); | |
3375 | } | |
3376 | } | |
3377 | ||
3378 | static struct intel_super * | |
3379 | imsm_thunderdome(struct intel_super **super_list, int len) | |
3380 | { | |
3381 | struct intel_super *super_table[len]; | |
3382 | struct intel_disk *disk_list = NULL; | |
3383 | struct intel_super *champion, *spare; | |
3384 | struct intel_super *s, **del; | |
3385 | int tbl_size = 0; | |
3386 | int conflict; | |
3387 | int i; | |
3388 | ||
3389 | memset(super_table, 0, sizeof(super_table)); | |
3390 | for (s = *super_list; s; s = s->next) | |
3391 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
3392 | ||
3393 | for (i = 0; i < tbl_size; i++) { | |
3394 | struct imsm_disk *d; | |
3395 | struct intel_disk *idisk; | |
3396 | struct imsm_super *mpb = super_table[i]->anchor; | |
3397 | ||
3398 | s = super_table[i]; | |
3399 | d = &s->disks->disk; | |
3400 | ||
3401 | /* 'd' must appear in merged disk list for its | |
3402 | * configuration to be valid | |
3403 | */ | |
3404 | idisk = disk_list_get(d->serial, disk_list); | |
3405 | if (idisk && idisk->owner == i) | |
3406 | s = validate_members(s, disk_list, i); | |
3407 | else | |
3408 | s = NULL; | |
3409 | ||
3410 | if (!s) | |
3411 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
3412 | __func__, mpb->family_num, | |
3413 | super_table[i]->disks->major, | |
3414 | super_table[i]->disks->minor); | |
3415 | super_table[i] = s; | |
3416 | } | |
3417 | ||
3418 | /* This is where the mdadm implementation differs from the Windows | |
3419 | * driver which has no strict concept of a container. We can only | |
3420 | * assemble one family from a container, so when returning a prodigal | |
3421 | * array member to this system the code will not be able to disambiguate | |
3422 | * the container contents that should be assembled ("foreign" versus | |
3423 | * "local"). It requires user intervention to set the orig_family_num | |
3424 | * to a new value to establish a new container. The Windows driver in | |
3425 | * this situation fixes up the volume name in place and manages the | |
3426 | * foreign array as an independent entity. | |
3427 | */ | |
3428 | s = NULL; | |
3429 | spare = NULL; | |
3430 | conflict = 0; | |
3431 | for (i = 0; i < tbl_size; i++) { | |
3432 | struct intel_super *tbl_ent = super_table[i]; | |
3433 | int is_spare = 0; | |
3434 | ||
3435 | if (!tbl_ent) | |
3436 | continue; | |
3437 | ||
3438 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
3439 | spare = tbl_ent; | |
3440 | is_spare = 1; | |
3441 | } | |
3442 | ||
3443 | if (s && !is_spare) { | |
3444 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
3445 | conflict++; | |
3446 | } else if (!s && !is_spare) | |
3447 | s = tbl_ent; | |
3448 | } | |
3449 | ||
3450 | if (!s) | |
3451 | s = spare; | |
3452 | if (!s) { | |
3453 | champion = NULL; | |
3454 | goto out; | |
3455 | } | |
3456 | champion = s; | |
3457 | ||
3458 | if (conflict) | |
3459 | fprintf(stderr, "Chose family %#x on '%s', " | |
3460 | "assemble conflicts to new container with '--update=uuid'\n", | |
3461 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
3462 | ||
3463 | /* collect all dl's onto 'champion', and update them to | |
3464 | * champion's version of the status | |
3465 | */ | |
3466 | for (s = *super_list; s; s = s->next) { | |
3467 | struct imsm_super *mpb = champion->anchor; | |
3468 | struct dl *dl = s->disks; | |
3469 | ||
3470 | if (s == champion) | |
3471 | continue; | |
3472 | ||
3473 | for (i = 0; i < mpb->num_disks; i++) { | |
3474 | struct imsm_disk *disk; | |
3475 | ||
3476 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
3477 | if (disk) { | |
3478 | dl->disk = *disk; | |
3479 | /* only set index on disks that are a member of | |
3480 | * a populated contianer, i.e. one with | |
3481 | * raid_devs | |
3482 | */ | |
3483 | if (is_failed(&dl->disk)) | |
3484 | dl->index = -2; | |
3485 | else if (is_spare(&dl->disk)) | |
3486 | dl->index = -1; | |
3487 | break; | |
3488 | } | |
3489 | } | |
3490 | ||
3491 | if (i >= mpb->num_disks) { | |
3492 | struct intel_disk *idisk; | |
3493 | ||
3494 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 3495 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
3496 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
3497 | dl->index = -1; | |
3498 | else { | |
3499 | dl->index = -2; | |
3500 | continue; | |
3501 | } | |
3502 | } | |
3503 | ||
3504 | dl->next = champion->disks; | |
3505 | champion->disks = dl; | |
3506 | s->disks = NULL; | |
3507 | } | |
3508 | ||
3509 | /* delete 'champion' from super_list */ | |
3510 | for (del = super_list; *del; ) { | |
3511 | if (*del == champion) { | |
3512 | *del = (*del)->next; | |
3513 | break; | |
3514 | } else | |
3515 | del = &(*del)->next; | |
3516 | } | |
3517 | champion->next = NULL; | |
3518 | ||
3519 | out: | |
3520 | while (disk_list) { | |
3521 | struct intel_disk *idisk = disk_list; | |
3522 | ||
3523 | disk_list = disk_list->next; | |
3524 | free(idisk); | |
3525 | } | |
3526 | ||
3527 | return champion; | |
3528 | } | |
3529 | ||
cdddbdbc | 3530 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
e1902a7b | 3531 | char *devname) |
cdddbdbc DW |
3532 | { |
3533 | struct mdinfo *sra; | |
a2b97981 DW |
3534 | struct intel_super *super_list = NULL; |
3535 | struct intel_super *super = NULL; | |
db575f3b | 3536 | int devnum = fd2devnum(fd); |
a2b97981 | 3537 | struct mdinfo *sd; |
db575f3b | 3538 | int retry; |
a2b97981 DW |
3539 | int err = 0; |
3540 | int i; | |
dab4a513 DW |
3541 | |
3542 | /* check if 'fd' an opened container */ | |
b526e52d | 3543 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
cdddbdbc DW |
3544 | if (!sra) |
3545 | return 1; | |
3546 | ||
3547 | if (sra->array.major_version != -1 || | |
3548 | sra->array.minor_version != -2 || | |
1602d52c AW |
3549 | strcmp(sra->text_version, "imsm") != 0) { |
3550 | err = 1; | |
3551 | goto error; | |
3552 | } | |
a2b97981 DW |
3553 | /* load all mpbs */ |
3554 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
49133e57 | 3555 | struct intel_super *s = alloc_super(); |
7a6ecd55 | 3556 | char nm[32]; |
a2b97981 | 3557 | int dfd; |
f2f5c343 | 3558 | int rv; |
a2b97981 DW |
3559 | |
3560 | err = 1; | |
3561 | if (!s) | |
3562 | goto error; | |
3563 | s->next = super_list; | |
3564 | super_list = s; | |
cdddbdbc | 3565 | |
a2b97981 | 3566 | err = 2; |
cdddbdbc | 3567 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3568 | dfd = dev_open(nm, O_RDWR); |
a2b97981 DW |
3569 | if (dfd < 0) |
3570 | goto error; | |
3571 | ||
d424212e | 3572 | rv = find_intel_hba_capability(dfd, s, devname); |
f2f5c343 LM |
3573 | /* no orom/efi or non-intel hba of the disk */ |
3574 | if (rv != 0) | |
3575 | goto error; | |
3576 | ||
e1902a7b | 3577 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
db575f3b DW |
3578 | |
3579 | /* retry the load if we might have raced against mdmon */ | |
a2b97981 | 3580 | if (err == 3 && mdmon_running(devnum)) |
db575f3b DW |
3581 | for (retry = 0; retry < 3; retry++) { |
3582 | usleep(3000); | |
e1902a7b | 3583 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
a2b97981 | 3584 | if (err != 3) |
db575f3b DW |
3585 | break; |
3586 | } | |
a2b97981 DW |
3587 | if (err) |
3588 | goto error; | |
cdddbdbc DW |
3589 | } |
3590 | ||
a2b97981 DW |
3591 | /* all mpbs enter, maybe one leaves */ |
3592 | super = imsm_thunderdome(&super_list, i); | |
3593 | if (!super) { | |
3594 | err = 1; | |
3595 | goto error; | |
cdddbdbc DW |
3596 | } |
3597 | ||
47ee5a45 DW |
3598 | if (find_missing(super) != 0) { |
3599 | free_imsm(super); | |
a2b97981 DW |
3600 | err = 2; |
3601 | goto error; | |
47ee5a45 | 3602 | } |
8e59f3d8 AK |
3603 | |
3604 | /* load migration record */ | |
3605 | err = load_imsm_migr_rec(super, NULL); | |
3606 | if (err) { | |
3607 | err = 4; | |
3608 | goto error; | |
3609 | } | |
e2f41b2c AK |
3610 | |
3611 | /* Check migration compatibility */ | |
3612 | if (check_mpb_migr_compatibility(super) != 0) { | |
3613 | fprintf(stderr, Name ": Unsupported migration detected"); | |
3614 | if (devname) | |
3615 | fprintf(stderr, " on %s\n", devname); | |
3616 | else | |
3617 | fprintf(stderr, " (IMSM).\n"); | |
3618 | ||
3619 | err = 5; | |
3620 | goto error; | |
3621 | } | |
3622 | ||
a2b97981 DW |
3623 | err = 0; |
3624 | ||
3625 | error: | |
3626 | while (super_list) { | |
3627 | struct intel_super *s = super_list; | |
3628 | ||
3629 | super_list = super_list->next; | |
3630 | free_imsm(s); | |
3631 | } | |
1602d52c | 3632 | sysfs_free(sra); |
a2b97981 DW |
3633 | |
3634 | if (err) | |
3635 | return err; | |
f7e7067b | 3636 | |
cdddbdbc | 3637 | *sbp = super; |
db575f3b | 3638 | st->container_dev = devnum; |
a2b97981 | 3639 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 3640 | st->ss = &super_imsm; |
cdddbdbc DW |
3641 | st->minor_version = 0; |
3642 | st->max_devs = IMSM_MAX_DEVICES; | |
3643 | } | |
cdddbdbc DW |
3644 | return 0; |
3645 | } | |
2b959fbf N |
3646 | |
3647 | static int load_container_imsm(struct supertype *st, int fd, char *devname) | |
3648 | { | |
3649 | return load_super_imsm_all(st, fd, &st->sb, devname); | |
3650 | } | |
cdddbdbc DW |
3651 | #endif |
3652 | ||
3653 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
3654 | { | |
3655 | struct intel_super *super; | |
3656 | int rv; | |
3657 | ||
691c6ee1 N |
3658 | if (test_partition(fd)) |
3659 | /* IMSM not allowed on partitions */ | |
3660 | return 1; | |
3661 | ||
37424f13 DW |
3662 | free_super_imsm(st); |
3663 | ||
49133e57 | 3664 | super = alloc_super(); |
cdddbdbc DW |
3665 | if (!super) { |
3666 | fprintf(stderr, | |
3667 | Name ": malloc of %zu failed.\n", | |
3668 | sizeof(*super)); | |
3669 | return 1; | |
3670 | } | |
ea2bc72b LM |
3671 | /* Load hba and capabilities if they exist. |
3672 | * But do not preclude loading metadata in case capabilities or hba are | |
3673 | * non-compliant and ignore_hw_compat is set. | |
3674 | */ | |
d424212e | 3675 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 3676 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 3677 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 LM |
3678 | if (devname) |
3679 | fprintf(stderr, | |
3680 | Name ": No OROM/EFI properties for %s\n", devname); | |
3681 | free_imsm(super); | |
3682 | return 2; | |
3683 | } | |
a2b97981 | 3684 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
3685 | |
3686 | if (rv) { | |
3687 | if (devname) | |
3688 | fprintf(stderr, | |
3689 | Name ": Failed to load all information " | |
3690 | "sections on %s\n", devname); | |
3691 | free_imsm(super); | |
3692 | return rv; | |
3693 | } | |
3694 | ||
3695 | st->sb = super; | |
3696 | if (st->ss == NULL) { | |
3697 | st->ss = &super_imsm; | |
3698 | st->minor_version = 0; | |
3699 | st->max_devs = IMSM_MAX_DEVICES; | |
3700 | } | |
8e59f3d8 AK |
3701 | |
3702 | /* load migration record */ | |
3703 | load_imsm_migr_rec(super, NULL); | |
3704 | ||
e2f41b2c AK |
3705 | /* Check for unsupported migration features */ |
3706 | if (check_mpb_migr_compatibility(super) != 0) { | |
3707 | fprintf(stderr, Name ": Unsupported migration detected"); | |
3708 | if (devname) | |
3709 | fprintf(stderr, " on %s\n", devname); | |
3710 | else | |
3711 | fprintf(stderr, " (IMSM).\n"); | |
3712 | return 3; | |
3713 | } | |
3714 | ||
cdddbdbc DW |
3715 | return 0; |
3716 | } | |
3717 | ||
ef6ffade DW |
3718 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
3719 | { | |
3720 | if (info->level == 1) | |
3721 | return 128; | |
3722 | return info->chunk_size >> 9; | |
3723 | } | |
3724 | ||
ff596308 | 3725 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains) |
ef6ffade DW |
3726 | { |
3727 | __u32 num_stripes; | |
3728 | ||
3729 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
ff596308 | 3730 | num_stripes /= num_domains; |
ef6ffade DW |
3731 | |
3732 | return num_stripes; | |
3733 | } | |
3734 | ||
fcfd9599 DW |
3735 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
3736 | { | |
4025c288 DW |
3737 | if (info->level == 1) |
3738 | return info->size * 2; | |
3739 | else | |
3740 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
fcfd9599 DW |
3741 | } |
3742 | ||
4d1313e9 DW |
3743 | static void imsm_update_version_info(struct intel_super *super) |
3744 | { | |
3745 | /* update the version and attributes */ | |
3746 | struct imsm_super *mpb = super->anchor; | |
3747 | char *version; | |
3748 | struct imsm_dev *dev; | |
3749 | struct imsm_map *map; | |
3750 | int i; | |
3751 | ||
3752 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
3753 | dev = get_imsm_dev(super, i); | |
3754 | map = get_imsm_map(dev, 0); | |
3755 | if (__le32_to_cpu(dev->size_high) > 0) | |
3756 | mpb->attributes |= MPB_ATTRIB_2TB; | |
3757 | ||
3758 | /* FIXME detect when an array spans a port multiplier */ | |
3759 | #if 0 | |
3760 | mpb->attributes |= MPB_ATTRIB_PM; | |
3761 | #endif | |
3762 | ||
3763 | if (mpb->num_raid_devs > 1 || | |
3764 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
3765 | version = MPB_VERSION_ATTRIBS; | |
3766 | switch (get_imsm_raid_level(map)) { | |
3767 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
3768 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
3769 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
3770 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
3771 | } | |
3772 | } else { | |
3773 | if (map->num_members >= 5) | |
3774 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
3775 | else if (dev->status == DEV_CLONE_N_GO) | |
3776 | version = MPB_VERSION_CNG; | |
3777 | else if (get_imsm_raid_level(map) == 5) | |
3778 | version = MPB_VERSION_RAID5; | |
3779 | else if (map->num_members >= 3) | |
3780 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
3781 | else if (get_imsm_raid_level(map) == 1) | |
3782 | version = MPB_VERSION_RAID1; | |
3783 | else | |
3784 | version = MPB_VERSION_RAID0; | |
3785 | } | |
3786 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
3787 | } | |
3788 | } | |
3789 | ||
aa534678 DW |
3790 | static int check_name(struct intel_super *super, char *name, int quiet) |
3791 | { | |
3792 | struct imsm_super *mpb = super->anchor; | |
3793 | char *reason = NULL; | |
3794 | int i; | |
3795 | ||
3796 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
3797 | reason = "must be 16 characters or less"; | |
3798 | ||
3799 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
3800 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3801 | ||
3802 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
3803 | reason = "already exists"; | |
3804 | break; | |
3805 | } | |
3806 | } | |
3807 | ||
3808 | if (reason && !quiet) | |
3809 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
3810 | ||
3811 | return !reason; | |
3812 | } | |
3813 | ||
8b353278 DW |
3814 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
3815 | unsigned long long size, char *name, | |
3816 | char *homehost, int *uuid) | |
cdddbdbc | 3817 | { |
c2c087e6 DW |
3818 | /* We are creating a volume inside a pre-existing container. |
3819 | * so st->sb is already set. | |
3820 | */ | |
3821 | struct intel_super *super = st->sb; | |
949c47a0 | 3822 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 3823 | struct intel_dev *dv; |
c2c087e6 DW |
3824 | struct imsm_dev *dev; |
3825 | struct imsm_vol *vol; | |
3826 | struct imsm_map *map; | |
3827 | int idx = mpb->num_raid_devs; | |
3828 | int i; | |
3829 | unsigned long long array_blocks; | |
2c092cad | 3830 | size_t size_old, size_new; |
ff596308 | 3831 | __u32 num_data_stripes; |
cdddbdbc | 3832 | |
88c32bb1 | 3833 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 3834 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 3835 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
3836 | return 0; |
3837 | } | |
3838 | ||
2c092cad DW |
3839 | /* ensure the mpb is large enough for the new data */ |
3840 | size_old = __le32_to_cpu(mpb->mpb_size); | |
3841 | size_new = disks_to_mpb_size(info->nr_disks); | |
3842 | if (size_new > size_old) { | |
3843 | void *mpb_new; | |
3844 | size_t size_round = ROUND_UP(size_new, 512); | |
3845 | ||
3846 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
3847 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
3848 | return 0; | |
3849 | } | |
8e59f3d8 AK |
3850 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
3851 | fprintf(stderr, Name | |
3852 | ": %s could not allocate migr_rec buffer\n", | |
3853 | __func__); | |
3854 | free(super->buf); | |
3855 | free(super); | |
3856 | return 0; | |
3857 | } | |
2c092cad DW |
3858 | memcpy(mpb_new, mpb, size_old); |
3859 | free(mpb); | |
3860 | mpb = mpb_new; | |
949c47a0 | 3861 | super->anchor = mpb_new; |
2c092cad DW |
3862 | mpb->mpb_size = __cpu_to_le32(size_new); |
3863 | memset(mpb_new + size_old, 0, size_round - size_old); | |
3864 | } | |
bf5a934a | 3865 | super->current_vol = idx; |
d23fe947 DW |
3866 | /* when creating the first raid device in this container set num_disks |
3867 | * to zero, i.e. delete this spare and add raid member devices in | |
3868 | * add_to_super_imsm_volume() | |
3869 | */ | |
3870 | if (super->current_vol == 0) | |
3871 | mpb->num_disks = 0; | |
5a038140 | 3872 | |
aa534678 DW |
3873 | if (!check_name(super, name, 0)) |
3874 | return 0; | |
ba2de7ba DW |
3875 | dv = malloc(sizeof(*dv)); |
3876 | if (!dv) { | |
3877 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
3878 | return 0; | |
3879 | } | |
1a2487c2 | 3880 | dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
949c47a0 | 3881 | if (!dev) { |
ba2de7ba | 3882 | free(dv); |
949c47a0 DW |
3883 | fprintf(stderr, Name": could not allocate raid device\n"); |
3884 | return 0; | |
3885 | } | |
1a2487c2 | 3886 | |
c2c087e6 | 3887 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
03bcbc65 DW |
3888 | if (info->level == 1) |
3889 | array_blocks = info_to_blocks_per_member(info); | |
3890 | else | |
3891 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
3892 | info->layout, info->chunk_size, | |
3893 | info->size*2); | |
979d38be DW |
3894 | /* round array size down to closest MB */ |
3895 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
3896 | ||
c2c087e6 DW |
3897 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
3898 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 3899 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
3900 | vol = &dev->vol; |
3901 | vol->migr_state = 0; | |
1484e727 | 3902 | set_migr_type(dev, MIGR_INIT); |
c2c087e6 | 3903 | vol->dirty = 0; |
f8f603f1 | 3904 | vol->curr_migr_unit = 0; |
a965f303 | 3905 | map = get_imsm_map(dev, 0); |
0dcecb2e | 3906 | map->pba_of_lba0 = __cpu_to_le32(super->create_offset); |
fcfd9599 | 3907 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade | 3908 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 3909 | map->failed_disk_num = ~0; |
c2c087e6 DW |
3910 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
3911 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 3912 | map->ddf = 1; |
ef6ffade DW |
3913 | |
3914 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
3915 | free(dev); |
3916 | free(dv); | |
ef6ffade DW |
3917 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
3918 | "in a raid1 volume\n"); | |
3919 | return 0; | |
3920 | } | |
81062a36 DW |
3921 | |
3922 | map->raid_level = info->level; | |
4d1313e9 | 3923 | if (info->level == 10) { |
c2c087e6 | 3924 | map->raid_level = 1; |
4d1313e9 | 3925 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
3926 | } else if (info->level == 1) |
3927 | map->num_domains = info->raid_disks; | |
3928 | else | |
ff596308 | 3929 | map->num_domains = 1; |
81062a36 | 3930 | |
ff596308 DW |
3931 | num_data_stripes = info_to_num_data_stripes(info, map->num_domains); |
3932 | map->num_data_stripes = __cpu_to_le32(num_data_stripes); | |
ef6ffade | 3933 | |
c2c087e6 DW |
3934 | map->num_members = info->raid_disks; |
3935 | for (i = 0; i < map->num_members; i++) { | |
3936 | /* initialized in add_to_super */ | |
4eb26970 | 3937 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 3938 | } |
949c47a0 | 3939 | mpb->num_raid_devs++; |
ba2de7ba DW |
3940 | |
3941 | dv->dev = dev; | |
3942 | dv->index = super->current_vol; | |
3943 | dv->next = super->devlist; | |
3944 | super->devlist = dv; | |
c2c087e6 | 3945 | |
4d1313e9 DW |
3946 | imsm_update_version_info(super); |
3947 | ||
c2c087e6 | 3948 | return 1; |
cdddbdbc DW |
3949 | } |
3950 | ||
bf5a934a DW |
3951 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
3952 | unsigned long long size, char *name, | |
3953 | char *homehost, int *uuid) | |
3954 | { | |
3955 | /* This is primarily called by Create when creating a new array. | |
3956 | * We will then get add_to_super called for each component, and then | |
3957 | * write_init_super called to write it out to each device. | |
3958 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
3959 | * array. | |
3960 | * To create on a pre-existing array a different method will be called. | |
3961 | * This one is just for fresh drives. | |
3962 | */ | |
3963 | struct intel_super *super; | |
3964 | struct imsm_super *mpb; | |
3965 | size_t mpb_size; | |
4d1313e9 | 3966 | char *version; |
bf5a934a | 3967 | |
bf5a934a | 3968 | if (st->sb) |
e683ca88 DW |
3969 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
3970 | ||
3971 | if (info) | |
3972 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
3973 | else | |
3974 | mpb_size = 512; | |
bf5a934a | 3975 | |
49133e57 | 3976 | super = alloc_super(); |
e683ca88 | 3977 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 3978 | free(super); |
e683ca88 DW |
3979 | super = NULL; |
3980 | } | |
3981 | if (!super) { | |
3982 | fprintf(stderr, Name | |
3983 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
3984 | return 0; |
3985 | } | |
8e59f3d8 AK |
3986 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
3987 | fprintf(stderr, Name | |
3988 | ": %s could not allocate migr_rec buffer\n", __func__); | |
3989 | free(super->buf); | |
3990 | free(super); | |
3991 | return 0; | |
3992 | } | |
e683ca88 | 3993 | memset(super->buf, 0, mpb_size); |
ef649044 | 3994 | mpb = super->buf; |
e683ca88 DW |
3995 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
3996 | st->sb = super; | |
3997 | ||
3998 | if (info == NULL) { | |
3999 | /* zeroing superblock */ | |
4000 | return 0; | |
4001 | } | |
bf5a934a | 4002 | |
4d1313e9 DW |
4003 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4004 | ||
4005 | version = (char *) mpb->sig; | |
4006 | strcpy(version, MPB_SIGNATURE); | |
4007 | version += strlen(MPB_SIGNATURE); | |
4008 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4009 | |
bf5a934a DW |
4010 | return 1; |
4011 | } | |
4012 | ||
0e600426 | 4013 | #ifndef MDASSEMBLE |
f20c3968 | 4014 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4015 | int fd, char *devname) |
4016 | { | |
4017 | struct intel_super *super = st->sb; | |
d23fe947 | 4018 | struct imsm_super *mpb = super->anchor; |
bf5a934a DW |
4019 | struct dl *dl; |
4020 | struct imsm_dev *dev; | |
4021 | struct imsm_map *map; | |
4eb26970 | 4022 | int slot; |
bf5a934a | 4023 | |
949c47a0 | 4024 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 4025 | map = get_imsm_map(dev, 0); |
bf5a934a | 4026 | |
208933a7 N |
4027 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
4028 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
4029 | devname); | |
4030 | return 1; | |
4031 | } | |
4032 | ||
efb30e7f DW |
4033 | if (fd == -1) { |
4034 | /* we're doing autolayout so grab the pre-marked (in | |
4035 | * validate_geometry) raid_disk | |
4036 | */ | |
4037 | for (dl = super->disks; dl; dl = dl->next) | |
4038 | if (dl->raiddisk == dk->raid_disk) | |
4039 | break; | |
4040 | } else { | |
4041 | for (dl = super->disks; dl ; dl = dl->next) | |
4042 | if (dl->major == dk->major && | |
4043 | dl->minor == dk->minor) | |
4044 | break; | |
4045 | } | |
d23fe947 | 4046 | |
208933a7 N |
4047 | if (!dl) { |
4048 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 4049 | return 1; |
