]>
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 | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
77 | #define MPB_ATTRIB_SUPPORTED MPB_ATTRIB_CHECKSUM_VERIFY | \ | |
78 | MPB_ATTRIB_2TB | \ | |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE | |
fe7ed8cb | 85 | |
8e59f3d8 | 86 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 87 | #define IMSM_RESERVED_SECTORS 4096 |
979d38be | 88 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
89 | |
90 | /* Disk configuration info. */ | |
91 | #define IMSM_MAX_DEVICES 255 | |
92 | struct imsm_disk { | |
93 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
94 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
95 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
f2f27e63 DW |
96 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
97 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
98 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 99 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb DW |
100 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
101 | #define IMSM_DISK_FILLERS 4 | |
cdddbdbc DW |
102 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ |
103 | }; | |
104 | ||
105 | /* RAID map configuration infos. */ | |
106 | struct imsm_map { | |
107 | __u32 pba_of_lba0; /* start address of partition */ | |
108 | __u32 blocks_per_member;/* blocks per member */ | |
109 | __u32 num_data_stripes; /* number of data stripes */ | |
110 | __u16 blocks_per_strip; | |
111 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
112 | #define IMSM_T_STATE_NORMAL 0 | |
113 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
114 | #define IMSM_T_STATE_DEGRADED 2 |
115 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
116 | __u8 raid_level; |
117 | #define IMSM_T_RAID0 0 | |
118 | #define IMSM_T_RAID1 1 | |
119 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
120 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
121 | __u8 num_domains; /* number of parity domains */ |
122 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 123 | __u8 ddf; |
cdddbdbc | 124 | __u32 filler[7]; /* expansion area */ |
7eef0453 | 125 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 126 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
127 | * top byte contains some flags |
128 | */ | |
cdddbdbc DW |
129 | } __attribute__ ((packed)); |
130 | ||
131 | struct imsm_vol { | |
f8f603f1 | 132 | __u32 curr_migr_unit; |
fe7ed8cb | 133 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 134 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
135 | #define MIGR_INIT 0 |
136 | #define MIGR_REBUILD 1 | |
137 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
138 | #define MIGR_GEN_MIGR 3 | |
139 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 140 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
141 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
142 | __u8 dirty; | |
fe7ed8cb DW |
143 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
144 | __u16 verify_errors; /* number of mismatches */ | |
145 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
146 | __u32 filler[4]; | |
cdddbdbc DW |
147 | struct imsm_map map[1]; |
148 | /* here comes another one if migr_state */ | |
149 | } __attribute__ ((packed)); | |
150 | ||
151 | struct imsm_dev { | |
fe7ed8cb | 152 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
153 | __u32 size_low; |
154 | __u32 size_high; | |
fe7ed8cb DW |
155 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
156 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
157 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
158 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
159 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
160 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
161 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
162 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
163 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
164 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
165 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
166 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
167 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
168 | __u32 status; /* Persistent RaidDev status */ |
169 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
170 | __u8 migr_priority; |
171 | __u8 num_sub_vols; | |
172 | __u8 tid; | |
173 | __u8 cng_master_disk; | |
174 | __u16 cache_policy; | |
175 | __u8 cng_state; | |
176 | __u8 cng_sub_state; | |
177 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
178 | __u32 filler[IMSM_DEV_FILLERS]; |
179 | struct imsm_vol vol; | |
180 | } __attribute__ ((packed)); | |
181 | ||
182 | struct imsm_super { | |
183 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
184 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
185 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
186 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
187 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
188 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
189 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
190 | __u8 num_disks; /* 0x38 Number of configured disks */ |
191 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
192 | __u8 error_log_pos; /* 0x3A */ |
193 | __u8 fill[1]; /* 0x3B */ | |
194 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
195 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
196 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
197 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
198 | #define IMSM_FILLERS 35 | |
199 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
200 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
201 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 202 | /* here comes BBM logs */ |
cdddbdbc DW |
203 | } __attribute__ ((packed)); |
204 | ||
604b746f JD |
205 | #define BBM_LOG_MAX_ENTRIES 254 |
206 | ||
207 | struct bbm_log_entry { | |
208 | __u64 defective_block_start; | |
209 | #define UNREADABLE 0xFFFFFFFF | |
210 | __u32 spare_block_offset; | |
211 | __u16 remapped_marked_count; | |
212 | __u16 disk_ordinal; | |
213 | } __attribute__ ((__packed__)); | |
214 | ||
215 | struct bbm_log { | |
216 | __u32 signature; /* 0xABADB10C */ | |
217 | __u32 entry_count; | |
218 | __u32 reserved_spare_block_count; /* 0 */ | |
219 | __u32 reserved; /* 0xFFFF */ | |
220 | __u64 first_spare_lba; | |
221 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
222 | } __attribute__ ((__packed__)); | |
223 | ||
224 | ||
cdddbdbc DW |
225 | #ifndef MDASSEMBLE |
226 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
227 | #endif | |
228 | ||
8e59f3d8 AK |
229 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
230 | ||
231 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
232 | ||
233 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must | |
234 | * be recovered using srcMap */ | |
235 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
236 | * already been migrated and must | |
237 | * be recovered from checkpoint area */ | |
238 | struct migr_record { | |
239 | __u32 rec_status; /* Status used to determine how to restart | |
240 | * migration in case it aborts | |
241 | * in some fashion */ | |
242 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
243 | __u32 family_num; /* Family number of MPB | |
244 | * containing the RaidDev | |
245 | * that is migrating */ | |
246 | __u32 ascending_migr; /* True if migrating in increasing | |
247 | * order of lbas */ | |
248 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
249 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
250 | * member disk | |
251 | * advances per unit-of-operation */ | |
252 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
253 | __u32 dest_1st_member_lba; /* First member lba on first | |
254 | * stripe of destination */ | |
255 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
256 | __u32 post_migr_vol_cap; /* Size of volume after | |
257 | * migration completes */ | |
258 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
259 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
260 | * migration ckpt record was read from | |
261 | * (for recovered migrations) */ | |
262 | } __attribute__ ((__packed__)); | |
263 | ||
1484e727 DW |
264 | static __u8 migr_type(struct imsm_dev *dev) |
265 | { | |
266 | if (dev->vol.migr_type == MIGR_VERIFY && | |
267 | dev->status & DEV_VERIFY_AND_FIX) | |
268 | return MIGR_REPAIR; | |
269 | else | |
270 | return dev->vol.migr_type; | |
271 | } | |
272 | ||
273 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
274 | { | |
275 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
276 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
277 | */ | |
278 | if (migr_type == MIGR_REPAIR) { | |
279 | dev->vol.migr_type = MIGR_VERIFY; | |
280 | dev->status |= DEV_VERIFY_AND_FIX; | |
281 | } else { | |
282 | dev->vol.migr_type = migr_type; | |
283 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
284 | } | |
285 | } | |
286 | ||
87eb16df | 287 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 288 | { |
87eb16df DW |
289 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
290 | } | |
cdddbdbc | 291 | |
87eb16df DW |
292 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
293 | { | |
294 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
295 | } |
296 | ||
ba2de7ba DW |
297 | struct intel_dev { |
298 | struct imsm_dev *dev; | |
299 | struct intel_dev *next; | |
f21e18ca | 300 | unsigned index; |
ba2de7ba DW |
301 | }; |
302 | ||
88654014 LM |
303 | struct intel_hba { |
304 | enum sys_dev_type type; | |
305 | char *path; | |
306 | char *pci_id; | |
307 | struct intel_hba *next; | |
308 | }; | |
309 | ||
1a64be56 LM |
310 | enum action { |
311 | DISK_REMOVE = 1, | |
312 | DISK_ADD | |
313 | }; | |
cdddbdbc DW |
314 | /* internal representation of IMSM metadata */ |
315 | struct intel_super { | |
316 | union { | |
949c47a0 DW |
317 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
318 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 319 | }; |
8e59f3d8 AK |
320 | union { |
321 | void *migr_rec_buf; /* buffer for I/O operations */ | |
322 | struct migr_record *migr_rec; /* migration record */ | |
323 | }; | |
949c47a0 | 324 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
325 | void *next_buf; /* for realloc'ing buf from the manager */ |
326 | size_t next_len; | |
c2c087e6 | 327 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 328 | int current_vol; /* index of raid device undergoing creation */ |
0dcecb2e | 329 | __u32 create_offset; /* common start for 'current_vol' */ |
148acb7b | 330 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 331 | struct intel_dev *devlist; |
cdddbdbc DW |
332 | struct dl { |
333 | struct dl *next; | |
334 | int index; | |
335 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
336 | int major, minor; | |
337 | char *devname; | |
b9f594fe | 338 | struct imsm_disk disk; |
cdddbdbc | 339 | int fd; |
0dcecb2e DW |
340 | int extent_cnt; |
341 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 342 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 343 | enum action action; |
ca0748fa | 344 | } *disks, *current_disk; |
1a64be56 LM |
345 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
346 | active */ | |
47ee5a45 | 347 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 348 | struct bbm_log *bbm_log; |
88654014 | 349 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 350 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
351 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
352 | }; | |
353 | ||
354 | struct intel_disk { | |
355 | struct imsm_disk disk; | |
356 | #define IMSM_UNKNOWN_OWNER (-1) | |
357 | int owner; | |
358 | struct intel_disk *next; | |
cdddbdbc DW |
359 | }; |
360 | ||
c2c087e6 DW |
361 | struct extent { |
362 | unsigned long long start, size; | |
363 | }; | |
364 | ||
694575e7 KW |
365 | /* definitions of reshape process types */ |
366 | enum imsm_reshape_type { | |
367 | CH_TAKEOVER, | |
b5347799 | 368 | CH_MIGRATION, |
694575e7 KW |
369 | }; |
370 | ||
88758e9d DW |
371 | /* definition of messages passed to imsm_process_update */ |
372 | enum imsm_update_type { | |
373 | update_activate_spare, | |
8273f55e | 374 | update_create_array, |
33414a01 | 375 | update_kill_array, |
aa534678 | 376 | update_rename_array, |
1a64be56 | 377 | update_add_remove_disk, |
78b10e66 | 378 | update_reshape_container_disks, |
48c5303a | 379 | update_reshape_migration, |
2d40f3a1 AK |
380 | update_takeover, |
381 | update_general_migration_checkpoint, | |
88758e9d DW |
382 | }; |
383 | ||
384 | struct imsm_update_activate_spare { | |
385 | enum imsm_update_type type; | |
d23fe947 | 386 | struct dl *dl; |
88758e9d DW |
387 | int slot; |
388 | int array; | |
389 | struct imsm_update_activate_spare *next; | |
390 | }; | |
391 | ||
78b10e66 N |
392 | struct geo_params { |
393 | int dev_id; | |
394 | char *dev_name; | |
395 | long long size; | |
396 | int level; | |
397 | int layout; | |
398 | int chunksize; | |
399 | int raid_disks; | |
400 | }; | |
401 | ||
bb025c2f KW |
402 | enum takeover_direction { |
403 | R10_TO_R0, | |
404 | R0_TO_R10 | |
405 | }; | |
406 | struct imsm_update_takeover { | |
407 | enum imsm_update_type type; | |
408 | int subarray; | |
409 | enum takeover_direction direction; | |
410 | }; | |
78b10e66 N |
411 | |
412 | struct imsm_update_reshape { | |
413 | enum imsm_update_type type; | |
414 | int old_raid_disks; | |
415 | int new_raid_disks; | |
48c5303a PC |
416 | |
417 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
418 | }; | |
419 | ||
420 | struct imsm_update_reshape_migration { | |
421 | enum imsm_update_type type; | |
422 | int old_raid_disks; | |
423 | int new_raid_disks; | |
424 | /* fields for array migration changes | |
425 | */ | |
426 | int subdev; | |
427 | int new_level; | |
428 | int new_layout; | |
4bba0439 | 429 | int new_chunksize; |
48c5303a | 430 | |
d195167d | 431 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
432 | }; |
433 | ||
2d40f3a1 AK |
434 | struct imsm_update_general_migration_checkpoint { |
435 | enum imsm_update_type type; | |
436 | __u32 curr_migr_unit; | |
437 | }; | |
438 | ||
54c2c1ea DW |
439 | struct disk_info { |
440 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
441 | }; | |
442 | ||
8273f55e DW |
443 | struct imsm_update_create_array { |
444 | enum imsm_update_type type; | |
8273f55e | 445 | int dev_idx; |
6a3e913e | 446 | struct imsm_dev dev; |
8273f55e DW |
447 | }; |
448 | ||
33414a01 DW |
449 | struct imsm_update_kill_array { |
450 | enum imsm_update_type type; | |
451 | int dev_idx; | |
452 | }; | |
453 | ||
aa534678 DW |
454 | struct imsm_update_rename_array { |
455 | enum imsm_update_type type; | |
456 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
457 | int dev_idx; | |
458 | }; | |
459 | ||
1a64be56 | 460 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
461 | enum imsm_update_type type; |
462 | }; | |
463 | ||
88654014 LM |
464 | |
465 | static const char *_sys_dev_type[] = { | |
466 | [SYS_DEV_UNKNOWN] = "Unknown", | |
467 | [SYS_DEV_SAS] = "SAS", | |
468 | [SYS_DEV_SATA] = "SATA" | |
469 | }; | |
470 | ||
471 | const char *get_sys_dev_type(enum sys_dev_type type) | |
472 | { | |
473 | if (type >= SYS_DEV_MAX) | |
474 | type = SYS_DEV_UNKNOWN; | |
475 | ||
476 | return _sys_dev_type[type]; | |
477 | } | |
478 | ||
479 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
480 | { | |
481 | struct intel_hba *result = malloc(sizeof(*result)); | |
482 | if (result) { | |
483 | result->type = device->type; | |
484 | result->path = strdup(device->path); | |
485 | result->next = NULL; | |
486 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
487 | result->pci_id++; | |
488 | } | |
489 | return result; | |
490 | } | |
491 | ||
492 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
493 | { | |
494 | struct intel_hba *result=NULL; | |
495 | for (result = hba; result; result = result->next) { | |
496 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
497 | break; | |
498 | } | |
499 | return result; | |
500 | } | |
501 | ||
b4cf4cba | 502 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
503 | { |
504 | struct intel_hba *hba; | |
505 | ||
506 | /* check if disk attached to Intel HBA */ | |
507 | hba = find_intel_hba(super->hba, device); | |
508 | if (hba != NULL) | |
509 | return 1; | |
510 | /* Check if HBA is already attached to super */ | |
511 | if (super->hba == NULL) { | |
512 | super->hba = alloc_intel_hba(device); | |
513 | return 1; | |
514 | } | |
515 | ||
516 | hba = super->hba; | |
517 | /* Intel metadata allows for all disks attached to the same type HBA. | |
518 | * Do not sypport odf HBA types mixing | |
519 | */ | |
520 | if (device->type != hba->type) | |
521 | return 2; | |
522 | ||
523 | while (hba->next) | |
524 | hba = hba->next; | |
525 | ||
526 | hba->next = alloc_intel_hba(device); | |
527 | return 1; | |
528 | } | |
529 | ||
530 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
531 | { | |
532 | struct sys_dev *list, *elem, *prev; | |
533 | char *disk_path; | |
534 | ||
535 | if ((list = find_intel_devices()) == NULL) | |
536 | return 0; | |
537 | ||
538 | if (fd < 0) | |
539 | disk_path = (char *) devname; | |
540 | else | |
541 | disk_path = diskfd_to_devpath(fd); | |
542 | ||
543 | if (!disk_path) { | |
544 | free_sys_dev(&list); | |
545 | return 0; | |
546 | } | |
547 | ||
548 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
549 | if (path_attached_to_hba(disk_path, elem->path)) { | |
550 | if (prev == NULL) | |
551 | list = list->next; | |
552 | else | |
553 | prev->next = elem->next; | |
554 | elem->next = NULL; | |
555 | if (disk_path != devname) | |
556 | free(disk_path); | |
557 | free_sys_dev(&list); | |
558 | return elem; | |
559 | } | |
560 | } | |
561 | if (disk_path != devname) | |
562 | free(disk_path); | |
563 | free_sys_dev(&list); | |
564 | ||
565 | return NULL; | |
566 | } | |
567 | ||
568 | ||
d424212e N |
569 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
570 | char *devname); | |
f2f5c343 | 571 | |
cdddbdbc DW |
572 | static struct supertype *match_metadata_desc_imsm(char *arg) |
573 | { | |
574 | struct supertype *st; | |
575 | ||
576 | if (strcmp(arg, "imsm") != 0 && | |
577 | strcmp(arg, "default") != 0 | |
578 | ) | |
579 | return NULL; | |
580 | ||
581 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
582 | if (!st) |
583 | return NULL; | |
ef609477 | 584 | memset(st, 0, sizeof(*st)); |
d1d599ea | 585 | st->container_dev = NoMdDev; |
cdddbdbc DW |
586 | st->ss = &super_imsm; |
587 | st->max_devs = IMSM_MAX_DEVICES; | |
588 | st->minor_version = 0; | |
589 | st->sb = NULL; | |
590 | return st; | |
591 | } | |
592 | ||
0e600426 | 593 | #ifndef MDASSEMBLE |
cdddbdbc DW |
594 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
595 | { | |
596 | return &mpb->sig[MPB_SIG_LEN]; | |
597 | } | |
9e2d750d | 598 | #endif |
cdddbdbc | 599 | |
949c47a0 DW |
600 | /* retrieve a disk directly from the anchor when the anchor is known to be |
601 | * up-to-date, currently only at load time | |
602 | */ | |
603 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 604 | { |
949c47a0 | 605 | if (index >= mpb->num_disks) |
cdddbdbc DW |
606 | return NULL; |
607 | return &mpb->disk[index]; | |
608 | } | |
609 | ||
95d07a2c LM |
610 | /* retrieve the disk description based on a index of the disk |
611 | * in the sub-array | |
612 | */ | |
613 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 614 | { |
b9f594fe DW |
615 | struct dl *d; |
616 | ||
617 | for (d = super->disks; d; d = d->next) | |
618 | if (d->index == index) | |
95d07a2c LM |
619 | return d; |
620 | ||
621 | return NULL; | |
622 | } | |
623 | /* retrieve a disk from the parsed metadata */ | |
624 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
625 | { | |
626 | struct dl *dl; | |
627 | ||
628 | dl = get_imsm_dl_disk(super, index); | |
629 | if (dl) | |
630 | return &dl->disk; | |
631 | ||
b9f594fe | 632 | return NULL; |
949c47a0 DW |
633 | } |
634 | ||
635 | /* generate a checksum directly from the anchor when the anchor is known to be | |