208933a7 | 4050 | } |
bf5a934a | 4051 | |
d23fe947 DW |
4052 | /* add a pristine spare to the metadata */ |
4053 | if (dl->index < 0) { | |
4054 | dl->index = super->anchor->num_disks; | |
4055 | super->anchor->num_disks++; | |
4056 | } | |
4eb26970 DW |
4057 | /* Check the device has not already been added */ |
4058 | slot = get_imsm_disk_slot(map, dl->index); | |
4059 | if (slot >= 0 && | |
98130f40 | 4060 | (get_imsm_ord_tbl_ent(dev, slot, -1) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
4061 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
4062 | devname); | |
4063 | return 1; | |
4064 | } | |
be73972f | 4065 | set_imsm_ord_tbl_ent(map, dk->number, dl->index); |
ee5aad5a | 4066 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 DW |
4067 | |
4068 | /* if we are creating the first raid device update the family number */ | |
4069 | if (super->current_vol == 0) { | |
4070 | __u32 sum; | |
4071 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
4072 | struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index); | |
4073 | ||
791b666a AW |
4074 | if (!_dev || !_disk) { |
4075 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
4076 | return 1; | |
4077 | } | |
d23fe947 DW |
4078 | *_dev = *dev; |
4079 | *_disk = dl->disk; | |
148acb7b DW |
4080 | sum = random32(); |
4081 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4082 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4083 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4084 | } |
f20c3968 DW |
4085 | |
4086 | return 0; | |
bf5a934a DW |
4087 | } |
4088 | ||
88654014 | 4089 | |
f20c3968 | 4090 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4091 | int fd, char *devname) |
cdddbdbc | 4092 | { |
c2c087e6 | 4093 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4094 | struct dl *dd; |
4095 | unsigned long long size; | |
f2f27e63 | 4096 | __u32 id; |
c2c087e6 DW |
4097 | int rv; |
4098 | struct stat stb; | |
4099 | ||
88654014 LM |
4100 | /* If we are on an RAID enabled platform check that the disk is |
4101 | * attached to the raid controller. | |
4102 | * We do not need to test disks attachment for container based additions, | |
4103 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4104 | */ |
d424212e | 4105 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4106 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4107 | if (rv != 0) { |
4108 | dprintf("capability: %p fd: %d ret: %d\n", | |
4109 | super->orom, fd, rv); | |
4110 | return 1; | |
88c32bb1 DW |
4111 | } |
4112 | ||
f20c3968 DW |
4113 | if (super->current_vol >= 0) |
4114 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4115 | |
c2c087e6 DW |
4116 | fstat(fd, &stb); |
4117 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 4118 | if (!dd) { |
c2c087e6 DW |
4119 | fprintf(stderr, |
4120 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 4121 | return 1; |
c2c087e6 DW |
4122 | } |
4123 | memset(dd, 0, sizeof(*dd)); | |
4124 | dd->major = major(stb.st_rdev); | |
4125 | dd->minor = minor(stb.st_rdev); | |
b9f594fe | 4126 | dd->index = -1; |
c2c087e6 | 4127 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 4128 | dd->fd = fd; |
689c9bf3 | 4129 | dd->e = NULL; |
1a64be56 | 4130 | dd->action = DISK_ADD; |
c2c087e6 | 4131 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4132 | if (rv) { |
c2c087e6 | 4133 | fprintf(stderr, |
0030e8d6 | 4134 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4135 | free(dd); |
0030e8d6 | 4136 | abort(); |
c2c087e6 DW |
4137 | } |
4138 | ||
c2c087e6 DW |
4139 | get_dev_size(fd, NULL, &size); |
4140 | size /= 512; | |
1f24f035 | 4141 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 4142 | dd->disk.total_blocks = __cpu_to_le32(size); |
ee5aad5a | 4143 | dd->disk.status = SPARE_DISK; |
c2c087e6 | 4144 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4145 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4146 | else |
b9f594fe | 4147 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4148 | |
4149 | if (st->update_tail) { | |
1a64be56 LM |
4150 | dd->next = super->disk_mgmt_list; |
4151 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4152 | } else { |
4153 | dd->next = super->disks; | |
4154 | super->disks = dd; | |
ceaf0ee1 | 4155 | super->updates_pending++; |
43dad3d6 | 4156 | } |
f20c3968 DW |
4157 | |
4158 | return 0; | |
cdddbdbc DW |
4159 | } |
4160 | ||
1a64be56 LM |
4161 | |
4162 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4163 | { | |
4164 | struct intel_super *super = st->sb; | |
4165 | struct dl *dd; | |
4166 | ||
4167 | /* remove from super works only in mdmon - for communication | |
4168 | * manager - monitor. Check if communication memory buffer | |
4169 | * is prepared. | |
4170 | */ | |
4171 | if (!st->update_tail) { | |
4172 | fprintf(stderr, | |
4173 | Name ": %s shall be used in mdmon context only" | |
4174 | "(line %d).\n", __func__, __LINE__); | |
4175 | return 1; | |
4176 | } | |
4177 | dd = malloc(sizeof(*dd)); | |
4178 | if (!dd) { | |
4179 | fprintf(stderr, | |
4180 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
4181 | return 1; | |
4182 | } | |
4183 | memset(dd, 0, sizeof(*dd)); | |
4184 | dd->major = dk->major; | |
4185 | dd->minor = dk->minor; | |
4186 | dd->index = -1; | |
4187 | dd->fd = -1; | |
4188 | dd->disk.status = SPARE_DISK; | |
4189 | dd->action = DISK_REMOVE; | |
4190 | ||
4191 | dd->next = super->disk_mgmt_list; | |
4192 | super->disk_mgmt_list = dd; | |
4193 | ||
4194 | ||
4195 | return 0; | |
4196 | } | |
4197 | ||
f796af5d DW |
4198 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
4199 | ||
4200 | static union { | |
4201 | char buf[512]; | |
4202 | struct imsm_super anchor; | |
4203 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 4204 | |
d23fe947 DW |
4205 | /* spare records have their own family number and do not have any defined raid |
4206 | * devices | |
4207 | */ | |
4208 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
4209 | { | |
d23fe947 | 4210 | struct imsm_super *mpb = super->anchor; |
f796af5d | 4211 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
4212 | __u32 sum; |
4213 | struct dl *d; | |
4214 | ||
f796af5d DW |
4215 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
4216 | spare->generation_num = __cpu_to_le32(1UL), | |
4217 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
4218 | spare->num_disks = 1, | |
4219 | spare->num_raid_devs = 0, | |
4220 | spare->cache_size = mpb->cache_size, | |
4221 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
4222 | ||
4223 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
4224 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
4225 | |
4226 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 4227 | if (d->index != -1) |
d23fe947 DW |
4228 | continue; |
4229 | ||
f796af5d DW |
4230 | spare->disk[0] = d->disk; |
4231 | sum = __gen_imsm_checksum(spare); | |
4232 | spare->family_num = __cpu_to_le32(sum); | |
4233 | spare->orig_family_num = 0; | |
4234 | sum = __gen_imsm_checksum(spare); | |
4235 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 4236 | |
f796af5d | 4237 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
4238 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4239 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 4240 | return 1; |
d23fe947 DW |
4241 | } |
4242 | if (doclose) { | |
4243 | close(d->fd); | |
4244 | d->fd = -1; | |
4245 | } | |
4246 | } | |
4247 | ||
e74255d9 | 4248 | return 0; |
d23fe947 DW |
4249 | } |
4250 | ||
146c6260 AK |
4251 | static int is_gen_migration(struct imsm_dev *dev); |
4252 | ||
36988a3d | 4253 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 4254 | { |
36988a3d | 4255 | struct intel_super *super = st->sb; |
949c47a0 | 4256 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4257 | struct dl *d; |
4258 | __u32 generation; | |
4259 | __u32 sum; | |
d23fe947 | 4260 | int spares = 0; |
949c47a0 | 4261 | int i; |
a48ac0a8 | 4262 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 4263 | int num_disks = 0; |
146c6260 | 4264 | int clear_migration_record = 1; |
cdddbdbc | 4265 | |
c2c087e6 DW |
4266 | /* 'generation' is incremented everytime the metadata is written */ |
4267 | generation = __le32_to_cpu(mpb->generation_num); | |
4268 | generation++; | |
4269 | mpb->generation_num = __cpu_to_le32(generation); | |
4270 | ||
148acb7b DW |
4271 | /* fix up cases where previous mdadm releases failed to set |
4272 | * orig_family_num | |
4273 | */ | |
4274 | if (mpb->orig_family_num == 0) | |
4275 | mpb->orig_family_num = mpb->family_num; | |
4276 | ||
d23fe947 | 4277 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 4278 | if (d->index == -1) |
d23fe947 | 4279 | spares++; |
36988a3d | 4280 | else { |
d23fe947 | 4281 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4282 | num_disks++; |
4283 | } | |
d23fe947 | 4284 | } |
36988a3d | 4285 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 4286 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4287 | num_disks++; |
4288 | } | |
4289 | mpb->num_disks = num_disks; | |
4290 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 4291 | |
949c47a0 DW |
4292 | for (i = 0; i < mpb->num_raid_devs; i++) { |
4293 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
4294 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
4295 | if (dev && dev2) { | |
4296 | imsm_copy_dev(dev, dev2); | |
4297 | mpb_size += sizeof_imsm_dev(dev, 0); | |
4298 | } | |
146c6260 AK |
4299 | if (is_gen_migration(dev2)) |
4300 | clear_migration_record = 0; | |
949c47a0 | 4301 | } |
a48ac0a8 DW |
4302 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
4303 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 4304 | |
c2c087e6 | 4305 | /* recalculate checksum */ |
949c47a0 | 4306 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
4307 | mpb->check_sum = __cpu_to_le32(sum); |
4308 | ||
146c6260 AK |
4309 | if (clear_migration_record) |
4310 | memset(super->migr_rec_buf, 0, 512); | |
4311 | ||
d23fe947 | 4312 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 4313 | for (d = super->disks; d ; d = d->next) { |
d23fe947 DW |
4314 | if (d->index < 0) |
4315 | continue; | |
f796af5d | 4316 | if (store_imsm_mpb(d->fd, mpb)) |
c2c087e6 DW |
4317 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4318 | __func__, d->major, d->minor, strerror(errno)); | |
146c6260 AK |
4319 | if (clear_migration_record) { |
4320 | unsigned long long dsize; | |
4321 | ||
4322 | get_dev_size(d->fd, NULL, &dsize); | |
4323 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
4324 | write(d->fd, super->migr_rec_buf, 512); | |
4325 | } | |
4326 | } | |
c2c087e6 DW |
4327 | if (doclose) { |
4328 | close(d->fd); | |
4329 | d->fd = -1; | |
4330 | } | |
4331 | } | |
4332 | ||
d23fe947 DW |
4333 | if (spares) |
4334 | return write_super_imsm_spares(super, doclose); | |
4335 | ||
e74255d9 | 4336 | return 0; |
c2c087e6 DW |
4337 | } |
4338 | ||
0e600426 | 4339 | |
9b1fb677 | 4340 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
4341 | { |
4342 | size_t len; | |
4343 | struct imsm_update_create_array *u; | |
4344 | struct intel_super *super = st->sb; | |
9b1fb677 | 4345 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
54c2c1ea DW |
4346 | struct imsm_map *map = get_imsm_map(dev, 0); |
4347 | struct disk_info *inf; | |
4348 | struct imsm_disk *disk; | |
4349 | int i; | |
43dad3d6 | 4350 | |
54c2c1ea DW |
4351 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
4352 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
4353 | u = malloc(len); |
4354 | if (!u) { | |
4355 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4356 | __func__); | |
4357 | return 1; | |
4358 | } | |
4359 | ||
4360 | u->type = update_create_array; | |
9b1fb677 | 4361 | u->dev_idx = dev_idx; |
43dad3d6 | 4362 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
4363 | inf = get_disk_info(u); |
4364 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 4365 | int idx = get_imsm_disk_idx(dev, i, -1); |
9b1fb677 | 4366 | |
54c2c1ea DW |
4367 | disk = get_imsm_disk(super, idx); |
4368 | serialcpy(inf[i].serial, disk->serial); | |
4369 | } | |
43dad3d6 DW |
4370 | append_metadata_update(st, u, len); |
4371 | ||
4372 | return 0; | |
4373 | } | |
4374 | ||
1a64be56 | 4375 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
4376 | { |
4377 | struct intel_super *super = st->sb; | |
4378 | size_t len; | |
1a64be56 | 4379 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 4380 | |
1a64be56 | 4381 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
4382 | return 0; |
4383 | ||
4384 | len = sizeof(*u); | |
4385 | u = malloc(len); | |
4386 | if (!u) { | |
4387 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4388 | __func__); | |
4389 | return 1; | |
4390 | } | |
4391 | ||
1a64be56 | 4392 | u->type = update_add_remove_disk; |
43dad3d6 DW |
4393 | append_metadata_update(st, u, len); |
4394 | ||
4395 | return 0; | |
4396 | } | |
4397 | ||
c2c087e6 DW |
4398 | static int write_init_super_imsm(struct supertype *st) |
4399 | { | |
9b1fb677 DW |
4400 | struct intel_super *super = st->sb; |
4401 | int current_vol = super->current_vol; | |
4402 | ||
4403 | /* we are done with current_vol reset it to point st at the container */ | |
4404 | super->current_vol = -1; | |
4405 | ||
8273f55e | 4406 | if (st->update_tail) { |
43dad3d6 DW |
4407 | /* queue the recently created array / added disk |
4408 | * as a metadata update */ | |
43dad3d6 | 4409 | int rv; |
8273f55e | 4410 | |
43dad3d6 | 4411 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 4412 | if (current_vol < 0) { |
1a64be56 LM |
4413 | /* in the mgmt (add/remove) disk case we are running |
4414 | * in mdmon context, so don't close fd's | |
43dad3d6 | 4415 | */ |
1a64be56 | 4416 | return mgmt_disk(st); |
43dad3d6 | 4417 | } else |
9b1fb677 | 4418 | rv = create_array(st, current_vol); |
8273f55e | 4419 | |
43dad3d6 | 4420 | return rv; |
d682f344 N |
4421 | } else { |
4422 | struct dl *d; | |
4423 | for (d = super->disks; d; d = d->next) | |
4424 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 4425 | return write_super_imsm(st, 1); |
d682f344 | 4426 | } |
cdddbdbc | 4427 | } |
0e600426 | 4428 | #endif |
cdddbdbc | 4429 | |
e683ca88 | 4430 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 4431 | { |
e683ca88 DW |
4432 | struct intel_super *super = st->sb; |
4433 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 4434 | |
e683ca88 | 4435 | if (!mpb) |
ad97895e DW |
4436 | return 1; |
4437 | ||
1799c9e8 | 4438 | #ifndef MDASSEMBLE |
e683ca88 | 4439 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
4440 | #else |
4441 | return 1; | |
4442 | #endif | |
cdddbdbc DW |
4443 | } |
4444 | ||
0e600426 N |
4445 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
4446 | { | |
4447 | return __le32_to_cpu(mpb->bbm_log_size); | |
4448 | } | |
4449 | ||
4450 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
4451 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
4452 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 4453 | unsigned long long size, char *dev, |
2c514b71 NB |
4454 | unsigned long long *freesize, |
4455 | int verbose) | |
cdddbdbc | 4456 | { |
c2c087e6 DW |
4457 | int fd; |
4458 | unsigned long long ldsize; | |
f2f5c343 LM |
4459 | struct intel_super *super=NULL; |
4460 | int rv = 0; | |
cdddbdbc | 4461 | |
c2c087e6 DW |
4462 | if (level != LEVEL_CONTAINER) |
4463 | return 0; | |
4464 | if (!dev) | |
4465 | return 1; | |
4466 | ||
4467 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4468 | if (fd < 0) { | |
2c514b71 NB |
4469 | if (verbose) |
4470 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
4471 | dev, strerror(errno)); | |
c2c087e6 DW |
4472 | return 0; |
4473 | } | |
4474 | if (!get_dev_size(fd, dev, &ldsize)) { | |
4475 | close(fd); | |
4476 | return 0; | |
4477 | } | |
f2f5c343 LM |
4478 | |
4479 | /* capabilities retrieve could be possible | |
4480 | * note that there is no fd for the disks in array. | |
4481 | */ | |
4482 | super = alloc_super(); | |
4483 | if (!super) { | |
4484 | fprintf(stderr, | |
4485 | Name ": malloc of %zu failed.\n", | |
4486 | sizeof(*super)); | |
4487 | close(fd); | |
4488 | return 0; | |
4489 | } | |
4490 | ||
d424212e | 4491 | rv = find_intel_hba_capability(fd, super, verbose ? dev : NULL); |
f2f5c343 LM |
4492 | if (rv != 0) { |
4493 | #if DEBUG | |
4494 | char str[256]; | |
4495 | fd2devname(fd, str); | |
4496 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
4497 | fd, str, super->orom, rv, raiddisks); | |
4498 | #endif | |
4499 | /* no orom/efi or non-intel hba of the disk */ | |
4500 | close(fd); | |
4501 | free_imsm(super); | |
4502 | return 0; | |
4503 | } | |
c2c087e6 | 4504 | close(fd); |
f2f5c343 LM |
4505 | if (super->orom && raiddisks > super->orom->tds) { |
4506 | if (verbose) | |
4507 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
4508 | " platform supported disks: %d\n", | |
4509 | raiddisks, super->orom->tds); | |
4510 | ||
4511 | free_imsm(super); | |
4512 | return 0; | |
4513 | } | |
c2c087e6 DW |
4514 | |
4515 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 4516 | free_imsm(super); |
c2c087e6 DW |
4517 | |
4518 | return 1; | |
cdddbdbc DW |
4519 | } |
4520 | ||
0dcecb2e DW |
4521 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
4522 | { | |
4523 | const unsigned long long base_start = e[*idx].start; | |
4524 | unsigned long long end = base_start + e[*idx].size; | |
4525 | int i; | |
4526 | ||
4527 | if (base_start == end) | |
4528 | return 0; | |
4529 | ||
4530 | *idx = *idx + 1; | |
4531 | for (i = *idx; i < num_extents; i++) { | |
4532 | /* extend overlapping extents */ | |
4533 | if (e[i].start >= base_start && | |
4534 | e[i].start <= end) { | |
4535 | if (e[i].size == 0) | |
4536 | return 0; | |
4537 | if (e[i].start + e[i].size > end) | |
4538 | end = e[i].start + e[i].size; | |
4539 | } else if (e[i].start > end) { | |
4540 | *idx = i; | |
4541 | break; | |
4542 | } | |
4543 | } | |
4544 | ||
4545 | return end - base_start; | |
4546 | } | |
4547 | ||
4548 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
4549 | { | |
4550 | /* build a composite disk with all known extents and generate a new | |
4551 | * 'maxsize' given the "all disks in an array must share a common start | |
4552 | * offset" constraint | |
4553 | */ | |
4554 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
4555 | struct dl *dl; | |
4556 | int i, j; | |
4557 | int start_extent; | |
4558 | unsigned long long pos; | |
b9d77223 | 4559 | unsigned long long start = 0; |
0dcecb2e DW |
4560 | unsigned long long maxsize; |
4561 | unsigned long reserve; | |
4562 | ||
4563 | if (!e) | |
a7dd165b | 4564 | return 0; |
0dcecb2e DW |
4565 | |
4566 | /* coalesce and sort all extents. also, check to see if we need to | |
4567 | * reserve space between member arrays | |
4568 | */ | |
4569 | j = 0; | |
4570 | for (dl = super->disks; dl; dl = dl->next) { | |
4571 | if (!dl->e) | |
4572 | continue; | |
4573 | for (i = 0; i < dl->extent_cnt; i++) | |
4574 | e[j++] = dl->e[i]; | |
4575 | } | |
4576 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
4577 | ||
4578 | /* merge extents */ | |
4579 | i = 0; | |
4580 | j = 0; | |
4581 | while (i < sum_extents) { | |
4582 | e[j].start = e[i].start; | |
4583 | e[j].size = find_size(e, &i, sum_extents); | |
4584 | j++; | |
4585 | if (e[j-1].size == 0) | |
4586 | break; | |
4587 | } | |
4588 | ||
4589 | pos = 0; | |
4590 | maxsize = 0; | |
4591 | start_extent = 0; | |
4592 | i = 0; | |
4593 | do { | |
4594 | unsigned long long esize; | |
4595 | ||
4596 | esize = e[i].start - pos; | |
4597 | if (esize >= maxsize) { | |
4598 | maxsize = esize; | |
4599 | start = pos; | |
4600 | start_extent = i; | |
4601 | } | |
4602 | pos = e[i].start + e[i].size; | |
4603 | i++; | |
4604 | } while (e[i-1].size); | |
4605 | free(e); | |
4606 | ||
a7dd165b DW |
4607 | if (maxsize == 0) |
4608 | return 0; | |
4609 | ||
4610 | /* FIXME assumes volume at offset 0 is the first volume in a | |
4611 | * container | |
4612 | */ | |
0dcecb2e DW |
4613 | if (start_extent > 0) |
4614 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
4615 | else | |
4616 | reserve = 0; | |
4617 | ||
4618 | if (maxsize < reserve) | |
a7dd165b | 4619 | return 0; |
0dcecb2e DW |
4620 | |
4621 | super->create_offset = ~((__u32) 0); | |
4622 | if (start + reserve > super->create_offset) | |
a7dd165b | 4623 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
4624 | super->create_offset = start + reserve; |
4625 | ||
4626 | return maxsize - reserve; | |
4627 | } | |
4628 | ||
88c32bb1 DW |
4629 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
4630 | { | |
4631 | if (level < 0 || level == 6 || level == 4) | |
4632 | return 0; | |
4633 | ||
4634 | /* if we have an orom prevent invalid raid levels */ | |
4635 | if (orom) | |
4636 | switch (level) { | |
4637 | case 0: return imsm_orom_has_raid0(orom); | |
4638 | case 1: | |
4639 | if (raiddisks > 2) | |
4640 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
4641 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
4642 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
4643 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
4644 | } |
4645 | else | |
4646 | return 1; /* not on an Intel RAID platform so anything goes */ | |
4647 | ||
4648 | return 0; | |
4649 | } | |
4650 | ||
73408129 | 4651 | |
35f81cbb | 4652 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) |
73408129 LM |
4653 | /* |
4654 | * validate volume parameters with OROM/EFI capabilities | |
4655 | */ | |
6592ce37 DW |
4656 | static int |
4657 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
c21e737b | 4658 | int raiddisks, int *chunk, int verbose) |
6592ce37 | 4659 | { |
73408129 LM |
4660 | #if DEBUG |
4661 | verbose = 1; | |
4662 | #endif | |
4663 | /* validate container capabilities */ | |
4664 | if (super->orom && raiddisks > super->orom->tds) { | |
4665 | if (verbose) | |
4666 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
4667 | " platform supported disks: %d\n", | |
4668 | raiddisks, super->orom->tds); | |
4669 | return 0; | |
4670 | } | |
4671 | ||
4672 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
4673 | if (super->orom && (!is_raid_level_supported(super->orom, level, | |
4674 | raiddisks))) { | |
6592ce37 DW |
4675 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
4676 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
4677 | return 0; | |
4678 | } | |
c21e737b CA |
4679 | if (super->orom && level != 1) { |
4680 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
4681 | *chunk = imsm_orom_default_chunk(super->orom); | |
4682 | else if (chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
4683 | pr_vrb(": platform does not support a chunk size of: " | |
4684 | "%d\n", *chunk); | |
4685 | return 0; | |
4686 | } | |
6592ce37 DW |
4687 | } |
4688 | if (layout != imsm_level_to_layout(level)) { | |
4689 | if (level == 5) | |
4690 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
4691 | else if (level == 10) | |
4692 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
4693 | else | |
4694 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
4695 | layout, level); | |
4696 | return 0; | |
4697 | } | |
6592ce37 DW |
4698 | return 1; |
4699 | } | |
4700 | ||
c2c087e6 DW |
4701 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
4702 | * FIX ME add ahci details | |
4703 | */ | |
8b353278 | 4704 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 4705 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 4706 | unsigned long long size, char *dev, |
2c514b71 NB |
4707 | unsigned long long *freesize, |
4708 | int verbose) | |
cdddbdbc | 4709 | { |
c2c087e6 DW |
4710 | struct stat stb; |
4711 | struct intel_super *super = st->sb; | |
a20d2ba5 | 4712 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4713 | struct dl *dl; |
4714 | unsigned long long pos = 0; | |
4715 | unsigned long long maxsize; | |
4716 | struct extent *e; | |
4717 | int i; | |
cdddbdbc | 4718 | |
88c32bb1 DW |
4719 | /* We must have the container info already read in. */ |
4720 | if (!super) | |
c2c087e6 DW |
4721 | return 0; |
4722 | ||
d54559f0 LM |
4723 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, verbose)) { |
4724 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
4725 | "Cannot proceed with the action(s).\n"); | |
c2c087e6 | 4726 | return 0; |
d54559f0 | 4727 | } |
c2c087e6 DW |
4728 | if (!dev) { |
4729 | /* General test: make sure there is space for | |
2da8544a DW |
4730 | * 'raiddisks' device extents of size 'size' at a given |
4731 | * offset | |
c2c087e6 | 4732 | */ |
e46273eb | 4733 | unsigned long long minsize = size; |
b7528a20 | 4734 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
4735 | int dcnt = 0; |
4736 | if (minsize == 0) | |
4737 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
4738 | for (dl = super->disks; dl ; dl = dl->next) { | |
4739 | int found = 0; | |
4740 | ||
bf5a934a | 4741 | pos = 0; |
c2c087e6 DW |
4742 | i = 0; |
4743 | e = get_extents(super, dl); | |
4744 | if (!e) continue; | |
4745 | do { | |
4746 | unsigned long long esize; | |
4747 | esize = e[i].start - pos; | |
4748 | if (esize >= minsize) | |
4749 | found = 1; | |
b7528a20 | 4750 | if (found && start_offset == MaxSector) { |
2da8544a DW |
4751 | start_offset = pos; |
4752 | break; | |
4753 | } else if (found && pos != start_offset) { | |
4754 | found = 0; | |
4755 | break; | |
4756 | } | |
c2c087e6 DW |
4757 | pos = e[i].start + e[i].size; |
4758 | i++; | |
4759 | } while (e[i-1].size); | |
4760 | if (found) | |
4761 | dcnt++; | |
4762 | free(e); | |
4763 | } | |
4764 | if (dcnt < raiddisks) { | |
2c514b71 NB |
4765 | if (verbose) |
4766 | fprintf(stderr, Name ": imsm: Not enough " | |
4767 | "devices with space for this array " | |
4768 | "(%d < %d)\n", | |
4769 | dcnt, raiddisks); | |
c2c087e6 DW |
4770 | return 0; |
4771 | } | |
4772 | return 1; | |
4773 | } | |
0dcecb2e | 4774 | |
c2c087e6 DW |
4775 | /* This device must be a member of the set */ |
4776 | if (stat(dev, &stb) < 0) | |
4777 | return 0; | |
4778 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
4779 | return 0; | |
4780 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
4781 | if (dl->major == (int)major(stb.st_rdev) && |
4782 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
4783 | break; |
4784 | } | |
4785 | if (!dl) { | |
2c514b71 NB |
4786 | if (verbose) |
4787 | fprintf(stderr, Name ": %s is not in the " | |
4788 | "same imsm set\n", dev); | |
c2c087e6 | 4789 | return 0; |
a20d2ba5 DW |
4790 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
4791 | /* If a volume is present then the current creation attempt | |
4792 | * cannot incorporate new spares because the orom may not | |
4793 | * understand this configuration (all member disks must be | |
4794 | * members of each array in the container). | |
4795 | */ | |
4796 | fprintf(stderr, Name ": %s is a spare and a volume" | |
4797 | " is already defined for this container\n", dev); | |
4798 | fprintf(stderr, Name ": The option-rom requires all member" | |
4799 | " disks to be a member of all volumes\n"); | |
4800 | return 0; | |
c2c087e6 | 4801 | } |
0dcecb2e DW |
4802 | |
4803 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
4804 | e = get_extents(super, dl); |
4805 | maxsize = 0; | |
4806 | i = 0; | |
0dcecb2e DW |
4807 | if (e) { |
4808 | do { | |
4809 | unsigned long long esize; | |
4810 | ||
4811 | esize = e[i].start - pos; | |
4812 | if (esize >= maxsize) | |
4813 | maxsize = esize; | |
4814 | pos = e[i].start + e[i].size; | |
4815 | i++; | |
4816 | } while (e[i-1].size); | |
4817 | dl->e = e; | |
4818 | dl->extent_cnt = i; | |
4819 | } else { | |
4820 | if (verbose) | |
4821 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
4822 | dev); | |
4823 | return 0; | |
4824 | } | |
4825 | if (maxsize < size) { | |
4826 | if (verbose) | |
4827 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
4828 | dev, maxsize, size); | |
4829 | return 0; | |
4830 | } | |
4831 | ||
4832 | /* count total number of extents for merge */ | |
4833 | i = 0; | |
4834 | for (dl = super->disks; dl; dl = dl->next) | |
4835 | if (dl->e) | |
4836 | i += dl->extent_cnt; | |
4837 | ||
4838 | maxsize = merge_extents(super, i); | |
a7dd165b | 4839 | if (maxsize < size || maxsize == 0) { |
0dcecb2e DW |
4840 | if (verbose) |
4841 | fprintf(stderr, Name ": not enough space after merge (%llu < %llu)\n", | |
4842 | maxsize, size); | |
4843 | return 0; | |
0dcecb2e DW |
4844 | } |
4845 | ||
c2c087e6 DW |
4846 | *freesize = maxsize; |
4847 | ||
4848 | return 1; | |
cdddbdbc DW |
4849 | } |
4850 | ||
efb30e7f DW |
4851 | static int reserve_space(struct supertype *st, int raiddisks, |
4852 | unsigned long long size, int chunk, | |
4853 | unsigned long long *freesize) | |
4854 | { | |
4855 | struct intel_super *super = st->sb; | |
4856 | struct imsm_super *mpb = super->anchor; | |
4857 | struct dl *dl; | |
4858 | int i; | |
4859 | int extent_cnt; | |
4860 | struct extent *e; | |
4861 | unsigned long long maxsize; | |
4862 | unsigned long long minsize; | |
4863 | int cnt; | |
4864 | int used; | |
4865 | ||
4866 | /* find the largest common start free region of the possible disks */ | |
4867 | used = 0; | |
4868 | extent_cnt = 0; | |
4869 | cnt = 0; | |
4870 | for (dl = super->disks; dl; dl = dl->next) { | |
4871 | dl->raiddisk = -1; | |
4872 | ||
4873 | if (dl->index >= 0) | |
4874 | used++; | |
4875 | ||
4876 | /* don't activate new spares if we are orom constrained | |
4877 | * and there is already a volume active in the container | |
4878 | */ | |
4879 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
4880 | continue; | |
4881 | ||
4882 | e = get_extents(super, dl); | |
4883 | if (!e) | |
4884 | continue; | |
4885 | for (i = 1; e[i-1].size; i++) | |
4886 | ; | |
4887 | dl->e = e; | |
4888 | dl->extent_cnt = i; | |
4889 | extent_cnt += i; | |
4890 | cnt++; | |
4891 | } | |
4892 | ||
4893 | maxsize = merge_extents(super, extent_cnt); | |
4894 | minsize = size; | |
4895 | if (size == 0) | |
612e59d8 CA |
4896 | /* chunk is in K */ |
4897 | minsize = chunk * 2; | |
efb30e7f DW |
4898 | |
4899 | if (cnt < raiddisks || | |
4900 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
4901 | maxsize < minsize || |
4902 | maxsize == 0) { | |
efb30e7f DW |
4903 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
4904 | return 0; /* No enough free spaces large enough */ | |
4905 | } | |
4906 | ||
4907 | if (size == 0) { | |
4908 | size = maxsize; | |
4909 | if (chunk) { | |
612e59d8 CA |
4910 | size /= 2 * chunk; |
4911 | size *= 2 * chunk; | |
efb30e7f DW |
4912 | } |
4913 | } | |
4914 | ||
4915 | cnt = 0; | |
4916 | for (dl = super->disks; dl; dl = dl->next) | |
4917 | if (dl->e) | |
4918 | dl->raiddisk = cnt++; | |
4919 | ||
4920 | *freesize = size; | |
4921 | ||
4922 | return 1; | |
4923 | } | |
4924 | ||
bf5a934a | 4925 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 4926 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
4927 | char *dev, unsigned long long *freesize, |
4928 | int verbose) | |
4929 | { | |
4930 | int fd, cfd; | |
4931 | struct mdinfo *sra; | |
20cbe8d2 | 4932 | int is_member = 0; |
bf5a934a | 4933 | |
d54559f0 LM |
4934 | /* load capability |
4935 | * if given unused devices create a container | |
bf5a934a DW |
4936 | * if given given devices in a container create a member volume |
4937 | */ | |
4938 | if (level == LEVEL_CONTAINER) { | |
4939 | /* Must be a fresh device to add to a container */ | |
4940 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
4941 | raiddisks, |
4942 | chunk?*chunk:0, size, | |
bf5a934a DW |
4943 | dev, freesize, |
4944 | verbose); | |
4945 | } | |
4946 | ||
8592f29d N |
4947 | if (!dev) { |
4948 | if (st->sb && freesize) { | |
efb30e7f DW |
4949 | /* we are being asked to automatically layout a |
4950 | * new volume based on the current contents of | |
4951 | * the container. If the the parameters can be | |
4952 | * satisfied reserve_space will record the disks, | |
4953 | * start offset, and size of the volume to be | |
4954 | * created. add_to_super and getinfo_super | |
4955 | * detect when autolayout is in progress. | |
4956 | */ | |
6592ce37 DW |
4957 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
4958 | raiddisks, chunk, | |
4959 | verbose)) | |
4960 | return 0; | |
c21e737b CA |
4961 | return reserve_space(st, raiddisks, size, |
4962 | chunk?*chunk:0, freesize); | |
8592f29d N |
4963 | } |
4964 | return 1; | |
4965 | } | |
bf5a934a DW |
4966 | if (st->sb) { |
4967 | /* creating in a given container */ | |
4968 | return validate_geometry_imsm_volume(st, level, layout, | |
4969 | raiddisks, chunk, size, | |
4970 | dev, freesize, verbose); | |
4971 | } | |
4972 | ||
bf5a934a DW |
4973 | /* This device needs to be a device in an 'imsm' container */ |
4974 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4975 | if (fd >= 0) { | |
4976 | if (verbose) | |
4977 | fprintf(stderr, | |
4978 | Name ": Cannot create this array on device %s\n", | |
4979 | dev); | |
4980 | close(fd); | |
4981 | return 0; | |
4982 | } | |
4983 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
4984 | if (verbose) | |
4985 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
4986 | dev, strerror(errno)); | |
4987 | return 0; | |
4988 | } | |
4989 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
4990 | cfd = open_container(fd); | |
20cbe8d2 | 4991 | close(fd); |
bf5a934a | 4992 | if (cfd < 0) { |
bf5a934a DW |
4993 | if (verbose) |
4994 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
4995 | dev); | |
4996 | return 0; | |
4997 | } | |
4998 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 4999 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
5000 | strcmp(sra->text_version, "imsm") == 0) |
5001 | is_member = 1; | |
5002 | sysfs_free(sra); | |
5003 | if (is_member) { | |
bf5a934a DW |
5004 | /* This is a member of a imsm container. Load the container |
5005 | * and try to create a volume | |
5006 | */ | |
5007 | struct intel_super *super; | |
5008 | ||
e1902a7b | 5009 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL) == 0) { |
bf5a934a DW |
5010 | st->sb = super; |
5011 | st->container_dev = fd2devnum(cfd); | |
5012 | close(cfd); | |
5013 | return validate_geometry_imsm_volume(st, level, layout, | |
5014 | raiddisks, chunk, | |
5015 | size, dev, | |
5016 | freesize, verbose); | |
5017 | } | |
20cbe8d2 | 5018 | } |
bf5a934a | 5019 | |
20cbe8d2 AW |
5020 | if (verbose) |
5021 | fprintf(stderr, Name ": failed container membership check\n"); | |
5022 | ||
5023 | close(cfd); | |
5024 | return 0; | |
bf5a934a | 5025 | } |
0bd16cf2 | 5026 | |
30f58b22 | 5027 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
5028 | { |
5029 | struct intel_super *super = st->sb; | |
5030 | ||
30f58b22 DW |
5031 | if (level && *level == UnSet) |
5032 | *level = LEVEL_CONTAINER; | |
5033 | ||
5034 | if (level && layout && *layout == UnSet) | |
5035 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 5036 | |
1d54f286 N |
5037 | if (chunk && (*chunk == UnSet || *chunk == 0) && |
5038 | super && super->orom) | |
30f58b22 | 5039 | *chunk = imsm_orom_default_chunk(super->orom); |
0bd16cf2 DJ |
5040 | } |
5041 | ||
33414a01 DW |
5042 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
5043 | ||
5044 | static int kill_subarray_imsm(struct supertype *st) | |
5045 | { | |
5046 | /* remove the subarray currently referenced by ->current_vol */ | |
5047 | __u8 i; | |
5048 | struct intel_dev **dp; | |
5049 | struct intel_super *super = st->sb; | |
5050 | __u8 current_vol = super->current_vol; | |
5051 | struct imsm_super *mpb = super->anchor; | |
5052 | ||
5053 | if (super->current_vol < 0) | |
5054 | return 2; | |
5055 | super->current_vol = -1; /* invalidate subarray cursor */ | |
5056 | ||
5057 | /* block deletions that would change the uuid of active subarrays | |
5058 | * | |
5059 | * FIXME when immutable ids are available, but note that we'll | |
5060 | * also need to fixup the invalidated/active subarray indexes in | |
5061 | * mdstat | |
5062 | */ | |
5063 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5064 | char subarray[4]; | |
5065 | ||
5066 | if (i < current_vol) | |
5067 | continue; | |
5068 | sprintf(subarray, "%u", i); | |
5069 | if (is_subarray_active(subarray, st->devname)) { | |
5070 | fprintf(stderr, | |
5071 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
5072 | current_vol, i); | |
5073 | ||
5074 | return 2; | |
5075 | } | |
5076 | } | |
5077 | ||
5078 | if (st->update_tail) { | |
5079 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
5080 | ||
5081 | if (!u) | |
5082 | return 2; | |
5083 | u->type = update_kill_array; | |
5084 | u->dev_idx = current_vol; | |
5085 | append_metadata_update(st, u, sizeof(*u)); | |
5086 | ||
5087 | return 0; | |
5088 | } | |
5089 | ||
5090 | for (dp = &super->devlist; *dp;) | |
5091 | if ((*dp)->index == current_vol) { | |
5092 | *dp = (*dp)->next; | |
5093 | } else { | |
5094 | handle_missing(super, (*dp)->dev); | |
5095 | if ((*dp)->index > current_vol) | |
5096 | (*dp)->index--; | |
5097 | dp = &(*dp)->next; | |
5098 | } | |
5099 | ||
5100 | /* no more raid devices, all active components are now spares, | |
5101 | * but of course failed are still failed | |
5102 | */ | |
5103 | if (--mpb->num_raid_devs == 0) { | |
5104 | struct dl *d; | |
5105 | ||
5106 | for (d = super->disks; d; d = d->next) | |
5107 | if (d->index > -2) { | |
5108 | d->index = -1; | |
5109 | d->disk.status = SPARE_DISK; | |
5110 | } | |
5111 | } | |
5112 | ||
5113 | super->updates_pending++; | |
5114 | ||
5115 | return 0; | |
5116 | } | |
aa534678 | 5117 | |
a951a4f7 | 5118 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 5119 | char *update, struct mddev_ident *ident) |
aa534678 DW |
5120 | { |
5121 | /* update the subarray currently referenced by ->current_vol */ | |
5122 | struct intel_super *super = st->sb; | |
5123 | struct imsm_super *mpb = super->anchor; | |
5124 | ||
aa534678 DW |
5125 | if (strcmp(update, "name") == 0) { |
5126 | char *name = ident->name; | |
a951a4f7 N |
5127 | char *ep; |
5128 | int vol; | |
aa534678 | 5129 | |
a951a4f7 | 5130 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
5131 | fprintf(stderr, |
5132 | Name ": Unable to update name of active subarray\n"); | |
5133 | return 2; | |
5134 | } | |
5135 | ||
5136 | if (!check_name(super, name, 0)) | |
5137 | return 2; | |
5138 | ||
a951a4f7 N |
5139 | vol = strtoul(subarray, &ep, 10); |
5140 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
5141 | return 2; | |
5142 | ||
aa534678 DW |
5143 | if (st->update_tail) { |
5144 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
5145 | ||
5146 | if (!u) | |
5147 | return 2; | |
5148 | u->type = update_rename_array; | |
a951a4f7 | 5149 | u->dev_idx = vol; |
aa534678 DW |
5150 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
5151 | append_metadata_update(st, u, sizeof(*u)); | |
5152 | } else { | |
5153 | struct imsm_dev *dev; | |
5154 | int i; | |
5155 | ||
a951a4f7 | 5156 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
5157 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
5158 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5159 | dev = get_imsm_dev(super, i); | |
5160 | handle_missing(super, dev); | |
5161 | } | |
5162 | super->updates_pending++; | |
5163 | } | |
5164 | } else | |
5165 | return 2; | |
5166 | ||
5167 | return 0; | |
5168 | } | |
bf5a934a | 5169 | |
28bce06f AK |
5170 | static int is_gen_migration(struct imsm_dev *dev) |
5171 | { | |
5172 | if (!dev->vol.migr_state) | |
5173 | return 0; | |
5174 | ||
5175 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
5176 | return 1; | |
5177 | ||
5178 | return 0; | |
5179 | } | |
71204a50 | 5180 | #endif /* MDASSEMBLE */ |
28bce06f | 5181 | |
1e5c6983 DW |
5182 | static int is_rebuilding(struct imsm_dev *dev) |
5183 | { | |
5184 | struct imsm_map *migr_map; | |
5185 | ||
5186 | if (!dev->vol.migr_state) | |
5187 | return 0; | |
5188 | ||
5189 | if (migr_type(dev) != MIGR_REBUILD) | |
5190 | return 0; | |
5191 | ||
5192 | migr_map = get_imsm_map(dev, 1); | |
5193 | ||
5194 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
5195 | return 1; | |
5196 | else | |
5197 | return 0; | |
5198 | } | |
5199 | ||
5200 | static void update_recovery_start(struct imsm_dev *dev, struct mdinfo *array) | |
5201 | { | |
5202 | struct mdinfo *rebuild = NULL; | |
5203 | struct mdinfo *d; | |
5204 | __u32 units; | |
5205 | ||
5206 | if (!is_rebuilding(dev)) | |
5207 | return; | |
5208 | ||
5209 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
5210 | for (d = array->devs; d; d = d->next) | |
5211 | if (d->recovery_start == 0) { | |
5212 | if (rebuild) | |
5213 | return; | |
5214 | rebuild = d; | |
5215 | } | |
5216 | ||
4363fd80 DW |
5217 | if (!rebuild) { |
5218 | /* (?) none of the disks are marked with | |
5219 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
5220 | * disk_ord_tbl was not correctly updated | |
5221 | */ | |
5222 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
5223 | return; | |
5224 | } | |
5225 | ||
1e5c6983 DW |
5226 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
5227 | rebuild->recovery_start = units * blocks_per_migr_unit(dev); | |
5228 | } | |
5229 | ||
5230 | ||
00bbdbda | 5231 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 5232 | { |
4f5bc454 DW |
5233 | /* Given a container loaded by load_super_imsm_all, |
5234 | * extract information about all the arrays into | |
5235 | * an mdinfo tree. | |
00bbdbda | 5236 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
5237 | * |
5238 | * For each imsm_dev create an mdinfo, fill it in, | |
5239 | * then look for matching devices in super->disks | |
5240 | * and create appropriate device mdinfo. | |
5241 | */ | |
5242 | struct intel_super *super = st->sb; | |
949c47a0 | 5243 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 5244 | struct mdinfo *rest = NULL; |
00bbdbda | 5245 | unsigned int i; |
a06d022d | 5246 | int bbm_errors = 0; |
abef11a3 AK |
5247 | struct dl *d; |
5248 | int spare_disks = 0; | |
cdddbdbc | 5249 | |
a06d022d KW |
5250 | /* check for bad blocks */ |
5251 | if (imsm_bbm_log_size(super->anchor)) | |
5252 | bbm_errors = 1; | |
604b746f | 5253 | |
abef11a3 AK |
5254 | /* count spare devices, not used in maps |
5255 | */ | |
5256 | for (d = super->disks; d; d = d->next) | |
5257 | if (d->index == -1) | |
5258 | spare_disks++; | |
5259 | ||
4f5bc454 | 5260 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
5261 | struct imsm_dev *dev; |
5262 | struct imsm_map *map; | |
86e3692b | 5263 | struct imsm_map *map2; |
4f5bc454 | 5264 | struct mdinfo *this; |
2db86302 | 5265 | int slot, chunk; |
00bbdbda N |
5266 | char *ep; |
5267 | ||
5268 | if (subarray && | |
5269 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
5270 | continue; | |
5271 | ||
5272 | dev = get_imsm_dev(super, i); | |
5273 | map = get_imsm_map(dev, 0); | |
86e3692b | 5274 | map2 = get_imsm_map(dev, 1); |
4f5bc454 | 5275 | |
1ce0101c DW |
5276 | /* do not publish arrays that are in the middle of an |
5277 | * unsupported migration | |
5278 | */ | |
5279 | if (dev->vol.migr_state && | |
28bce06f | 5280 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
5281 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
5282 | " unsupported migration in progress\n", | |
5283 | dev->volume); | |
5284 | continue; | |
5285 | } | |
2db86302 LM |
5286 | /* do not publish arrays that are not support by controller's |
5287 | * OROM/EFI | |
5288 | */ | |
1ce0101c | 5289 | |
2db86302 | 5290 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
7b0bbd0f | 5291 | #ifndef MDASSEMBLE |
2db86302 LM |
5292 | if (!validate_geometry_imsm_orom(super, |
5293 | get_imsm_raid_level(map), /* RAID level */ | |
5294 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
5295 | map->num_members, /* raid disks */ | |
5296 | &chunk, | |
5297 | 1 /* verbose */)) { | |
5298 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
5299 | "Cannot proceed with the action(s).\n"); | |
5300 | continue; | |
5301 | } | |
7b0bbd0f | 5302 | #endif /* MDASSEMBLE */ |
4f5bc454 | 5303 | this = malloc(sizeof(*this)); |
0fbd635c | 5304 | if (!this) { |
cf1be220 | 5305 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
5306 | sizeof(*this)); |
5307 | break; | |
5308 | } | |
4f5bc454 DW |
5309 | memset(this, 0, sizeof(*this)); |
5310 | this->next = rest; | |
4f5bc454 | 5311 | |
301406c9 | 5312 | super->current_vol = i; |
a5d85af7 | 5313 | getinfo_super_imsm_volume(st, this, NULL); |
4f5bc454 | 5314 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 5315 | unsigned long long recovery_start; |
4f5bc454 DW |
5316 | struct mdinfo *info_d; |
5317 | struct dl *d; | |
5318 | int idx; | |
9a1608e5 | 5319 | int skip; |
7eef0453 | 5320 | __u32 ord; |
4f5bc454 | 5321 | |
9a1608e5 | 5322 | skip = 0; |
98130f40 | 5323 | idx = get_imsm_disk_idx(dev, slot, 0); |
196b0d44 | 5324 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
4f5bc454 DW |
5325 | for (d = super->disks; d ; d = d->next) |
5326 | if (d->index == idx) | |
0fbd635c | 5327 | break; |
4f5bc454 | 5328 | |
1e5c6983 | 5329 | recovery_start = MaxSector; |
4f5bc454 | 5330 | if (d == NULL) |
9a1608e5 | 5331 | skip = 1; |
25ed7e59 | 5332 | if (d && is_failed(&d->disk)) |
9a1608e5 | 5333 | skip = 1; |
7eef0453 | 5334 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 5335 | recovery_start = 0; |
9a1608e5 DW |
5336 | |
5337 | /* | |
5338 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
5339 | * reset resync start to avoid a dirty-degraded |
5340 | * situation when performing the intial sync | |
9a1608e5 DW |
5341 | * |
5342 | * FIXME handle dirty degraded | |
5343 | */ | |
1e5c6983 | 5344 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 5345 | this->resync_start = MaxSector; |
9a1608e5 DW |
5346 | if (skip) |
5347 | continue; | |
4f5bc454 | 5348 | |
1e5c6983 | 5349 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
5350 | if (!info_d) { |
5351 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 5352 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
5353 | info_d = this->devs; |
5354 | while (info_d) { | |
5355 | struct mdinfo *d = info_d->next; | |
5356 | ||
5357 | free(info_d); | |
5358 | info_d = d; | |
5359 | } | |
9a1608e5 DW |
5360 | free(this); |
5361 | this = rest; | |
5362 | break; | |
5363 | } | |
4f5bc454 DW |
5364 | info_d->next = this->devs; |
5365 | this->devs = info_d; | |
5366 | ||
4f5bc454 DW |
5367 | info_d->disk.number = d->index; |
5368 | info_d->disk.major = d->major; | |
5369 | info_d->disk.minor = d->minor; | |
5370 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 5371 | info_d->recovery_start = recovery_start; |
86e3692b AK |
5372 | if (map2) { |
5373 | if (slot < map2->num_members) | |
5374 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5375 | else |
5376 | this->array.spare_disks++; | |
86e3692b AK |
5377 | } else { |
5378 | if (slot < map->num_members) | |
5379 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5380 | else |
5381 | this->array.spare_disks++; | |
86e3692b | 5382 | } |
1e5c6983 DW |
5383 | if (info_d->recovery_start == MaxSector) |
5384 | this->array.working_disks++; | |
4f5bc454 DW |
5385 | |
5386 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5387 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
5388 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
4f5bc454 | 5389 | } |
1e5c6983 DW |
5390 | /* now that the disk list is up-to-date fixup recovery_start */ |
5391 | update_recovery_start(dev, this); | |
abef11a3 | 5392 | this->array.spare_disks += spare_disks; |
9a1608e5 | 5393 | rest = this; |
4f5bc454 DW |
5394 | } |
5395 | ||
a06d022d KW |
5396 | /* if array has bad blocks, set suitable bit in array status */ |
5397 | if (bbm_errors) | |
5398 | rest->array.state |= (1<<MD_SB_BBM_ERRORS); | |
5399 | ||
4f5bc454 | 5400 | return rest; |
cdddbdbc DW |
5401 | } |
5402 | ||
845dea95 | 5403 | |