636 | * up-to-date, currently only at load or write_super after coalescing | |
637 | */ | |
638 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
639 | { |
640 | __u32 end = mpb->mpb_size / sizeof(end); | |
641 | __u32 *p = (__u32 *) mpb; | |
642 | __u32 sum = 0; | |
643 | ||
97f734fd N |
644 | while (end--) { |
645 | sum += __le32_to_cpu(*p); | |
646 | p++; | |
647 | } | |
cdddbdbc DW |
648 | |
649 | return sum - __le32_to_cpu(mpb->check_sum); | |
650 | } | |
651 | ||
a965f303 DW |
652 | static size_t sizeof_imsm_map(struct imsm_map *map) |
653 | { | |
654 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
655 | } | |
656 | ||
657 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 658 | { |
5e7b0330 AK |
659 | /* A device can have 2 maps if it is in the middle of a migration. |
660 | * If second_map is: | |
661 | * 0 - we return the first map | |
662 | * 1 - we return the second map if it exists, else NULL | |
663 | * -1 - we return the second map if it exists, else the first | |
664 | */ | |
a965f303 DW |
665 | struct imsm_map *map = &dev->vol.map[0]; |
666 | ||
5e7b0330 | 667 | if (second_map == 1 && !dev->vol.migr_state) |
a965f303 | 668 | return NULL; |
5e7b0330 AK |
669 | else if (second_map == 1 || |
670 | (second_map < 0 && dev->vol.migr_state)) { | |
a965f303 DW |
671 | void *ptr = map; |
672 | ||
673 | return ptr + sizeof_imsm_map(map); | |
674 | } else | |
675 | return map; | |
5e7b0330 | 676 | |
a965f303 | 677 | } |
cdddbdbc | 678 | |
3393c6af DW |
679 | /* return the size of the device. |
680 | * migr_state increases the returned size if map[0] were to be duplicated | |
681 | */ | |
682 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
683 | { |
684 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
685 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
686 | |
687 | /* migrating means an additional map */ | |
a965f303 DW |
688 | if (dev->vol.migr_state) |
689 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
690 | else if (migr_state) |
691 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
692 | |
693 | return size; | |
694 | } | |
695 | ||
54c2c1ea DW |
696 | #ifndef MDASSEMBLE |
697 | /* retrieve disk serial number list from a metadata update */ | |
698 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
699 | { | |
700 | void *u = update; | |
701 | struct disk_info *inf; | |
702 | ||
703 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
704 | sizeof_imsm_dev(&update->dev, 0); | |
705 | ||
706 | return inf; | |
707 | } | |
708 | #endif | |
709 | ||
949c47a0 | 710 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
711 | { |
712 | int offset; | |
713 | int i; | |
714 | void *_mpb = mpb; | |
715 | ||
949c47a0 | 716 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
717 | return NULL; |
718 | ||
719 | /* devices start after all disks */ | |
720 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
721 | ||
722 | for (i = 0; i <= index; i++) | |
723 | if (i == index) | |
724 | return _mpb + offset; | |
725 | else | |
3393c6af | 726 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
727 | |
728 | return NULL; | |
729 | } | |
730 | ||
949c47a0 DW |
731 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
732 | { | |
ba2de7ba DW |
733 | struct intel_dev *dv; |
734 | ||
949c47a0 DW |
735 | if (index >= super->anchor->num_raid_devs) |
736 | return NULL; | |
ba2de7ba DW |
737 | for (dv = super->devlist; dv; dv = dv->next) |
738 | if (dv->index == index) | |
739 | return dv->dev; | |
740 | return NULL; | |
949c47a0 DW |
741 | } |
742 | ||
98130f40 AK |
743 | /* |
744 | * for second_map: | |
745 | * == 0 get first map | |
746 | * == 1 get second map | |
747 | * == -1 than get map according to the current migr_state | |
748 | */ | |
749 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
750 | int slot, | |
751 | int second_map) | |
7eef0453 DW |
752 | { |
753 | struct imsm_map *map; | |
754 | ||
5e7b0330 | 755 | map = get_imsm_map(dev, second_map); |
7eef0453 | 756 | |
ff077194 DW |
757 | /* top byte identifies disk under rebuild */ |
758 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
759 | } | |
760 | ||
761 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 762 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 763 | { |
98130f40 | 764 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
765 | |
766 | return ord_to_idx(ord); | |
7eef0453 DW |
767 | } |
768 | ||
be73972f DW |
769 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
770 | { | |
771 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
772 | } | |
773 | ||
f21e18ca | 774 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
775 | { |
776 | int slot; | |
777 | __u32 ord; | |
778 | ||
779 | for (slot = 0; slot < map->num_members; slot++) { | |
780 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
781 | if (ord_to_idx(ord) == idx) | |
782 | return slot; | |
783 | } | |
784 | ||
785 | return -1; | |
786 | } | |
787 | ||
cdddbdbc DW |
788 | static int get_imsm_raid_level(struct imsm_map *map) |
789 | { | |
790 | if (map->raid_level == 1) { | |
791 | if (map->num_members == 2) | |
792 | return 1; | |
793 | else | |
794 | return 10; | |
795 | } | |
796 | ||
797 | return map->raid_level; | |
798 | } | |
799 | ||
c2c087e6 DW |
800 | static int cmp_extent(const void *av, const void *bv) |
801 | { | |
802 | const struct extent *a = av; | |
803 | const struct extent *b = bv; | |
804 | if (a->start < b->start) | |
805 | return -1; | |
806 | if (a->start > b->start) | |
807 | return 1; | |
808 | return 0; | |
809 | } | |
810 | ||
0dcecb2e | 811 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 812 | { |
c2c087e6 | 813 | int memberships = 0; |
620b1713 | 814 | int i; |
c2c087e6 | 815 | |
949c47a0 DW |
816 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
817 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 818 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 819 | |
620b1713 DW |
820 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
821 | memberships++; | |
c2c087e6 | 822 | } |
0dcecb2e DW |
823 | |
824 | return memberships; | |
825 | } | |
826 | ||
827 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
828 | { | |
829 | /* find a list of used extents on the given physical device */ | |
830 | struct extent *rv, *e; | |
620b1713 | 831 | int i; |
0dcecb2e DW |
832 | int memberships = count_memberships(dl, super); |
833 | __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
834 | ||
c2c087e6 DW |
835 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
836 | if (!rv) | |
837 | return NULL; | |
838 | e = rv; | |
839 | ||
949c47a0 DW |
840 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
841 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 842 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 843 | |
620b1713 DW |
844 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
845 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
846 | e->size = __le32_to_cpu(map->blocks_per_member); | |
847 | e++; | |
c2c087e6 DW |
848 | } |
849 | } | |
850 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
851 | ||
14e8215b DW |
852 | /* determine the start of the metadata |
853 | * when no raid devices are defined use the default | |
854 | * ...otherwise allow the metadata to truncate the value | |
855 | * as is the case with older versions of imsm | |
856 | */ | |
857 | if (memberships) { | |
858 | struct extent *last = &rv[memberships - 1]; | |
859 | __u32 remainder; | |
860 | ||
861 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
862 | (last->start + last->size); | |
dda5855f DW |
863 | /* round down to 1k block to satisfy precision of the kernel |
864 | * 'size' interface | |
865 | */ | |
866 | remainder &= ~1UL; | |
867 | /* make sure remainder is still sane */ | |
f21e18ca | 868 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 869 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
870 | if (reservation > remainder) |
871 | reservation = remainder; | |
872 | } | |
873 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
874 | e->size = 0; |
875 | return rv; | |
876 | } | |
877 | ||
14e8215b DW |
878 | /* try to determine how much space is reserved for metadata from |
879 | * the last get_extents() entry, otherwise fallback to the | |
880 | * default | |
881 | */ | |
882 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
883 | { | |
884 | struct extent *e; | |
885 | int i; | |
886 | __u32 rv; | |
887 | ||
888 | /* for spares just return a minimal reservation which will grow | |
889 | * once the spare is picked up by an array | |
890 | */ | |
891 | if (dl->index == -1) | |
892 | return MPB_SECTOR_CNT; | |
893 | ||
894 | e = get_extents(super, dl); | |
895 | if (!e) | |
896 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
897 | ||
898 | /* scroll to last entry */ | |
899 | for (i = 0; e[i].size; i++) | |
900 | continue; | |
901 | ||
902 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
903 | ||
904 | free(e); | |
905 | ||
906 | return rv; | |
907 | } | |
908 | ||
25ed7e59 DW |
909 | static int is_spare(struct imsm_disk *disk) |
910 | { | |
911 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
912 | } | |
913 | ||
914 | static int is_configured(struct imsm_disk *disk) | |
915 | { | |
916 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
917 | } | |
918 | ||
919 | static int is_failed(struct imsm_disk *disk) | |
920 | { | |
921 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
922 | } | |
923 | ||
80e7f8c3 AC |
924 | /* Return minimum size of a spare that can be used in this array*/ |
925 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
926 | { | |
927 | struct intel_super *super = st->sb; | |
928 | struct dl *dl; | |
929 | struct extent *e; | |
930 | int i; | |
931 | unsigned long long rv = 0; | |
932 | ||
933 | if (!super) | |
934 | return rv; | |
935 | /* find first active disk in array */ | |
936 | dl = super->disks; | |
937 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
938 | dl = dl->next; | |
939 | if (!dl) | |
940 | return rv; | |
941 | /* find last lba used by subarrays */ | |
942 | e = get_extents(super, dl); | |
943 | if (!e) | |
944 | return rv; | |
945 | for (i = 0; e[i].size; i++) | |
946 | continue; | |
947 | if (i > 0) | |
948 | rv = e[i-1].start + e[i-1].size; | |
949 | free(e); | |
950 | /* add the amount of space needed for metadata */ | |
951 | rv = rv + MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
952 | return rv * 512; | |
953 | } | |
954 | ||
1799c9e8 | 955 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
956 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
957 | struct imsm_dev *dev); | |
1e5c6983 | 958 | |
c47b0ff6 AK |
959 | static void print_imsm_dev(struct intel_super *super, |
960 | struct imsm_dev *dev, | |
961 | char *uuid, | |
962 | int disk_idx) | |
cdddbdbc DW |
963 | { |
964 | __u64 sz; | |
0d80bb2f | 965 | int slot, i; |
a965f303 | 966 | struct imsm_map *map = get_imsm_map(dev, 0); |
dd8bcb3b | 967 | struct imsm_map *map2 = get_imsm_map(dev, 1); |
b10b37b8 | 968 | __u32 ord; |
cdddbdbc DW |
969 | |
970 | printf("\n"); | |
1e7bc0ed | 971 | printf("[%.16s]:\n", dev->volume); |
44470971 | 972 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
973 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
974 | if (map2) | |
975 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
976 | printf("\n"); | |
977 | printf(" Members : %d", map->num_members); | |
978 | if (map2) | |
979 | printf(" <-- %d", map2->num_members); | |
980 | printf("\n"); | |
0d80bb2f DW |
981 | printf(" Slots : ["); |
982 | for (i = 0; i < map->num_members; i++) { | |
dd8bcb3b | 983 | ord = get_imsm_ord_tbl_ent(dev, i, 0); |
0d80bb2f DW |
984 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
985 | } | |
dd8bcb3b AK |
986 | printf("]"); |
987 | if (map2) { | |
988 | printf(" <-- ["); | |
989 | for (i = 0; i < map2->num_members; i++) { | |
990 | ord = get_imsm_ord_tbl_ent(dev, i, 1); | |
991 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); | |
992 | } | |
993 | printf("]"); | |
994 | } | |
995 | printf("\n"); | |
7095bccb AK |
996 | printf(" Failed disk : "); |
997 | if (map->failed_disk_num == 0xff) | |
998 | printf("none"); | |
999 | else | |
1000 | printf("%i", map->failed_disk_num); | |
1001 | printf("\n"); | |
620b1713 DW |
1002 | slot = get_imsm_disk_slot(map, disk_idx); |
1003 | if (slot >= 0) { | |
98130f40 | 1004 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
b10b37b8 DW |
1005 | printf(" This Slot : %d%s\n", slot, |
1006 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1007 | } else | |
cdddbdbc DW |
1008 | printf(" This Slot : ?\n"); |
1009 | sz = __le32_to_cpu(dev->size_high); | |
1010 | sz <<= 32; | |
1011 | sz += __le32_to_cpu(dev->size_low); | |
1012 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1013 | human_size(sz * 512)); | |
1014 | sz = __le32_to_cpu(map->blocks_per_member); | |
1015 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
1016 | human_size(sz * 512)); | |
1017 | printf(" Sector Offset : %u\n", | |
1018 | __le32_to_cpu(map->pba_of_lba0)); | |
1019 | printf(" Num Stripes : %u\n", | |
1020 | __le32_to_cpu(map->num_data_stripes)); | |
dd8bcb3b | 1021 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1022 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1023 | if (map2) |
1024 | printf(" <-- %u KiB", | |
1025 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1026 | printf("\n"); | |
cdddbdbc | 1027 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1028 | printf(" Migrate State : "); |
1484e727 DW |
1029 | if (dev->vol.migr_state) { |
1030 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1031 | printf("initialize\n"); |
1484e727 | 1032 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1033 | printf("rebuild\n"); |
1484e727 | 1034 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1035 | printf("check\n"); |
1484e727 | 1036 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1037 | printf("general migration\n"); |
1484e727 | 1038 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1039 | printf("state change\n"); |
1484e727 | 1040 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1041 | printf("repair\n"); |
1484e727 | 1042 | else |
8655a7b1 DW |
1043 | printf("<unknown:%d>\n", migr_type(dev)); |
1044 | } else | |
1045 | printf("idle\n"); | |
3393c6af DW |
1046 | printf(" Map State : %s", map_state_str[map->map_state]); |
1047 | if (dev->vol.migr_state) { | |
1048 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1e5c6983 | 1049 | |
b10b37b8 | 1050 | printf(" <-- %s", map_state_str[map->map_state]); |
1e5c6983 DW |
1051 | printf("\n Checkpoint : %u (%llu)", |
1052 | __le32_to_cpu(dev->vol.curr_migr_unit), | |
c47b0ff6 | 1053 | (unsigned long long)blocks_per_migr_unit(super, dev)); |
3393c6af DW |
1054 | } |
1055 | printf("\n"); | |
cdddbdbc | 1056 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1057 | } |
1058 | ||
14e8215b | 1059 | static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved) |
cdddbdbc | 1060 | { |
949c47a0 | 1061 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); |
1f24f035 | 1062 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1063 | __u64 sz; |
1064 | ||
d362da3d | 1065 | if (index < 0 || !disk) |
e9d82038 DW |
1066 | return; |
1067 | ||
cdddbdbc | 1068 | printf("\n"); |
1f24f035 | 1069 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
cdddbdbc | 1070 | printf(" Disk%02d Serial : %s\n", index, str); |
25ed7e59 DW |
1071 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1072 | is_configured(disk) ? " active" : "", | |
1073 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1074 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
14e8215b | 1075 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
1076 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1077 | human_size(sz * 512)); | |
1078 | } | |
1079 | ||
520e69e2 AK |
1080 | static int is_gen_migration(struct imsm_dev *dev); |
1081 | ||
1082 | void examine_migr_rec_imsm(struct intel_super *super) | |
1083 | { | |
1084 | struct migr_record *migr_rec = super->migr_rec; | |
1085 | struct imsm_super *mpb = super->anchor; | |
1086 | int i; | |
1087 | ||
1088 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1089 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1090 | if (is_gen_migration(dev) == 0) | |
1091 | continue; | |
1092 | ||
1093 | printf("\nMigration Record Information:"); | |
1094 | if (super->disks->index > 1) { | |
1095 | printf(" Empty\n "); | |
1096 | printf("Examine one of first two disks in array\n"); | |
1097 | break; | |
1098 | } | |
1099 | printf("\n Status : "); | |
1100 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1101 | printf("Normal\n"); | |
1102 | else | |
1103 | printf("Contains Data\n"); | |
1104 | printf(" Current Unit : %u\n", | |
1105 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1106 | printf(" Family : %u\n", | |
1107 | __le32_to_cpu(migr_rec->family_num)); | |
1108 | printf(" Ascending : %u\n", | |
1109 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1110 | printf(" Blocks Per Unit : %u\n", | |
1111 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1112 | printf(" Dest. Depth Per Unit : %u\n", | |
1113 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1114 | printf(" Checkpoint Area pba : %u\n", | |
1115 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1116 | printf(" First member lba : %u\n", | |
1117 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1118 | printf(" Total Number of Units : %u\n", | |
1119 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1120 | printf(" Size of volume : %u\n", | |
1121 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1122 | printf(" Expansion space for LBA64 : %u\n", | |
1123 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1124 | printf(" Record was read from : %u\n", | |
1125 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1126 | ||
1127 | break; | |
1128 | } | |
1129 | } | |
9e2d750d | 1130 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1131 | /******************************************************************************* |
1132 | * function: imsm_check_attributes | |
1133 | * Description: Function checks if features represented by attributes flags | |
1134 | * are supported by mdadm. | |
1135 | * Parameters: | |
1136 | * attributes - Attributes read from metadata | |
1137 | * Returns: | |
1138 | * 0 - passed attributes contains unsupported features flags | |
1139 | * 1 - all features are supported | |
1140 | ******************************************************************************/ | |
1141 | static int imsm_check_attributes(__u32 attributes) | |
1142 | { | |
1143 | int ret_val = 1; | |
1144 | __u32 not_supported = (MPB_ATTRIB_SUPPORTED)^0xffffffff; | |
1145 | ||
1146 | not_supported &= attributes; | |
1147 | if (not_supported) { | |
1148 | fprintf(stderr, Name "(IMSM): Unsupported attributes : %x\n", not_supported); | |
1149 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1150 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1151 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1152 | } | |
1153 | if (not_supported & MPB_ATTRIB_2TB) { | |
1154 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1155 | not_supported ^= MPB_ATTRIB_2TB; | |
1156 | } | |
1157 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1158 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1159 | not_supported ^= MPB_ATTRIB_RAID0; | |
1160 | } | |
1161 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1162 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1163 | not_supported ^= MPB_ATTRIB_RAID1; | |
1164 | } | |
1165 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1166 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1167 | not_supported ^= MPB_ATTRIB_RAID10; | |
1168 | } | |
1169 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1170 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1171 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1172 | } | |
1173 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1174 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1175 | not_supported ^= MPB_ATTRIB_RAID5; | |