fb49eef2 | 5404 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed) |
c2a1e7da | 5405 | { |
a965f303 | 5406 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
5407 | |
5408 | if (!failed) | |
3393c6af DW |
5409 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
5410 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
5411 | |
5412 | switch (get_imsm_raid_level(map)) { | |
5413 | case 0: | |
5414 | return IMSM_T_STATE_FAILED; | |
5415 | break; | |
5416 | case 1: | |
5417 | if (failed < map->num_members) | |
5418 | return IMSM_T_STATE_DEGRADED; | |
5419 | else | |
5420 | return IMSM_T_STATE_FAILED; | |
5421 | break; | |
5422 | case 10: | |
5423 | { | |
5424 | /** | |
c92a2527 DW |
5425 | * check to see if any mirrors have failed, otherwise we |
5426 | * are degraded. Even numbered slots are mirrored on | |
5427 | * slot+1 | |
c2a1e7da | 5428 | */ |
c2a1e7da | 5429 | int i; |
d9b420a5 N |
5430 | /* gcc -Os complains that this is unused */ |
5431 | int insync = insync; | |
c2a1e7da DW |
5432 | |
5433 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 5434 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
c92a2527 DW |
5435 | int idx = ord_to_idx(ord); |
5436 | struct imsm_disk *disk; | |
c2a1e7da | 5437 | |
c92a2527 DW |
5438 | /* reset the potential in-sync count on even-numbered |
5439 | * slots. num_copies is always 2 for imsm raid10 | |
5440 | */ | |
5441 | if ((i & 1) == 0) | |
5442 | insync = 2; | |
c2a1e7da | 5443 | |
c92a2527 | 5444 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 5445 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 5446 | insync--; |
c2a1e7da | 5447 | |
c92a2527 DW |
5448 | /* no in-sync disks left in this mirror the |
5449 | * array has failed | |
5450 | */ | |
5451 | if (insync == 0) | |
5452 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
5453 | } |
5454 | ||
5455 | return IMSM_T_STATE_DEGRADED; | |
5456 | } | |
5457 | case 5: | |
5458 | if (failed < 2) | |
5459 | return IMSM_T_STATE_DEGRADED; | |
5460 | else | |
5461 | return IMSM_T_STATE_FAILED; | |
5462 | break; | |
5463 | default: | |
5464 | break; | |
5465 | } | |
5466 | ||
5467 | return map->map_state; | |
5468 | } | |
5469 | ||
ff077194 | 5470 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev) |
c2a1e7da DW |
5471 | { |
5472 | int i; | |
5473 | int failed = 0; | |
5474 | struct imsm_disk *disk; | |
ff077194 | 5475 | struct imsm_map *map = get_imsm_map(dev, 0); |
0556e1a2 DW |
5476 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
5477 | __u32 ord; | |
5478 | int idx; | |
c2a1e7da | 5479 | |
0556e1a2 DW |
5480 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
5481 | * disks that are being rebuilt. New failures are recorded to | |
5482 | * map[0]. So we look through all the disks we started with and | |
5483 | * see if any failures are still present, or if any new ones | |
5484 | * have arrived | |
5485 | * | |
5486 | * FIXME add support for online capacity expansion and | |
5487 | * raid-level-migration | |
5488 | */ | |
5489 | for (i = 0; i < prev->num_members; i++) { | |
5490 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); | |
5491 | ord |= __le32_to_cpu(map->disk_ord_tbl[i]); | |
5492 | idx = ord_to_idx(ord); | |
c2a1e7da | 5493 | |
949c47a0 | 5494 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 5495 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
fcb84475 | 5496 | failed++; |
c2a1e7da DW |
5497 | } |
5498 | ||
5499 | return failed; | |
845dea95 NB |
5500 | } |
5501 | ||
97b4d0e9 DW |
5502 | #ifndef MDASSEMBLE |
5503 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
5504 | char *inst) | |
5505 | { | |
5506 | struct intel_super *super = c->sb; | |
5507 | struct imsm_super *mpb = super->anchor; | |
5508 | ||
5509 | if (atoi(inst) >= mpb->num_raid_devs) { | |
5510 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
5511 | __func__, atoi(inst)); | |
5512 | return -ENODEV; | |
5513 | } | |
5514 | ||
5515 | dprintf("imsm: open_new %s\n", inst); | |
5516 | a->info.container_member = atoi(inst); | |
5517 | return 0; | |
5518 | } | |
5519 | ||
0c046afd DW |
5520 | static int is_resyncing(struct imsm_dev *dev) |
5521 | { | |
5522 | struct imsm_map *migr_map; | |
5523 | ||
5524 | if (!dev->vol.migr_state) | |
5525 | return 0; | |
5526 | ||
1484e727 DW |
5527 | if (migr_type(dev) == MIGR_INIT || |
5528 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
5529 | return 1; |
5530 | ||
4c9bc37b AK |
5531 | if (migr_type(dev) == MIGR_GEN_MIGR) |
5532 | return 0; | |
5533 | ||
0c046afd DW |
5534 | migr_map = get_imsm_map(dev, 1); |
5535 | ||
4c9bc37b AK |
5536 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
5537 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
5538 | return 1; |
5539 | else | |
5540 | return 0; | |
5541 | } | |
5542 | ||
0556e1a2 DW |
5543 | /* return true if we recorded new information */ |
5544 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 5545 | { |
0556e1a2 DW |
5546 | __u32 ord; |
5547 | int slot; | |
5548 | struct imsm_map *map; | |
5549 | ||
5550 | /* new failures are always set in map[0] */ | |
5551 | map = get_imsm_map(dev, 0); | |
5552 | ||
5553 | slot = get_imsm_disk_slot(map, idx); | |
5554 | if (slot < 0) | |
5555 | return 0; | |
5556 | ||
5557 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 5558 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
5559 | return 0; |
5560 | ||
f2f27e63 | 5561 | disk->status |= FAILED_DISK; |
0556e1a2 | 5562 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
f21e18ca | 5563 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
5564 | map->failed_disk_num = slot; |
5565 | return 1; | |
5566 | } | |
5567 | ||
5568 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
5569 | { | |
5570 | mark_failure(dev, disk, idx); | |
5571 | ||
5572 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
5573 | return; | |
5574 | ||
47ee5a45 DW |
5575 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
5576 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
5577 | } | |
5578 | ||
33414a01 DW |
5579 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
5580 | { | |
5581 | __u8 map_state; | |
5582 | struct dl *dl; | |
5583 | int failed; | |
5584 | ||
5585 | if (!super->missing) | |
5586 | return; | |
5587 | failed = imsm_count_failed(super, dev); | |
5588 | map_state = imsm_check_degraded(super, dev, failed); | |
5589 | ||
5590 | dprintf("imsm: mark missing\n"); | |
5591 | end_migration(dev, map_state); | |
5592 | for (dl = super->missing; dl; dl = dl->next) | |
5593 | mark_missing(dev, &dl->disk, dl->index); | |
5594 | super->updates_pending++; | |
5595 | } | |
5596 | ||
70bdf0dc AK |
5597 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev) |
5598 | { | |
5599 | int used_disks = imsm_num_data_members(dev, 0); | |
5600 | unsigned long long array_blocks; | |
5601 | struct imsm_map *map; | |
5602 | ||
5603 | if (used_disks == 0) { | |
5604 | /* when problems occures | |
5605 | * return current array_blocks value | |
5606 | */ | |
5607 | array_blocks = __le32_to_cpu(dev->size_high); | |
5608 | array_blocks = array_blocks << 32; | |
5609 | array_blocks += __le32_to_cpu(dev->size_low); | |
5610 | ||
5611 | return array_blocks; | |
5612 | } | |
5613 | ||
5614 | /* set array size in metadata | |
5615 | */ | |
5616 | map = get_imsm_map(dev, 0); | |
5617 | array_blocks = map->blocks_per_member * used_disks; | |
5618 | ||
5619 | /* round array size down to closest MB | |
5620 | */ | |
5621 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
5622 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
5623 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
5624 | ||
5625 | return array_blocks; | |
5626 | } | |
5627 | ||
28bce06f AK |
5628 | static void imsm_set_disk(struct active_array *a, int n, int state); |
5629 | ||
0e2d1a4e AK |
5630 | static void imsm_progress_container_reshape(struct intel_super *super) |
5631 | { | |
5632 | /* if no device has a migr_state, but some device has a | |
5633 | * different number of members than the previous device, start | |
5634 | * changing the number of devices in this device to match | |
5635 | * previous. | |
5636 | */ | |
5637 | struct imsm_super *mpb = super->anchor; | |
5638 | int prev_disks = -1; | |
5639 | int i; | |
1dfaa380 | 5640 | int copy_map_size; |
0e2d1a4e AK |
5641 | |
5642 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5643 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5644 | struct imsm_map *map = get_imsm_map(dev, 0); | |
5645 | struct imsm_map *map2; | |
5646 | int prev_num_members; | |
0e2d1a4e AK |
5647 | |
5648 | if (dev->vol.migr_state) | |
5649 | return; | |
5650 | ||
5651 | if (prev_disks == -1) | |
5652 | prev_disks = map->num_members; | |
5653 | if (prev_disks == map->num_members) | |
5654 | continue; | |
5655 | ||
5656 | /* OK, this array needs to enter reshape mode. | |
5657 | * i.e it needs a migr_state | |
5658 | */ | |
5659 | ||
1dfaa380 | 5660 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
5661 | prev_num_members = map->num_members; |
5662 | map->num_members = prev_disks; | |
5663 | dev->vol.migr_state = 1; | |
5664 | dev->vol.curr_migr_unit = 0; | |
5665 | dev->vol.migr_type = MIGR_GEN_MIGR; | |
5666 | for (i = prev_num_members; | |
5667 | i < map->num_members; i++) | |
5668 | set_imsm_ord_tbl_ent(map, i, i); | |
5669 | map2 = get_imsm_map(dev, 1); | |
5670 | /* Copy the current map */ | |
1dfaa380 | 5671 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
5672 | map2->num_members = prev_num_members; |
5673 | ||
70bdf0dc | 5674 | imsm_set_array_size(dev); |
0e2d1a4e AK |
5675 | super->updates_pending++; |
5676 | } | |
5677 | } | |
5678 | ||
aad6f216 | 5679 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
5680 | * states are handled in imsm_set_disk() with one exception, when a |
5681 | * resync is stopped due to a new failure this routine will set the | |
5682 | * 'degraded' state for the array. | |
5683 | */ | |
01f157d7 | 5684 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
5685 | { |
5686 | int inst = a->info.container_member; | |
5687 | struct intel_super *super = a->container->sb; | |
949c47a0 | 5688 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5689 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd DW |
5690 | int failed = imsm_count_failed(super, dev); |
5691 | __u8 map_state = imsm_check_degraded(super, dev, failed); | |
1e5c6983 | 5692 | __u32 blocks_per_unit; |
a862209d | 5693 | |
1af97990 AK |
5694 | if (dev->vol.migr_state && |
5695 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
5696 | /* array state change is blocked due to reshape action | |
aad6f216 N |
5697 | * We might need to |
5698 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
5699 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
5700 | * - update curr_migr_unit | |
1af97990 | 5701 | */ |
aad6f216 N |
5702 | if (a->curr_action == reshape) { |
5703 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 5704 | goto mark_checkpoint; |
aad6f216 N |
5705 | } else { |
5706 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
5707 | /* for some reason we aborted the reshape. | |
5708 | * Better clean up | |
5709 | */ | |
5710 | struct imsm_map *map2 = get_imsm_map(dev, 1); | |
5711 | dev->vol.migr_state = 0; | |
5712 | dev->vol.migr_type = 0; | |
5713 | dev->vol.curr_migr_unit = 0; | |
5714 | memcpy(map, map2, sizeof_imsm_map(map2)); | |
5715 | super->updates_pending++; | |
5716 | } | |
5717 | if (a->last_checkpoint >= a->info.component_size) { | |
5718 | unsigned long long array_blocks; | |
5719 | int used_disks; | |
e154ced3 | 5720 | struct mdinfo *mdi; |
aad6f216 | 5721 | |
9653001d | 5722 | used_disks = imsm_num_data_members(dev, 0); |
d55adef9 AK |
5723 | if (used_disks > 0) { |
5724 | array_blocks = | |
5725 | map->blocks_per_member * | |
5726 | used_disks; | |
5727 | /* round array size down to closest MB | |
5728 | */ | |
5729 | array_blocks = (array_blocks | |
5730 | >> SECT_PER_MB_SHIFT) | |
5731 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
5732 | a->info.custom_array_size = array_blocks; |
5733 | /* encourage manager to update array | |
5734 | * size | |
5735 | */ | |
e154ced3 | 5736 | |
d55adef9 | 5737 | a->check_reshape = 1; |
633b5610 | 5738 | } |
e154ced3 AK |
5739 | /* finalize online capacity expansion/reshape */ |
5740 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
5741 | imsm_set_disk(a, | |
5742 | mdi->disk.raid_disk, | |
5743 | mdi->curr_state); | |
5744 | ||
0e2d1a4e | 5745 | imsm_progress_container_reshape(super); |
e154ced3 | 5746 | } |
aad6f216 | 5747 | } |
1af97990 AK |
5748 | } |
5749 | ||
47ee5a45 | 5750 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
5751 | if (consistent == 2) |
5752 | handle_missing(super, dev); | |
1e5c6983 | 5753 | |
0c046afd | 5754 | if (consistent == 2 && |
b7941fd6 | 5755 | (!is_resync_complete(&a->info) || |
0c046afd DW |
5756 | map_state != IMSM_T_STATE_NORMAL || |
5757 | dev->vol.migr_state)) | |
01f157d7 | 5758 | consistent = 0; |
272906ef | 5759 | |
b7941fd6 | 5760 | if (is_resync_complete(&a->info)) { |
0c046afd | 5761 | /* complete intialization / resync, |
0556e1a2 DW |
5762 | * recovery and interrupted recovery is completed in |
5763 | * ->set_disk | |
0c046afd DW |
5764 | */ |
5765 | if (is_resyncing(dev)) { | |
5766 | dprintf("imsm: mark resync done\n"); | |
f8f603f1 | 5767 | end_migration(dev, map_state); |
115c3803 | 5768 | super->updates_pending++; |
484240d8 | 5769 | a->last_checkpoint = 0; |
115c3803 | 5770 | } |
0c046afd DW |
5771 | } else if (!is_resyncing(dev) && !failed) { |
5772 | /* mark the start of the init process if nothing is failed */ | |
b7941fd6 | 5773 | dprintf("imsm: mark resync start\n"); |
1484e727 | 5774 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 5775 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 5776 | else |
8e59f3d8 | 5777 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 5778 | super->updates_pending++; |
115c3803 | 5779 | } |
a862209d | 5780 | |
633b5610 | 5781 | mark_checkpoint: |
1e5c6983 DW |
5782 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
5783 | blocks_per_unit = blocks_per_migr_unit(dev); | |
4f0a7acc | 5784 | if (blocks_per_unit) { |
1e5c6983 DW |
5785 | __u32 units32; |
5786 | __u64 units; | |
5787 | ||
4f0a7acc | 5788 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
5789 | units32 = units; |
5790 | ||
5791 | /* check that we did not overflow 32-bits, and that | |
5792 | * curr_migr_unit needs updating | |
5793 | */ | |
5794 | if (units32 == units && | |
bfd80a56 | 5795 | units32 != 0 && |
1e5c6983 DW |
5796 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
5797 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
5798 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
5799 | super->updates_pending++; | |
5800 | } | |
5801 | } | |
f8f603f1 | 5802 | |
3393c6af | 5803 | /* mark dirty / clean */ |
0c046afd | 5804 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 5805 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
5806 | if (consistent) |
5807 | dev->vol.dirty = 0; | |
5808 | else | |
5809 | dev->vol.dirty = 1; | |
a862209d DW |
5810 | super->updates_pending++; |
5811 | } | |
28bce06f | 5812 | |
01f157d7 | 5813 | return consistent; |
a862209d DW |
5814 | } |
5815 | ||
8d45d196 | 5816 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 5817 | { |
8d45d196 DW |
5818 | int inst = a->info.container_member; |
5819 | struct intel_super *super = a->container->sb; | |
949c47a0 | 5820 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5821 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 | 5822 | struct imsm_disk *disk; |
0c046afd | 5823 | int failed; |
b10b37b8 | 5824 | __u32 ord; |
0c046afd | 5825 | __u8 map_state; |
8d45d196 DW |
5826 | |
5827 | if (n > map->num_members) | |
5828 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
5829 | n, map->num_members - 1); | |
5830 | ||
5831 | if (n < 0) | |
5832 | return; | |
5833 | ||
4e6e574a | 5834 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 5835 | |
98130f40 | 5836 | ord = get_imsm_ord_tbl_ent(dev, n, -1); |
b10b37b8 | 5837 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 5838 | |
5802a811 | 5839 | /* check for new failures */ |
0556e1a2 DW |
5840 | if (state & DS_FAULTY) { |
5841 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
5842 | super->updates_pending++; | |
8d45d196 | 5843 | } |
47ee5a45 | 5844 | |
19859edc | 5845 | /* check if in_sync */ |
0556e1a2 | 5846 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
b10b37b8 DW |
5847 | struct imsm_map *migr_map = get_imsm_map(dev, 1); |
5848 | ||
5849 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
5850 | super->updates_pending++; |
5851 | } | |
8d45d196 | 5852 | |
0c046afd DW |
5853 | failed = imsm_count_failed(super, dev); |
5854 | map_state = imsm_check_degraded(super, dev, failed); | |
5802a811 | 5855 | |
0c046afd DW |
5856 | /* check if recovery complete, newly degraded, or failed */ |
5857 | if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) { | |
f8f603f1 | 5858 | end_migration(dev, map_state); |
0556e1a2 DW |
5859 | map = get_imsm_map(dev, 0); |
5860 | map->failed_disk_num = ~0; | |
0c046afd | 5861 | super->updates_pending++; |
484240d8 | 5862 | a->last_checkpoint = 0; |
0c046afd DW |
5863 | } else if (map_state == IMSM_T_STATE_DEGRADED && |
5864 | map->map_state != map_state && | |
5865 | !dev->vol.migr_state) { | |
5866 | dprintf("imsm: mark degraded\n"); | |
5867 | map->map_state = map_state; | |
5868 | super->updates_pending++; | |
484240d8 | 5869 | a->last_checkpoint = 0; |
0c046afd DW |
5870 | } else if (map_state == IMSM_T_STATE_FAILED && |
5871 | map->map_state != map_state) { | |
5872 | dprintf("imsm: mark failed\n"); | |
f8f603f1 | 5873 | end_migration(dev, map_state); |
0c046afd | 5874 | super->updates_pending++; |
484240d8 | 5875 | a->last_checkpoint = 0; |
28bce06f AK |
5876 | } else if (is_gen_migration(dev)) { |
5877 | dprintf("imsm: Detected General Migration in state: "); | |
5878 | if (map_state == IMSM_T_STATE_NORMAL) { | |
5879 | end_migration(dev, map_state); | |
5880 | map = get_imsm_map(dev, 0); | |
5881 | map->failed_disk_num = ~0; | |
5882 | dprintf("normal\n"); | |
5883 | } else { | |
5884 | if (map_state == IMSM_T_STATE_DEGRADED) { | |
5885 | printf("degraded\n"); | |
5886 | end_migration(dev, map_state); | |
5887 | } else { | |
5888 | dprintf("failed\n"); | |
5889 | } | |
5890 | map->map_state = map_state; | |
5891 | } | |
5892 | super->updates_pending++; | |
5802a811 | 5893 | } |
845dea95 NB |
5894 | } |
5895 | ||
f796af5d | 5896 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 5897 | { |
f796af5d | 5898 | void *buf = mpb; |
c2a1e7da DW |
5899 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
5900 | unsigned long long dsize; | |
5901 | unsigned long long sectors; | |
5902 | ||
5903 | get_dev_size(fd, NULL, &dsize); | |
5904 | ||
272f648f DW |
5905 | if (mpb_size > 512) { |
5906 | /* -1 to account for anchor */ | |
5907 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 5908 | |
272f648f DW |
5909 | /* write the extended mpb to the sectors preceeding the anchor */ |
5910 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
5911 | return 1; | |
c2a1e7da | 5912 | |
f21e18ca N |
5913 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
5914 | != 512 * sectors) | |
272f648f DW |
5915 | return 1; |
5916 | } | |
c2a1e7da | 5917 | |
272f648f DW |
5918 | /* first block is stored on second to last sector of the disk */ |
5919 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
5920 | return 1; |
5921 | ||
f796af5d | 5922 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
5923 | return 1; |
5924 | ||
c2a1e7da DW |
5925 | return 0; |
5926 | } | |
5927 | ||
2e735d19 | 5928 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 5929 | { |
2e735d19 | 5930 | struct intel_super *super = container->sb; |
c2a1e7da | 5931 | |
1a64be56 | 5932 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
5933 | if (!super->updates_pending) |
5934 | return; | |
5935 | ||
36988a3d | 5936 | write_super_imsm(container, 0); |
c2a1e7da DW |
5937 | |
5938 | super->updates_pending = 0; | |
845dea95 NB |
5939 | } |
5940 | ||
272906ef DW |
5941 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
5942 | { | |
5943 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 5944 | int i = get_imsm_disk_idx(dev, idx, -1); |
272906ef DW |
5945 | struct dl *dl; |
5946 | ||
5947 | for (dl = super->disks; dl; dl = dl->next) | |
5948 | if (dl->index == i) | |
5949 | break; | |
5950 | ||
25ed7e59 | 5951 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
5952 | dl = NULL; |
5953 | ||
5954 | if (dl) | |
5955 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
5956 | ||
5957 | return dl; | |
5958 | } | |
5959 | ||
a20d2ba5 | 5960 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
5961 | struct active_array *a, int activate_new, |
5962 | struct mdinfo *additional_test_list) | |
272906ef DW |
5963 | { |
5964 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 5965 | int idx = get_imsm_disk_idx(dev, slot, -1); |
a20d2ba5 DW |
5966 | struct imsm_super *mpb = super->anchor; |
5967 | struct imsm_map *map; | |
272906ef DW |
5968 | unsigned long long pos; |
5969 | struct mdinfo *d; | |
5970 | struct extent *ex; | |
a20d2ba5 | 5971 | int i, j; |
272906ef | 5972 | int found; |
569cc43f DW |
5973 | __u32 array_start = 0; |
5974 | __u32 array_end = 0; | |
272906ef | 5975 | struct dl *dl; |
6c932028 | 5976 | struct mdinfo *test_list; |
272906ef DW |
5977 | |
5978 | for (dl = super->disks; dl; dl = dl->next) { | |
5979 | /* If in this array, skip */ | |
5980 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
5981 | if (d->state_fd >= 0 && |
5982 | d->disk.major == dl->major && | |
272906ef | 5983 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
5984 | dprintf("%x:%x already in array\n", |
5985 | dl->major, dl->minor); | |
272906ef DW |
5986 | break; |
5987 | } | |
5988 | if (d) | |
5989 | continue; | |
6c932028 AK |
5990 | test_list = additional_test_list; |
5991 | while (test_list) { | |
5992 | if (test_list->disk.major == dl->major && | |
5993 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
5994 | dprintf("%x:%x already in additional test list\n", |
5995 | dl->major, dl->minor); | |
5996 | break; | |
5997 | } | |
6c932028 | 5998 | test_list = test_list->next; |
8ba77d32 | 5999 | } |
6c932028 | 6000 | if (test_list) |
8ba77d32 | 6001 | continue; |
272906ef | 6002 | |
e553d2a4 | 6003 | /* skip in use or failed drives */ |
25ed7e59 | 6004 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
6005 | dl->index == -2) { |
6006 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 6007 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
6008 | continue; |
6009 | } | |
6010 | ||
a20d2ba5 DW |
6011 | /* skip pure spares when we are looking for partially |
6012 | * assimilated drives | |
6013 | */ | |
6014 | if (dl->index == -1 && !activate_new) | |
6015 | continue; | |
6016 | ||
272906ef | 6017 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 6018 | * It needs to be able to cover all member volumes |
272906ef DW |
6019 | */ |
6020 | ex = get_extents(super, dl); | |
6021 | if (!ex) { | |
6022 | dprintf("cannot get extents\n"); | |
6023 | continue; | |
6024 | } | |
a20d2ba5 DW |
6025 | for (i = 0; i < mpb->num_raid_devs; i++) { |
6026 | dev = get_imsm_dev(super, i); | |
6027 | map = get_imsm_map(dev, 0); | |
272906ef | 6028 | |
a20d2ba5 DW |
6029 | /* check if this disk is already a member of |
6030 | * this array | |
272906ef | 6031 | */ |
620b1713 | 6032 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
6033 | continue; |
6034 | ||
6035 | found = 0; | |
6036 | j = 0; | |
6037 | pos = 0; | |
6038 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
329c8278 DW |
6039 | array_end = array_start + |
6040 | __le32_to_cpu(map->blocks_per_member) - 1; | |
a20d2ba5 DW |
6041 | |
6042 | do { | |
6043 | /* check that we can start at pba_of_lba0 with | |
6044 | * blocks_per_member of space | |
6045 | */ | |
329c8278 | 6046 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
6047 | found = 1; |
6048 | break; | |
6049 | } | |
6050 | pos = ex[j].start + ex[j].size; | |
6051 | j++; | |
6052 | } while (ex[j-1].size); | |
6053 | ||
6054 | if (!found) | |
272906ef | 6055 | break; |
a20d2ba5 | 6056 | } |
272906ef DW |
6057 | |
6058 | free(ex); | |
a20d2ba5 | 6059 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
6060 | dprintf("%x:%x does not have %u to %u available\n", |
6061 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
6062 | /* No room */ |
6063 | continue; | |
a20d2ba5 DW |
6064 | } |
6065 | return dl; | |
272906ef DW |
6066 | } |
6067 | ||
6068 | return dl; | |
6069 | } | |
6070 | ||
95d07a2c LM |
6071 | |
6072 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
6073 | { | |
6074 | struct imsm_dev *dev2; | |
6075 | struct imsm_map *map; | |
6076 | struct dl *idisk; | |
6077 | int slot; | |
6078 | int idx; | |
6079 | __u8 state; | |
6080 | ||
6081 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
6082 | if (dev2) { | |
6083 | state = imsm_check_degraded(cont->sb, dev2, failed); | |
6084 | if (state == IMSM_T_STATE_FAILED) { | |
6085 | map = get_imsm_map(dev2, 0); | |
6086 | if (!map) | |
6087 | return 1; | |
6088 | for (slot = 0; slot < map->num_members; slot++) { | |
6089 | /* | |
6090 | * Check if failed disks are deleted from intel | |
6091 | * disk list or are marked to be deleted | |
6092 | */ | |
98130f40 | 6093 | idx = get_imsm_disk_idx(dev2, slot, -1); |
95d07a2c LM |
6094 | idisk = get_imsm_dl_disk(cont->sb, idx); |
6095 | /* | |
6096 | * Do not rebuild the array if failed disks | |
6097 | * from failed sub-array are not removed from | |
6098 | * container. | |
6099 | */ | |
6100 | if (idisk && | |
6101 | is_failed(&idisk->disk) && | |
6102 | (idisk->action != DISK_REMOVE)) | |
6103 | return 0; | |
6104 | } | |
6105 | } | |
6106 | } | |
6107 | return 1; | |
6108 | } | |
6109 | ||
88758e9d DW |
6110 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
6111 | struct metadata_update **updates) | |
6112 | { | |
6113 | /** | |
d23fe947 DW |
6114 | * Find a device with unused free space and use it to replace a |
6115 | * failed/vacant region in an array. We replace failed regions one a | |
6116 | * array at a time. The result is that a new spare disk will be added | |
6117 | * to the first failed array and after the monitor has finished | |
6118 | * propagating failures the remainder will be consumed. | |
88758e9d | 6119 | * |
d23fe947 DW |
6120 | * FIXME add a capability for mdmon to request spares from another |
6121 | * container. | |
88758e9d DW |
6122 | */ |
6123 | ||
6124 | struct intel_super *super = a->container->sb; | |
88758e9d | 6125 | int inst = a->info.container_member; |
949c47a0 | 6126 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 6127 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
6128 | int failed = a->info.array.raid_disks; |
6129 | struct mdinfo *rv = NULL; | |
6130 | struct mdinfo *d; | |
6131 | struct mdinfo *di; | |
6132 | struct metadata_update *mu; | |
6133 | struct dl *dl; | |
6134 | struct imsm_update_activate_spare *u; | |
6135 | int num_spares = 0; | |
6136 | int i; | |
95d07a2c | 6137 | int allowed; |
88758e9d DW |
6138 | |
6139 | for (d = a->info.devs ; d ; d = d->next) { | |
6140 | if ((d->curr_state & DS_FAULTY) && | |
6141 | d->state_fd >= 0) | |
6142 | /* wait for Removal to happen */ | |
6143 | return NULL; | |
6144 | if (d->state_fd >= 0) | |
6145 | failed--; | |
6146 | } | |
6147 | ||
6148 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
6149 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 AK |
6150 | |
6151 | if (dev->vol.migr_state && | |
6152 | dev->vol.migr_type == MIGR_GEN_MIGR) | |
6153 | /* No repair during migration */ | |
6154 | return NULL; | |
6155 | ||
89c67882 AK |
6156 | if (a->info.array.level == 4) |
6157 | /* No repair for takeovered array | |
6158 | * imsm doesn't support raid4 | |
6159 | */ | |
6160 | return NULL; | |
6161 | ||
fb49eef2 | 6162 | if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
6163 | return NULL; |
6164 | ||
95d07a2c LM |
6165 | /* |
6166 | * If there are any failed disks check state of the other volume. | |
6167 | * Block rebuild if the another one is failed until failed disks | |
6168 | * are removed from container. | |
6169 | */ | |
6170 | if (failed) { | |
6171 | dprintf("found failed disks in %s, check if there another" | |
6172 | "failed sub-array.\n", | |
6173 | dev->volume); | |
6174 | /* check if states of the other volumes allow for rebuild */ | |
6175 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
6176 | if (i != inst) { | |
6177 | allowed = imsm_rebuild_allowed(a->container, | |
6178 | i, failed); | |
6179 | if (!allowed) | |
6180 | return NULL; | |
6181 | } | |
6182 | } | |
6183 | } | |
6184 | ||
88758e9d | 6185 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
6186 | for (i = 0; i < a->info.array.raid_disks; i++) { |
6187 | for (d = a->info.devs ; d ; d = d->next) | |
6188 | if (d->disk.raid_disk == i) | |
6189 | break; | |
6190 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
6191 | if (d && (d->state_fd >= 0)) | |
6192 | continue; | |
6193 | ||
272906ef | 6194 | /* |
a20d2ba5 DW |
6195 | * OK, this device needs recovery. Try to re-add the |
6196 | * previous occupant of this slot, if this fails see if | |
6197 | * we can continue the assimilation of a spare that was | |
6198 | * partially assimilated, finally try to activate a new | |
6199 | * spare. | |
272906ef DW |
6200 | */ |
6201 | dl = imsm_readd(super, i, a); | |
6202 | if (!dl) | |
8ba77d32 | 6203 | dl = imsm_add_spare(super, i, a, 0, NULL); |
a20d2ba5 | 6204 | if (!dl) |
8ba77d32 | 6205 | dl = imsm_add_spare(super, i, a, 1, NULL); |
272906ef DW |
6206 | if (!dl) |
6207 | continue; | |
6208 | ||
6209 | /* found a usable disk with enough space */ | |
6210 | di = malloc(sizeof(*di)); | |
79244939 DW |
6211 | if (!di) |
6212 | continue; | |
272906ef DW |
6213 | memset(di, 0, sizeof(*di)); |
6214 | ||
6215 | /* dl->index will be -1 in the case we are activating a | |
6216 | * pristine spare. imsm_process_update() will create a | |
6217 | * new index in this case. Once a disk is found to be | |
6218 | * failed in all member arrays it is kicked from the | |
6219 | * metadata | |
6220 | */ | |
6221 | di->disk.number = dl->index; | |
d23fe947 | 6222 | |
272906ef DW |
6223 | /* (ab)use di->devs to store a pointer to the device |
6224 | * we chose | |
6225 | */ | |
6226 | di->devs = (struct mdinfo *) dl; | |
6227 | ||
6228 | di->disk.raid_disk = i; | |
6229 | di->disk.major = dl->major; | |
6230 | di->disk.minor = dl->minor; | |
6231 | di->disk.state = 0; | |
d23534e4 | 6232 | di->recovery_start = 0; |
272906ef DW |
6233 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); |
6234 | di->component_size = a->info.component_size; | |
6235 | di->container_member = inst; | |
148acb7b | 6236 | super->random = random32(); |
272906ef DW |
6237 | di->next = rv; |
6238 | rv = di; | |
6239 | num_spares++; | |
6240 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
6241 | i, di->data_offset); | |
88758e9d | 6242 | |
272906ef | 6243 | break; |
88758e9d DW |
6244 | } |
6245 | ||
6246 | if (!rv) | |
6247 | /* No spares found */ | |
6248 | return rv; | |
6249 | /* Now 'rv' has a list of devices to return. | |
6250 | * Create a metadata_update record to update the | |
6251 | * disk_ord_tbl for the array | |
6252 | */ | |
6253 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
6254 | if (mu) { |
6255 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
6256 | if (mu->buf == NULL) { | |
6257 | free(mu); | |
6258 | mu = NULL; | |
6259 | } | |
6260 | } | |
6261 | if (!mu) { | |
6262 | while (rv) { | |
6263 | struct mdinfo *n = rv->next; | |
6264 | ||
6265 | free(rv); | |
6266 | rv = n; | |
6267 | } | |
6268 | return NULL; | |
6269 | } | |
6270 | ||
88758e9d | 6271 | mu->space = NULL; |
cb23f1f4 | 6272 | mu->space_list = NULL; |
88758e9d DW |
6273 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
6274 | mu->next = *updates; | |
6275 | u = (struct imsm_update_activate_spare *) mu->buf; | |
6276 | ||
6277 | for (di = rv ; di ; di = di->next) { | |
6278 | u->type = update_activate_spare; | |
d23fe947 DW |
6279 | u->dl = (struct dl *) di->devs; |
6280 | di->devs = NULL; | |
88758e9d DW |
6281 | u->slot = di->disk.raid_disk; |
6282 | u->array = inst; | |
6283 | u->next = u + 1; | |
6284 | u++; | |
6285 | } | |
6286 | (u-1)->next = NULL; | |
6287 | *updates = mu; | |
6288 | ||
6289 | return rv; | |
6290 | } | |
6291 | ||
54c2c1ea | 6292 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 6293 | { |
54c2c1ea DW |
6294 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
6295 | struct imsm_map *map = get_imsm_map(dev, 0); | |
6296 | struct imsm_map *new_map = get_imsm_map(&u->dev, 0); | |
6297 | struct disk_info *inf = get_disk_info(u); | |
6298 | struct imsm_disk *disk; | |
8273f55e DW |
6299 | int i; |
6300 | int j; | |
8273f55e | 6301 | |
54c2c1ea | 6302 | for (i = 0; i < map->num_members; i++) { |
98130f40 | 6303 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, -1)); |
54c2c1ea DW |
6304 | for (j = 0; j < new_map->num_members; j++) |
6305 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
6306 | return 1; |
6307 | } | |
6308 | ||
6309 | return 0; | |
6310 | } | |
6311 | ||
1a64be56 LM |
6312 | |
6313 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
6314 | { | |
6315 | struct dl *dl = NULL; | |
6316 | for (dl = super->disks; dl; dl = dl->next) | |
6317 | if ((dl->major == major) && (dl->minor == minor)) | |
6318 | return dl; | |
6319 | return NULL; | |
6320 | } | |
6321 | ||
6322 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
6323 | { | |
6324 | struct dl *prev = NULL; | |
6325 | struct dl *dl; | |
6326 | ||
6327 | prev = NULL; | |
6328 | for (dl = super->disks; dl; dl = dl->next) { | |
6329 | if ((dl->major == major) && (dl->minor == minor)) { | |
6330 | /* remove */ | |
6331 | if (prev) | |
6332 | prev->next = dl->next; | |
6333 | else | |
6334 | super->disks = dl->next; | |
6335 | dl->next = NULL; | |
6336 | __free_imsm_disk(dl); | |
6337 | dprintf("%s: removed %x:%x\n", | |
6338 | __func__, major, minor); | |
6339 | break; | |
6340 | } | |
6341 | prev = dl; | |
6342 | } | |
6343 | return 0; | |
6344 | } | |
6345 | ||
f21e18ca | 6346 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 6347 | |
1a64be56 LM |
6348 | static int add_remove_disk_update(struct intel_super *super) |
6349 | { | |
6350 | int check_degraded = 0; | |
6351 | struct dl *disk = NULL; | |
6352 | /* add/remove some spares to/from the metadata/contrainer */ | |
6353 | while (super->disk_mgmt_list) { | |
6354 | struct dl *disk_cfg; | |
6355 | ||
6356 | disk_cfg = super->disk_mgmt_list; | |
6357 | super->disk_mgmt_list = disk_cfg->next; | |
6358 | disk_cfg->next = NULL; | |
6359 | ||
6360 | if (disk_cfg->action == DISK_ADD) { | |
6361 | disk_cfg->next = super->disks; | |
6362 | super->disks = disk_cfg; | |
6363 | check_degraded = 1; | |
6364 | dprintf("%s: added %x:%x\n", | |
6365 | __func__, disk_cfg->major, | |
6366 | disk_cfg->minor); | |
6367 | } else if (disk_cfg->action == DISK_REMOVE) { | |
6368 | dprintf("Disk remove action processed: %x.%x\n", | |
6369 | disk_cfg->major, disk_cfg->minor); | |
6370 | disk = get_disk_super(super, | |
6371 | disk_cfg->major, | |
6372 | disk_cfg->minor); | |
6373 | if (disk) { | |
6374 | /* store action status */ | |
6375 | disk->action = DISK_REMOVE; | |
6376 | /* remove spare disks only */ | |
6377 | if (disk->index == -1) { | |
6378 | remove_disk_super(super, | |
6379 | disk_cfg->major, | |
6380 | disk_cfg->minor); | |
6381 | } | |
6382 | } | |
6383 | /* release allocate disk structure */ | |
6384 | __free_imsm_disk(disk_cfg); | |
6385 | } | |
6386 | } | |
6387 | return check_degraded; | |
6388 | } | |
6389 | ||
a29911da PC |
6390 | |
6391 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
6392 | struct intel_super *super, | |
6393 | void ***space_list) | |
6394 | { | |
6395 | struct intel_dev *id; | |
6396 | void **tofree = NULL; | |
6397 | int ret_val = 0; | |
6398 | ||
6399 | dprintf("apply_reshape_migration_update()\n"); | |
6400 | if ((u->subdev < 0) || | |
6401 | (u->subdev > 1)) { | |
6402 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
6403 | return ret_val; | |
6404 | } | |
6405 | if ((space_list == NULL) || (*space_list == NULL)) { | |
6406 | dprintf("imsm: Error: Memory is not allocated\n"); | |
6407 | return ret_val; | |
6408 | } | |
6409 | ||
6410 | for (id = super->devlist ; id; id = id->next) { | |
6411 | if (id->index == (unsigned)u->subdev) { | |
6412 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
6413 | struct imsm_map *map; | |
6414 | struct imsm_dev *new_dev = | |
6415 | (struct imsm_dev *)*space_list; | |
6416 | struct imsm_map *migr_map = get_imsm_map(dev, 1); | |
6417 | int to_state; | |
6418 | struct dl *new_disk; | |
6419 | ||
6420 | if (new_dev == NULL) | |
6421 | return ret_val; | |
6422 | *space_list = **space_list; | |
6423 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
6424 | map = get_imsm_map(new_dev, 0); | |
6425 | if (migr_map) { | |
6426 | dprintf("imsm: Error: migration in progress"); | |
6427 | return ret_val; | |
6428 | } | |
6429 | ||
6430 | to_state = map->map_state; | |
6431 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
6432 | map->num_members++; | |
6433 | /* this should not happen */ | |
6434 | if (u->new_disks[0] < 0) { | |
6435 | map->failed_disk_num = | |
6436 | map->num_members - 1; | |
6437 | to_state = IMSM_T_STATE_DEGRADED; | |
6438 | } else | |
6439 | to_state = IMSM_T_STATE_NORMAL; | |
6440 | } | |
8e59f3d8 | 6441 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
6442 | if (u->new_level > -1) |
6443 | map->raid_level = u->new_level; | |
6444 | migr_map = get_imsm_map(new_dev, 1); | |
6445 | if ((u->new_level == 5) && | |
6446 | (migr_map->raid_level == 0)) { | |
6447 | int ord = map->num_members - 1; | |
6448 | migr_map->num_members--; | |
6449 | if (u->new_disks[0] < 0) | |
6450 | ord |= IMSM_ORD_REBUILD; | |
6451 | set_imsm_ord_tbl_ent(map, | |
6452 | map->num_members - 1, | |
6453 | ord); | |
6454 | } | |
6455 | id->dev = new_dev; | |
6456 | tofree = (void **)dev; | |
6457 | ||
4bba0439 PC |
6458 | /* update chunk size |
6459 | */ | |
6460 | if (u->new_chunksize > 0) | |
6461 | map->blocks_per_strip = | |
6462 | __cpu_to_le16(u->new_chunksize * 2); | |
6463 | ||
a29911da PC |
6464 | /* add disk |
6465 | */ | |
6466 | if ((u->new_level != 5) || | |
6467 | (migr_map->raid_level != 0) || | |
6468 | (migr_map->raid_level == map->raid_level)) | |
6469 | goto skip_disk_add; | |
6470 | ||
6471 | if (u->new_disks[0] >= 0) { | |
6472 | /* use passes spare | |
6473 | */ | |
6474 | new_disk = get_disk_super(super, | |
6475 | major(u->new_disks[0]), | |
6476 | minor(u->new_disks[0])); | |
6477 | dprintf("imsm: new disk for reshape is: %i:%i " | |
6478 | "(%p, index = %i)\n", | |
6479 | major(u->new_disks[0]), | |
6480 | minor(u->new_disks[0]), | |
6481 | new_disk, new_disk->index); | |
6482 | if (new_disk == NULL) | |
6483 | goto error_disk_add; | |
6484 | ||
6485 | new_disk->index = map->num_members - 1; | |
6486 | /* slot to fill in autolayout | |
6487 | */ | |
6488 | new_disk->raiddisk = new_disk->index; | |
6489 | new_disk->disk.status |= CONFIGURED_DISK; | |
6490 | new_disk->disk.status &= ~SPARE_DISK; | |
6491 | } else | |
6492 | goto error_disk_add; | |
6493 | ||
6494 | skip_disk_add: | |
6495 | *tofree = *space_list; | |
6496 | /* calculate new size | |
6497 | */ | |
6498 | imsm_set_array_size(new_dev); | |
6499 | ||
6500 | ret_val = 1; | |
6501 | } | |
6502 | } | |
6503 | ||
6504 | if (tofree) | |
6505 | *space_list = tofree; | |
6506 | return ret_val; | |
6507 | ||
6508 | error_disk_add: | |
6509 | dprintf("Error: imsm: Cannot find disk.\n"); | |
6510 | return ret_val; | |
6511 | } | |
6512 | ||
6513 | ||
2e5dc010 N |
6514 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
6515 | struct intel_super *super, | |
6516 | void ***space_list) | |
6517 | { | |
6518 | struct dl *new_disk; | |
6519 | struct intel_dev *id; | |
6520 | int i; | |
6521 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 6522 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
6523 | void **tofree = NULL; |
6524 | int devices_to_reshape = 1; | |
6525 | struct imsm_super *mpb = super->anchor; | |
6526 | int ret_val = 0; | |
d098291a | 6527 | unsigned int dev_id; |
2e5dc010 | 6528 | |
ed7333bd | 6529 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
6530 | |
6531 | /* enable spares to use in array */ | |
6532 | for (i = 0; i < delta_disks; i++) { | |
6533 | new_disk = get_disk_super(super, | |
6534 | major(u->new_disks[i]), | |
6535 | minor(u->new_disks[i])); | |
ed7333bd AK |
6536 | dprintf("imsm: new disk for reshape is: %i:%i " |
6537 | "(%p, index = %i)\n", | |
2e5dc010 N |
6538 | major(u->new_disks[i]), minor(u->new_disks[i]), |
6539 | new_disk, new_disk->index); | |
6540 | if ((new_disk == NULL) || | |
6541 | ((new_disk->index >= 0) && | |
6542 | (new_disk->index < u->old_raid_disks))) | |
6543 | goto update_reshape_exit; | |
ee4beede | 6544 | new_disk->index = disk_count++; |
2e5dc010 N |
6545 | /* slot to fill in autolayout |
6546 | */ | |
6547 | new_disk->raiddisk = new_disk->index; | |
6548 | new_disk->disk.status |= | |
6549 | CONFIGURED_DISK; | |
6550 | new_disk->disk.status &= ~SPARE_DISK; | |
6551 | } | |
6552 | ||
ed7333bd AK |
6553 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
6554 | mpb->num_raid_devs); | |
2e5dc010 N |
6555 | /* manage changes in volume |
6556 | */ | |
d098291a | 6557 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
6558 | void **sp = *space_list; |
6559 | struct imsm_dev *newdev; | |
6560 | struct imsm_map *newmap, *oldmap; | |
6561 | ||
d098291a AK |
6562 | for (id = super->devlist ; id; id = id->next) { |
6563 | if (id->index == dev_id) | |
6564 | break; | |
6565 | } | |
6566 | if (id == NULL) | |
6567 | break; | |
2e5dc010 N |
6568 | if (!sp) |
6569 | continue; | |
6570 | *space_list = *sp; | |
6571 | newdev = (void*)sp; | |
6572 | /* Copy the dev, but not (all of) the map */ | |
6573 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
6574 | oldmap = get_imsm_map(id->dev, 0); | |
6575 | newmap = get_imsm_map(newdev, 0); | |
6576 | /* Copy the current map */ | |
6577 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
6578 | /* update one device only | |
6579 | */ | |
6580 | if (devices_to_reshape) { | |
ed7333bd AK |
6581 | dprintf("imsm: modifying subdev: %i\n", |
6582 | id->index); | |
2e5dc010 N |
6583 | devices_to_reshape--; |
6584 | newdev->vol.migr_state = 1; | |
6585 | newdev->vol.curr_migr_unit = 0; | |
6586 | newdev->vol.migr_type = MIGR_GEN_MIGR; | |
6587 | newmap->num_members = u->new_raid_disks; | |
6588 | for (i = 0; i < delta_disks; i++) { | |
6589 | set_imsm_ord_tbl_ent(newmap, | |
6590 | u->old_raid_disks + i, | |
6591 | u->old_raid_disks + i); | |
6592 | } | |
6593 | /* New map is correct, now need to save old map | |
6594 | */ | |
6595 | newmap = get_imsm_map(newdev, 1); | |
6596 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
6597 | ||
70bdf0dc | 6598 | imsm_set_array_size(newdev); |
2e5dc010 N |
6599 | } |
6600 | ||
6601 | sp = (void **)id->dev; | |
6602 | id->dev = newdev; | |
6603 | *sp = tofree; | |
6604 | tofree = sp; | |
8e59f3d8 AK |
6605 | |
6606 | /* Clear migration record */ | |
6607 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 6608 | } |
819bc634 AK |
6609 | if (tofree) |
6610 | *space_list = tofree; | |
2e5dc010 N |
6611 | ret_val = 1; |
6612 | ||
6613 | update_reshape_exit: | |
6614 | ||
6615 | return ret_val; | |
6616 | } | |
6617 | ||
bb025c2f | 6618 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
6619 | struct intel_super *super, |
6620 | void ***space_list) | |
bb025c2f KW |
6621 | { |
6622 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
6623 | struct intel_dev *dv; |
6624 | struct imsm_dev *dev_new; | |
bb025c2f KW |
6625 | struct imsm_map *map; |
6626 | struct dl *dm, *du; | |
8ca6df95 | 6627 | int i; |
bb025c2f KW |
6628 | |
6629 | for (dv = super->devlist; dv; dv = dv->next) | |
6630 | if (dv->index == (unsigned int)u->subarray) { | |
6631 | dev = dv->dev; | |
6632 | break; | |
6633 | } | |
6634 | ||
6635 | if (dev == NULL) | |
6636 | return 0; | |
6637 | ||
6638 | map = get_imsm_map(dev, 0); | |
6639 | ||
6640 | if (u->direction == R10_TO_R0) { | |
43d5ec18 KW |
6641 | /* Number of failed disks must be half of initial disk number */ |
6642 | if (imsm_count_failed(super, dev) != (map->num_members / 2)) | |
6643 | return 0; | |
6644 | ||
bb025c2f KW |
6645 | /* iterate through devices to mark removed disks as spare */ |
6646 | for (dm = super->disks; dm; dm = dm->next) { | |
6647 | if (dm->disk.status & FAILED_DISK) { | |
6648 | int idx = dm->index; | |
6649 | /* update indexes on the disk list */ | |
6650 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
6651 | the index values will end up being correct.... NB */ | |
6652 | for (du = super->disks; du; du = du->next) | |
6653 | if (du->index > idx) | |
6654 | du->index--; | |
6655 | /* mark as spare disk */ | |
6656 | dm->disk.status = SPARE_DISK; | |
6657 | dm->index = -1; | |
6658 | } | |
6659 | } | |
bb025c2f KW |
6660 | /* update map */ |
6661 | map->num_members = map->num_members / 2; | |
6662 | map->map_state = IMSM_T_STATE_NORMAL; | |
6663 | map->num_domains = 1; | |
6664 | map->raid_level = 0; | |
6665 | map->failed_disk_num = -1; | |
6666 | } | |
6667 | ||
8ca6df95 KW |
6668 | if (u->direction == R0_TO_R10) { |
6669 | void **space; | |
6670 | /* update slots in current disk list */ | |
6671 | for (dm = super->disks; dm; dm = dm->next) { | |
6672 | if (dm->index >= 0) | |
6673 | dm->index *= 2; | |
6674 | } | |
6675 | /* create new *missing* disks */ | |
6676 | for (i = 0; i < map->num_members; i++) { | |
6677 | space = *space_list; | |
6678 | if (!space) | |
6679 | continue; | |
6680 | *space_list = *space; | |
6681 | du = (void *)space; | |
6682 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
6683 | du->fd = -1; |
6684 | du->minor = 0; | |
6685 | du->major = 0; | |
6686 | du->index = (i * 2) + 1; | |
6687 | sprintf((char *)du->disk.serial, | |
6688 | " MISSING_%d", du->index); | |
6689 | sprintf((char *)du->serial, | |
6690 | "MISSING_%d", du->index); | |
6691 | du->next = super->missing; | |
6692 | super->missing = du; | |
6693 | } | |
6694 | /* create new dev and map */ | |
6695 | space = *space_list; | |
6696 | if (!space) | |
6697 | return 0; | |
6698 | *space_list = *space; | |
6699 | dev_new = (void *)space; | |
6700 | memcpy(dev_new, dev, sizeof(*dev)); | |
6701 | /* update new map */ | |
6702 | map = get_imsm_map(dev_new, 0); | |
8ca6df95 | 6703 | map->num_members = map->num_members * 2; |
1a2487c2 | 6704 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
6705 | map->num_domains = 2; |
6706 | map->raid_level = 1; | |
6707 | /* replace dev<->dev_new */ | |
6708 | dv->dev = dev_new; | |
6709 | } | |
bb025c2f KW |
6710 | /* update disk order table */ |
6711 | for (du = super->disks; du; du = du->next) | |
6712 | if (du->index >= 0) | |
6713 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 6714 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
6715 | if (du->index >= 0) { |
6716 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
6717 | mark_missing(dev_new, &du->disk, du->index); | |
6718 | } | |
bb025c2f KW |
6719 | |
6720 | return 1; | |
6721 | } | |
6722 | ||
e8319a19 DW |
6723 | static void imsm_process_update(struct supertype *st, |
6724 | struct metadata_update *update) | |
6725 | { | |
6726 | /** | |
6727 | * crack open the metadata_update envelope to find the update record | |
6728 | * update can be one of: | |
d195167d AK |
6729 | * update_reshape_container_disks - all the arrays in the container |