1176 | } | |
1177 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1178 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1179 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1180 | } | |
1181 | if (not_supported & MPB_ATTRIB_BBM) { | |
1182 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1183 | not_supported ^= MPB_ATTRIB_BBM; | |
1184 | } | |
1185 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1186 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1187 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1188 | } | |
1189 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1190 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1191 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1192 | } | |
1193 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1194 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1195 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1196 | } | |
1197 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1198 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1199 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1200 | } | |
1201 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1202 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1203 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1204 | } | |
1205 | ||
1206 | if (not_supported) | |
1207 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1208 | ||
1209 | ret_val = 0; | |
1210 | } | |
1211 | ||
1212 | return ret_val; | |
1213 | } | |
1214 | ||
9e2d750d | 1215 | #ifndef MDASSEMBLE |
a5d85af7 | 1216 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1217 | |
cdddbdbc DW |
1218 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1219 | { | |
1220 | struct intel_super *super = st->sb; | |
949c47a0 | 1221 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1222 | char str[MAX_SIGNATURE_LENGTH]; |
1223 | int i; | |
27fd6274 DW |
1224 | struct mdinfo info; |
1225 | char nbuf[64]; | |
cdddbdbc | 1226 | __u32 sum; |
14e8215b | 1227 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1228 | struct dl *dl; |
27fd6274 | 1229 | |
cdddbdbc DW |
1230 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1231 | printf(" Magic : %s\n", str); | |
1232 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1233 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1234 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1235 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1236 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1237 | printf(" Attributes : "); |
1238 | if (imsm_check_attributes(mpb->attributes)) | |
1239 | printf("All supported\n"); | |
1240 | else | |
1241 | printf("not supported\n"); | |
a5d85af7 | 1242 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1243 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1244 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1245 | sum = __le32_to_cpu(mpb->check_sum); |
1246 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1247 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1248 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1249 | printf(" Disks : %d\n", mpb->num_disks); |
1250 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
14e8215b | 1251 | print_imsm_disk(mpb, super->disks->index, reserved); |
604b746f JD |
1252 | if (super->bbm_log) { |
1253 | struct bbm_log *log = super->bbm_log; | |
1254 | ||
1255 | printf("\n"); | |
1256 | printf("Bad Block Management Log:\n"); | |
1257 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1258 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1259 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1260 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1261 | printf(" First Spare : %llx\n", |
1262 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1263 | } |
44470971 DW |
1264 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1265 | struct mdinfo info; | |
1266 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1267 | ||
1268 | super->current_vol = i; | |
a5d85af7 | 1269 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1270 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1271 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1272 | } |
cdddbdbc DW |
1273 | for (i = 0; i < mpb->num_disks; i++) { |
1274 | if (i == super->disks->index) | |
1275 | continue; | |
14e8215b | 1276 | print_imsm_disk(mpb, i, reserved); |
cdddbdbc | 1277 | } |
94827db3 N |
1278 | for (dl = super->disks ; dl; dl = dl->next) { |
1279 | struct imsm_disk *disk; | |
1280 | char str[MAX_RAID_SERIAL_LEN + 1]; | |
1281 | __u64 sz; | |
1282 | ||
1283 | if (dl->index >= 0) | |
1284 | continue; | |
1285 | ||
1286 | disk = &dl->disk; | |
1287 | printf("\n"); | |
1288 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); | |
1289 | printf(" Disk Serial : %s\n", str); | |
1290 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", | |
1291 | is_configured(disk) ? " active" : "", | |
1292 | is_failed(disk) ? " failed" : ""); | |
1293 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
1294 | sz = __le32_to_cpu(disk->total_blocks) - reserved; | |
1295 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, | |
1296 | human_size(sz * 512)); | |
1297 | } | |
520e69e2 AK |
1298 | |
1299 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1300 | } |
1301 | ||
061f2c6a | 1302 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1303 | { |
27fd6274 | 1304 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1305 | struct mdinfo info; |
1306 | char nbuf[64]; | |
1e7bc0ed | 1307 | struct intel_super *super = st->sb; |
1e7bc0ed | 1308 | |
0d5a423f DW |
1309 | if (!super->anchor->num_raid_devs) { |
1310 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1311 | return; |
0d5a423f | 1312 | } |
ff54de6e | 1313 | |
a5d85af7 | 1314 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1315 | fname_from_uuid(st, &info, nbuf, ':'); |
1316 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1317 | } | |
1318 | ||
1319 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1320 | { | |
1321 | /* We just write a generic IMSM ARRAY entry */ | |
1322 | struct mdinfo info; | |
1323 | char nbuf[64]; | |
1324 | char nbuf1[64]; | |
1325 | struct intel_super *super = st->sb; | |
1326 | int i; | |
1327 | ||
1328 | if (!super->anchor->num_raid_devs) | |
1329 | return; | |
1330 | ||
a5d85af7 | 1331 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1332 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1333 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1334 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1335 | ||
1336 | super->current_vol = i; | |
a5d85af7 | 1337 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1338 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1339 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1340 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1341 | } |
cdddbdbc DW |
1342 | } |
1343 | ||
9d84c8ea DW |
1344 | static void export_examine_super_imsm(struct supertype *st) |
1345 | { | |
1346 | struct intel_super *super = st->sb; | |
1347 | struct imsm_super *mpb = super->anchor; | |
1348 | struct mdinfo info; | |
1349 | char nbuf[64]; | |
1350 | ||
a5d85af7 | 1351 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1352 | fname_from_uuid(st, &info, nbuf, ':'); |
1353 | printf("MD_METADATA=imsm\n"); | |
1354 | printf("MD_LEVEL=container\n"); | |
1355 | printf("MD_UUID=%s\n", nbuf+5); | |
1356 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1357 | } | |
1358 | ||
cdddbdbc DW |
1359 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1360 | { | |
3ebe00a1 DW |
1361 | struct mdinfo info; |
1362 | char nbuf[64]; | |
1363 | ||
a5d85af7 | 1364 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1365 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1366 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1367 | } |
1368 | ||
1369 | static void brief_detail_super_imsm(struct supertype *st) | |
1370 | { | |
ff54de6e N |
1371 | struct mdinfo info; |
1372 | char nbuf[64]; | |
a5d85af7 | 1373 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1374 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1375 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1376 | } |
d665cc31 DW |
1377 | |
1378 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1379 | static void fd2devname(int fd, char *name); | |
1380 | ||
120dc887 | 1381 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1382 | { |
120dc887 LM |
1383 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1384 | * controller, as well as non-connected ports | |
d665cc31 DW |
1385 | */ |
1386 | int hba_len = strlen(hba_path) + 1; | |
1387 | struct dirent *ent; | |
1388 | DIR *dir; | |
1389 | char *path = NULL; | |
1390 | int err = 0; | |
1391 | unsigned long port_mask = (1 << port_count) - 1; | |
1392 | ||
f21e18ca | 1393 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1394 | if (verbose) |
1395 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1396 | return 2; | |
1397 | } | |
1398 | ||
1399 | /* scroll through /sys/dev/block looking for devices attached to | |
1400 | * this hba | |
1401 | */ | |
1402 | dir = opendir("/sys/dev/block"); | |
1403 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1404 | int fd; | |
1405 | char model[64]; | |
1406 | char vendor[64]; | |
1407 | char buf[1024]; | |
1408 | int major, minor; | |
1409 | char *device; | |
1410 | char *c; | |
1411 | int port; | |
1412 | int type; | |
1413 | ||
1414 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1415 | continue; | |
1416 | path = devt_to_devpath(makedev(major, minor)); | |
1417 | if (!path) | |
1418 | continue; | |
1419 | if (!path_attached_to_hba(path, hba_path)) { | |
1420 | free(path); | |
1421 | path = NULL; | |
1422 | continue; | |
1423 | } | |
1424 | ||
1425 | /* retrieve the scsi device type */ | |
1426 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1427 | if (verbose) | |
1428 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1429 | err = 2; | |
1430 | break; | |
1431 | } | |
1432 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1433 | if (load_sys(device, buf) != 0) { | |
1434 | if (verbose) | |
1435 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1436 | path); | |
1437 | err = 2; | |
1438 | free(device); | |
1439 | break; | |
1440 | } | |
1441 | type = strtoul(buf, NULL, 10); | |
1442 | ||
1443 | /* if it's not a disk print the vendor and model */ | |
1444 | if (!(type == 0 || type == 7 || type == 14)) { | |
1445 | vendor[0] = '\0'; | |
1446 | model[0] = '\0'; | |
1447 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1448 | if (load_sys(device, buf) == 0) { | |
1449 | strncpy(vendor, buf, sizeof(vendor)); | |
1450 | vendor[sizeof(vendor) - 1] = '\0'; | |
1451 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1452 | while (isspace(*c) || *c == '\0') | |
1453 | *c-- = '\0'; | |
1454 | ||
1455 | } | |
1456 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1457 | if (load_sys(device, buf) == 0) { | |
1458 | strncpy(model, buf, sizeof(model)); | |
1459 | model[sizeof(model) - 1] = '\0'; | |
1460 | c = (char *) &model[sizeof(model) - 1]; | |
1461 | while (isspace(*c) || *c == '\0') | |
1462 | *c-- = '\0'; | |
1463 | } | |
1464 | ||
1465 | if (vendor[0] && model[0]) | |
1466 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1467 | else | |
1468 | switch (type) { /* numbers from hald/linux/device.c */ | |
1469 | case 1: sprintf(buf, "tape"); break; | |
1470 | case 2: sprintf(buf, "printer"); break; | |
1471 | case 3: sprintf(buf, "processor"); break; | |
1472 | case 4: | |
1473 | case 5: sprintf(buf, "cdrom"); break; | |
1474 | case 6: sprintf(buf, "scanner"); break; | |
1475 | case 8: sprintf(buf, "media_changer"); break; | |
1476 | case 9: sprintf(buf, "comm"); break; | |
1477 | case 12: sprintf(buf, "raid"); break; | |
1478 | default: sprintf(buf, "unknown"); | |
1479 | } | |
1480 | } else | |
1481 | buf[0] = '\0'; | |
1482 | free(device); | |
1483 | ||
1484 | /* chop device path to 'host%d' and calculate the port number */ | |
1485 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1486 | if (!c) { |
1487 | if (verbose) | |
1488 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1489 | err = 2; | |
1490 | break; | |
1491 | } | |
d665cc31 DW |
1492 | *c = '\0'; |
1493 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1494 | port -= host_base; | |
1495 | else { | |
1496 | if (verbose) { | |
1497 | *c = '/'; /* repair the full string */ | |
1498 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1499 | path); | |
1500 | } | |
1501 | err = 2; | |
1502 | break; | |
1503 | } | |
1504 | ||
1505 | /* mark this port as used */ | |
1506 | port_mask &= ~(1 << port); | |
1507 | ||
1508 | /* print out the device information */ | |
1509 | if (buf[0]) { | |
1510 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1511 | continue; | |
1512 | } | |
1513 | ||
1514 | fd = dev_open(ent->d_name, O_RDONLY); | |
1515 | if (fd < 0) | |
1516 | printf(" Port%d : - disk info unavailable -\n", port); | |
1517 | else { | |
1518 | fd2devname(fd, buf); | |
1519 | printf(" Port%d : %s", port, buf); | |
1520 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1521 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1522 | else |
664d5325 | 1523 | printf(" ()\n"); |
d665cc31 DW |
1524 | } |
1525 | close(fd); | |
1526 | free(path); | |
1527 | path = NULL; | |
1528 | } | |
1529 | if (path) | |
1530 | free(path); | |
1531 | if (dir) | |
1532 | closedir(dir); | |
1533 | if (err == 0) { | |
1534 | int i; | |
1535 | ||
1536 | for (i = 0; i < port_count; i++) | |
1537 | if (port_mask & (1 << i)) | |
1538 | printf(" Port%d : - no device attached -\n", i); | |
1539 | } | |
1540 | ||
1541 | return err; | |
1542 | } | |
1543 | ||
120dc887 LM |
1544 | static void print_found_intel_controllers(struct sys_dev *elem) |
1545 | { | |
1546 | for (; elem; elem = elem->next) { | |
1547 | fprintf(stderr, Name ": found Intel(R) "); | |
1548 | if (elem->type == SYS_DEV_SATA) | |
1549 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1550 | else if (elem->type == SYS_DEV_SAS) |
1551 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1552 | fprintf(stderr, "RAID controller"); |
1553 | if (elem->pci_id) | |
1554 | fprintf(stderr, " at %s", elem->pci_id); | |
1555 | fprintf(stderr, ".\n"); | |
1556 | } | |
1557 | fflush(stderr); | |
1558 | } | |
1559 | ||
120dc887 LM |
1560 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1561 | { | |
1562 | struct dirent *ent; | |
1563 | DIR *dir; | |
1564 | int host_base = -1; | |
1565 | ||
1566 | *port_count = 0; | |
1567 | if ((dir = opendir(hba_path)) == NULL) | |
1568 | return -1; | |
1569 | ||
1570 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1571 | int host; | |
1572 | ||
1573 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1574 | continue; | |
1575 | if (*port_count == 0) | |
1576 | host_base = host; | |
1577 | else if (host < host_base) | |
1578 | host_base = host; | |
1579 | ||
1580 | if (host + 1 > *port_count + host_base) | |
1581 | *port_count = host + 1 - host_base; | |
1582 | } | |
1583 | closedir(dir); | |
1584 | return host_base; | |
1585 | } | |
1586 | ||
a891a3c2 LM |
1587 | static void print_imsm_capability(const struct imsm_orom *orom) |
1588 | { | |
1589 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1590 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1591 | orom->hotfix_ver, orom->build); | |
1592 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1593 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1594 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1595 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1596 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1597 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1598 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1599 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1600 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1601 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1602 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1603 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1604 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1605 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1606 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1607 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1608 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1609 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1610 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1611 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1612 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1613 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1614 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
1615 | printf(" Max Disks : %d\n", orom->tds); | |
1616 | printf(" Max Volumes : %d\n", orom->vpa); | |
1617 | return; | |
1618 | } | |
1619 | ||
5615172f | 1620 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1621 | { |
1622 | /* There are two components to imsm platform support, the ahci SATA | |
1623 | * controller and the option-rom. To find the SATA controller we | |
1624 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1625 | * controller with the Intel vendor id is present. This approach | |
1626 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1627 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1628 | * detect platform raid capabilities. The option-rom resides in a | |
1629 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1630 | * platform capabilities. If raid support is disabled in the BIOS the | |
1631 | * option-rom capability structure will not be available. | |
1632 | */ | |
1633 | const struct imsm_orom *orom; | |
1634 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1635 | int host_base = 0; |
1636 | int port_count = 0; | |
120dc887 | 1637 | int result=0; |
d665cc31 | 1638 | |
5615172f | 1639 | if (enumerate_only) { |
a891a3c2 | 1640 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1641 | return 0; |
a891a3c2 LM |
1642 | list = find_intel_devices(); |
1643 | if (!list) | |
1644 | return 2; | |
1645 | for (hba = list; hba; hba = hba->next) { | |
1646 | orom = find_imsm_capability(hba->type); | |
1647 | if (!orom) { | |
1648 | result = 2; | |
1649 | break; | |
1650 | } | |
1651 | } | |
1652 | free_sys_dev(&list); | |
1653 | return result; | |
5615172f DW |
1654 | } |
1655 | ||
155cbb4c LM |
1656 | list = find_intel_devices(); |
1657 | if (!list) { | |
d665cc31 | 1658 | if (verbose) |
155cbb4c LM |
1659 | fprintf(stderr, Name ": no active Intel(R) RAID " |
1660 | "controller found.\n"); | |
d665cc31 DW |
1661 | free_sys_dev(&list); |
1662 | return 2; | |
1663 | } else if (verbose) | |
155cbb4c | 1664 | print_found_intel_controllers(list); |
d665cc31 | 1665 | |
a891a3c2 LM |
1666 | for (hba = list; hba; hba = hba->next) { |
1667 | orom = find_imsm_capability(hba->type); | |
1668 | if (!orom) | |
1669 | fprintf(stderr, Name ": imsm capabilities not found for controller: %s (type %s)\n", | |
1670 | hba->path, get_sys_dev_type(hba->type)); | |
1671 | else | |
1672 | print_imsm_capability(orom); | |
d665cc31 DW |
1673 | } |
1674 | ||
120dc887 LM |
1675 | for (hba = list; hba; hba = hba->next) { |
1676 | printf(" I/O Controller : %s (%s)\n", | |
1677 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1678 | |
120dc887 LM |
1679 | if (hba->type == SYS_DEV_SATA) { |
1680 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1681 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1682 | if (verbose) | |
1683 | fprintf(stderr, Name ": failed to enumerate " | |
1684 | "ports on SATA controller at %s.", hba->pci_id); | |
1685 | result |= 2; | |
1686 | } | |
1687 | } | |
d665cc31 | 1688 | } |
155cbb4c | 1689 | |
120dc887 LM |
1690 | free_sys_dev(&list); |
1691 | return result; | |
d665cc31 | 1692 | } |
cdddbdbc DW |
1693 | #endif |
1694 | ||
1695 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1696 | { | |
5115ca67 DW |
1697 | /* the imsm metadata format does not specify any host |
1698 | * identification information. We return -1 since we can never | |