6730 | * are being reshaped to have more devices. We need to mark | |
6731 | * the arrays for general migration and convert selected spares | |
6732 | * into active devices. | |
6733 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
6734 | * device in an array, update the disk_ord_tbl. If this disk is |
6735 | * present in all member arrays then also clear the SPARE_DISK | |
6736 | * flag | |
d195167d AK |
6737 | * update_create_array |
6738 | * update_kill_array | |
6739 | * update_rename_array | |
6740 | * update_add_remove_disk | |
e8319a19 DW |
6741 | */ |
6742 | struct intel_super *super = st->sb; | |
4d7b1503 | 6743 | struct imsm_super *mpb; |
e8319a19 DW |
6744 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
6745 | ||
4d7b1503 DW |
6746 | /* update requires a larger buf but the allocation failed */ |
6747 | if (super->next_len && !super->next_buf) { | |
6748 | super->next_len = 0; | |
6749 | return; | |
6750 | } | |
6751 | ||
6752 | if (super->next_buf) { | |
6753 | memcpy(super->next_buf, super->buf, super->len); | |
6754 | free(super->buf); | |
6755 | super->len = super->next_len; | |
6756 | super->buf = super->next_buf; | |
6757 | ||
6758 | super->next_len = 0; | |
6759 | super->next_buf = NULL; | |
6760 | } | |
6761 | ||
6762 | mpb = super->anchor; | |
6763 | ||
e8319a19 | 6764 | switch (type) { |
bb025c2f KW |
6765 | case update_takeover: { |
6766 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
6767 | if (apply_takeover_update(u, super, &update->space_list)) { |
6768 | imsm_update_version_info(super); | |
bb025c2f | 6769 | super->updates_pending++; |
1a2487c2 | 6770 | } |
bb025c2f KW |
6771 | break; |
6772 | } | |
6773 | ||
78b10e66 | 6774 | case update_reshape_container_disks: { |
d195167d | 6775 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
6776 | if (apply_reshape_container_disks_update( |
6777 | u, super, &update->space_list)) | |
6778 | super->updates_pending++; | |
78b10e66 N |
6779 | break; |
6780 | } | |
48c5303a | 6781 | case update_reshape_migration: { |
a29911da PC |
6782 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
6783 | if (apply_reshape_migration_update( | |
6784 | u, super, &update->space_list)) | |
6785 | super->updates_pending++; | |
48c5303a PC |
6786 | break; |
6787 | } | |
e8319a19 DW |
6788 | case update_activate_spare: { |
6789 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 6790 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 6791 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd | 6792 | struct imsm_map *migr_map; |
e8319a19 DW |
6793 | struct active_array *a; |
6794 | struct imsm_disk *disk; | |
0c046afd | 6795 | __u8 to_state; |
e8319a19 | 6796 | struct dl *dl; |
e8319a19 | 6797 | unsigned int found; |
0c046afd | 6798 | int failed; |
98130f40 | 6799 | int victim = get_imsm_disk_idx(dev, u->slot, -1); |
e8319a19 DW |
6800 | int i; |
6801 | ||
6802 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 6803 | if (dl == u->dl) |
e8319a19 DW |
6804 | break; |
6805 | ||
6806 | if (!dl) { | |
6807 | fprintf(stderr, "error: imsm_activate_spare passed " | |
1f24f035 DW |
6808 | "an unknown disk (index: %d)\n", |
6809 | u->dl->index); | |
e8319a19 DW |
6810 | return; |
6811 | } | |
6812 | ||
6813 | super->updates_pending++; | |
0c046afd DW |
6814 | /* count failures (excluding rebuilds and the victim) |
6815 | * to determine map[0] state | |
6816 | */ | |
6817 | failed = 0; | |
6818 | for (i = 0; i < map->num_members; i++) { | |
6819 | if (i == u->slot) | |
6820 | continue; | |
98130f40 AK |
6821 | disk = get_imsm_disk(super, |
6822 | get_imsm_disk_idx(dev, i, -1)); | |
25ed7e59 | 6823 | if (!disk || is_failed(disk)) |
0c046afd DW |
6824 | failed++; |
6825 | } | |
6826 | ||
d23fe947 DW |
6827 | /* adding a pristine spare, assign a new index */ |
6828 | if (dl->index < 0) { | |
6829 | dl->index = super->anchor->num_disks; | |
6830 | super->anchor->num_disks++; | |
6831 | } | |
d23fe947 | 6832 | disk = &dl->disk; |
f2f27e63 DW |
6833 | disk->status |= CONFIGURED_DISK; |
6834 | disk->status &= ~SPARE_DISK; | |
e8319a19 | 6835 | |
0c046afd DW |
6836 | /* mark rebuild */ |
6837 | to_state = imsm_check_degraded(super, dev, failed); | |
6838 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8e59f3d8 | 6839 | migrate(dev, super, to_state, MIGR_REBUILD); |
0c046afd DW |
6840 | migr_map = get_imsm_map(dev, 1); |
6841 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); | |
6842 | set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD); | |
6843 | ||
148acb7b DW |
6844 | /* update the family_num to mark a new container |
6845 | * generation, being careful to record the existing | |
6846 | * family_num in orig_family_num to clean up after | |
6847 | * earlier mdadm versions that neglected to set it. | |
6848 | */ | |
6849 | if (mpb->orig_family_num == 0) | |
6850 | mpb->orig_family_num = mpb->family_num; | |
6851 | mpb->family_num += super->random; | |
6852 | ||
e8319a19 DW |
6853 | /* count arrays using the victim in the metadata */ |
6854 | found = 0; | |
6855 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 6856 | dev = get_imsm_dev(super, a->info.container_member); |
620b1713 DW |
6857 | map = get_imsm_map(dev, 0); |
6858 | ||
6859 | if (get_imsm_disk_slot(map, victim) >= 0) | |
6860 | found++; | |
e8319a19 DW |
6861 | } |
6862 | ||
24565c9a | 6863 | /* delete the victim if it is no longer being |
e8319a19 DW |
6864 | * utilized anywhere |
6865 | */ | |
e8319a19 | 6866 | if (!found) { |
ae6aad82 | 6867 | struct dl **dlp; |
24565c9a | 6868 | |
47ee5a45 DW |
6869 | /* We know that 'manager' isn't touching anything, |
6870 | * so it is safe to delete | |
6871 | */ | |
24565c9a | 6872 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) |
ae6aad82 DW |
6873 | if ((*dlp)->index == victim) |
6874 | break; | |
47ee5a45 DW |
6875 | |
6876 | /* victim may be on the missing list */ | |
6877 | if (!*dlp) | |
6878 | for (dlp = &super->missing; *dlp; dlp = &(*dlp)->next) | |
6879 | if ((*dlp)->index == victim) | |
6880 | break; | |
24565c9a | 6881 | imsm_delete(super, dlp, victim); |
e8319a19 | 6882 | } |
8273f55e DW |
6883 | break; |
6884 | } | |
6885 | case update_create_array: { | |
6886 | /* someone wants to create a new array, we need to be aware of | |
6887 | * a few races/collisions: | |
6888 | * 1/ 'Create' called by two separate instances of mdadm | |
6889 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
6890 | * devices that have since been assimilated via | |
6891 | * activate_spare. | |
6892 | * In the event this update can not be carried out mdadm will | |
6893 | * (FIX ME) notice that its update did not take hold. | |
6894 | */ | |
6895 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 6896 | struct intel_dev *dv; |
8273f55e DW |
6897 | struct imsm_dev *dev; |
6898 | struct imsm_map *map, *new_map; | |
6899 | unsigned long long start, end; | |
6900 | unsigned long long new_start, new_end; | |
6901 | int i; | |
54c2c1ea DW |
6902 | struct disk_info *inf; |
6903 | struct dl *dl; | |
8273f55e DW |
6904 | |
6905 | /* handle racing creates: first come first serve */ | |
6906 | if (u->dev_idx < mpb->num_raid_devs) { | |
6907 | dprintf("%s: subarray %d already defined\n", | |
6908 | __func__, u->dev_idx); | |
ba2de7ba | 6909 | goto create_error; |
8273f55e DW |
6910 | } |
6911 | ||
6912 | /* check update is next in sequence */ | |
6913 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
6914 | dprintf("%s: can not create array %d expected index %d\n", |
6915 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 6916 | goto create_error; |
8273f55e DW |
6917 | } |
6918 | ||
a965f303 | 6919 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
6920 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
6921 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
54c2c1ea | 6922 | inf = get_disk_info(u); |
8273f55e DW |
6923 | |
6924 | /* handle activate_spare versus create race: | |
6925 | * check to make sure that overlapping arrays do not include | |
6926 | * overalpping disks | |
6927 | */ | |
6928 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 6929 | dev = get_imsm_dev(super, i); |
a965f303 | 6930 | map = get_imsm_map(dev, 0); |
8273f55e DW |
6931 | start = __le32_to_cpu(map->pba_of_lba0); |
6932 | end = start + __le32_to_cpu(map->blocks_per_member); | |
6933 | if ((new_start >= start && new_start <= end) || | |
6934 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
6935 | /* overlap */; |
6936 | else | |
6937 | continue; | |
6938 | ||
6939 | if (disks_overlap(super, i, u)) { | |
8273f55e | 6940 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 6941 | goto create_error; |
8273f55e DW |
6942 | } |
6943 | } | |
8273f55e | 6944 | |
949c47a0 DW |
6945 | /* check that prepare update was successful */ |
6946 | if (!update->space) { | |
6947 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 6948 | goto create_error; |
949c47a0 DW |
6949 | } |
6950 | ||
54c2c1ea DW |
6951 | /* check that all disks are still active before committing |
6952 | * changes. FIXME: could we instead handle this by creating a | |
6953 | * degraded array? That's probably not what the user expects, | |
6954 | * so better to drop this update on the floor. | |
6955 | */ | |
6956 | for (i = 0; i < new_map->num_members; i++) { | |
6957 | dl = serial_to_dl(inf[i].serial, super); | |
6958 | if (!dl) { | |
6959 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 6960 | goto create_error; |
54c2c1ea | 6961 | } |
949c47a0 DW |
6962 | } |
6963 | ||
8273f55e | 6964 | super->updates_pending++; |
54c2c1ea DW |
6965 | |
6966 | /* convert spares to members and fixup ord_tbl */ | |
6967 | for (i = 0; i < new_map->num_members; i++) { | |
6968 | dl = serial_to_dl(inf[i].serial, super); | |
6969 | if (dl->index == -1) { | |
6970 | dl->index = mpb->num_disks; | |
6971 | mpb->num_disks++; | |
6972 | dl->disk.status |= CONFIGURED_DISK; | |
6973 | dl->disk.status &= ~SPARE_DISK; | |
6974 | } | |
6975 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
6976 | } | |
6977 | ||
ba2de7ba DW |
6978 | dv = update->space; |
6979 | dev = dv->dev; | |
949c47a0 DW |
6980 | update->space = NULL; |
6981 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
6982 | dv->index = u->dev_idx; |
6983 | dv->next = super->devlist; | |
6984 | super->devlist = dv; | |
8273f55e | 6985 | mpb->num_raid_devs++; |
8273f55e | 6986 | |
4d1313e9 | 6987 | imsm_update_version_info(super); |
8273f55e | 6988 | break; |
ba2de7ba DW |
6989 | create_error: |
6990 | /* mdmon knows how to release update->space, but not | |
6991 | * ((struct intel_dev *) update->space)->dev | |
6992 | */ | |
6993 | if (update->space) { | |
6994 | dv = update->space; | |
6995 | free(dv->dev); | |
6996 | } | |
8273f55e | 6997 | break; |
e8319a19 | 6998 | } |
33414a01 DW |
6999 | case update_kill_array: { |
7000 | struct imsm_update_kill_array *u = (void *) update->buf; | |
7001 | int victim = u->dev_idx; | |
7002 | struct active_array *a; | |
7003 | struct intel_dev **dp; | |
7004 | struct imsm_dev *dev; | |
7005 | ||
7006 | /* sanity check that we are not affecting the uuid of | |
7007 | * active arrays, or deleting an active array | |
7008 | * | |
7009 | * FIXME when immutable ids are available, but note that | |
7010 | * we'll also need to fixup the invalidated/active | |
7011 | * subarray indexes in mdstat | |
7012 | */ | |
7013 | for (a = st->arrays; a; a = a->next) | |
7014 | if (a->info.container_member >= victim) | |
7015 | break; | |
7016 | /* by definition if mdmon is running at least one array | |
7017 | * is active in the container, so checking | |
7018 | * mpb->num_raid_devs is just extra paranoia | |
7019 | */ | |
7020 | dev = get_imsm_dev(super, victim); | |
7021 | if (a || !dev || mpb->num_raid_devs == 1) { | |
7022 | dprintf("failed to delete subarray-%d\n", victim); | |
7023 | break; | |
7024 | } | |
7025 | ||
7026 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 7027 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
7028 | *dp = (*dp)->next; |
7029 | } else { | |
f21e18ca | 7030 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
7031 | (*dp)->index--; |
7032 | dp = &(*dp)->next; | |
7033 | } | |
7034 | mpb->num_raid_devs--; | |
7035 | super->updates_pending++; | |
7036 | break; | |
7037 | } | |
aa534678 DW |
7038 | case update_rename_array: { |
7039 | struct imsm_update_rename_array *u = (void *) update->buf; | |
7040 | char name[MAX_RAID_SERIAL_LEN+1]; | |
7041 | int target = u->dev_idx; | |
7042 | struct active_array *a; | |
7043 | struct imsm_dev *dev; | |
7044 | ||
7045 | /* sanity check that we are not affecting the uuid of | |
7046 | * an active array | |
7047 | */ | |
7048 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
7049 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
7050 | for (a = st->arrays; a; a = a->next) | |
7051 | if (a->info.container_member == target) | |
7052 | break; | |
7053 | dev = get_imsm_dev(super, u->dev_idx); | |
7054 | if (a || !dev || !check_name(super, name, 1)) { | |
7055 | dprintf("failed to rename subarray-%d\n", target); | |
7056 | break; | |
7057 | } | |
7058 | ||
cdbe98cd | 7059 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
7060 | super->updates_pending++; |
7061 | break; | |
7062 | } | |
1a64be56 | 7063 | case update_add_remove_disk: { |
43dad3d6 | 7064 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
7065 | * being added, check teh status of add_remove_disk |
7066 | * if discs has been added. | |
7067 | */ | |
7068 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 7069 | struct active_array *a; |
072b727f DW |
7070 | |
7071 | super->updates_pending++; | |
1a64be56 | 7072 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
7073 | a->check_degraded = 1; |
7074 | } | |
43dad3d6 | 7075 | break; |
e8319a19 | 7076 | } |
1a64be56 LM |
7077 | default: |
7078 | fprintf(stderr, "error: unsuported process update type:" | |
7079 | "(type: %d)\n", type); | |
7080 | } | |
e8319a19 | 7081 | } |
88758e9d | 7082 | |
bc0b9d34 PC |
7083 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
7084 | ||
8273f55e DW |
7085 | static void imsm_prepare_update(struct supertype *st, |
7086 | struct metadata_update *update) | |
7087 | { | |
949c47a0 | 7088 | /** |
4d7b1503 DW |
7089 | * Allocate space to hold new disk entries, raid-device entries or a new |
7090 | * mpb if necessary. The manager synchronously waits for updates to | |
7091 | * complete in the monitor, so new mpb buffers allocated here can be | |
7092 | * integrated by the monitor thread without worrying about live pointers | |
7093 | * in the manager thread. | |
8273f55e | 7094 | */ |
949c47a0 | 7095 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
7096 | struct intel_super *super = st->sb; |
7097 | struct imsm_super *mpb = super->anchor; | |
7098 | size_t buf_len; | |
7099 | size_t len = 0; | |
949c47a0 DW |
7100 | |
7101 | switch (type) { | |
abedf5fc KW |
7102 | case update_takeover: { |
7103 | struct imsm_update_takeover *u = (void *)update->buf; | |
7104 | if (u->direction == R0_TO_R10) { | |
7105 | void **tail = (void **)&update->space_list; | |
7106 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
7107 | struct imsm_map *map = get_imsm_map(dev, 0); | |
7108 | int num_members = map->num_members; | |
7109 | void *space; | |
7110 | int size, i; | |
7111 | int err = 0; | |
7112 | /* allocate memory for added disks */ | |
7113 | for (i = 0; i < num_members; i++) { | |
7114 | size = sizeof(struct dl); | |
7115 | space = malloc(size); | |
7116 | if (!space) { | |
7117 | err++; | |
7118 | break; | |
7119 | } | |
7120 | *tail = space; | |
7121 | tail = space; | |
7122 | *tail = NULL; | |
7123 | } | |
7124 | /* allocate memory for new device */ | |
7125 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
7126 | (num_members * sizeof(__u32)); | |
7127 | space = malloc(size); | |
7128 | if (!space) | |
7129 | err++; | |
7130 | else { | |
7131 | *tail = space; | |
7132 | tail = space; | |
7133 | *tail = NULL; | |
7134 | } | |
7135 | if (!err) { | |
7136 | len = disks_to_mpb_size(num_members * 2); | |
7137 | } else { | |
7138 | /* if allocation didn't success, free buffer */ | |
7139 | while (update->space_list) { | |
7140 | void **sp = update->space_list; | |
7141 | update->space_list = *sp; | |
7142 | free(sp); | |
7143 | } | |
7144 | } | |
7145 | } | |
7146 | ||
7147 | break; | |
7148 | } | |
78b10e66 | 7149 | case update_reshape_container_disks: { |
d195167d AK |
7150 | /* Every raid device in the container is about to |
7151 | * gain some more devices, and we will enter a | |
7152 | * reconfiguration. | |
7153 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
7154 | * will now hold 2 of them. | |
7155 | * Thus we need new 'struct imsm_dev' allocations sized | |
7156 | * as sizeof_imsm_dev but with more devices in both maps. | |
7157 | */ | |
7158 | struct imsm_update_reshape *u = (void *)update->buf; | |
7159 | struct intel_dev *dl; | |
7160 | void **space_tail = (void**)&update->space_list; | |
7161 | ||
7162 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7163 | ||
7164 | for (dl = super->devlist; dl; dl = dl->next) { | |
7165 | int size = sizeof_imsm_dev(dl->dev, 1); | |
7166 | void *s; | |
d677e0b8 AK |
7167 | if (u->new_raid_disks > u->old_raid_disks) |
7168 | size += sizeof(__u32)*2* | |
7169 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
7170 | s = malloc(size); |
7171 | if (!s) | |
7172 | break; | |
7173 | *space_tail = s; | |
7174 | space_tail = s; | |
7175 | *space_tail = NULL; | |
7176 | } | |
7177 | ||
7178 | len = disks_to_mpb_size(u->new_raid_disks); | |
7179 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
7180 | break; |
7181 | } | |
48c5303a | 7182 | case update_reshape_migration: { |
bc0b9d34 PC |
7183 | /* for migration level 0->5 we need to add disks |
7184 | * so the same as for container operation we will copy | |
7185 | * device to the bigger location. | |
7186 | * in memory prepared device and new disk area are prepared | |
7187 | * for usage in process update | |
7188 | */ | |
7189 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
7190 | struct intel_dev *id; | |
7191 | void **space_tail = (void **)&update->space_list; | |
7192 | int size; | |
7193 | void *s; | |
7194 | int current_level = -1; | |
7195 | ||
7196 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7197 | ||
7198 | /* add space for bigger array in update | |
7199 | */ | |
7200 | for (id = super->devlist; id; id = id->next) { | |
7201 | if (id->index == (unsigned)u->subdev) { | |
7202 | size = sizeof_imsm_dev(id->dev, 1); | |
7203 | if (u->new_raid_disks > u->old_raid_disks) | |
7204 | size += sizeof(__u32)*2* | |
7205 | (u->new_raid_disks - u->old_raid_disks); | |
7206 | s = malloc(size); | |
7207 | if (!s) | |
7208 | break; | |
7209 | *space_tail = s; | |
7210 | space_tail = s; | |
7211 | *space_tail = NULL; | |
7212 | break; | |
7213 | } | |
7214 | } | |
7215 | if (update->space_list == NULL) | |
7216 | break; | |
7217 | ||
7218 | /* add space for disk in update | |
7219 | */ | |
7220 | size = sizeof(struct dl); | |
7221 | s = malloc(size); | |
7222 | if (!s) { | |
7223 | free(update->space_list); | |
7224 | update->space_list = NULL; | |
7225 | break; | |
7226 | } | |
7227 | *space_tail = s; | |
7228 | space_tail = s; | |
7229 | *space_tail = NULL; | |
7230 | ||
7231 | /* add spare device to update | |
7232 | */ | |
7233 | for (id = super->devlist ; id; id = id->next) | |
7234 | if (id->index == (unsigned)u->subdev) { | |
7235 | struct imsm_dev *dev; | |
7236 | struct imsm_map *map; | |
7237 | ||
7238 | dev = get_imsm_dev(super, u->subdev); | |
7239 | map = get_imsm_map(dev, 0); | |
7240 | current_level = map->raid_level; | |
7241 | break; | |
7242 | } | |
7243 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
7244 | struct mdinfo *spares; | |
7245 | ||
7246 | spares = get_spares_for_grow(st); | |
7247 | if (spares) { | |
7248 | struct dl *dl; | |
7249 | struct mdinfo *dev; | |
7250 | ||
7251 | dev = spares->devs; | |
7252 | if (dev) { | |
7253 | u->new_disks[0] = | |
7254 | makedev(dev->disk.major, | |
7255 | dev->disk.minor); | |
7256 | dl = get_disk_super(super, | |
7257 | dev->disk.major, | |
7258 | dev->disk.minor); | |
7259 | dl->index = u->old_raid_disks; | |
7260 | dev = dev->next; | |
7261 | } | |
7262 | sysfs_free(spares); | |
7263 | } | |
7264 | } | |
7265 | len = disks_to_mpb_size(u->new_raid_disks); | |
7266 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
7267 | break; |
7268 | } | |
949c47a0 DW |
7269 | case update_create_array: { |
7270 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 7271 | struct intel_dev *dv; |
54c2c1ea DW |
7272 | struct imsm_dev *dev = &u->dev; |
7273 | struct imsm_map *map = get_imsm_map(dev, 0); | |
7274 | struct dl *dl; | |
7275 | struct disk_info *inf; | |
7276 | int i; | |
7277 | int activate = 0; | |
949c47a0 | 7278 | |
54c2c1ea DW |
7279 | inf = get_disk_info(u); |
7280 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
7281 | /* allocate a new super->devlist entry */ |
7282 | dv = malloc(sizeof(*dv)); | |
7283 | if (dv) { | |
7284 | dv->dev = malloc(len); | |
7285 | if (dv->dev) | |
7286 | update->space = dv; | |
7287 | else { | |
7288 | free(dv); | |
7289 | update->space = NULL; | |
7290 | } | |
7291 | } | |
949c47a0 | 7292 | |
54c2c1ea DW |
7293 | /* count how many spares will be converted to members */ |
7294 | for (i = 0; i < map->num_members; i++) { | |
7295 | dl = serial_to_dl(inf[i].serial, super); | |
7296 | if (!dl) { | |
7297 | /* hmm maybe it failed?, nothing we can do about | |
7298 | * it here | |
7299 | */ | |
7300 | continue; | |
7301 | } | |
7302 | if (count_memberships(dl, super) == 0) | |
7303 | activate++; | |
7304 | } | |
7305 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
7306 | break; |
7307 | default: | |
7308 | break; | |
7309 | } | |
7310 | } | |
8273f55e | 7311 | |
4d7b1503 DW |
7312 | /* check if we need a larger metadata buffer */ |
7313 | if (super->next_buf) | |
7314 | buf_len = super->next_len; | |
7315 | else | |
7316 | buf_len = super->len; | |
7317 | ||
7318 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
7319 | /* ok we need a larger buf than what is currently allocated | |
7320 | * if this allocation fails process_update will notice that | |
7321 | * ->next_len is set and ->next_buf is NULL | |
7322 | */ | |
7323 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
7324 | if (super->next_buf) | |
7325 | free(super->next_buf); | |
7326 | ||
7327 | super->next_len = buf_len; | |
1f45a8ad DW |
7328 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
7329 | memset(super->next_buf, 0, buf_len); | |
7330 | else | |
4d7b1503 DW |
7331 | super->next_buf = NULL; |
7332 | } | |
8273f55e DW |
7333 | } |
7334 | ||
ae6aad82 | 7335 | /* must be called while manager is quiesced */ |
f21e18ca | 7336 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
7337 | { |
7338 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
7339 | struct dl *iter; |
7340 | struct imsm_dev *dev; | |
7341 | struct imsm_map *map; | |
24565c9a DW |
7342 | int i, j, num_members; |
7343 | __u32 ord; | |
ae6aad82 | 7344 | |
24565c9a DW |
7345 | dprintf("%s: deleting device[%d] from imsm_super\n", |
7346 | __func__, index); | |
ae6aad82 DW |
7347 | |
7348 | /* shift all indexes down one */ | |
7349 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 7350 | if (iter->index > (int)index) |
ae6aad82 | 7351 | iter->index--; |
47ee5a45 | 7352 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 7353 | if (iter->index > (int)index) |
47ee5a45 | 7354 | iter->index--; |
ae6aad82 DW |
7355 | |
7356 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7357 | dev = get_imsm_dev(super, i); | |
7358 | map = get_imsm_map(dev, 0); | |
24565c9a DW |
7359 | num_members = map->num_members; |
7360 | for (j = 0; j < num_members; j++) { | |
7361 | /* update ord entries being careful not to propagate | |
7362 | * ord-flags to the first map | |