1699 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1700 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1701 | * exclude member disks that do not belong, and we rely on |
1702 | * mdadm.conf to specify the arrays that should be assembled. | |
1703 | * Auto-assembly may still pick up "foreign" arrays. | |
1704 | */ | |
cdddbdbc | 1705 | |
9362c1c8 | 1706 | return -1; |
cdddbdbc DW |
1707 | } |
1708 | ||
1709 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1710 | { | |
51006d85 N |
1711 | /* The uuid returned here is used for: |
1712 | * uuid to put into bitmap file (Create, Grow) | |
1713 | * uuid for backup header when saving critical section (Grow) | |
1714 | * comparing uuids when re-adding a device into an array | |
1715 | * In these cases the uuid required is that of the data-array, | |
1716 | * not the device-set. | |
1717 | * uuid to recognise same set when adding a missing device back | |
1718 | * to an array. This is a uuid for the device-set. | |
1719 | * | |
1720 | * For each of these we can make do with a truncated | |
1721 | * or hashed uuid rather than the original, as long as | |
1722 | * everyone agrees. | |
1723 | * In each case the uuid required is that of the data-array, | |
1724 | * not the device-set. | |
43dad3d6 | 1725 | */ |
51006d85 N |
1726 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1727 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1728 | * - the orig_family_num of the container |
51006d85 N |
1729 | * - the index number of the volume |
1730 | * - the 'serial' number of the volume. | |
1731 | * Hopefully these are all constant. | |
1732 | */ | |
1733 | struct intel_super *super = st->sb; | |
43dad3d6 | 1734 | |
51006d85 N |
1735 | char buf[20]; |
1736 | struct sha1_ctx ctx; | |
1737 | struct imsm_dev *dev = NULL; | |
148acb7b | 1738 | __u32 family_num; |
51006d85 | 1739 | |
148acb7b DW |
1740 | /* some mdadm versions failed to set ->orig_family_num, in which |
1741 | * case fall back to ->family_num. orig_family_num will be | |
1742 | * fixed up with the first metadata update. | |
1743 | */ | |
1744 | family_num = super->anchor->orig_family_num; | |
1745 | if (family_num == 0) | |
1746 | family_num = super->anchor->family_num; | |
51006d85 | 1747 | sha1_init_ctx(&ctx); |
92bd8f8d | 1748 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1749 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1750 | if (super->current_vol >= 0) |
1751 | dev = get_imsm_dev(super, super->current_vol); | |
1752 | if (dev) { | |
1753 | __u32 vol = super->current_vol; | |
1754 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1755 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1756 | } | |
1757 | sha1_finish_ctx(&ctx, buf); | |
1758 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1759 | } |
1760 | ||
0d481d37 | 1761 | #if 0 |
4f5bc454 DW |
1762 | static void |
1763 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1764 | { |
cdddbdbc DW |
1765 | __u8 *v = get_imsm_version(mpb); |
1766 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1767 | char major[] = { 0, 0, 0 }; | |
1768 | char minor[] = { 0 ,0, 0 }; | |
1769 | char patch[] = { 0, 0, 0 }; | |
1770 | char *ver_parse[] = { major, minor, patch }; | |
1771 | int i, j; | |
1772 | ||
1773 | i = j = 0; | |
1774 | while (*v != '\0' && v < end) { | |
1775 | if (*v != '.' && j < 2) | |
1776 | ver_parse[i][j++] = *v; | |
1777 | else { | |
1778 | i++; | |
1779 | j = 0; | |
1780 | } | |
1781 | v++; | |
1782 | } | |
1783 | ||
4f5bc454 DW |
1784 | *m = strtol(minor, NULL, 0); |
1785 | *p = strtol(patch, NULL, 0); | |
1786 | } | |
0d481d37 | 1787 | #endif |
4f5bc454 | 1788 | |
1e5c6983 DW |
1789 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1790 | { | |
1791 | /* migr_strip_size when repairing or initializing parity */ | |
1792 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1793 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1794 | ||
1795 | switch (get_imsm_raid_level(map)) { | |
1796 | case 5: | |
1797 | case 10: | |
1798 | return chunk; | |
1799 | default: | |
1800 | return 128*1024 >> 9; | |
1801 | } | |
1802 | } | |
1803 | ||
1804 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1805 | { | |
1806 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1807 | * this is different than migr_strip_size_resync(), but it's good | |
1808 | * to be compatible | |
1809 | */ | |
1810 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1811 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1812 | ||
1813 | switch (get_imsm_raid_level(map)) { | |
1814 | case 1: | |
1815 | case 10: | |
1816 | if (map->num_members % map->num_domains == 0) | |
1817 | return 128*1024 >> 9; | |
1818 | else | |
1819 | return chunk; | |
1820 | case 5: | |
1821 | return max((__u32) 64*1024 >> 9, chunk); | |
1822 | default: | |
1823 | return 128*1024 >> 9; | |
1824 | } | |
1825 | } | |
1826 | ||
1827 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
1828 | { | |
1829 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1830 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1831 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); | |
1832 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
1833 | ||
1834 | return max((__u32) 1, hi_chunk / lo_chunk); | |
1835 | } | |
1836 | ||
1837 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
1838 | { | |
1839 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1840 | int level = get_imsm_raid_level(lo); | |
1841 | ||
1842 | if (level == 1 || level == 10) { | |
1843 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1844 | ||
1845 | return hi->num_domains; | |
1846 | } else | |
1847 | return num_stripes_per_unit_resync(dev); | |
1848 | } | |
1849 | ||
98130f40 | 1850 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
1851 | { |
1852 | /* named 'imsm_' because raid0, raid1 and raid10 | |
1853 | * counter-intuitively have the same number of data disks | |
1854 | */ | |
98130f40 | 1855 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
1856 | |
1857 | switch (get_imsm_raid_level(map)) { | |
1858 | case 0: | |
1859 | case 1: | |
1860 | case 10: | |
1861 | return map->num_members; | |
1862 | case 5: | |
1863 | return map->num_members - 1; | |
1864 | default: | |
1865 | dprintf("%s: unsupported raid level\n", __func__); | |
1866 | return 0; | |
1867 | } | |
1868 | } | |
1869 | ||
1870 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
1871 | { | |
1872 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1873 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1874 | ||
1875 | switch(get_imsm_raid_level(map)) { | |
1876 | case 1: | |
1877 | case 10: | |
1878 | return chunk * map->num_domains; | |
1879 | case 5: | |
1880 | return chunk * map->num_members; | |
1881 | default: | |
1882 | return chunk; | |
1883 | } | |
1884 | } | |
1885 | ||
1886 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
1887 | { | |
1888 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1889 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1890 | __u32 strip = block / chunk; | |
1891 | ||
1892 | switch (get_imsm_raid_level(map)) { | |
1893 | case 1: | |
1894 | case 10: { | |
1895 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
1896 | __u32 vol_stripe = vol_strip / map->num_members; | |
1897 | ||
1898 | return vol_stripe * chunk + block % chunk; | |
1899 | } case 5: { | |
1900 | __u32 stripe = strip / (map->num_members - 1); | |
1901 | ||
1902 | return stripe * chunk + block % chunk; | |
1903 | } | |
1904 | default: | |
1905 | return 0; | |
1906 | } | |
1907 | } | |
1908 | ||
c47b0ff6 AK |
1909 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1910 | struct imsm_dev *dev) | |
1e5c6983 DW |
1911 | { |
1912 | /* calculate the conversion factor between per member 'blocks' | |
1913 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
1914 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
1915 | */ | |
1916 | if (!dev->vol.migr_state) | |
1917 | return 0; | |
1918 | ||
1919 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
1920 | case MIGR_GEN_MIGR: { |
1921 | struct migr_record *migr_rec = super->migr_rec; | |
1922 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
1923 | } | |
1e5c6983 DW |
1924 | case MIGR_VERIFY: |
1925 | case MIGR_REPAIR: | |
1926 | case MIGR_INIT: { | |
1927 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1928 | __u32 stripes_per_unit; | |
1929 | __u32 blocks_per_unit; | |
1930 | __u32 parity_depth; | |
1931 | __u32 migr_chunk; | |
1932 | __u32 block_map; | |
1933 | __u32 block_rel; | |
1934 | __u32 segment; | |
1935 | __u32 stripe; | |
1936 | __u8 disks; | |
1937 | ||
1938 | /* yes, this is really the translation of migr_units to | |
1939 | * per-member blocks in the 'resync' case | |
1940 | */ | |
1941 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
1942 | migr_chunk = migr_strip_blocks_resync(dev); | |
98130f40 | 1943 | disks = imsm_num_data_members(dev, 0); |
1e5c6983 | 1944 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 1945 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
1946 | segment = blocks_per_unit / stripe; |
1947 | block_rel = blocks_per_unit - segment * stripe; | |
1948 | parity_depth = parity_segment_depth(dev); | |
1949 | block_map = map_migr_block(dev, block_rel); | |
1950 | return block_map + parity_depth * segment; | |
1951 | } | |
1952 | case MIGR_REBUILD: { | |
1953 | __u32 stripes_per_unit; | |
1954 | __u32 migr_chunk; | |
1955 | ||
1956 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
1957 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
1958 | return migr_chunk * stripes_per_unit; | |
1959 | } | |
1e5c6983 DW |
1960 | case MIGR_STATE_CHANGE: |
1961 | default: | |
1962 | return 0; | |
1963 | } | |
1964 | } | |
1965 | ||
c2c087e6 DW |
1966 | static int imsm_level_to_layout(int level) |
1967 | { | |
1968 | switch (level) { | |
1969 | case 0: | |
1970 | case 1: | |
1971 | return 0; | |
1972 | case 5: | |
1973 | case 6: | |
a380c027 | 1974 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 1975 | case 10: |
c92a2527 | 1976 | return 0x102; |
c2c087e6 | 1977 | } |
a18a888e | 1978 | return UnSet; |
c2c087e6 DW |
1979 | } |
1980 | ||
8e59f3d8 AK |
1981 | /******************************************************************************* |
1982 | * Function: read_imsm_migr_rec | |
1983 | * Description: Function reads imsm migration record from last sector of disk | |
1984 | * Parameters: | |
1985 | * fd : disk descriptor | |
1986 | * super : metadata info | |
1987 | * Returns: | |
1988 | * 0 : success, | |
1989 | * -1 : fail | |
1990 | ******************************************************************************/ | |
1991 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
1992 | { | |
1993 | int ret_val = -1; | |
1994 | unsigned long long dsize; | |
1995 | ||
1996 | get_dev_size(fd, NULL, &dsize); | |
1997 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
1998 | fprintf(stderr, | |
1999 | Name ": Cannot seek to anchor block: %s\n", | |
2000 | strerror(errno)); | |
2001 | goto out; | |
2002 | } | |
2003 | if (read(fd, super->migr_rec_buf, 512) != 512) { | |
2004 | fprintf(stderr, | |
2005 | Name ": Cannot read migr record block: %s\n", | |
2006 | strerror(errno)); | |
2007 | goto out; | |
2008 | } | |
2009 | ret_val = 0; | |
2010 | ||
2011 | out: | |
2012 | return ret_val; | |
2013 | } | |
2014 | ||
2015 | /******************************************************************************* | |
2016 | * Function: load_imsm_migr_rec | |
2017 | * Description: Function reads imsm migration record (it is stored at the last | |
2018 | * sector of disk) | |
2019 | * Parameters: | |
2020 | * super : imsm internal array info | |
2021 | * info : general array info | |
2022 | * Returns: | |
2023 | * 0 : success | |
2024 | * -1 : fail | |
2025 | ******************************************************************************/ | |
2026 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2027 | { | |
2028 | struct mdinfo *sd; | |
2029 | struct dl *dl = NULL; | |
2030 | char nm[30]; | |
2031 | int retval = -1; | |
2032 | int fd = -1; | |
2033 | ||
2034 | if (info) { | |
2035 | for (sd = info->devs ; sd ; sd = sd->next) { | |
2036 | /* read only from one of the first two slots */ | |
2037 | if ((sd->disk.raid_disk > 1) || | |
2038 | (sd->disk.raid_disk < 0)) | |
2039 | continue; | |
2040 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
2041 | fd = dev_open(nm, O_RDONLY); | |
2042 | if (fd >= 0) | |
2043 | break; | |
2044 | } | |
2045 | } | |
2046 | if (fd < 0) { | |
2047 | for (dl = super->disks; dl; dl = dl->next) { | |
2048 | /* read only from one of the first two slots */ | |
2049 | if (dl->index > 1) | |
2050 | continue; | |
2051 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2052 | fd = dev_open(nm, O_RDONLY); | |
2053 | if (fd >= 0) | |
2054 | break; | |
2055 | } | |
2056 | } | |
2057 | if (fd < 0) | |
2058 | goto out; | |
2059 | retval = read_imsm_migr_rec(fd, super); | |
2060 | ||
2061 | out: | |
2062 | if (fd >= 0) | |
2063 | close(fd); | |
2064 | return retval; | |
2065 | } | |
2066 | ||
9e2d750d | 2067 | #ifndef MDASSEMBLE |
c17608ea AK |
2068 | /******************************************************************************* |
2069 | * function: imsm_create_metadata_checkpoint_update | |
2070 | * Description: It creates update for checkpoint change. | |
2071 | * Parameters: | |
2072 | * super : imsm internal array info | |
2073 | * u : pointer to prepared update | |
2074 | * Returns: | |
2075 | * Uptate length. | |
2076 | * If length is equal to 0, input pointer u contains no update | |
2077 | ******************************************************************************/ | |
2078 | static int imsm_create_metadata_checkpoint_update( | |
2079 | struct intel_super *super, | |
2080 | struct imsm_update_general_migration_checkpoint **u) | |
2081 | { | |
2082 | ||
2083 | int update_memory_size = 0; | |
2084 | ||
2085 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2086 | ||
2087 | if (u == NULL) | |
2088 | return 0; | |
2089 | *u = NULL; | |
2090 | ||
2091 | /* size of all update data without anchor */ | |
2092 | update_memory_size = | |
2093 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2094 | ||
2095 | *u = calloc(1, update_memory_size); | |
2096 | if (*u == NULL) { | |
2097 | dprintf("error: cannot get memory for " | |
2098 | "imsm_create_metadata_checkpoint_update update\n"); | |
2099 | return 0; | |
2100 | } | |
2101 | (*u)->type = update_general_migration_checkpoint; | |
2102 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2103 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2104 | (*u)->curr_migr_unit); | |
2105 | ||
2106 | return update_memory_size; | |
2107 | } | |
2108 | ||
2109 | ||
2110 | static void imsm_update_metadata_locally(struct supertype *st, | |
2111 | void *buf, int len); | |
2112 | ||
687629c2 AK |
2113 | /******************************************************************************* |
2114 | * Function: write_imsm_migr_rec | |
2115 | * Description: Function writes imsm migration record | |
2116 | * (at the last sector of disk) | |
2117 | * Parameters: | |
2118 | * super : imsm internal array info | |
2119 | * Returns: | |
2120 | * 0 : success | |
2121 | * -1 : if fail | |
2122 | ******************************************************************************/ | |
2123 | static int write_imsm_migr_rec(struct supertype *st) | |
2124 | { | |
2125 | struct intel_super *super = st->sb; | |
2126 | unsigned long long dsize; | |
2127 | char nm[30]; | |
2128 | int fd = -1; | |
2129 | int retval = -1; | |
2130 | struct dl *sd; | |
c17608ea AK |
2131 | int len; |
2132 | struct imsm_update_general_migration_checkpoint *u; | |
687629c2 AK |
2133 | |
2134 | for (sd = super->disks ; sd ; sd = sd->next) { | |
2135 | /* write to 2 first slots only */ | |
2136 | if ((sd->index < 0) || (sd->index > 1)) | |
2137 | continue; | |
2138 | sprintf(nm, "%d:%d", sd->major, sd->minor); | |
2139 | fd = dev_open(nm, O_RDWR); | |
2140 | if (fd < 0) | |
2141 | continue; | |
2142 | get_dev_size(fd, NULL, &dsize); | |
2143 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
2144 | fprintf(stderr, | |
2145 | Name ": Cannot seek to anchor block: %s\n", | |
2146 | strerror(errno)); | |
2147 | goto out; | |
2148 | } | |
2149 | if (write(fd, super->migr_rec_buf, 512) != 512) { | |
2150 | fprintf(stderr, | |
2151 | Name ": Cannot write migr record block: %s\n", | |
2152 | strerror(errno)); | |
2153 | goto out; | |
2154 | } | |
2155 | close(fd); | |
2156 | fd = -1; | |
2157 | } | |
c17608ea AK |
2158 | /* update checkpoint information in metadata */ |
2159 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2160 | ||
2161 | if (len <= 0) { | |
2162 | dprintf("imsm: Cannot prepare update\n"); | |
2163 | goto out; | |
2164 | } | |
2165 | /* update metadata locally */ | |
2166 | imsm_update_metadata_locally(st, u, len); | |
2167 | /* and possibly remotely */ | |
2168 | if (st->update_tail) { | |
2169 | append_metadata_update(st, u, len); | |
2170 | /* during reshape we do all work inside metadata handler | |
2171 | * manage_reshape(), so metadata update has to be triggered | |
2172 | * insida it | |
2173 | */ | |
2174 | flush_metadata_updates(st); | |
2175 | st->update_tail = &st->updates; | |
2176 | } else | |
2177 | free(u); | |
687629c2 AK |
2178 | |
2179 | retval = 0; | |
2180 | out: | |
2181 | if (fd >= 0) | |
2182 | close(fd); | |
2183 | return retval; | |
2184 | } | |
9e2d750d | 2185 | #endif /* MDASSEMBLE */ |
687629c2 | 2186 | |
a5d85af7 | 2187 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2188 | { |
2189 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2190 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2191 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 2192 | struct imsm_map *map = get_imsm_map(dev, 0); |
81ac8b4d | 2193 | struct imsm_map *prev_map = get_imsm_map(dev, 1); |
b335e593 | 2194 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2195 | struct dl *dl; |
e207da2f | 2196 | char *devname; |
139dae11 | 2197 | unsigned int component_size_alligment; |
a5d85af7 | 2198 | int map_disks = info->array.raid_disks; |
bf5a934a | 2199 | |
95eeceeb | 2200 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2201 | if (prev_map) |
2202 | map_to_analyse = prev_map; | |
2203 | ||
ca0748fa | 2204 | dl = super->current_disk; |
9894ec0d | 2205 | |
bf5a934a | 2206 | info->container_member = super->current_vol; |
cd0430a1 | 2207 | info->array.raid_disks = map->num_members; |
b335e593 | 2208 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2209 | info->array.layout = imsm_level_to_layout(info->array.level); |
2210 | info->array.md_minor = -1; | |
2211 | info->array.ctime = 0; | |
2212 | info->array.utime = 0; | |
b335e593 AK |
2213 | info->array.chunk_size = |
2214 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2215 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2216 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2217 | info->custom_array_size <<= 32; | |
2218 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3f83228a N |
2219 | if (prev_map && map->map_state == prev_map->map_state) { |
2220 | info->reshape_active = 1; | |
b335e593 AK |
2221 | info->new_level = get_imsm_raid_level(map); |
2222 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2223 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2224 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2225 | if (info->delta_disks) { |
2226 | /* this needs to be applied to every array | |
2227 | * in the container. | |
2228 | */ | |
2229 | info->reshape_active = 2; | |
2230 | } | |
3f83228a N |
2231 | /* We shape information that we give to md might have to be |
2232 | * modify to cope with md's requirement for reshaping arrays. | |
2233 | * For example, when reshaping a RAID0, md requires it to be | |
2234 | * presented as a degraded RAID4. | |
2235 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2236 | * the array as already being RAID5, but the 'before' layout | |
2237 | * is a RAID4-like layout. | |
2238 | */ | |
2239 | switch (info->array.level) { | |
2240 | case 0: | |
2241 | switch(info->new_level) { | |
2242 | case 0: | |
2243 | /* conversion is happening as RAID4 */ | |