7363 | */ | |
98130f40 | 7364 | ord = get_imsm_ord_tbl_ent(dev, j, -1); |
ae6aad82 | 7365 | |
24565c9a DW |
7366 | if (ord_to_idx(ord) <= index) |
7367 | continue; | |
ae6aad82 | 7368 | |
24565c9a DW |
7369 | map = get_imsm_map(dev, 0); |
7370 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); | |
7371 | map = get_imsm_map(dev, 1); | |
7372 | if (map) | |
7373 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
7374 | } |
7375 | } | |
7376 | ||
7377 | mpb->num_disks--; | |
7378 | super->updates_pending++; | |
24565c9a DW |
7379 | if (*dlp) { |
7380 | struct dl *dl = *dlp; | |
7381 | ||
7382 | *dlp = (*dlp)->next; | |
7383 | __free_imsm_disk(dl); | |
7384 | } | |
ae6aad82 DW |
7385 | } |
7386 | ||
687629c2 AK |
7387 | /******************************************************************************* |
7388 | * Function: open_backup_targets | |
7389 | * Description: Function opens file descriptors for all devices given in | |
7390 | * info->devs | |
7391 | * Parameters: | |
7392 | * info : general array info | |
7393 | * raid_disks : number of disks | |
7394 | * raid_fds : table of device's file descriptors | |
7395 | * Returns: | |
7396 | * 0 : success | |
7397 | * -1 : fail | |
7398 | ******************************************************************************/ | |
7399 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds) | |
7400 | { | |
7401 | struct mdinfo *sd; | |
7402 | ||
7403 | for (sd = info->devs ; sd ; sd = sd->next) { | |
7404 | char *dn; | |
7405 | ||
7406 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
7407 | dprintf("disk is faulty!!\n"); | |
7408 | continue; | |
7409 | } | |
7410 | ||
7411 | if ((sd->disk.raid_disk >= raid_disks) || | |
7412 | (sd->disk.raid_disk < 0)) | |
7413 | continue; | |
7414 | ||
7415 | dn = map_dev(sd->disk.major, | |
7416 | sd->disk.minor, 1); | |
7417 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
7418 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
7419 | fprintf(stderr, "cannot open component\n"); | |
7420 | return -1; | |
7421 | } | |
7422 | } | |
7423 | return 0; | |
7424 | } | |
7425 | ||
7426 | /******************************************************************************* | |
7427 | * Function: init_migr_record_imsm | |
7428 | * Description: Function inits imsm migration record | |
7429 | * Parameters: | |
7430 | * super : imsm internal array info | |
7431 | * dev : device under migration | |
7432 | * info : general array info to find the smallest device | |
7433 | * Returns: | |
7434 | * none | |
7435 | ******************************************************************************/ | |
7436 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
7437 | struct mdinfo *info) | |
7438 | { | |
7439 | struct intel_super *super = st->sb; | |
7440 | struct migr_record *migr_rec = super->migr_rec; | |
7441 | int new_data_disks; | |
7442 | unsigned long long dsize, dev_sectors; | |
7443 | long long unsigned min_dev_sectors = -1LLU; | |
7444 | struct mdinfo *sd; | |
7445 | char nm[30]; | |
7446 | int fd; | |
7447 | struct imsm_map *map_dest = get_imsm_map(dev, 0); | |
7448 | struct imsm_map *map_src = get_imsm_map(dev, 1); | |
7449 | unsigned long long num_migr_units; | |
7450 | ||
7451 | unsigned long long array_blocks = | |
7452 | (((unsigned long long)__le32_to_cpu(dev->size_high)) << 32) + | |
7453 | __le32_to_cpu(dev->size_low); | |
7454 | ||
7455 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
7456 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
7457 | ||
7458 | /* only ascending reshape supported now */ | |
7459 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
7460 | ||
7461 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
7462 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
7463 | migr_rec->dest_depth_per_unit *= map_dest->blocks_per_strip; | |
7464 | new_data_disks = imsm_num_data_members(dev, 0); | |
7465 | migr_rec->blocks_per_unit = | |
7466 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
7467 | migr_rec->dest_depth_per_unit = | |
7468 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
7469 | ||
7470 | num_migr_units = | |
7471 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
7472 | ||
7473 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
7474 | num_migr_units++; | |
7475 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
7476 | ||
7477 | migr_rec->post_migr_vol_cap = dev->size_low; | |
7478 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
7479 | ||
7480 | ||
7481 | /* Find the smallest dev */ | |
7482 | for (sd = info->devs ; sd ; sd = sd->next) { | |
7483 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
7484 | fd = dev_open(nm, O_RDONLY); | |
7485 | if (fd < 0) | |
7486 | continue; | |
7487 | get_dev_size(fd, NULL, &dsize); | |
7488 | dev_sectors = dsize / 512; | |
7489 | if (dev_sectors < min_dev_sectors) | |
7490 | min_dev_sectors = dev_sectors; | |
7491 | close(fd); | |
7492 | } | |
7493 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
7494 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
7495 | ||
7496 | write_imsm_migr_rec(st); | |
7497 | ||
7498 | return; | |
7499 | } | |
7500 | ||
7501 | /******************************************************************************* | |
7502 | * Function: save_backup_imsm | |
7503 | * Description: Function saves critical data stripes to Migration Copy Area | |
7504 | * and updates the current migration unit status. | |
7505 | * Use restore_stripes() to form a destination stripe, | |
7506 | * and to write it to the Copy Area. | |
7507 | * Parameters: | |
7508 | * st : supertype information | |
7509 | * info : general array info | |
7510 | * buf : input buffer | |
7511 | * write_offset : address of data to backup | |
7512 | * length : length of data to backup (blocks_per_unit) | |
7513 | * Returns: | |
7514 | * 0 : success | |
7515 | *, -1 : fail | |
7516 | ******************************************************************************/ | |
7517 | int save_backup_imsm(struct supertype *st, | |
7518 | struct imsm_dev *dev, | |
7519 | struct mdinfo *info, | |
7520 | void *buf, | |
7521 | int new_data, | |
7522 | int length) | |
7523 | { | |
7524 | int rv = -1; | |
7525 | struct intel_super *super = st->sb; | |
7526 | unsigned long long *target_offsets = NULL; | |
7527 | int *targets = NULL; | |
7528 | int i; | |
7529 | struct imsm_map *map_dest = get_imsm_map(dev, 0); | |
7530 | int new_disks = map_dest->num_members; | |
7531 | ||
7532 | targets = malloc(new_disks * sizeof(int)); | |
7533 | if (!targets) | |
7534 | goto abort; | |
7535 | ||
7536 | target_offsets = malloc(new_disks * sizeof(unsigned long long)); | |
7537 | if (!target_offsets) | |
7538 | goto abort; | |
7539 | ||
7540 | for (i = 0; i < new_disks; i++) { | |
7541 | targets[i] = -1; | |
7542 | target_offsets[i] = (unsigned long long) | |
7543 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
7544 | } | |
7545 | ||
7546 | if (open_backup_targets(info, new_disks, targets)) | |
7547 | goto abort; | |
7548 | ||
7549 | if (restore_stripes(targets, /* list of dest devices */ | |
7550 | target_offsets, /* migration record offsets */ | |
7551 | new_disks, | |
7552 | info->new_chunk, | |
7553 | info->new_level, | |
7554 | info->new_layout, | |
7555 | -1, /* source backup file descriptor */ | |
7556 | 0, /* input buf offset | |
7557 | * always 0 buf is already offset */ | |
7558 | 0, | |
7559 | length, | |
7560 | buf) != 0) { | |
7561 | fprintf(stderr, Name ": Error restoring stripes\n"); | |
7562 | goto abort; | |
7563 | } | |
7564 | ||
7565 | rv = 0; | |
7566 | ||
7567 | abort: | |
7568 | if (targets) { | |
7569 | for (i = 0; i < new_disks; i++) | |
7570 | if (targets[i] >= 0) | |
7571 | close(targets[i]); | |
7572 | free(targets); | |
7573 | } | |
7574 | free(target_offsets); | |
7575 | ||
7576 | return rv; | |
7577 | } | |
7578 | ||
7579 | /******************************************************************************* | |
7580 | * Function: save_checkpoint_imsm | |
7581 | * Description: Function called for current unit status update | |
7582 | * in the migration record. It writes it to disk. | |
7583 | * Parameters: | |
7584 | * super : imsm internal array info | |
7585 | * info : general array info | |
7586 | * Returns: | |
7587 | * 0: success | |
7588 | * 1: failure | |
7589 | ******************************************************************************/ | |
7590 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
7591 | { | |
7592 | struct intel_super *super = st->sb; | |
7593 | load_imsm_migr_rec(super, info); | |
7594 | if (__le32_to_cpu(super->migr_rec->blocks_per_unit) == 0) { | |
7595 | dprintf("ERROR: blocks_per_unit = 0!!!\n"); | |
7596 | return 1; | |
7597 | } | |
7598 | ||
7599 | super->migr_rec->curr_migr_unit = | |
7600 | __cpu_to_le32(info->reshape_progress / | |
7601 | __le32_to_cpu(super->migr_rec->blocks_per_unit)); | |
7602 | super->migr_rec->rec_status = __cpu_to_le32(state); | |
7603 | super->migr_rec->dest_1st_member_lba = | |
7604 | __cpu_to_le32((__le32_to_cpu(super->migr_rec->curr_migr_unit)) | |
7605 | * __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
7606 | if (write_imsm_migr_rec(st) < 0) { | |
7607 | dprintf("imsm: Cannot write migration record " | |
7608 | "outside backup area\n"); | |
7609 | return 1; | |
7610 | } | |
7611 | ||
7612 | return 0; | |
7613 | } | |
7614 | ||
2cda7640 ML |
7615 | static char disk_by_path[] = "/dev/disk/by-path/"; |
7616 | ||
7617 | static const char *imsm_get_disk_controller_domain(const char *path) | |
7618 | { | |
2cda7640 | 7619 | char disk_path[PATH_MAX]; |
96234762 LM |
7620 | char *drv=NULL; |
7621 | struct stat st; | |
2cda7640 | 7622 | |
96234762 LM |
7623 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
7624 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
7625 | if (stat(disk_path, &st) == 0) { | |
7626 | struct sys_dev* hba; | |
7627 | char *path=NULL; | |
7628 | ||
7629 | path = devt_to_devpath(st.st_rdev); | |
7630 | if (path == NULL) | |
7631 | return "unknown"; | |
7632 | hba = find_disk_attached_hba(-1, path); | |
7633 | if (hba && hba->type == SYS_DEV_SAS) | |
7634 | drv = "isci"; | |
7635 | else if (hba && hba->type == SYS_DEV_SATA) | |
7636 | drv = "ahci"; | |
7637 | else | |
7638 | drv = "unknown"; | |
7639 | dprintf("path: %s hba: %s attached: %s\n", | |
7640 | path, (hba) ? hba->path : "NULL", drv); | |
7641 | free(path); | |
7642 | if (hba) | |
7643 | free_sys_dev(&hba); | |
2cda7640 | 7644 | } |
96234762 | 7645 | return drv; |
2cda7640 ML |
7646 | } |
7647 | ||
78b10e66 N |
7648 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
7649 | { | |
7650 | char subdev_name[20]; | |
7651 | struct mdstat_ent *mdstat; | |
7652 | ||
7653 | sprintf(subdev_name, "%d", subdev); | |
7654 | mdstat = mdstat_by_subdev(subdev_name, container); | |
7655 | if (!mdstat) | |
7656 | return -1; | |
7657 | ||
7658 | *minor = mdstat->devnum; | |
7659 | free_mdstat(mdstat); | |
7660 | return 0; | |
7661 | } | |
7662 | ||
7663 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
7664 | struct geo_params *geo, | |
7665 | int *old_raid_disks) | |
7666 | { | |
694575e7 KW |
7667 | /* currently we only support increasing the number of devices |
7668 | * for a container. This increases the number of device for each | |
7669 | * member array. They must all be RAID0 or RAID5. | |
7670 | */ | |
78b10e66 N |
7671 | int ret_val = 0; |
7672 | struct mdinfo *info, *member; | |
7673 | int devices_that_can_grow = 0; | |
7674 | ||
7675 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
7676 | "st->devnum = (%i)\n", | |
7677 | st->devnum); | |
7678 | ||
7679 | if (geo->size != -1 || | |
7680 | geo->level != UnSet || | |
7681 | geo->layout != UnSet || | |
7682 | geo->chunksize != 0 || | |
7683 | geo->raid_disks == UnSet) { | |
7684 | dprintf("imsm: Container operation is allowed for " | |
7685 | "raid disks number change only.\n"); | |
7686 | return ret_val; | |
7687 | } | |
7688 | ||
7689 | info = container_content_imsm(st, NULL); | |
7690 | for (member = info; member; member = member->next) { | |
7691 | int result; | |
7692 | int minor; | |
7693 | ||
7694 | dprintf("imsm: checking device_num: %i\n", | |
7695 | member->container_member); | |
7696 | ||
d7d205bd | 7697 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
7698 | /* we work on container for Online Capacity Expansion |
7699 | * only so raid_disks has to grow | |
7700 | */ | |
7701 | dprintf("imsm: for container operation raid disks " | |
7702 | "increase is required\n"); | |
7703 | break; | |
7704 | } | |
7705 | ||
7706 | if ((info->array.level != 0) && | |
7707 | (info->array.level != 5)) { | |
7708 | /* we cannot use this container with other raid level | |
7709 | */ | |
690aae1a | 7710 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
7711 | " raid level (%i) detected\n", |
7712 | info->array.level); | |
7713 | break; | |
7714 | } else { | |
7715 | /* check for platform support | |
7716 | * for this raid level configuration | |
7717 | */ | |
7718 | struct intel_super *super = st->sb; | |
7719 | if (!is_raid_level_supported(super->orom, | |
7720 | member->array.level, | |
7721 | geo->raid_disks)) { | |
690aae1a | 7722 | dprintf("platform does not support raid%d with" |
78b10e66 N |
7723 | " %d disk%s\n", |
7724 | info->array.level, | |
7725 | geo->raid_disks, | |
7726 | geo->raid_disks > 1 ? "s" : ""); | |
7727 | break; | |
7728 | } | |
2a4a08e7 AK |
7729 | /* check if component size is aligned to chunk size |
7730 | */ | |
7731 | if (info->component_size % | |
7732 | (info->array.chunk_size/512)) { | |
7733 | dprintf("Component size is not aligned to " | |
7734 | "chunk size\n"); | |
7735 | break; | |
7736 | } | |
78b10e66 N |
7737 | } |
7738 | ||
7739 | if (*old_raid_disks && | |
7740 | info->array.raid_disks != *old_raid_disks) | |
7741 | break; | |
7742 | *old_raid_disks = info->array.raid_disks; | |
7743 | ||
7744 | /* All raid5 and raid0 volumes in container | |
7745 | * have to be ready for Online Capacity Expansion | |
7746 | * so they need to be assembled. We have already | |
7747 | * checked that no recovery etc is happening. | |
7748 | */ | |
7749 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
7750 | st->container_dev, | |
7751 | &minor); | |
7752 | if (result < 0) { | |
7753 | dprintf("imsm: cannot find array\n"); | |
7754 | break; | |
7755 | } | |
7756 | devices_that_can_grow++; | |
7757 | } | |
7758 | sysfs_free(info); | |
7759 | if (!member && devices_that_can_grow) | |
7760 | ret_val = 1; | |
7761 | ||
7762 | if (ret_val) | |
7763 | dprintf("\tContainer operation allowed\n"); | |
7764 | else | |
7765 | dprintf("\tError: %i\n", ret_val); | |
7766 | ||
7767 | return ret_val; | |
7768 | } | |
7769 | ||
7770 | /* Function: get_spares_for_grow | |
7771 | * Description: Allocates memory and creates list of spare devices | |
7772 | * avaliable in container. Checks if spare drive size is acceptable. | |
7773 | * Parameters: Pointer to the supertype structure | |
7774 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
7775 | * NULL if fail | |
7776 | */ | |
7777 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
7778 | { | |
78b10e66 | 7779 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 7780 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
7781 | } |
7782 | ||
7783 | /****************************************************************************** | |
7784 | * function: imsm_create_metadata_update_for_reshape | |
7785 | * Function creates update for whole IMSM container. | |
7786 | * | |
7787 | ******************************************************************************/ | |
7788 | static int imsm_create_metadata_update_for_reshape( | |
7789 | struct supertype *st, | |
7790 | struct geo_params *geo, | |
7791 | int old_raid_disks, | |
7792 | struct imsm_update_reshape **updatep) | |
7793 | { | |
7794 | struct intel_super *super = st->sb; | |
7795 | struct imsm_super *mpb = super->anchor; | |
7796 | int update_memory_size = 0; | |
7797 | struct imsm_update_reshape *u = NULL; | |
7798 | struct mdinfo *spares = NULL; | |
7799 | int i; | |
7800 | int delta_disks = 0; | |
bbd24d86 | 7801 | struct mdinfo *dev; |
78b10e66 N |
7802 | |
7803 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
7804 | geo->raid_disks); | |
7805 | ||
7806 | delta_disks = geo->raid_disks - old_raid_disks; | |
7807 | ||
7808 | /* size of all update data without anchor */ | |
7809 | update_memory_size = sizeof(struct imsm_update_reshape); | |
7810 | ||
7811 | /* now add space for spare disks that we need to add. */ | |
7812 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
7813 | ||
7814 | u = calloc(1, update_memory_size); | |
7815 | if (u == NULL) { | |
7816 | dprintf("error: " | |
7817 | "cannot get memory for imsm_update_reshape update\n"); | |
7818 | return 0; | |
7819 | } | |
7820 | u->type = update_reshape_container_disks; | |
7821 | u->old_raid_disks = old_raid_disks; | |
7822 | u->new_raid_disks = geo->raid_disks; | |
7823 | ||
7824 | /* now get spare disks list | |
7825 | */ | |
7826 | spares = get_spares_for_grow(st); | |
7827 | ||
7828 | if (spares == NULL | |
7829 | || delta_disks > spares->array.spare_disks) { | |
e14e5960 KW |
7830 | fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices " |
7831 | "for %s.\n", geo->dev_name); | |
78b10e66 N |
7832 | goto abort; |
7833 | } | |
7834 | ||
7835 | /* we have got spares | |
7836 | * update disk list in imsm_disk list table in anchor | |
7837 | */ | |
7838 | dprintf("imsm: %i spares are available.\n\n", | |
7839 | spares->array.spare_disks); | |
7840 | ||
bbd24d86 | 7841 | dev = spares->devs; |
78b10e66 | 7842 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
7843 | struct dl *dl; |
7844 | ||
bbd24d86 AK |
7845 | if (dev == NULL) |
7846 | break; | |
78b10e66 N |
7847 | u->new_disks[i] = makedev(dev->disk.major, |
7848 | dev->disk.minor); | |
7849 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
7850 | dl->index = mpb->num_disks; |
7851 | mpb->num_disks++; | |
bbd24d86 | 7852 | dev = dev->next; |
78b10e66 | 7853 | } |
78b10e66 N |
7854 | |
7855 | abort: | |
7856 | /* free spares | |
7857 | */ | |
7858 | sysfs_free(spares); | |
7859 | ||
d677e0b8 | 7860 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 7861 | if (i == delta_disks) { |
d677e0b8 | 7862 | dprintf(" OK\n"); |
78b10e66 N |
7863 | *updatep = u; |
7864 | return update_memory_size; | |
7865 | } | |
7866 | free(u); | |
d677e0b8 | 7867 | dprintf(" Error\n"); |
78b10e66 N |
7868 | |
7869 | return 0; | |
7870 | } | |
7871 | ||
48c5303a PC |
7872 | /****************************************************************************** |
7873 | * function: imsm_create_metadata_update_for_migration() | |
7874 | * Creates update for IMSM array. | |
7875 | * | |
7876 | ******************************************************************************/ | |
7877 | static int imsm_create_metadata_update_for_migration( | |
7878 | struct supertype *st, | |
7879 | struct geo_params *geo, | |
7880 | struct imsm_update_reshape_migration **updatep) | |
7881 | { | |
7882 | struct intel_super *super = st->sb; | |
7883 | int update_memory_size = 0; | |
7884 | struct imsm_update_reshape_migration *u = NULL; | |
7885 | struct imsm_dev *dev; | |
7886 | int previous_level = -1; | |
7887 | ||
7888 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
7889 | " New Level = %i\n", geo->level); | |
7890 | ||
7891 | /* size of all update data without anchor */ | |
7892 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
7893 | ||
7894 | u = calloc(1, update_memory_size); | |
7895 | if (u == NULL) { | |
7896 | dprintf("error: cannot get memory for " | |
7897 | "imsm_create_metadata_update_for_migration\n"); | |
7898 | return 0; | |
7899 | } | |
7900 | u->type = update_reshape_migration; | |
7901 | u->subdev = super->current_vol; | |
7902 | u->new_level = geo->level; | |
7903 | u->new_layout = geo->layout; | |
7904 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
7905 | u->new_disks[0] = -1; | |
4bba0439 | 7906 | u->new_chunksize = -1; |
48c5303a PC |
7907 | |
7908 | dev = get_imsm_dev(super, u->subdev); | |
7909 | if (dev) { | |
7910 | struct imsm_map *map; | |
7911 | ||
7912 | map = get_imsm_map(dev, 0); | |
4bba0439 PC |
7913 | if (map) { |
7914 | int current_chunk_size = | |
7915 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
7916 | ||
7917 | if (geo->chunksize != current_chunk_size) { | |
7918 | u->new_chunksize = geo->chunksize / 1024; | |
7919 | dprintf("imsm: " | |
7920 | "chunk size change from %i to %i\n", | |
7921 | current_chunk_size, u->new_chunksize); | |
7922 | } | |
48c5303a | 7923 | previous_level = map->raid_level; |
4bba0439 | 7924 | } |
48c5303a PC |
7925 | } |
7926 | if ((geo->level == 5) && (previous_level == 0)) { | |
7927 | struct mdinfo *spares = NULL; | |
7928 | ||
7929 | u->new_raid_disks++; | |
7930 | spares = get_spares_for_grow(st); | |
7931 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
7932 | free(u); | |
7933 | sysfs_free(spares); | |
7934 | update_memory_size = 0; | |
7935 | dprintf("error: cannot get spare device " | |
7936 | "for requested migration"); | |
7937 | return 0; | |
7938 | } | |
7939 | sysfs_free(spares); | |
7940 | } | |
7941 | dprintf("imsm: reshape update preparation : OK\n"); | |
7942 | *updatep = u; | |
7943 | ||
7944 | return update_memory_size; | |
7945 | } | |
7946 | ||
8dd70bce AK |
7947 | static void imsm_update_metadata_locally(struct supertype *st, |
7948 | void *buf, int len) | |
7949 | { | |
7950 | struct metadata_update mu; | |
7951 | ||
7952 | mu.buf = buf; | |
7953 | mu.len = len; | |
7954 | mu.space = NULL; | |
7955 | mu.space_list = NULL; | |
7956 | mu.next = NULL; | |
7957 | imsm_prepare_update(st, &mu); | |
7958 | imsm_process_update(st, &mu); | |
7959 | ||
7960 | while (mu.space_list) { | |
7961 | void **space = mu.space_list; | |
7962 | mu.space_list = *space; | |
7963 | free(space); | |
7964 | } | |
7965 | } | |
78b10e66 | 7966 | |
471bceb6 | 7967 | /*************************************************************************** |
694575e7 | 7968 | * Function: imsm_analyze_change |
471bceb6 KW |
7969 | * Description: Function analyze change for single volume |
7970 | * and validate if transition is supported | |
694575e7 KW |
7971 | * Parameters: Geometry parameters, supertype structure |
7972 | * Returns: Operation type code on success, -1 if fail | |
471bceb6 KW |
7973 | ****************************************************************************/ |
7974 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
7975 | struct geo_params *geo) | |
694575e7 | 7976 | { |
471bceb6 KW |
7977 | struct mdinfo info; |
7978 | int change = -1; | |
7979 | int check_devs = 0; | |
c21e737b | 7980 | int chunk; |
471bceb6 KW |
7981 | |
7982 | getinfo_super_imsm_volume(st, &info, NULL); | |
7983 | ||
7984 | if ((geo->level != info.array.level) && | |
7985 | (geo->level >= 0) && | |
7986 | (geo->level != UnSet)) { | |
7987 | switch (info.array.level) { | |
7988 | case 0: | |
7989 | if (geo->level == 5) { | |
b5347799 | 7990 | change = CH_MIGRATION; |
471bceb6 KW |
7991 | check_devs = 1; |
7992 | } | |
7993 | if (geo->level == 10) { | |
7994 | change = CH_TAKEOVER; | |
7995 | check_devs = 1; | |
7996 | } | |
dfe77a9e KW |
7997 | break; |
7998 | case 1: | |
7999 | if (geo->level == 0) { | |
8000 | change = CH_TAKEOVER; | |
8001 | check_devs = 1; | |
8002 | } | |
471bceb6 | 8003 | break; |
471bceb6 KW |
8004 | case 10: |
8005 | if (geo->level == 0) { | |
8006 | change = CH_TAKEOVER; | |
8007 | check_devs = 1; | |
8008 | } | |
8009 | break; | |
8010 | } | |
8011 | if (change == -1) { | |
8012 | fprintf(stderr, | |
8013 | Name " Error. Level Migration from %d to %d " | |
8014 | "not supported!\n", | |
8015 | info.array.level, geo->level); | |
8016 | goto analyse_change_exit; | |
8017 | } | |
8018 | } else | |
8019 | geo->level = info.array.level; | |
8020 | ||
8021 | if ((geo->layout != info.array.layout) | |