2244 | info->array.level = 4; | |
2245 | info->array.raid_disks += 1; | |
2246 | break; | |
2247 | case 5: | |
2248 | /* conversion is happening as RAID5 */ | |
2249 | info->array.level = 5; | |
2250 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2251 | info->delta_disks -= 1; |
2252 | break; | |
2253 | default: | |
2254 | /* FIXME error message */ | |
2255 | info->array.level = UnSet; | |
2256 | break; | |
2257 | } | |
2258 | break; | |
2259 | } | |
b335e593 AK |
2260 | } else { |
2261 | info->new_level = UnSet; | |
2262 | info->new_layout = UnSet; | |
2263 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2264 | info->delta_disks = 0; |
b335e593 | 2265 | } |
ca0748fa | 2266 | |
efb30e7f DW |
2267 | if (dl) { |
2268 | info->disk.major = dl->major; | |
2269 | info->disk.minor = dl->minor; | |
ca0748fa | 2270 | info->disk.number = dl->index; |
656b6b5a N |
2271 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2272 | dl->index); | |
efb30e7f | 2273 | } |
bf5a934a | 2274 | |
b335e593 AK |
2275 | info->data_offset = __le32_to_cpu(map_to_analyse->pba_of_lba0); |
2276 | info->component_size = | |
2277 | __le32_to_cpu(map_to_analyse->blocks_per_member); | |
139dae11 AK |
2278 | |
2279 | /* check component size aligment | |
2280 | */ | |
2281 | component_size_alligment = | |
2282 | info->component_size % (info->array.chunk_size/512); | |
2283 | ||
2284 | if (component_size_alligment && | |
2285 | (info->array.level != 1) && (info->array.level != UnSet)) { | |
2286 | dprintf("imsm: reported component size alligned from %llu ", | |
2287 | info->component_size); | |
2288 | info->component_size -= component_size_alligment; | |
2289 | dprintf("to %llu (%i).\n", | |
2290 | info->component_size, component_size_alligment); | |
2291 | } | |
2292 | ||
301406c9 | 2293 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2294 | info->recovery_start = MaxSector; |
bf5a934a | 2295 | |
d2e6d5d6 | 2296 | info->reshape_progress = 0; |
b6796ce1 | 2297 | info->resync_start = MaxSector; |
b335e593 AK |
2298 | if (map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2299 | dev->vol.dirty) { | |
301406c9 | 2300 | info->resync_start = 0; |
b6796ce1 AK |
2301 | } |
2302 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2303 | switch (migr_type(dev)) { |
2304 | case MIGR_REPAIR: | |
2305 | case MIGR_INIT: { | |
c47b0ff6 AK |
2306 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2307 | dev); | |
1e5c6983 DW |
2308 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2309 | ||
2310 | info->resync_start = blocks_per_unit * units; | |
2311 | break; | |
2312 | } | |
d2e6d5d6 | 2313 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2314 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2315 | dev); | |
2316 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2317 | unsigned long long array_blocks; |
2318 | int used_disks; | |
d2e6d5d6 | 2319 | |
befb629b AK |
2320 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2321 | (units < | |
2322 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2323 | (super->migr_rec->rec_status == | |
2324 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2325 | units++; | |
2326 | ||
d2e6d5d6 | 2327 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2328 | |
d2e6d5d6 AK |
2329 | dprintf("IMSM: General Migration checkpoint : %llu " |
2330 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2331 | (unsigned long long)units, |
2332 | (unsigned long long)blocks_per_unit, | |
2333 | info->reshape_progress); | |
75156c46 AK |
2334 | |
2335 | used_disks = imsm_num_data_members(dev, 1); | |
2336 | if (used_disks > 0) { | |
2337 | array_blocks = map->blocks_per_member * | |
2338 | used_disks; | |
2339 | /* round array size down to closest MB | |
2340 | */ | |
2341 | info->custom_array_size = (array_blocks | |
2342 | >> SECT_PER_MB_SHIFT) | |
2343 | << SECT_PER_MB_SHIFT; | |
2344 | } | |
d2e6d5d6 | 2345 | } |
1e5c6983 DW |
2346 | case MIGR_VERIFY: |
2347 | /* we could emulate the checkpointing of | |
2348 | * 'sync_action=check' migrations, but for now | |
2349 | * we just immediately complete them | |
2350 | */ | |
2351 | case MIGR_REBUILD: | |
2352 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2353 | case MIGR_STATE_CHANGE: |
2354 | /* FIXME handle other migrations */ | |
2355 | default: | |
2356 | /* we are not dirty, so... */ | |
2357 | info->resync_start = MaxSector; | |
2358 | } | |
b6796ce1 | 2359 | } |
301406c9 DW |
2360 | |
2361 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2362 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2363 | |
f35f2525 N |
2364 | info->array.major_version = -1; |
2365 | info->array.minor_version = -2; | |
e207da2f AW |
2366 | devname = devnum2devname(st->container_dev); |
2367 | *info->text_version = '\0'; | |
2368 | if (devname) | |
2369 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2370 | free(devname); | |
a67dd8cc | 2371 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2372 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2373 | |
2374 | if (dmap) { | |
2375 | int i, j; | |
2376 | for (i=0; i<map_disks; i++) { | |
2377 | dmap[i] = 0; | |
2378 | if (i < info->array.raid_disks) { | |
2379 | struct imsm_disk *dsk; | |
98130f40 | 2380 | j = get_imsm_disk_idx(dev, i, -1); |
a5d85af7 N |
2381 | dsk = get_imsm_disk(super, j); |
2382 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2383 | dmap[i] = 1; | |
2384 | } | |
2385 | } | |
2386 | } | |
81ac8b4d | 2387 | } |
bf5a934a | 2388 | |
97b4d0e9 DW |
2389 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed); |
2390 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev); | |
2391 | ||
2392 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2393 | { | |
2394 | struct dl *d; | |
2395 | ||
2396 | for (d = super->missing; d; d = d->next) | |
2397 | if (d->index == index) | |
2398 | return &d->disk; | |
2399 | return NULL; | |
2400 | } | |
2401 | ||
a5d85af7 | 2402 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2403 | { |
2404 | struct intel_super *super = st->sb; | |
4f5bc454 | 2405 | struct imsm_disk *disk; |
a5d85af7 | 2406 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2407 | int max_enough = -1; |
2408 | int i; | |
2409 | struct imsm_super *mpb; | |
4f5bc454 | 2410 | |
bf5a934a | 2411 | if (super->current_vol >= 0) { |
a5d85af7 | 2412 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2413 | return; |
2414 | } | |
95eeceeb | 2415 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2416 | |
2417 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2418 | * spares | |
2419 | */ | |
2420 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2421 | info->array.level = LEVEL_CONTAINER; |
2422 | info->array.layout = 0; | |
2423 | info->array.md_minor = -1; | |
c2c087e6 | 2424 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2425 | info->array.utime = 0; |
2426 | info->array.chunk_size = 0; | |
2427 | ||
2428 | info->disk.major = 0; | |
2429 | info->disk.minor = 0; | |
cdddbdbc | 2430 | info->disk.raid_disk = -1; |
c2c087e6 | 2431 | info->reshape_active = 0; |
f35f2525 N |
2432 | info->array.major_version = -1; |
2433 | info->array.minor_version = -2; | |
c2c087e6 | 2434 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2435 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2436 | info->disk.number = -1; |
2437 | info->disk.state = 0; | |
c5afc314 | 2438 | info->name[0] = 0; |
921d9e16 | 2439 | info->recovery_start = MaxSector; |
c2c087e6 | 2440 | |
97b4d0e9 | 2441 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2442 | mpb = super->anchor; |
97b4d0e9 | 2443 | |
ab3cb6b3 N |
2444 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2445 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2446 | int failed, enough, j, missing = 0; | |
2447 | struct imsm_map *map; | |
2448 | __u8 state; | |
97b4d0e9 | 2449 | |
ab3cb6b3 N |
2450 | failed = imsm_count_failed(super, dev); |
2451 | state = imsm_check_degraded(super, dev, failed); | |
2452 | map = get_imsm_map(dev, dev->vol.migr_state); | |
2453 | ||
2454 | /* any newly missing disks? | |
2455 | * (catches single-degraded vs double-degraded) | |
2456 | */ | |
2457 | for (j = 0; j < map->num_members; j++) { | |
98130f40 | 2458 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
ab3cb6b3 N |
2459 | __u32 idx = ord_to_idx(ord); |
2460 | ||
2461 | if (!(ord & IMSM_ORD_REBUILD) && | |
2462 | get_imsm_missing(super, idx)) { | |
2463 | missing = 1; | |
2464 | break; | |
2465 | } | |
97b4d0e9 | 2466 | } |
ab3cb6b3 N |
2467 | |
2468 | if (state == IMSM_T_STATE_FAILED) | |
2469 | enough = -1; | |
2470 | else if (state == IMSM_T_STATE_DEGRADED && | |
2471 | (state != map->map_state || missing)) | |
2472 | enough = 0; | |
2473 | else /* we're normal, or already degraded */ | |
2474 | enough = 1; | |
2475 | ||
2476 | /* in the missing/failed disk case check to see | |
2477 | * if at least one array is runnable | |
2478 | */ | |
2479 | max_enough = max(max_enough, enough); | |
2480 | } | |
2481 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2482 | info->container_enough = max_enough; | |
97b4d0e9 | 2483 | |
4a04ec6c | 2484 | if (super->disks) { |
14e8215b DW |
2485 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2486 | ||
b9f594fe | 2487 | disk = &super->disks->disk; |
14e8215b DW |
2488 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
2489 | info->component_size = reserved; | |
25ed7e59 | 2490 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2491 | /* we don't change info->disk.raid_disk here because |
2492 | * this state will be finalized in mdmon after we have | |
2493 | * found the 'most fresh' version of the metadata | |
2494 | */ | |
25ed7e59 DW |
2495 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2496 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2497 | } |
a575e2a7 DW |
2498 | |
2499 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2500 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2501 | */ | |
2502 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2503 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2504 | else |
2505 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2506 | |
2507 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2508 | if (map) { | |
2509 | int i; | |
2510 | for (i = 0; i < map_disks; i++) | |
2511 | map[i] = 1; | |
2512 | } | |
2513 | ||
cdddbdbc DW |
2514 | } |
2515 | ||
5c4cd5da AC |
2516 | /* allocates memory and fills disk in mdinfo structure |
2517 | * for each disk in array */ | |
2518 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2519 | { | |
2520 | struct mdinfo *mddev = NULL; | |
2521 | struct intel_super *super = st->sb; | |
2522 | struct imsm_disk *disk; | |
2523 | int count = 0; | |
2524 | struct dl *dl; | |
2525 | if (!super || !super->disks) | |
2526 | return NULL; | |
2527 | dl = super->disks; | |
2528 | mddev = malloc(sizeof(*mddev)); | |
2529 | if (!mddev) { | |
2530 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2531 | return NULL; | |
2532 | } | |
2533 | memset(mddev, 0, sizeof(*mddev)); | |
2534 | while (dl) { | |
2535 | struct mdinfo *tmp; | |
2536 | disk = &dl->disk; | |
2537 | tmp = malloc(sizeof(*tmp)); | |
2538 | if (!tmp) { | |
2539 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2540 | if (mddev) | |
2541 | sysfs_free(mddev); | |
2542 | return NULL; | |
2543 | } | |
2544 | memset(tmp, 0, sizeof(*tmp)); | |
2545 | if (mddev->devs) | |
2546 | tmp->next = mddev->devs; | |
2547 | mddev->devs = tmp; | |
2548 | tmp->disk.number = count++; | |
2549 | tmp->disk.major = dl->major; | |
2550 | tmp->disk.minor = dl->minor; | |
2551 | tmp->disk.state = is_configured(disk) ? | |
2552 | (1 << MD_DISK_ACTIVE) : 0; | |
2553 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2554 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2555 | tmp->disk.raid_disk = -1; | |
2556 | dl = dl->next; | |
2557 | } | |
2558 | return mddev; | |
2559 | } | |
2560 | ||
cdddbdbc DW |
2561 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2562 | char *update, char *devname, int verbose, | |
2563 | int uuid_set, char *homehost) | |
2564 | { | |
f352c545 DW |
2565 | /* For 'assemble' and 'force' we need to return non-zero if any |
2566 | * change was made. For others, the return value is ignored. | |
2567 | * Update options are: | |
2568 | * force-one : This device looks a bit old but needs to be included, | |
2569 | * update age info appropriately. | |
2570 | * assemble: clear any 'faulty' flag to allow this device to | |
2571 | * be assembled. | |
2572 | * force-array: Array is degraded but being forced, mark it clean | |
2573 | * if that will be needed to assemble it. | |
2574 | * | |
2575 | * newdev: not used ???? | |
2576 | * grow: Array has gained a new device - this is currently for | |
2577 | * linear only | |
2578 | * resync: mark as dirty so a resync will happen. | |
2579 | * name: update the name - preserving the homehost | |
6e46bf34 | 2580 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2581 | * |
2582 | * Following are not relevant for this imsm: | |
2583 | * sparc2.2 : update from old dodgey metadata | |
2584 | * super-minor: change the preferred_minor number | |
2585 | * summaries: update redundant counters. | |
f352c545 DW |
2586 | * homehost: update the recorded homehost |
2587 | * _reshape_progress: record new reshape_progress position. | |
2588 | */ | |
6e46bf34 DW |
2589 | int rv = 1; |
2590 | struct intel_super *super = st->sb; | |
2591 | struct imsm_super *mpb; | |
f352c545 | 2592 | |
6e46bf34 DW |
2593 | /* we can only update container info */ |
2594 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2595 | return 1; | |
2596 | ||
2597 | mpb = super->anchor; | |
2598 | ||
2599 | if (strcmp(update, "uuid") == 0 && uuid_set && !info->update_private) | |
1e2b2765 | 2600 | rv = -1; |
6e46bf34 DW |
2601 | else if (strcmp(update, "uuid") == 0 && uuid_set && info->update_private) { |
2602 | mpb->orig_family_num = *((__u32 *) info->update_private); | |
2603 | rv = 0; | |
2604 | } else if (strcmp(update, "uuid") == 0) { | |
2605 | __u32 *new_family = malloc(sizeof(*new_family)); | |
2606 | ||
2607 | /* update orig_family_number with the incoming random | |
2608 | * data, report the new effective uuid, and store the | |
2609 | * new orig_family_num for future updates. | |
2610 | */ | |
2611 | if (new_family) { | |
2612 | memcpy(&mpb->orig_family_num, info->uuid, sizeof(__u32)); | |
2613 | uuid_from_super_imsm(st, info->uuid); | |
2614 | *new_family = mpb->orig_family_num; | |
2615 | info->update_private = new_family; | |
2616 | rv = 0; | |
2617 | } | |
2618 | } else if (strcmp(update, "assemble") == 0) | |
2619 | rv = 0; | |
2620 | else | |
1e2b2765 | 2621 | rv = -1; |
f352c545 | 2622 | |
6e46bf34 DW |
2623 | /* successful update? recompute checksum */ |
2624 | if (rv == 0) | |
2625 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2626 | |
2627 | return rv; | |
cdddbdbc DW |
2628 | } |
2629 | ||
c2c087e6 | 2630 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2631 | { |
c2c087e6 | 2632 | size_t size; |
cdddbdbc | 2633 | |
c2c087e6 DW |
2634 | size = sizeof(struct imsm_super); |
2635 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2636 | size += 2 * sizeof(struct imsm_dev); | |
2637 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2638 | size += (4 - 2) * sizeof(struct imsm_map); | |
2639 | /* 4 possible disk_ord_tbl's */ | |
2640 | size += 4 * (disks - 1) * sizeof(__u32); | |
2641 | ||
2642 | return size; | |
2643 | } | |
2644 | ||
2645 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
2646 | { | |
2647 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
2648 | return 0; | |
2649 | ||
2650 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
2651 | } |
2652 | ||
ba2de7ba DW |
2653 | static void free_devlist(struct intel_super *super) |
2654 | { | |
2655 | struct intel_dev *dv; | |
2656 | ||
2657 | while (super->devlist) { | |
2658 | dv = super->devlist->next; | |
2659 | free(super->devlist->dev); | |
2660 | free(super->devlist); | |
2661 | super->devlist = dv; | |
2662 | } | |
2663 | } | |
2664 | ||
2665 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
2666 | { | |
2667 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
2668 | } | |
2669 | ||
cdddbdbc DW |
2670 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
2671 | { | |
2672 | /* | |
2673 | * return: | |
2674 | * 0 same, or first was empty, and second was copied | |
2675 | * 1 second had wrong number | |
2676 | * 2 wrong uuid | |
2677 | * 3 wrong other info | |
2678 | */ | |
2679 | struct intel_super *first = st->sb; | |
2680 | struct intel_super *sec = tst->sb; | |
2681 | ||
2682 | if (!first) { | |
2683 | st->sb = tst->sb; | |
2684 | tst->sb = NULL; | |
2685 | return 0; | |
2686 | } | |
8603ea6f LM |
2687 | /* in platform dependent environment test if the disks |
2688 | * use the same Intel hba | |
2689 | */ | |
2690 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
2691 | if (!first->hba || !sec->hba || |
2692 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
2693 | fprintf(stderr, |
2694 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
2695 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
2696 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
2697 | return 3; |
2698 | } | |
2699 | } | |
cdddbdbc | 2700 | |
d23fe947 DW |
2701 | /* if an anchor does not have num_raid_devs set then it is a free |
2702 | * floating spare | |
2703 | */ | |
2704 | if (first->anchor->num_raid_devs > 0 && | |
2705 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
2706 | /* Determine if these disks might ever have been |
2707 | * related. Further disambiguation can only take place | |
2708 | * in load_super_imsm_all | |
2709 | */ | |
2710 | __u32 first_family = first->anchor->orig_family_num; | |
2711 | __u32 sec_family = sec->anchor->orig_family_num; | |
2712 | ||
f796af5d DW |
2713 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
2714 | MAX_SIGNATURE_LENGTH) != 0) | |
2715 | return 3; | |
2716 | ||
a2b97981 DW |
2717 | if (first_family == 0) |
2718 | first_family = first->anchor->family_num; | |
2719 | if (sec_family == 0) | |
2720 | sec_family = sec->anchor->family_num; | |
2721 | ||
2722 | if (first_family != sec_family) | |
d23fe947 | 2723 | return 3; |
f796af5d | 2724 | |
d23fe947 | 2725 | } |
cdddbdbc | 2726 | |
f796af5d | 2727 | |
3e372e5a DW |
2728 | /* if 'first' is a spare promote it to a populated mpb with sec's |
2729 | * family number | |
2730 | */ | |
2731 | if (first->anchor->num_raid_devs == 0 && | |
2732 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 2733 | int i; |
ba2de7ba DW |
2734 | struct intel_dev *dv; |
2735 | struct imsm_dev *dev; | |
78d30f94 DW |
2736 | |
2737 | /* we need to copy raid device info from sec if an allocation | |
2738 | * fails here we don't associate the spare | |
2739 | */ | |
2740 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
2741 | dv = malloc(sizeof(*dv)); |
2742 | if (!dv) | |
2743 | break; | |
2744 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
2745 | if (!dev) { | |
2746 | free(dv); | |
2747 | break; | |
78d30f94 | 2748 | } |
ba2de7ba DW |
2749 | dv->dev = dev; |
2750 | dv->index = i; | |
2751 | dv->next = first->devlist; | |
2752 | first->devlist = dv; | |
78d30f94 | 2753 | } |
709743c5 | 2754 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
2755 | /* allocation failure */ |
2756 | free_devlist(first); | |
2757 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
2758 | return 3; | |
78d30f94 | 2759 | } |
3e372e5a | 2760 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 2761 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 2762 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 2763 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
2764 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
2765 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
2766 | } |
2767 | ||
cdddbdbc DW |
2768 | return 0; |
2769 | } | |
2770 | ||
0030e8d6 DW |
2771 | static void fd2devname(int fd, char *name) |
2772 | { | |
2773 | struct stat st; | |
2774 | char path[256]; | |
33a6535d | 2775 | char dname[PATH_MAX]; |
0030e8d6 DW |
2776 | char *nm; |
2777 | int rv; | |
2778 | ||
2779 | name[0] = '\0'; | |
2780 | if (fstat(fd, &st) != 0) | |
2781 | return; | |
2782 | sprintf(path, "/sys/dev/block/%d:%d", | |
2783 | major(st.st_rdev), minor(st.st_rdev)); | |
2784 | ||
2785 | rv = readlink(path, dname, sizeof(dname)); | |
2786 | if (rv <= 0) | |
2787 | return; | |
2788 | ||
2789 | dname[rv] = '\0'; | |
2790 | nm = strrchr(dname, '/'); | |
2791 | nm++; | |
2792 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
2793 | } | |
2794 | ||
cdddbdbc DW |
2795 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
2796 | ||
2797 | static int imsm_read_serial(int fd, char *devname, | |
2798 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
2799 | { | |
2800 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
2801 | int rv; |
2802 | int rsp_len; | |
1f24f035 | 2803 | int len; |
316e2bf4 DW |
2804 | char *dest; |
2805 | char *src; | |
2806 | char *rsp_buf; | |
2807 | int i; | |
cdddbdbc DW |
2808 | |
2809 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 2810 | |
f9ba0ff1 DW |
2811 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
2812 | ||
40ebbb9c | 2813 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
2814 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
2815 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
2816 | return 0; |
2817 | } | |
2818 | ||
cdddbdbc DW |
2819 | if (rv != 0) { |
2820 | if (devname) | |
2821 | fprintf(stderr, | |
2822 | Name ": Failed to retrieve serial for %s\n", | |
2823 | devname); | |
2824 | return rv; | |
2825 | } | |
2826 | ||
2827 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
2828 | if (!rsp_len) { |
2829 | if (devname) | |
2830 | fprintf(stderr, | |
2831 | Name ": Failed to retrieve serial for %s\n", | |
2832 | devname); | |
2833 | return 2; | |
2834 | } | |
1f24f035 | 2835 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 2836 | |
316e2bf4 DW |
2837 | /* trim all whitespace and non-printable characters and convert |
2838 | * ':' to ';' | |
2839 | */ | |
2840 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
2841 | src = &rsp_buf[i]; | |
2842 | if (*src > 0x20) { | |
2843 | /* ':' is reserved for use in placeholder serial | |
2844 | * numbers for missing disks | |
2845 | */ | |
2846 | if (*src == ':') | |
2847 | *dest++ = ';'; | |
2848 | else | |
2849 | *dest++ = *src; | |
2850 | } | |
2851 | } | |
2852 | len = dest - rsp_buf; | |
2853 | dest = rsp_buf; | |
2854 | ||
2855 | /* truncate leading characters */ | |
2856 | if (len > MAX_RAID_SERIAL_LEN) { | |
2857 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 2858 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 2859 | } |
5c3db629 | 2860 | |
5c3db629 | 2861 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 2862 | memcpy(serial, dest, len); |
cdddbdbc DW |
2863 | |
2864 | return 0; | |
2865 | } | |
2866 | ||
1f24f035 DW |
2867 | static int serialcmp(__u8 *s1, __u8 *s2) |
2868 | { | |
2869 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
2870 | } | |
2871 | ||
2872 | static void serialcpy(__u8 *dest, __u8 *src) | |
2873 | { | |
2874 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
2875 | } | |
2876 | ||
1799c9e8 | 2877 | #ifndef MDASSEMBLE |
54c2c1ea DW |
2878 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
2879 | { | |
2880 | struct dl *dl; | |
2881 | ||
2882 | for (dl = super->disks; dl; dl = dl->next) | |
2883 | if (serialcmp(dl->serial, serial) == 0) | |
2884 | break; | |
2885 | ||
2886 | return dl; | |
2887 | } | |
1799c9e8 | 2888 | #endif |
54c2c1ea | 2889 | |
a2b97981 DW |
2890 | static struct imsm_disk * |
2891 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
2892 | { | |
2893 | int i; | |
2894 | ||
2895 | for (i = 0; i < mpb->num_disks; i++) { | |
2896 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
2897 | ||
2898 | if (serialcmp(disk->serial, serial) == 0) { | |
2899 | if (idx) | |
2900 | *idx = i; | |
2901 | return disk; | |
2902 | } | |
2903 | } | |
2904 | ||
2905 | return NULL; | |
2906 | } | |
2907 | ||
cdddbdbc DW |
2908 | static int |
2909 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
2910 | { | |
a2b97981 | 2911 | struct imsm_disk *disk; |
cdddbdbc DW |
2912 | struct dl *dl; |
2913 | struct stat stb; | |
cdddbdbc | 2914 | int rv; |
a2b97981 | 2915 | char name[40]; |
d23fe947 DW |
2916 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
2917 | ||
2918 | rv = imsm_read_serial(fd, devname, serial); | |
2919 | ||
2920 | if (rv != 0) | |
2921 | return 2; | |
2922 | ||
a2b97981 | 2923 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 2924 | if (!dl) { |
cdddbdbc DW |
2925 | if (devname) |
2926 | fprintf(stderr, | |
2927 | Name ": failed to allocate disk buffer for %s\n", | |
2928 | devname); | |
2929 | return 2; | |
2930 | } | |
cdddbdbc | 2931 | |
a2b97981 DW |
2932 | fstat(fd, &stb); |
2933 | dl->major = major(stb.st_rdev); | |
2934 | dl->minor = minor(stb.st_rdev); | |
2935 | dl->next = super->disks; | |
2936 | dl->fd = keep_fd ? fd : -1; | |
2937 | assert(super->disks == NULL); | |
2938 | super->disks = dl; | |
2939 | serialcpy(dl->serial, serial); | |
2940 | dl->index = -2; | |
2941 | dl->e = NULL; | |
2942 | fd2devname(fd, name); | |
2943 | if (devname) | |
2944 | dl->devname = strdup(devname); | |
2945 | else | |
2946 | dl->devname = strdup(name); | |
cdddbdbc | 2947 | |
d23fe947 | 2948 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
2949 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
2950 | if (disk) { | |
2951 | dl->disk = *disk; | |
2952 | /* only set index on disks that are a member of a | |
2953 | * populated contianer, i.e. one with raid_devs | |
2954 | */ | |
2955 | if (is_failed(&dl->disk)) | |
3f6efecc | 2956 | dl->index = -2; |
a2b97981 DW |
2957 | else if (is_spare(&dl->disk)) |
2958 | dl->index = -1; | |
3f6efecc DW |
2959 | } |
2960 | ||
949c47a0 DW |
2961 | return 0; |
2962 | } | |
2963 | ||
0e600426 | 2964 | #ifndef MDASSEMBLE |
0c046afd DW |
2965 | /* When migrating map0 contains the 'destination' state while map1 |
2966 | * contains the current state. When not migrating map0 contains the | |
2967 | * current state. This routine assumes that map[0].map_state is set to | |
2968 | * the current array state before being called. | |
2969 | * | |
2970 | * Migration is indicated by one of the following states | |
2971 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 2972 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 2973 | * map1state=unitialized) |
1484e727 | 2974 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 2975 | * map1state=normal) |
e3bba0e0 | 2976 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 2977 | * map1state=degraded) |
8e59f3d8 AK |
2978 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
2979 | * map1state=normal) | |
0c046afd | 2980 | */ |
8e59f3d8 AK |
2981 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
2982 | __u8 to_state, int migr_type) | |
3393c6af | 2983 | { |
0c046afd | 2984 | struct imsm_map *dest; |
3393c6af DW |
2985 | struct imsm_map *src = get_imsm_map(dev, 0); |
2986 | ||
0c046afd | 2987 | dev->vol.migr_state = 1; |
1484e727 | 2988 | set_migr_type(dev, migr_type); |
f8f603f1 | 2989 | dev->vol.curr_migr_unit = 0; |
0c046afd DW |
2990 | dest = get_imsm_map(dev, 1); |
2991 | ||
0556e1a2 | 2992 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 2993 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
2994 | if ((migr_type == MIGR_REBUILD) || |
2995 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
2996 | __u32 ord; |
2997 | int i; | |
2998 | ||
2999 | for (i = 0; i < src->num_members; i++) { | |
3000 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3001 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3002 | } | |
3003 | } | |
3004 | ||
8e59f3d8 AK |
3005 | if (migr_type == MIGR_GEN_MIGR) |
3006 | /* Clear migration record */ | |
3007 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3008 | ||
0c046afd | 3009 | src->map_state = to_state; |
949c47a0 | 3010 | } |
f8f603f1 DW |
3011 | |
3012 | static void end_migration(struct imsm_dev *dev, __u8 map_state) | |
3013 | { | |
3014 | struct imsm_map *map = get_imsm_map(dev, 0); | |
0556e1a2 | 3015 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
28bce06f | 3016 | int i, j; |
0556e1a2 DW |
3017 | |
3018 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3019 | * completed in the last migration. | |
3020 | * | |
28bce06f | 3021 | * FIXME add support for raid-level-migration |
0556e1a2 DW |
3022 | */ |
3023 | for (i = 0; i < prev->num_members; i++) | |
28bce06f AK |
3024 | for (j = 0; j < map->num_members; j++) |
3025 | /* during online capacity expansion | |
3026 | * disks position can be changed if takeover is used | |
3027 | */ | |
3028 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3029 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3030 | map->disk_ord_tbl[j] |= prev->disk_ord_tbl[i]; | |
3031 | break; | |
3032 | } | |
f8f603f1 DW |
3033 | |
3034 | dev->vol.migr_state = 0; | |
28bce06f | 3035 | dev->vol.migr_type = 0; |
f8f603f1 DW |
3036 | dev->vol.curr_migr_unit = 0; |
3037 | map->map_state = map_state; | |
3038 | } | |
0e600426 | 3039 | #endif |
949c47a0 DW |
3040 | |
3041 | static int parse_raid_devices(struct intel_super *super) | |
3042 | { | |
3043 | int i; | |
3044 | struct imsm_dev *dev_new; | |
4d7b1503 | 3045 | size_t len, len_migr; |
401d313b | 3046 | size_t max_len = 0; |
4d7b1503 DW |
3047 | size_t space_needed = 0; |
3048 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3049 | |
3050 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3051 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3052 | struct intel_dev *dv; |
949c47a0 | 3053 | |
4d7b1503 DW |
3054 | len = sizeof_imsm_dev(dev_iter, 0); |
3055 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3056 | if (len_migr > len) | |
3057 | space_needed += len_migr - len; | |
3058 | ||
ba2de7ba DW |
3059 | dv = malloc(sizeof(*dv)); |
3060 | if (!dv) | |
3061 | return 1; | |
401d313b AK |
3062 | if (max_len < len_migr) |
3063 | max_len = len_migr; | |
3064 | if (max_len > len_migr) | |
3065 | space_needed += max_len - len_migr; | |
3066 | dev_new = malloc(max_len); | |
ba2de7ba DW |
3067 | if (!dev_new) { |
3068 | free(dv); | |
949c47a0 | 3069 | return 1; |
ba2de7ba | 3070 | } |
949c47a0 | 3071 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3072 | dv->dev = dev_new; |
3073 | dv->index = i; | |
3074 | dv->next = super->devlist; | |
3075 | super->devlist = dv; | |
949c47a0 | 3076 | } |
cdddbdbc | 3077 | |
4d7b1503 DW |
3078 | /* ensure that super->buf is large enough when all raid devices |
3079 | * are migrating | |
3080 | */ | |
3081 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3082 | void *buf; | |
3083 | ||
3084 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3085 | if (posix_memalign(&buf, 512, len) != 0) | |
3086 | return 1; | |
3087 | ||
1f45a8ad DW |
3088 | memcpy(buf, super->buf, super->len); |
3089 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3090 | free(super->buf); |
3091 | super->buf = buf; | |
3092 | super->len = len; | |
3093 | } | |
3094 | ||
cdddbdbc DW |
3095 | return 0; |
3096 | } | |
3097 | ||
604b746f JD |
3098 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3099 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3100 | { | |
3101 | void *ptr = NULL; | |
3102 | ||
3103 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3104 | ptr = mpb; | |
3105 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3106 | } | |
3107 | ||
3108 | return ptr; | |
3109 | } | |
3110 | ||
e2f41b2c AK |
3111 | /******************************************************************************* |
3112 | * Function: check_mpb_migr_compatibility | |
3113 | * Description: Function checks for unsupported migration features: | |
3114 | * - migration optimization area (pba_of_lba0) | |
3115 | * - descending reshape (ascending_migr) | |
3116 | * Parameters: | |
3117 | * super : imsm metadata information | |
3118 | * Returns: | |
3119 | * 0 : migration is compatible | |
3120 | * -1 : migration is not compatible | |
3121 | ******************************************************************************/ | |
3122 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3123 | { | |
3124 | struct imsm_map *map0, *map1; | |
3125 | struct migr_record *migr_rec = super->migr_rec; | |
3126 | int i; | |
3127 | ||
3128 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3129 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3130 | ||
3131 | if (dev_iter && | |
3132 | dev_iter->vol.migr_state == 1 && | |
3133 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3134 | /* This device is migrating */ | |
3135 | map0 = get_imsm_map(dev_iter, 0); | |
3136 | map1 = get_imsm_map(dev_iter, 1); | |
3137 | if (map0->pba_of_lba0 != map1->pba_of_lba0) | |
3138 | /* migration optimization area was used */ | |
3139 | return -1; | |
3140 | if (migr_rec->ascending_migr == 0 | |
3141 | && migr_rec->dest_depth_per_unit > 0) | |
3142 | /* descending reshape not supported yet */ | |
3143 | return -1; | |
3144 | } | |
3145 | } | |
3146 | return 0; | |
3147 | } | |
3148 | ||
d23fe947 | 3149 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3150 | |
cdddbdbc | 3151 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3152 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3153 | */ |
3154 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3155 | { | |
3156 | unsigned long long dsize; | |
cdddbdbc DW |
3157 | unsigned long long sectors; |
3158 | struct stat; | |
6416d527 | 3159 | struct imsm_super *anchor; |
cdddbdbc DW |
3160 | __u32 check_sum; |
3161 | ||
cdddbdbc | 3162 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3163 | if (dsize < 1024) { |
3164 | if (devname) | |
3165 | fprintf(stderr, | |
3166 | Name ": %s: device to small for imsm\n", | |
3167 | devname); | |
3168 | return 1; | |
3169 | } | |
cdddbdbc DW |
3170 | |
3171 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3172 | if (devname) | |
2e062e82 AK |
3173 | fprintf(stderr, Name |
3174 | ": Cannot seek to anchor block on %s: %s\n", | |
cdddbdbc DW |
3175 | devname, strerror(errno)); |
3176 | return 1; | |
3177 | } | |
3178 | ||
949c47a0 | 3179 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
3180 | if (devname) |
3181 | fprintf(stderr, | |
3182 | Name ": Failed to allocate imsm anchor buffer" | |
3183 | " on %s\n", devname); | |
3184 | return 1; | |
3185 | } | |
949c47a0 | 3186 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
3187 | if (devname) |
3188 | fprintf(stderr, | |
3189 | Name ": Cannot read anchor block on %s: %s\n", | |
3190 | devname, strerror(errno)); | |
6416d527 | 3191 | free(anchor); |
cdddbdbc DW |
3192 | return 1; |
3193 | } | |
3194 | ||
6416d527 | 3195 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
3196 | if (devname) |
3197 | fprintf(stderr, | |
3198 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 3199 | free(anchor); |
cdddbdbc DW |
3200 | return 2; |
3201 | } | |
3202 | ||
d23fe947 | 3203 | __free_imsm(super, 0); |
f2f5c343 LM |
3204 | /* reload capability and hba */ |
3205 | ||
3206 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3207 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3208 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3209 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
3210 | if (devname) |
3211 | fprintf(stderr, | |
3212 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 3213 | super->len); |
6416d527 | 3214 | free(anchor); |
cdddbdbc DW |
3215 | return 2; |
3216 | } | |
949c47a0 | 3217 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3218 | |
6416d527 NB |
3219 | sectors = mpb_sectors(anchor) - 1; |
3220 | free(anchor); | |
8e59f3d8 AK |
3221 | |
3222 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { | |
3223 | fprintf(stderr, Name | |
3224 | ": %s could not allocate migr_rec buffer\n", __func__); | |
3225 | free(super->buf); | |
3226 | return 2; | |
3227 | } | |
3228 | ||
949c47a0 | 3229 | if (!sectors) { |
ecf45690 DW |
3230 | check_sum = __gen_imsm_checksum(super->anchor); |
3231 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3232 | if (devname) | |
3233 | fprintf(stderr, | |
3234 | Name ": IMSM checksum %x != %x on %s\n", | |
3235 | check_sum, | |
3236 | __le32_to_cpu(super->anchor->check_sum), | |
3237 | devname); | |
3238 | return 2; | |
3239 | } | |
3240 | ||
a2b97981 | 3241 | return 0; |
949c47a0 | 3242 | } |
cdddbdbc DW |
3243 | |
3244 | /* read the extended mpb */ | |
3245 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3246 | if (devname) | |
3247 | fprintf(stderr, | |
3248 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
3249 | devname, strerror(errno)); | |
3250 | return 1; | |
3251 | } | |
3252 | ||
f21e18ca | 3253 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
3254 | if (devname) |
3255 | fprintf(stderr, | |
3256 | Name ": Cannot read extended mpb on %s: %s\n", | |
3257 | devname, strerror(errno)); | |
3258 | return 2; | |
3259 | } | |
3260 | ||
949c47a0 DW |
3261 | check_sum = __gen_imsm_checksum(super->anchor); |
3262 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
3263 | if (devname) |
3264 | fprintf(stderr, | |
3265 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 3266 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 3267 | devname); |
db575f3b | 3268 | return 3; |
cdddbdbc DW |
3269 | } |
3270 | ||
604b746f JD |
3271 | /* FIXME the BBM log is disk specific so we cannot use this global |
3272 | * buffer for all disks. Ok for now since we only look at the global | |
3273 | * bbm_log_size parameter to gate assembly | |
3274 | */ | |
3275 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3276 | ||
a2b97981 DW |
3277 | return 0; |
3278 | } | |
3279 | ||
8e59f3d8 AK |
3280 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3281 | ||
a2b97981 DW |
3282 | static int |
3283 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3284 | { | |
3285 | int err; | |
3286 | ||
3287 | err = load_imsm_mpb(fd, super, devname); | |
3288 | if (err) | |
3289 | return err; | |
3290 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3291 | if (err) | |
3292 | return err; | |
3293 | err = parse_raid_devices(super); | |
4d7b1503 | 3294 | |
a2b97981 | 3295 | return err; |
cdddbdbc DW |
3296 | } |
3297 | ||
ae6aad82 DW |
3298 | static void __free_imsm_disk(struct dl *d) |
3299 | { | |
3300 | if (d->fd >= 0) | |
3301 | close(d->fd); | |
3302 | if (d->devname) | |
3303 | free(d->devname); | |
0dcecb2e DW |
3304 | if (d->e) |
3305 | free(d->e); | |
ae6aad82 DW |
3306 | free(d); |
3307 | ||
3308 | } | |
1a64be56 | 3309 | |
cdddbdbc DW |
3310 | static void free_imsm_disks(struct intel_super *super) |
3311 | { | |
47ee5a45 | 3312 | struct dl *d; |
cdddbdbc | 3313 | |
47ee5a45 DW |
3314 | while (super->disks) { |
3315 | d = super->disks; | |
cdddbdbc | 3316 | super->disks = d->next; |
ae6aad82 | 3317 | __free_imsm_disk(d); |
cdddbdbc | 3318 | } |
cb82edca AK |
3319 | while (super->disk_mgmt_list) { |
3320 | d = super->disk_mgmt_list; | |
3321 | super->disk_mgmt_list = d->next; | |
3322 | __free_imsm_disk(d); | |
3323 | } | |
47ee5a45 DW |
3324 | while (super->missing) { |
3325 | d = super->missing; | |
3326 | super->missing = d->next; | |
3327 | __free_imsm_disk(d); | |
3328 | } | |
3329 | ||
cdddbdbc DW |
3330 | } |
3331 | ||
9ca2c81c | 3332 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3333 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3334 | { |
88654014 LM |
3335 | struct intel_hba *elem, *next; |
3336 | ||
9ca2c81c | 3337 | if (super->buf) { |
949c47a0 | 3338 | free(super->buf); |
9ca2c81c DW |
3339 | super->buf = NULL; |
3340 | } | |
f2f5c343 LM |
3341 | /* unlink capability description */ |
3342 | super->orom = NULL; | |
8e59f3d8 AK |
3343 | if (super->migr_rec_buf) { |
3344 | free(super->migr_rec_buf); | |
3345 | super->migr_rec_buf = NULL; | |
3346 | } | |
d23fe947 DW |
3347 | if (free_disks) |
3348 | free_imsm_disks(super); | |
ba2de7ba | 3349 | free_devlist(super); |
88654014 LM |
3350 | elem = super->hba; |
3351 | while (elem) { | |
3352 | if (elem->path) | |
3353 | free((void *)elem->path); | |
3354 | next = elem->next; | |
3355 | free(elem); | |
3356 | elem = next; | |