8022 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 8023 | change = CH_MIGRATION; |
471bceb6 KW |
8024 | if ((info.array.layout == 0) |
8025 | && (info.array.level == 5) | |
8026 | && (geo->layout == 5)) { | |
8027 | /* reshape 5 -> 4 */ | |
8028 | } else if ((info.array.layout == 5) | |
8029 | && (info.array.level == 5) | |
8030 | && (geo->layout == 0)) { | |
8031 | /* reshape 4 -> 5 */ | |
8032 | geo->layout = 0; | |
8033 | geo->level = 5; | |
8034 | } else { | |
8035 | fprintf(stderr, | |
8036 | Name " Error. Layout Migration from %d to %d " | |
8037 | "not supported!\n", | |
8038 | info.array.layout, geo->layout); | |
8039 | change = -1; | |
8040 | goto analyse_change_exit; | |
8041 | } | |
8042 | } else | |
8043 | geo->layout = info.array.layout; | |
8044 | ||
8045 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
8046 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 8047 | change = CH_MIGRATION; |
471bceb6 KW |
8048 | else |
8049 | geo->chunksize = info.array.chunk_size; | |
8050 | ||
c21e737b | 8051 | chunk = geo->chunksize / 1024; |
471bceb6 KW |
8052 | if (!validate_geometry_imsm(st, |
8053 | geo->level, | |
8054 | geo->layout, | |
8055 | geo->raid_disks, | |
c21e737b | 8056 | &chunk, |
471bceb6 KW |
8057 | geo->size, |
8058 | 0, 0, 1)) | |
8059 | change = -1; | |
8060 | ||
8061 | if (check_devs) { | |
8062 | struct intel_super *super = st->sb; | |
8063 | struct imsm_super *mpb = super->anchor; | |
8064 | ||
8065 | if (mpb->num_raid_devs > 1) { | |
8066 | fprintf(stderr, | |
8067 | Name " Error. Cannot perform operation on %s" | |
8068 | "- for this operation it MUST be single " | |
8069 | "array in container\n", | |
8070 | geo->dev_name); | |
8071 | change = -1; | |
8072 | } | |
8073 | } | |
8074 | ||
8075 | analyse_change_exit: | |
8076 | ||
8077 | return change; | |
694575e7 KW |
8078 | } |
8079 | ||
bb025c2f KW |
8080 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
8081 | { | |
8082 | struct intel_super *super = st->sb; | |
8083 | struct imsm_update_takeover *u; | |
8084 | ||
8085 | u = malloc(sizeof(struct imsm_update_takeover)); | |
8086 | if (u == NULL) | |
8087 | return 1; | |
8088 | ||
8089 | u->type = update_takeover; | |
8090 | u->subarray = super->current_vol; | |
8091 | ||
8092 | /* 10->0 transition */ | |
8093 | if (geo->level == 0) | |
8094 | u->direction = R10_TO_R0; | |
8095 | ||
0529c688 KW |
8096 | /* 0->10 transition */ |
8097 | if (geo->level == 10) | |
8098 | u->direction = R0_TO_R10; | |
8099 | ||
bb025c2f KW |
8100 | /* update metadata locally */ |
8101 | imsm_update_metadata_locally(st, u, | |
8102 | sizeof(struct imsm_update_takeover)); | |
8103 | /* and possibly remotely */ | |
8104 | if (st->update_tail) | |
8105 | append_metadata_update(st, u, | |
8106 | sizeof(struct imsm_update_takeover)); | |
8107 | else | |
8108 | free(u); | |
8109 | ||
8110 | return 0; | |
8111 | } | |
8112 | ||
6dc0be30 AK |
8113 | static int warn_user_about_risk(void) |
8114 | { | |
8115 | int rv = 0; | |
8116 | ||
8117 | fprintf(stderr, | |
8118 | "\nThis is an experimental feature. Data on the RAID volume(s) " | |
8119 | "can be lost!!!\n\n" | |
8120 | "To continue command execution please make sure that\n" | |
8121 | "the grow process will not be interrupted. Use safe power\n" | |
8122 | "supply to avoid unexpected system reboot. Make sure that\n" | |
8123 | "reshaped container is not assembled automatically during\n" | |
8124 | "system boot.\n" | |
8125 | "If reshape is interrupted, assemble array manually\n" | |
8126 | "using e.g. '-Ac' option and up to date mdadm.conf file.\n" | |
8127 | "Assembly in scan mode is not possible in such case.\n" | |
8128 | "Growing container with boot array is not possible.\n" | |
8129 | "If boot array reshape is interrupted, whole file system\n" | |
8130 | "can be lost.\n\n"); | |
8131 | rv = ask("Do you want to continue? "); | |
8132 | fprintf(stderr, "\n"); | |
8133 | ||
8134 | return rv; | |
8135 | } | |
8136 | ||
78b10e66 N |
8137 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
8138 | int layout, int chunksize, int raid_disks, | |
41784c88 AK |
8139 | int delta_disks, char *backup, char *dev, |
8140 | int verbose) | |
78b10e66 | 8141 | { |
78b10e66 N |
8142 | int ret_val = 1; |
8143 | struct geo_params geo; | |
8144 | ||
8145 | dprintf("imsm: reshape_super called.\n"); | |
8146 | ||
71204a50 | 8147 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
8148 | |
8149 | geo.dev_name = dev; | |
694575e7 | 8150 | geo.dev_id = st->devnum; |
78b10e66 N |
8151 | geo.size = size; |
8152 | geo.level = level; | |
8153 | geo.layout = layout; | |
8154 | geo.chunksize = chunksize; | |
8155 | geo.raid_disks = raid_disks; | |
41784c88 AK |
8156 | if (delta_disks != UnSet) |
8157 | geo.raid_disks += delta_disks; | |
78b10e66 N |
8158 | |
8159 | dprintf("\tfor level : %i\n", geo.level); | |
8160 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
8161 | ||
8162 | if (experimental() == 0) | |
8163 | return ret_val; | |
8164 | ||
78b10e66 | 8165 | if (st->container_dev == st->devnum) { |
694575e7 KW |
8166 | /* On container level we can only increase number of devices. */ |
8167 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 8168 | int old_raid_disks = 0; |
6dc0be30 AK |
8169 | |
8170 | /* this warning will be removed when imsm checkpointing | |
8171 | * will be implemented, and restoring from check-point | |
8172 | * operation will be transparent for reboot process | |
8173 | */ | |
8174 | if (warn_user_about_risk() == 0) | |
8175 | return ret_val; | |
8176 | ||
78b10e66 N |
8177 | if (imsm_reshape_is_allowed_on_container( |
8178 | st, &geo, &old_raid_disks)) { | |
8179 | struct imsm_update_reshape *u = NULL; | |
8180 | int len; | |
8181 | ||
8182 | len = imsm_create_metadata_update_for_reshape( | |
8183 | st, &geo, old_raid_disks, &u); | |
8184 | ||
ed08d51c AK |
8185 | if (len <= 0) { |
8186 | dprintf("imsm: Cannot prepare update\n"); | |
8187 | goto exit_imsm_reshape_super; | |
8188 | } | |
8189 | ||
8dd70bce AK |
8190 | ret_val = 0; |
8191 | /* update metadata locally */ | |
8192 | imsm_update_metadata_locally(st, u, len); | |
8193 | /* and possibly remotely */ | |
8194 | if (st->update_tail) | |
8195 | append_metadata_update(st, u, len); | |
8196 | else | |
ed08d51c | 8197 | free(u); |
8dd70bce | 8198 | |
694575e7 | 8199 | } else { |
e7ff7e40 AK |
8200 | fprintf(stderr, Name ": (imsm) Operation " |
8201 | "is not allowed on this container\n"); | |
694575e7 KW |
8202 | } |
8203 | } else { | |
8204 | /* On volume level we support following operations | |
471bceb6 KW |
8205 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
8206 | * - chunk size migration | |
8207 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
8208 | */ | |
8209 | struct intel_super *super = st->sb; | |
8210 | struct intel_dev *dev = super->devlist; | |
8211 | int change, devnum; | |
694575e7 | 8212 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
8213 | /* find requested device */ |
8214 | while (dev) { | |
8215 | imsm_find_array_minor_by_subdev(dev->index, st->container_dev, &devnum); | |
8216 | if (devnum == geo.dev_id) | |
8217 | break; | |
8218 | dev = dev->next; | |
8219 | } | |
8220 | if (dev == NULL) { | |
8221 | fprintf(stderr, Name " Cannot find %s (%i) subarray\n", | |
8222 | geo.dev_name, geo.dev_id); | |
8223 | goto exit_imsm_reshape_super; | |
8224 | } | |
8225 | super->current_vol = dev->index; | |
694575e7 KW |
8226 | change = imsm_analyze_change(st, &geo); |
8227 | switch (change) { | |
471bceb6 | 8228 | case CH_TAKEOVER: |
bb025c2f | 8229 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 8230 | break; |
48c5303a PC |
8231 | case CH_MIGRATION: { |
8232 | struct imsm_update_reshape_migration *u = NULL; | |
8233 | int len = | |
8234 | imsm_create_metadata_update_for_migration( | |
8235 | st, &geo, &u); | |
8236 | if (len < 1) { | |
8237 | dprintf("imsm: " | |
8238 | "Cannot prepare update\n"); | |
8239 | break; | |
8240 | } | |
471bceb6 | 8241 | ret_val = 0; |
48c5303a PC |
8242 | /* update metadata locally */ |
8243 | imsm_update_metadata_locally(st, u, len); | |
8244 | /* and possibly remotely */ | |
8245 | if (st->update_tail) | |
8246 | append_metadata_update(st, u, len); | |
8247 | else | |
8248 | free(u); | |
8249 | } | |
8250 | break; | |
471bceb6 KW |
8251 | default: |
8252 | ret_val = 1; | |
694575e7 | 8253 | } |
694575e7 | 8254 | } |
78b10e66 | 8255 | |
ed08d51c | 8256 | exit_imsm_reshape_super: |
78b10e66 N |
8257 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
8258 | return ret_val; | |
8259 | } | |
2cda7640 | 8260 | |
eee67a47 AK |
8261 | /******************************************************************************* |
8262 | * Function: wait_for_reshape_imsm | |
8263 | * Description: Function writes new sync_max value and waits until | |
8264 | * reshape process reach new position | |
8265 | * Parameters: | |
8266 | * sra : general array info | |
8267 | * to_complete : new sync_max position | |
8268 | * ndata : number of disks in new array's layout | |
8269 | * Returns: | |
8270 | * 0 : success, | |
8271 | * 1 : there is no reshape in progress, | |
8272 | * -1 : fail | |
8273 | ******************************************************************************/ | |
8274 | int wait_for_reshape_imsm(struct mdinfo *sra, unsigned long long to_complete, | |
8275 | int ndata) | |
8276 | { | |
8277 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
8278 | unsigned long long completed; | |
8279 | ||
8280 | struct timeval timeout; | |
8281 | ||
8282 | if (fd < 0) | |
8283 | return 1; | |
8284 | ||
8285 | sysfs_fd_get_ll(fd, &completed); | |
8286 | ||
8287 | if (to_complete == 0) {/* reshape till the end of array */ | |
8288 | sysfs_set_str(sra, NULL, "sync_max", "max"); | |
8289 | to_complete = MaxSector; | |
8290 | } else { | |
8291 | if (completed > to_complete) | |
8292 | return -1; | |
8293 | if (sysfs_set_num(sra, NULL, "sync_max", | |
8294 | to_complete / ndata) != 0) { | |
8295 | close(fd); | |
8296 | return -1; | |
8297 | } | |
8298 | } | |
8299 | ||
8300 | /* FIXME should not need a timeout at all */ | |
8301 | timeout.tv_sec = 30; | |
8302 | timeout.tv_usec = 0; | |
8303 | do { | |
8304 | char action[20]; | |
8305 | fd_set rfds; | |
8306 | FD_ZERO(&rfds); | |
8307 | FD_SET(fd, &rfds); | |
8308 | select(fd+1, NULL, NULL, &rfds, &timeout); | |
8309 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
8310 | close(fd); | |
8311 | return 1; | |
8312 | } | |
8313 | if (sysfs_get_str(sra, NULL, "sync_action", | |
8314 | action, 20) > 0 && | |
8315 | strncmp(action, "reshape", 7) != 0) | |
8316 | break; | |
8317 | } while (completed < to_complete); | |
8318 | close(fd); | |
8319 | return 0; | |
8320 | ||
8321 | } | |
8322 | ||
b915c95f AK |
8323 | /******************************************************************************* |
8324 | * Function: check_degradation_change | |
8325 | * Description: Check that array hasn't become failed. | |
8326 | * Parameters: | |
8327 | * info : for sysfs access | |
8328 | * sources : source disks descriptors | |
8329 | * degraded: previous degradation level | |
8330 | * Returns: | |
8331 | * degradation level | |
8332 | ******************************************************************************/ | |
8333 | int check_degradation_change(struct mdinfo *info, | |
8334 | int *sources, | |
8335 | int degraded) | |
8336 | { | |
8337 | unsigned long long new_degraded; | |
8338 | sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
8339 | if (new_degraded != (unsigned long long)degraded) { | |
8340 | /* check each device to ensure it is still working */ | |
8341 | struct mdinfo *sd; | |
8342 | new_degraded = 0; | |
8343 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8344 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
8345 | continue; | |
8346 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
8347 | char sbuf[20]; | |
8348 | if (sysfs_get_str(info, | |
8349 | sd, "state", sbuf, 20) < 0 || | |
8350 | strstr(sbuf, "faulty") || | |
8351 | strstr(sbuf, "in_sync") == NULL) { | |
8352 | /* this device is dead */ | |
8353 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
8354 | if (sd->disk.raid_disk >= 0 && | |
8355 | sources[sd->disk.raid_disk] >= 0) { | |
8356 | close(sources[ | |
8357 | sd->disk.raid_disk]); | |
8358 | sources[sd->disk.raid_disk] = | |
8359 | -1; | |
8360 | } | |
8361 | new_degraded++; | |
8362 | } | |
8363 | } | |
8364 | } | |
8365 | } | |
8366 | ||
8367 | return new_degraded; | |
8368 | } | |
8369 | ||
10f22854 AK |
8370 | /******************************************************************************* |
8371 | * Function: imsm_manage_reshape | |
8372 | * Description: Function finds array under reshape and it manages reshape | |
8373 | * process. It creates stripes backups (if required) and sets | |
8374 | * checheckpoits. | |
8375 | * Parameters: | |
8376 | * afd : Backup handle (nattive) - not used | |
8377 | * sra : general array info | |
8378 | * reshape : reshape parameters - not used | |
8379 | * st : supertype structure | |
8380 | * blocks : size of critical section [blocks] | |
8381 | * fds : table of source device descriptor | |
8382 | * offsets : start of array (offest per devices) | |
8383 | * dests : not used | |
8384 | * destfd : table of destination device descriptor | |
8385 | * destoffsets : table of destination offsets (per device) | |
8386 | * Returns: | |
8387 | * 1 : success, reshape is done | |
8388 | * 0 : fail | |
8389 | ******************************************************************************/ | |
999b4972 N |
8390 | static int imsm_manage_reshape( |
8391 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 8392 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
8393 | int *fds, unsigned long long *offsets, |
8394 | int dests, int *destfd, unsigned long long *destoffsets) | |
8395 | { | |
10f22854 AK |
8396 | int ret_val = 0; |
8397 | struct intel_super *super = st->sb; | |
8398 | struct intel_dev *dv = NULL; | |
8399 | struct imsm_dev *dev = NULL; | |
8400 | struct imsm_map *map_src, *map_dest; | |
8401 | int migr_vol_qan = 0; | |
8402 | int ndata, odata; /* [bytes] */ | |
8403 | int chunk; /* [bytes] */ | |
8404 | struct migr_record *migr_rec; | |
8405 | char *buf = NULL; | |
8406 | unsigned int buf_size; /* [bytes] */ | |
8407 | unsigned long long max_position; /* array size [bytes] */ | |
8408 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
8409 | unsigned long long old_data_stripe_length; | |
8410 | unsigned long long new_data_stripe_length; | |
8411 | unsigned long long start_src; /* [bytes] */ | |
8412 | unsigned long long start; /* [bytes] */ | |
8413 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 8414 | int degraded = 0; |
10f22854 AK |
8415 | |
8416 | if (!fds || !offsets || !destfd || !destoffsets || !sra) | |
8417 | goto abort; | |
8418 | ||
8419 | /* Find volume during the reshape */ | |
8420 | for (dv = super->devlist; dv; dv = dv->next) { | |
8421 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
8422 | && dv->dev->vol.migr_state == 1) { | |
8423 | dev = dv->dev; | |
8424 | migr_vol_qan++; | |
8425 | } | |
8426 | } | |
8427 | /* Only one volume can migrate at the same time */ | |
8428 | if (migr_vol_qan != 1) { | |
8429 | fprintf(stderr, Name " : %s", migr_vol_qan ? | |
8430 | "Number of migrating volumes greater than 1\n" : | |
8431 | "There is no volume during migrationg\n"); | |
8432 | goto abort; | |
8433 | } | |
8434 | ||
8435 | map_src = get_imsm_map(dev, 1); | |
8436 | if (map_src == NULL) | |
8437 | goto abort; | |
8438 | map_dest = get_imsm_map(dev, 0); | |
8439 | ||
8440 | ndata = imsm_num_data_members(dev, 0); | |
8441 | odata = imsm_num_data_members(dev, 1); | |
8442 | ||
8443 | chunk = map_src->blocks_per_strip * 512; | |
8444 | old_data_stripe_length = odata * chunk; | |
8445 | ||
8446 | migr_rec = super->migr_rec; | |
8447 | ||
8448 | /* [bytes] */ | |
8449 | sra->new_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; | |
8450 | sra->new_level = map_dest->raid_level; | |
8451 | new_data_stripe_length = sra->new_chunk * ndata; | |
8452 | ||
8453 | /* initialize migration record for start condition */ | |
8454 | if (sra->reshape_progress == 0) | |
8455 | init_migr_record_imsm(st, dev, sra); | |
8456 | ||
8457 | /* size for data */ | |
8458 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
8459 | /* extend buffer size for parity disk */ | |
8460 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
8461 | /* add space for stripe aligment */ | |
8462 | buf_size += old_data_stripe_length; | |
8463 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
8464 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
8465 | goto abort; | |
8466 | } | |
8467 | ||
8468 | max_position = | |
8469 | __le32_to_cpu(migr_rec->post_migr_vol_cap) + | |
8470 | ((unsigned long long)__le32_to_cpu( | |
8471 | migr_rec->post_migr_vol_cap_hi) << 32); | |
8472 | ||
8473 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
8474 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
8475 | /* current reshape position [blocks] */ | |
8476 | unsigned long long current_position = | |
8477 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
8478 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
8479 | unsigned long long border; | |
8480 | ||
b915c95f AK |
8481 | /* Check that array hasn't become failed. |
8482 | */ | |
8483 | degraded = check_degradation_change(sra, fds, degraded); | |
8484 | if (degraded > 1) { | |
8485 | dprintf("imsm: Abort reshape due to degradation" | |
8486 | " level (%i)\n", degraded); | |
8487 | goto abort; | |
8488 | } | |
8489 | ||
10f22854 AK |
8490 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
8491 | ||
8492 | if ((current_position + next_step) > max_position) | |
8493 | next_step = max_position - current_position; | |
8494 | ||
8495 | start = (map_src->pba_of_lba0 + dev->reserved_blocks + | |
8496 | current_position) * 512; | |
8497 | ||
8498 | /* allign reading start to old geometry */ | |
8499 | start_buf_shift = start % old_data_stripe_length; | |
8500 | start_src = start - start_buf_shift; | |
8501 | ||
8502 | border = (start_src / odata) - (start / ndata); | |
8503 | border /= 512; | |
8504 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
8505 | /* save critical stripes to buf | |
8506 | * start - start address of current unit | |
8507 | * to backup [bytes] | |
8508 | * start_src - start address of current unit | |
8509 | * to backup alligned to source array | |
8510 | * [bytes] | |
8511 | */ | |
8512 | unsigned long long next_step_filler = 0; | |
8513 | unsigned long long copy_length = next_step * 512; | |
8514 | ||
8515 | /* allign copy area length to stripe in old geometry */ | |
8516 | next_step_filler = ((copy_length + start_buf_shift) | |
8517 | % old_data_stripe_length); | |
8518 | if (next_step_filler) | |
8519 | next_step_filler = (old_data_stripe_length | |
8520 | - next_step_filler); | |
8521 | dprintf("save_stripes() parameters: start = %llu," | |
8522 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
8523 | "\tstart_in_buf_shift = %llu," | |
8524 | "\tnext_step_filler = %llu\n", | |
8525 | start, start_src, copy_length, | |
8526 | start_buf_shift, next_step_filler); | |
8527 | ||
8528 | if (save_stripes(fds, offsets, map_src->num_members, | |
8529 | chunk, sra->array.level, | |
8530 | sra->array.layout, 0, NULL, start_src, | |
8531 | copy_length + | |
8532 | next_step_filler + start_buf_shift, | |
8533 | buf)) { | |
8534 | dprintf("imsm: Cannot save stripes" | |
8535 | " to buffer\n"); | |
8536 | goto abort; | |
8537 | } | |
8538 | /* Convert data to destination format and store it | |
8539 | * in backup general migration area | |
8540 | */ | |
8541 | if (save_backup_imsm(st, dev, sra, | |
8542 | buf + start_buf_shift, | |
8543 | ndata, copy_length)) { | |
8544 | dprintf("imsm: Cannot save stripes to " | |
8545 | "target devices\n"); | |
8546 | goto abort; | |
8547 | } | |
8548 | if (save_checkpoint_imsm(st, sra, | |
8549 | UNIT_SRC_IN_CP_AREA)) { | |
8550 | dprintf("imsm: Cannot write checkpoint to " | |
8551 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
8552 | goto abort; | |
8553 | } | |
8554 | /* decrease backup_blocks */ | |
8555 | if (backup_blocks > (unsigned long)next_step) | |
8556 | backup_blocks -= next_step; | |
8557 | else | |
8558 | backup_blocks = 0; | |
8559 | } | |
8560 | /* When data backed up, checkpoint stored, | |
8561 | * kick the kernel to reshape unit of data | |
8562 | */ | |
8563 | next_step = next_step + sra->reshape_progress; | |
8564 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); | |
8565 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
8566 | ||
8567 | /* wait until reshape finish */ | |
8568 | if (wait_for_reshape_imsm(sra, next_step, ndata) < 0) | |
8569 | dprintf("wait_for_reshape_imsm returned error," | |
8570 | " but we ignore it!\n"); | |
8571 | ||
8572 | sra->reshape_progress = next_step; | |
8573 | ||
8574 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL)) { | |
8575 | dprintf("imsm: Cannot write checkpoint to " | |
8576 | "migration record (UNIT_SRC_NORMAL)\n"); | |
8577 | goto abort; | |
8578 | } | |
8579 | ||
8580 | } | |
8581 | ||
8582 | /* return '1' if done */ | |
8583 | ret_val = 1; | |
8584 | abort: | |
8585 | free(buf); | |
8586 | abort_reshape(sra); | |
8587 | ||
8588 | return ret_val; | |
999b4972 | 8589 | } |
71204a50 | 8590 | #endif /* MDASSEMBLE */ |
999b4972 | 8591 | |
cdddbdbc DW |
8592 | struct superswitch super_imsm = { |
8593 | #ifndef MDASSEMBLE | |
8594 | .examine_super = examine_super_imsm, | |
8595 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 8596 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 8597 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
8598 | .detail_super = detail_super_imsm, |
8599 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 8600 | .write_init_super = write_init_super_imsm, |
0e600426 N |
8601 | .validate_geometry = validate_geometry_imsm, |
8602 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 8603 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 8604 | .detail_platform = detail_platform_imsm, |
33414a01 | 8605 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 8606 | .update_subarray = update_subarray_imsm, |
2b959fbf | 8607 | .load_container = load_container_imsm, |
71204a50 N |
8608 | .default_geometry = default_geometry_imsm, |
8609 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
8610 | .reshape_super = imsm_reshape_super, | |
8611 | .manage_reshape = imsm_manage_reshape, | |
cdddbdbc DW |
8612 | #endif |
8613 | .match_home = match_home_imsm, | |
8614 | .uuid_from_super= uuid_from_super_imsm, | |
8615 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 8616 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
8617 | .update_super = update_super_imsm, |
8618 | ||
8619 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 8620 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
8621 | |
8622 | .compare_super = compare_super_imsm, | |
8623 | ||
8624 | .load_super = load_super_imsm, | |
bf5a934a | 8625 | .init_super = init_super_imsm, |
e683ca88 | 8626 | .store_super = store_super_imsm, |
cdddbdbc DW |
8627 | .free_super = free_super_imsm, |
8628 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 8629 | .container_content = container_content_imsm, |
cdddbdbc | 8630 | |
cdddbdbc | 8631 | .external = 1, |
4cce4069 | 8632 | .name = "imsm", |
845dea95 | 8633 | |
0e600426 | 8634 | #ifndef MDASSEMBLE |
845dea95 NB |
8635 | /* for mdmon */ |
8636 | .open_new = imsm_open_new, | |
ed9d66aa | 8637 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
8638 | .set_disk = imsm_set_disk, |
8639 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 8640 | .activate_spare = imsm_activate_spare, |
e8319a19 | 8641 | .process_update = imsm_process_update, |
8273f55e | 8642 | .prepare_update = imsm_prepare_update, |
0e600426 | 8643 | #endif /* MDASSEMBLE */ |
cdddbdbc | 8644 | }; |