88c32bb1 | 3357 | } |
88654014 | 3358 | super->hba = NULL; |
cdddbdbc DW |
3359 | } |
3360 | ||
9ca2c81c DW |
3361 | static void free_imsm(struct intel_super *super) |
3362 | { | |
d23fe947 | 3363 | __free_imsm(super, 1); |
9ca2c81c DW |
3364 | free(super); |
3365 | } | |
cdddbdbc DW |
3366 | |
3367 | static void free_super_imsm(struct supertype *st) | |
3368 | { | |
3369 | struct intel_super *super = st->sb; | |
3370 | ||
3371 | if (!super) | |
3372 | return; | |
3373 | ||
3374 | free_imsm(super); | |
3375 | st->sb = NULL; | |
3376 | } | |
3377 | ||
49133e57 | 3378 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
3379 | { |
3380 | struct intel_super *super = malloc(sizeof(*super)); | |
3381 | ||
3382 | if (super) { | |
3383 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 3384 | super->current_vol = -1; |
0dcecb2e | 3385 | super->create_offset = ~((__u32 ) 0); |
c2c087e6 | 3386 | } |
c2c087e6 DW |
3387 | return super; |
3388 | } | |
3389 | ||
f0f5a016 LM |
3390 | /* |
3391 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3392 | */ | |
d424212e | 3393 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3394 | { |
3395 | struct sys_dev *hba_name; | |
3396 | int rv = 0; | |
3397 | ||
3398 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3399 | super->orom = NULL; |
f0f5a016 LM |
3400 | super->hba = NULL; |
3401 | return 0; | |
3402 | } | |
3403 | hba_name = find_disk_attached_hba(fd, NULL); | |
3404 | if (!hba_name) { | |
d424212e | 3405 | if (devname) |
f0f5a016 LM |
3406 | fprintf(stderr, |
3407 | Name ": %s is not attached to Intel(R) RAID controller.\n", | |
d424212e | 3408 | devname); |
f0f5a016 LM |
3409 | return 1; |
3410 | } | |
3411 | rv = attach_hba_to_super(super, hba_name); | |
3412 | if (rv == 2) { | |
d424212e N |
3413 | if (devname) { |
3414 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3415 | |
f0f5a016 LM |
3416 | fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID " |
3417 | "controller (%s),\n" | |
3418 | " but the container is assigned to Intel(R) " | |
3419 | "%s RAID controller (", | |
d424212e | 3420 | devname, |
f0f5a016 LM |
3421 | hba_name->path, |
3422 | hba_name->pci_id ? : "Err!", | |
3423 | get_sys_dev_type(hba_name->type)); | |
3424 | ||
f0f5a016 LM |
3425 | while (hba) { |
3426 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3427 | if (hba->next) | |
3428 | fprintf(stderr, ", "); | |
3429 | hba = hba->next; | |
3430 | } | |
3431 | ||
3432 | fprintf(stderr, ").\n" | |
3433 | " Mixing devices attached to different controllers " | |
3434 | "is not allowed.\n"); | |
3435 | } | |
3436 | free_sys_dev(&hba_name); | |
3437 | return 2; | |
3438 | } | |
f2f5c343 | 3439 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3440 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3441 | if (!super->orom) |
3442 | return 3; | |
f0f5a016 LM |
3443 | return 0; |
3444 | } | |
3445 | ||
cdddbdbc | 3446 | #ifndef MDASSEMBLE |
47ee5a45 DW |
3447 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3448 | * disks that have disappeared from the system. This routine relies on | |
3449 | * the mpb being uptodate, which it is at load time. | |
3450 | */ | |
3451 | static int find_missing(struct intel_super *super) | |
3452 | { | |
3453 | int i; | |
3454 | struct imsm_super *mpb = super->anchor; | |
3455 | struct dl *dl; | |
3456 | struct imsm_disk *disk; | |
47ee5a45 DW |
3457 | |
3458 | for (i = 0; i < mpb->num_disks; i++) { | |
3459 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3460 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3461 | if (dl) |
3462 | continue; | |
47ee5a45 DW |
3463 | |
3464 | dl = malloc(sizeof(*dl)); | |
3465 | if (!dl) | |
3466 | return 1; | |
3467 | dl->major = 0; | |
3468 | dl->minor = 0; | |
3469 | dl->fd = -1; | |
3470 | dl->devname = strdup("missing"); | |
3471 | dl->index = i; | |
3472 | serialcpy(dl->serial, disk->serial); | |
3473 | dl->disk = *disk; | |
689c9bf3 | 3474 | dl->e = NULL; |
47ee5a45 DW |
3475 | dl->next = super->missing; |
3476 | super->missing = dl; | |
3477 | } | |
3478 | ||
3479 | return 0; | |
3480 | } | |
3481 | ||
a2b97981 DW |
3482 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3483 | { | |
3484 | struct intel_disk *idisk = disk_list; | |
3485 | ||
3486 | while (idisk) { | |
3487 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3488 | break; | |
3489 | idisk = idisk->next; | |
3490 | } | |
3491 | ||
3492 | return idisk; | |
3493 | } | |
3494 | ||
3495 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3496 | struct intel_super *super, | |
3497 | struct intel_disk **disk_list) | |
3498 | { | |
3499 | struct imsm_disk *d = &super->disks->disk; | |
3500 | struct imsm_super *mpb = super->anchor; | |
3501 | int i, j; | |
3502 | ||
3503 | for (i = 0; i < tbl_size; i++) { | |
3504 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3505 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3506 | ||
3507 | if (tbl_mpb->family_num == mpb->family_num) { | |
3508 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3509 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3510 | __func__, super->disks->major, | |
3511 | super->disks->minor, | |
3512 | table[i]->disks->major, | |
3513 | table[i]->disks->minor); | |
3514 | break; | |
3515 | } | |
3516 | ||
3517 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3518 | is_configured(d) == is_configured(tbl_d)) && | |
3519 | tbl_mpb->generation_num < mpb->generation_num) { | |
3520 | /* current version of the mpb is a | |
3521 | * better candidate than the one in | |
3522 | * super_table, but copy over "cross | |
3523 | * generational" status | |
3524 | */ | |
3525 | struct intel_disk *idisk; | |
3526 | ||
3527 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3528 | __func__, super->disks->major, | |
3529 | super->disks->minor, | |
3530 | table[i]->disks->major, | |
3531 | table[i]->disks->minor); | |
3532 | ||
3533 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3534 | if (idisk && is_failed(&idisk->disk)) | |
3535 | tbl_d->status |= FAILED_DISK; | |
3536 | break; | |
3537 | } else { | |
3538 | struct intel_disk *idisk; | |
3539 | struct imsm_disk *disk; | |
3540 | ||
3541 | /* tbl_mpb is more up to date, but copy | |
3542 | * over cross generational status before | |
3543 | * returning | |
3544 | */ | |
3545 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3546 | if (disk && is_failed(disk)) | |
3547 | d->status |= FAILED_DISK; | |
3548 | ||
3549 | idisk = disk_list_get(d->serial, *disk_list); | |
3550 | if (idisk) { | |
3551 | idisk->owner = i; | |
3552 | if (disk && is_configured(disk)) | |
3553 | idisk->disk.status |= CONFIGURED_DISK; | |
3554 | } | |
3555 | ||
3556 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3557 | __func__, super->disks->major, | |
3558 | super->disks->minor, | |
3559 | table[i]->disks->major, | |
3560 | table[i]->disks->minor); | |
3561 | ||
3562 | return tbl_size; | |
3563 | } | |
3564 | } | |
3565 | } | |
3566 | ||
3567 | if (i >= tbl_size) | |
3568 | table[tbl_size++] = super; | |
3569 | else | |
3570 | table[i] = super; | |
3571 | ||
3572 | /* update/extend the merged list of imsm_disk records */ | |
3573 | for (j = 0; j < mpb->num_disks; j++) { | |
3574 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3575 | struct intel_disk *idisk; | |
3576 | ||
3577 | idisk = disk_list_get(disk->serial, *disk_list); | |
3578 | if (idisk) { | |
3579 | idisk->disk.status |= disk->status; | |
3580 | if (is_configured(&idisk->disk) || | |
3581 | is_failed(&idisk->disk)) | |
3582 | idisk->disk.status &= ~(SPARE_DISK); | |
3583 | } else { | |
3584 | idisk = calloc(1, sizeof(*idisk)); | |
3585 | if (!idisk) | |
3586 | return -1; | |
3587 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
3588 | idisk->disk = *disk; | |
3589 | idisk->next = *disk_list; | |
3590 | *disk_list = idisk; | |
3591 | } | |
3592 | ||
3593 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3594 | idisk->owner = i; | |
3595 | } | |
3596 | ||
3597 | return tbl_size; | |
3598 | } | |
3599 | ||
3600 | static struct intel_super * | |
3601 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3602 | const int owner) | |
3603 | { | |
3604 | struct imsm_super *mpb = super->anchor; | |
3605 | int ok_count = 0; | |
3606 | int i; | |
3607 | ||
3608 | for (i = 0; i < mpb->num_disks; i++) { | |
3609 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3610 | struct intel_disk *idisk; | |
3611 | ||
3612 | idisk = disk_list_get(disk->serial, disk_list); | |
3613 | if (idisk) { | |
3614 | if (idisk->owner == owner || | |
3615 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3616 | ok_count++; | |
3617 | else | |
3618 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3619 | __func__, disk->serial, idisk->owner, | |
3620 | owner); | |
3621 | } else { | |
3622 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
3623 | __func__, __le32_to_cpu(mpb->family_num), i, | |
3624 | disk->serial); | |
3625 | break; | |
3626 | } | |
3627 | } | |
3628 | ||
3629 | if (ok_count == mpb->num_disks) | |
3630 | return super; | |
3631 | return NULL; | |
3632 | } | |
3633 | ||
3634 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
3635 | { | |
3636 | struct intel_super *s; | |
3637 | ||
3638 | for (s = super_list; s; s = s->next) { | |
3639 | if (family_num != s->anchor->family_num) | |
3640 | continue; | |
3641 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
3642 | __le32_to_cpu(family_num), s->disks->devname); | |
3643 | } | |
3644 | } | |
3645 | ||
3646 | static struct intel_super * | |
3647 | imsm_thunderdome(struct intel_super **super_list, int len) | |
3648 | { | |
3649 | struct intel_super *super_table[len]; | |
3650 | struct intel_disk *disk_list = NULL; | |
3651 | struct intel_super *champion, *spare; | |
3652 | struct intel_super *s, **del; | |
3653 | int tbl_size = 0; | |
3654 | int conflict; | |
3655 | int i; | |
3656 | ||
3657 | memset(super_table, 0, sizeof(super_table)); | |
3658 | for (s = *super_list; s; s = s->next) | |
3659 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
3660 | ||
3661 | for (i = 0; i < tbl_size; i++) { | |
3662 | struct imsm_disk *d; | |
3663 | struct intel_disk *idisk; | |
3664 | struct imsm_super *mpb = super_table[i]->anchor; | |
3665 | ||
3666 | s = super_table[i]; | |
3667 | d = &s->disks->disk; | |
3668 | ||
3669 | /* 'd' must appear in merged disk list for its | |
3670 | * configuration to be valid | |
3671 | */ | |
3672 | idisk = disk_list_get(d->serial, disk_list); | |
3673 | if (idisk && idisk->owner == i) | |
3674 | s = validate_members(s, disk_list, i); | |
3675 | else | |
3676 | s = NULL; | |
3677 | ||
3678 | if (!s) | |
3679 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
3680 | __func__, mpb->family_num, | |
3681 | super_table[i]->disks->major, | |
3682 | super_table[i]->disks->minor); | |
3683 | super_table[i] = s; | |
3684 | } | |
3685 | ||
3686 | /* This is where the mdadm implementation differs from the Windows | |
3687 | * driver which has no strict concept of a container. We can only | |
3688 | * assemble one family from a container, so when returning a prodigal | |
3689 | * array member to this system the code will not be able to disambiguate | |
3690 | * the container contents that should be assembled ("foreign" versus | |
3691 | * "local"). It requires user intervention to set the orig_family_num | |
3692 | * to a new value to establish a new container. The Windows driver in | |
3693 | * this situation fixes up the volume name in place and manages the | |
3694 | * foreign array as an independent entity. | |
3695 | */ | |
3696 | s = NULL; | |
3697 | spare = NULL; | |
3698 | conflict = 0; | |
3699 | for (i = 0; i < tbl_size; i++) { | |
3700 | struct intel_super *tbl_ent = super_table[i]; | |
3701 | int is_spare = 0; | |
3702 | ||
3703 | if (!tbl_ent) | |
3704 | continue; | |
3705 | ||
3706 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
3707 | spare = tbl_ent; | |
3708 | is_spare = 1; | |
3709 | } | |
3710 | ||
3711 | if (s && !is_spare) { | |
3712 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
3713 | conflict++; | |
3714 | } else if (!s && !is_spare) | |
3715 | s = tbl_ent; | |
3716 | } | |
3717 | ||
3718 | if (!s) | |
3719 | s = spare; | |
3720 | if (!s) { | |
3721 | champion = NULL; | |
3722 | goto out; | |
3723 | } | |
3724 | champion = s; | |
3725 | ||
3726 | if (conflict) | |
3727 | fprintf(stderr, "Chose family %#x on '%s', " | |
3728 | "assemble conflicts to new container with '--update=uuid'\n", | |
3729 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
3730 | ||
3731 | /* collect all dl's onto 'champion', and update them to | |
3732 | * champion's version of the status | |
3733 | */ | |
3734 | for (s = *super_list; s; s = s->next) { | |
3735 | struct imsm_super *mpb = champion->anchor; | |
3736 | struct dl *dl = s->disks; | |
3737 | ||
3738 | if (s == champion) | |
3739 | continue; | |
3740 | ||
3741 | for (i = 0; i < mpb->num_disks; i++) { | |
3742 | struct imsm_disk *disk; | |
3743 | ||
3744 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
3745 | if (disk) { | |
3746 | dl->disk = *disk; | |
3747 | /* only set index on disks that are a member of | |
3748 | * a populated contianer, i.e. one with | |
3749 | * raid_devs | |
3750 | */ | |
3751 | if (is_failed(&dl->disk)) | |
3752 | dl->index = -2; | |
3753 | else if (is_spare(&dl->disk)) | |
3754 | dl->index = -1; | |
3755 | break; | |
3756 | } | |
3757 | } | |
3758 | ||
3759 | if (i >= mpb->num_disks) { | |
3760 | struct intel_disk *idisk; | |
3761 | ||
3762 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 3763 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
3764 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
3765 | dl->index = -1; | |
3766 | else { | |
3767 | dl->index = -2; | |
3768 | continue; | |
3769 | } | |
3770 | } | |
3771 | ||
3772 | dl->next = champion->disks; | |
3773 | champion->disks = dl; | |
3774 | s->disks = NULL; | |
3775 | } | |
3776 | ||
3777 | /* delete 'champion' from super_list */ | |
3778 | for (del = super_list; *del; ) { | |
3779 | if (*del == champion) { | |
3780 | *del = (*del)->next; | |
3781 | break; | |
3782 | } else | |
3783 | del = &(*del)->next; | |
3784 | } | |
3785 | champion->next = NULL; | |
3786 | ||
3787 | out: | |
3788 | while (disk_list) { | |
3789 | struct intel_disk *idisk = disk_list; | |
3790 | ||
3791 | disk_list = disk_list->next; | |
3792 | free(idisk); | |
3793 | } | |
3794 | ||
3795 | return champion; | |
3796 | } | |
3797 | ||
cdddbdbc | 3798 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
e1902a7b | 3799 | char *devname) |
cdddbdbc DW |
3800 | { |
3801 | struct mdinfo *sra; | |
a2b97981 DW |
3802 | struct intel_super *super_list = NULL; |
3803 | struct intel_super *super = NULL; | |
db575f3b | 3804 | int devnum = fd2devnum(fd); |
a2b97981 | 3805 | struct mdinfo *sd; |
db575f3b | 3806 | int retry; |
a2b97981 DW |
3807 | int err = 0; |
3808 | int i; | |
dab4a513 DW |
3809 | |
3810 | /* check if 'fd' an opened container */ | |
b526e52d | 3811 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
cdddbdbc DW |
3812 | if (!sra) |
3813 | return 1; | |
3814 | ||
3815 | if (sra->array.major_version != -1 || | |
3816 | sra->array.minor_version != -2 || | |
1602d52c AW |
3817 | strcmp(sra->text_version, "imsm") != 0) { |
3818 | err = 1; | |
3819 | goto error; | |
3820 | } | |
a2b97981 DW |
3821 | /* load all mpbs */ |
3822 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
49133e57 | 3823 | struct intel_super *s = alloc_super(); |
7a6ecd55 | 3824 | char nm[32]; |
a2b97981 | 3825 | int dfd; |
f2f5c343 | 3826 | int rv; |
a2b97981 DW |
3827 | |
3828 | err = 1; | |
3829 | if (!s) | |
3830 | goto error; | |
3831 | s->next = super_list; | |
3832 | super_list = s; | |
cdddbdbc | 3833 | |
a2b97981 | 3834 | err = 2; |
cdddbdbc | 3835 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3836 | dfd = dev_open(nm, O_RDWR); |
a2b97981 DW |
3837 | if (dfd < 0) |
3838 | goto error; | |
3839 | ||
d424212e | 3840 | rv = find_intel_hba_capability(dfd, s, devname); |
f2f5c343 LM |
3841 | /* no orom/efi or non-intel hba of the disk */ |
3842 | if (rv != 0) | |
3843 | goto error; | |
3844 | ||
e1902a7b | 3845 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
db575f3b DW |
3846 | |
3847 | /* retry the load if we might have raced against mdmon */ | |
a2b97981 | 3848 | if (err == 3 && mdmon_running(devnum)) |
db575f3b DW |
3849 | for (retry = 0; retry < 3; retry++) { |
3850 | usleep(3000); | |
e1902a7b | 3851 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
a2b97981 | 3852 | if (err != 3) |
db575f3b DW |
3853 | break; |
3854 | } | |
a2b97981 DW |
3855 | if (err) |
3856 | goto error; | |
cdddbdbc DW |
3857 | } |
3858 | ||
a2b97981 DW |
3859 | /* all mpbs enter, maybe one leaves */ |
3860 | super = imsm_thunderdome(&super_list, i); | |
3861 | if (!super) { | |
3862 | err = 1; | |
3863 | goto error; | |
cdddbdbc DW |
3864 | } |
3865 | ||
47ee5a45 DW |
3866 | if (find_missing(super) != 0) { |
3867 | free_imsm(super); | |
a2b97981 DW |
3868 | err = 2; |
3869 | goto error; | |
47ee5a45 | 3870 | } |
8e59f3d8 AK |
3871 | |
3872 | /* load migration record */ | |
3873 | err = load_imsm_migr_rec(super, NULL); | |
3874 | if (err) { | |
3875 | err = 4; | |
3876 | goto error; | |
3877 | } | |
e2f41b2c AK |
3878 | |
3879 | /* Check migration compatibility */ | |
3880 | if (check_mpb_migr_compatibility(super) != 0) { | |
3881 | fprintf(stderr, Name ": Unsupported migration detected"); | |
3882 | if (devname) | |
3883 | fprintf(stderr, " on %s\n", devname); | |
3884 | else | |
3885 | fprintf(stderr, " (IMSM).\n"); | |
3886 | ||
3887 | err = 5; | |
3888 | goto error; | |
3889 | } | |
3890 | ||
a2b97981 DW |
3891 | err = 0; |
3892 | ||
3893 | error: | |
3894 | while (super_list) { | |
3895 | struct intel_super *s = super_list; | |
3896 | ||
3897 | super_list = super_list->next; | |
3898 | free_imsm(s); | |
3899 | } | |
1602d52c | 3900 | sysfs_free(sra); |
a2b97981 DW |
3901 | |
3902 | if (err) | |
3903 | return err; | |
f7e7067b | 3904 | |
cdddbdbc | 3905 | *sbp = super; |
db575f3b | 3906 | st->container_dev = devnum; |
a2b97981 | 3907 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 3908 | st->ss = &super_imsm; |
cdddbdbc DW |
3909 | st->minor_version = 0; |
3910 | st->max_devs = IMSM_MAX_DEVICES; | |
3911 | } | |
cdddbdbc DW |
3912 | return 0; |
3913 | } | |
2b959fbf N |
3914 | |
3915 | static int load_container_imsm(struct supertype *st, int fd, char *devname) | |
3916 | { | |
3917 | return load_super_imsm_all(st, fd, &st->sb, devname); | |
3918 | } | |
cdddbdbc DW |
3919 | #endif |
3920 | ||
3921 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
3922 | { | |
3923 | struct intel_super *super; | |
3924 | int rv; | |
3925 | ||
691c6ee1 N |
3926 | if (test_partition(fd)) |
3927 | /* IMSM not allowed on partitions */ | |
3928 | return 1; | |
3929 | ||
37424f13 DW |
3930 | free_super_imsm(st); |
3931 | ||
49133e57 | 3932 | super = alloc_super(); |
cdddbdbc DW |
3933 | if (!super) { |
3934 | fprintf(stderr, | |
3935 | Name ": malloc of %zu failed.\n", | |
3936 | sizeof(*super)); | |
3937 | return 1; | |
3938 | } | |
ea2bc72b LM |
3939 | /* Load hba and capabilities if they exist. |
3940 | * But do not preclude loading metadata in case capabilities or hba are | |
3941 | * non-compliant and ignore_hw_compat is set. | |
3942 | */ | |
d424212e | 3943 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 3944 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 3945 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 LM |
3946 | if (devname) |
3947 | fprintf(stderr, | |
3948 | Name ": No OROM/EFI properties for %s\n", devname); | |
3949 | free_imsm(super); | |
3950 | return 2; | |
3951 | } | |
a2b97981 | 3952 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
3953 | |
3954 | if (rv) { | |
3955 | if (devname) | |
3956 | fprintf(stderr, | |
3957 | Name ": Failed to load all information " | |
3958 | "sections on %s\n", devname); | |
3959 | free_imsm(super); | |
3960 | return rv; | |
3961 | } | |
3962 | ||
3963 | st->sb = super; | |
3964 | if (st->ss == NULL) { | |
3965 | st->ss = &super_imsm; | |
3966 | st->minor_version = 0; | |
3967 | st->max_devs = IMSM_MAX_DEVICES; | |
3968 | } | |
8e59f3d8 AK |
3969 | |
3970 | /* load migration record */ | |
2e062e82 AK |
3971 | if (load_imsm_migr_rec(super, NULL) == 0) { |
3972 | /* Check for unsupported migration features */ | |
3973 | if (check_mpb_migr_compatibility(super) != 0) { | |
3974 | fprintf(stderr, | |
3975 | Name ": Unsupported migration detected"); | |
3976 | if (devname) | |
3977 | fprintf(stderr, " on %s\n", devname); | |
3978 | else | |
3979 | fprintf(stderr, " (IMSM).\n"); | |
3980 | return 3; | |
3981 | } | |
e2f41b2c AK |
3982 | } |
3983 | ||
cdddbdbc DW |
3984 | return 0; |
3985 | } | |
3986 | ||
ef6ffade DW |
3987 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
3988 | { | |
3989 | if (info->level == 1) | |
3990 | return 128; | |
3991 | return info->chunk_size >> 9; | |
3992 | } | |
3993 | ||
ff596308 | 3994 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains) |
ef6ffade DW |
3995 | { |
3996 | __u32 num_stripes; | |
3997 | ||
3998 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
ff596308 | 3999 | num_stripes /= num_domains; |
ef6ffade DW |
4000 | |
4001 | return num_stripes; | |
4002 | } | |
4003 | ||
fcfd9599 DW |
4004 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
4005 | { | |
4025c288 DW |
4006 | if (info->level == 1) |
4007 | return info->size * 2; | |
4008 | else | |
4009 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
fcfd9599 DW |
4010 | } |
4011 | ||
4d1313e9 DW |
4012 | static void imsm_update_version_info(struct intel_super *super) |
4013 | { | |
4014 | /* update the version and attributes */ | |
4015 | struct imsm_super *mpb = super->anchor; | |
4016 | char *version; | |
4017 | struct imsm_dev *dev; | |
4018 | struct imsm_map *map; | |
4019 | int i; | |
4020 | ||
4021 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4022 | dev = get_imsm_dev(super, i); | |
4023 | map = get_imsm_map(dev, 0); | |
4024 | if (__le32_to_cpu(dev->size_high) > 0) | |
4025 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4026 | ||
4027 | /* FIXME detect when an array spans a port multiplier */ | |
4028 | #if 0 | |
4029 | mpb->attributes |= MPB_ATTRIB_PM; | |
4030 | #endif | |
4031 | ||
4032 | if (mpb->num_raid_devs > 1 || | |
4033 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4034 | version = MPB_VERSION_ATTRIBS; | |
4035 | switch (get_imsm_raid_level(map)) { | |
4036 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4037 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4038 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4039 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4040 | } | |
4041 | } else { | |
4042 | if (map->num_members >= 5) | |
4043 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4044 | else if (dev->status == DEV_CLONE_N_GO) | |
4045 | version = MPB_VERSION_CNG; | |
4046 | else if (get_imsm_raid_level(map) == 5) | |
4047 | version = MPB_VERSION_RAID5; | |
4048 | else if (map->num_members >= 3) | |
4049 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4050 | else if (get_imsm_raid_level(map) == 1) | |
4051 | version = MPB_VERSION_RAID1; | |
4052 | else | |
4053 | version = MPB_VERSION_RAID0; | |
4054 | } | |
4055 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4056 | } | |
4057 | } | |
4058 | ||
aa534678 DW |
4059 | static int check_name(struct intel_super *super, char *name, int quiet) |
4060 | { | |
4061 | struct imsm_super *mpb = super->anchor; | |
4062 | char *reason = NULL; | |
4063 | int i; | |
4064 | ||
4065 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4066 | reason = "must be 16 characters or less"; | |
4067 | ||
4068 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4069 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4070 | ||
4071 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4072 | reason = "already exists"; | |
4073 | break; | |
4074 | } | |
4075 | } | |
4076 | ||
4077 | if (reason && !quiet) | |
4078 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
4079 | ||
4080 | return !reason; | |
4081 | } | |
4082 | ||
8b353278 DW |
4083 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4084 | unsigned long long size, char *name, | |
4085 | char *homehost, int *uuid) | |
cdddbdbc | 4086 | { |
c2c087e6 DW |
4087 | /* We are creating a volume inside a pre-existing container. |
4088 | * so st->sb is already set. | |
4089 | */ | |
4090 | struct intel_super *super = st->sb; | |
949c47a0 | 4091 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4092 | struct intel_dev *dv; |
c2c087e6 DW |
4093 | struct imsm_dev *dev; |
4094 | struct imsm_vol *vol; | |
4095 | struct imsm_map *map; | |
4096 | int idx = mpb->num_raid_devs; | |
4097 | int i; | |
4098 | unsigned long long array_blocks; | |
2c092cad | 4099 | size_t size_old, size_new; |
ff596308 | 4100 | __u32 num_data_stripes; |
cdddbdbc | 4101 | |
88c32bb1 | 4102 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 4103 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 4104 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4105 | return 0; |
4106 | } | |
4107 | ||
2c092cad DW |
4108 | /* ensure the mpb is large enough for the new data */ |
4109 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4110 | size_new = disks_to_mpb_size(info->nr_disks); | |
4111 | if (size_new > size_old) { | |
4112 | void *mpb_new; | |
4113 | size_t size_round = ROUND_UP(size_new, 512); | |
4114 | ||
4115 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
4116 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
4117 | return 0; | |
4118 | } | |
8e59f3d8 AK |
4119 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
4120 | fprintf(stderr, Name | |
4121 | ": %s could not allocate migr_rec buffer\n", | |
4122 | __func__); | |
4123 | free(super->buf); | |
4124 | free(super); | |
4125 | return 0; | |
4126 | } | |
2c092cad DW |
4127 | memcpy(mpb_new, mpb, size_old); |
4128 | free(mpb); | |
4129 | mpb = mpb_new; | |
949c47a0 | 4130 | super->anchor = mpb_new; |
2c092cad DW |
4131 | mpb->mpb_size = __cpu_to_le32(size_new); |
4132 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4133 | } | |
bf5a934a | 4134 | super->current_vol = idx; |
d23fe947 DW |
4135 | /* when creating the first raid device in this container set num_disks |
4136 | * to zero, i.e. delete this spare and add raid member devices in | |
4137 | * add_to_super_imsm_volume() | |
4138 | */ | |
4139 | if (super->current_vol == 0) | |
4140 | mpb->num_disks = 0; | |
5a038140 | 4141 | |
aa534678 DW |
4142 | if (!check_name(super, name, 0)) |
4143 | return 0; | |
ba2de7ba DW |
4144 | dv = malloc(sizeof(*dv)); |
4145 | if (!dv) { | |
4146 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
4147 | return 0; | |
4148 | } | |
1a2487c2 | 4149 | dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
949c47a0 | 4150 | if (!dev) { |
ba2de7ba | 4151 | free(dv); |
949c47a0 DW |
4152 | fprintf(stderr, Name": could not allocate raid device\n"); |
4153 | return 0; | |
4154 | } | |
1a2487c2 | 4155 | |
c2c087e6 | 4156 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
03bcbc65 DW |
4157 | if (info->level == 1) |
4158 | array_blocks = info_to_blocks_per_member(info); | |
4159 | else | |
4160 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
4161 | info->layout, info->chunk_size, | |
4162 | info->size*2); | |
979d38be DW |
4163 | /* round array size down to closest MB */ |
4164 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4165 | ||
c2c087e6 DW |
4166 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4167 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4168 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4169 | vol = &dev->vol; |
4170 | vol->migr_state = 0; | |
1484e727 | 4171 | set_migr_type(dev, MIGR_INIT); |
c2c087e6 | 4172 | vol->dirty = 0; |
f8f603f1 | 4173 | vol->curr_migr_unit = 0; |
a965f303 | 4174 | map = get_imsm_map(dev, 0); |
0dcecb2e | 4175 | map->pba_of_lba0 = __cpu_to_le32(super->create_offset); |
fcfd9599 | 4176 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade | 4177 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4178 | map->failed_disk_num = ~0; |
c2c087e6 DW |
4179 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
4180 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4181 | map->ddf = 1; |
ef6ffade DW |
4182 | |
4183 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4184 | free(dev); |
4185 | free(dv); | |
ef6ffade DW |
4186 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
4187 | "in a raid1 volume\n"); | |
4188 | return 0; | |
4189 | } | |
81062a36 DW |
4190 | |
4191 | map->raid_level = info->level; | |
4d1313e9 | 4192 | if (info->level == 10) { |
c2c087e6 | 4193 | map->raid_level = 1; |
4d1313e9 | 4194 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4195 | } else if (info->level == 1) |
4196 | map->num_domains = info->raid_disks; | |
4197 | else | |
ff596308 | 4198 | map->num_domains = 1; |
81062a36 | 4199 | |
ff596308 DW |
4200 | num_data_stripes = info_to_num_data_stripes(info, map->num_domains); |
4201 | map->num_data_stripes = __cpu_to_le32(num_data_stripes); | |
ef6ffade | 4202 | |
c2c087e6 DW |
4203 | map->num_members = info->raid_disks; |
4204 | for (i = 0; i < map->num_members; i++) { | |
4205 | /* initialized in add_to_super */ | |
4eb26970 | 4206 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4207 | } |
949c47a0 | 4208 | mpb->num_raid_devs++; |
ba2de7ba DW |
4209 | |
4210 | dv->dev = dev; | |
4211 | dv->index = super->current_vol; | |
4212 | dv->next = super->devlist; | |
4213 | super->devlist = dv; | |
c2c087e6 | 4214 | |
4d1313e9 DW |
4215 | imsm_update_version_info(super); |
4216 | ||
c2c087e6 | 4217 | return 1; |
cdddbdbc DW |
4218 | } |
4219 | ||
bf5a934a DW |
4220 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4221 | unsigned long long size, char *name, | |
4222 | char *homehost, int *uuid) | |
4223 | { | |
4224 | /* This is primarily called by Create when creating a new array. | |
4225 | * We will then get add_to_super called for each component, and then | |
4226 | * write_init_super called to write it out to each device. | |
4227 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4228 | * array. | |
4229 | * To create on a pre-existing array a different method will be called. | |
4230 | * This one is just for fresh drives. | |
4231 | */ | |
4232 | struct intel_super *super; | |
4233 | struct imsm_super *mpb; | |
4234 | size_t mpb_size; | |
4d1313e9 | 4235 | char *version; |
bf5a934a | 4236 | |
bf5a934a | 4237 | if (st->sb) |
e683ca88 DW |
4238 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
4239 | ||
4240 | if (info) | |
4241 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4242 | else | |
4243 | mpb_size = 512; | |
bf5a934a | 4244 | |
49133e57 | 4245 | super = alloc_super(); |
e683ca88 | 4246 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4247 | free(super); |
e683ca88 DW |
4248 | super = NULL; |
4249 | } | |
4250 | if (!super) { | |
4251 | fprintf(stderr, Name | |
4252 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
4253 | return 0; |
4254 | } | |
8e59f3d8 AK |
4255 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
4256 | fprintf(stderr, Name | |
4257 | ": %s could not allocate migr_rec buffer\n", __func__); | |
4258 | free(super->buf); | |
4259 | free(super); | |
4260 | return 0; | |
4261 | } | |
e683ca88 | 4262 | memset(super->buf, 0, mpb_size); |
ef649044 | 4263 | mpb = super->buf; |
e683ca88 DW |
4264 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4265 | st->sb = super; | |
4266 | ||
4267 | if (info == NULL) { | |
4268 | /* zeroing superblock */ | |
4269 | return 0; | |
4270 | } | |
bf5a934a | 4271 | |
4d1313e9 DW |
4272 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4273 | ||
4274 | version = (char *) mpb->sig; | |
4275 | strcpy(version, MPB_SIGNATURE); | |
4276 | version += strlen(MPB_SIGNATURE); | |
4277 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4278 | |
bf5a934a DW |
4279 | return 1; |
4280 | } | |
4281 | ||
0e600426 | 4282 | #ifndef MDASSEMBLE |
f20c3968 | 4283 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4284 | int fd, char *devname) |
4285 | { | |
4286 | struct intel_super *super = st->sb; | |
d23fe947 | 4287 | struct imsm_super *mpb = super->anchor; |
bf5a934a DW |
4288 | struct dl *dl; |
4289 | struct imsm_dev *dev; | |
4290 | struct imsm_map *map; | |
4eb26970 | 4291 | int slot; |
bf5a934a | 4292 | |
949c47a0 | 4293 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 4294 | map = get_imsm_map(dev, 0); |
bf5a934a | 4295 | |
208933a7 N |
4296 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
4297 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
4298 | devname); | |
4299 | return 1; | |
4300 | } | |
4301 | ||
efb30e7f DW |
4302 | if (fd == -1) { |
4303 | /* we're doing autolayout so grab the pre-marked (in | |
4304 | * validate_geometry) raid_disk | |
4305 | */ | |
4306 | for (dl = super->disks; dl; dl = dl->next) | |
4307 | if (dl->raiddisk == dk->raid_disk) | |
4308 | break; | |
4309 | } else { | |
4310 | for (dl = super->disks; dl ; dl = dl->next) | |
4311 | if (dl->major == dk->major && | |
4312 | dl->minor == dk->minor) | |
4313 | break; | |
4314 | } | |
d23fe947 | 4315 | |
208933a7 N |
4316 | if (!dl) { |
4317 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 4318 | return 1; |
208933a7 | 4319 | } |
bf5a934a | 4320 | |
d23fe947 DW |
4321 | /* add a pristine spare to the metadata */ |
4322 | if (dl->index < 0) { | |
4323 | dl->index = super->anchor->num_disks; | |
4324 | super->anchor->num_disks++; | |
4325 | } | |
4eb26970 DW |
4326 | /* Check the device has not already been added */ |
4327 | slot = get_imsm_disk_slot(map, dl->index); | |
4328 | if (slot >= 0 && | |
98130f40 | 4329 | (get_imsm_ord_tbl_ent(dev, slot, -1) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
4330 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
4331 | devname); | |
4332 | return 1; | |
4333 | } | |
656b6b5a | 4334 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4335 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 DW |
4336 | |
4337 | /* if we are creating the first raid device update the family number */ | |
4338 | if (super->current_vol == 0) { | |
4339 | __u32 sum; | |
4340 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
4341 | struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index); | |
4342 | ||
791b666a AW |
4343 | if (!_dev || !_disk) { |
4344 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
4345 | return 1; | |
4346 | } | |
d23fe947 DW |
4347 | *_dev = *dev; |
4348 | *_disk = dl->disk; | |
148acb7b DW |
4349 | sum = random32(); |
4350 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4351 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4352 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4353 | } |
ca0748fa | 4354 | super->current_disk = dl; |
f20c3968 | 4355 | return 0; |
bf5a934a DW |
4356 | } |
4357 | ||
88654014 | 4358 | |
f20c3968 | 4359 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4360 | int fd, char *devname) |
cdddbdbc | 4361 | { |
c2c087e6 | 4362 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4363 | struct dl *dd; |
4364 | unsigned long long size; | |
f2f27e63 | 4365 | __u32 id; |
c2c087e6 DW |
4366 | int rv; |
4367 | struct stat stb; | |
4368 | ||
88654014 LM |
4369 | /* If we are on an RAID enabled platform check that the disk is |
4370 | * attached to the raid controller. | |
4371 | * We do not need to test disks attachment for container based additions, | |
4372 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4373 | */ |
d424212e | 4374 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4375 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4376 | if (rv != 0) { |
4377 | dprintf("capability: %p fd: %d ret: %d\n", | |
4378 | super->orom, fd, rv); | |
4379 | return 1; | |
88c32bb1 DW |
4380 | } |
4381 | ||
f20c3968 DW |
4382 | if (super->current_vol >= 0) |
4383 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4384 | |
c2c087e6 DW |
4385 | fstat(fd, &stb); |
4386 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 4387 | if (!dd) { |
c2c087e6 DW |
4388 | fprintf(stderr, |
4389 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 4390 | return 1; |
c2c087e6 DW |
4391 | } |
4392 | memset(dd, 0, sizeof(*dd)); | |
4393 | dd->major = major(stb.st_rdev); | |
4394 | dd->minor = minor(stb.st_rdev); | |
b9f594fe | 4395 | dd->index = -1; |
c2c087e6 | 4396 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 4397 | dd->fd = fd; |
689c9bf3 | 4398 | dd->e = NULL; |
1a64be56 | 4399 | dd->action = DISK_ADD; |
c2c087e6 | 4400 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4401 | if (rv) { |
c2c087e6 | 4402 | fprintf(stderr, |
0030e8d6 | 4403 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4404 | free(dd); |
0030e8d6 | 4405 | abort(); |
c2c087e6 DW |
4406 | } |
4407 | ||
c2c087e6 DW |
4408 | get_dev_size(fd, NULL, &size); |
4409 | size /= 512; | |
1f24f035 | 4410 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 4411 | dd->disk.total_blocks = __cpu_to_le32(size); |
ee5aad5a | 4412 | dd->disk.status = SPARE_DISK; |
c2c087e6 | 4413 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4414 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4415 | else |
b9f594fe | 4416 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4417 | |
4418 | if (st->update_tail) { | |
1a64be56 LM |
4419 | dd->next = super->disk_mgmt_list; |
4420 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4421 | } else { |
4422 | dd->next = super->disks; | |
4423 | super->disks = dd; | |
ceaf0ee1 | 4424 | super->updates_pending++; |
43dad3d6 | 4425 | } |
f20c3968 DW |
4426 | |
4427 | return 0; | |
cdddbdbc DW |
4428 | } |
4429 | ||
1a64be56 LM |
4430 | |
4431 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4432 | { | |
4433 | struct intel_super *super = st->sb; | |
4434 | struct dl *dd; | |
4435 | ||
4436 | /* remove from super works only in mdmon - for communication | |
4437 | * manager - monitor. Check if communication memory buffer | |
4438 | * is prepared. | |
4439 | */ | |
4440 | if (!st->update_tail) { | |
4441 | fprintf(stderr, | |
4442 | Name ": %s shall be used in mdmon context only" | |
4443 | "(line %d).\n", __func__, __LINE__); | |
4444 | return 1; | |
4445 | } | |
4446 | dd = malloc(sizeof(*dd)); | |
4447 | if (!dd) { | |
4448 | fprintf(stderr, | |
4449 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
4450 | return 1; | |
4451 | } | |
4452 | memset(dd, 0, sizeof(*dd)); | |
4453 | dd->major = dk->major; | |
4454 | dd->minor = dk->minor; | |
4455 | dd->index = -1; | |
4456 | dd->fd = -1; | |
4457 | dd->disk.status = SPARE_DISK; | |
4458 | dd->action = DISK_REMOVE; | |
4459 | ||
4460 | dd->next = super->disk_mgmt_list; | |
4461 | super->disk_mgmt_list = dd; | |
4462 | ||
4463 | ||
4464 | return 0; | |
4465 | } | |
4466 | ||
f796af5d DW |
4467 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
4468 | ||
4469 | static union { | |
4470 | char buf[512]; | |
4471 | struct imsm_super anchor; | |
4472 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 4473 | |
d23fe947 DW |
4474 | /* spare records have their own family number and do not have any defined raid |
4475 | * devices | |
4476 | */ | |
4477 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
4478 | { | |
d23fe947 | 4479 | struct imsm_super *mpb = super->anchor; |
f796af5d | 4480 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
4481 | __u32 sum; |
4482 | struct dl *d; | |
4483 | ||
f796af5d DW |
4484 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
4485 | spare->generation_num = __cpu_to_le32(1UL), | |
4486 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
4487 | spare->num_disks = 1, | |
4488 | spare->num_raid_devs = 0, | |
4489 | spare->cache_size = mpb->cache_size, | |
4490 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
4491 | ||
4492 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
4493 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
4494 | |
4495 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 4496 | if (d->index != -1) |
d23fe947 DW |
4497 | continue; |
4498 | ||
f796af5d DW |
4499 | spare->disk[0] = d->disk; |
4500 | sum = __gen_imsm_checksum(spare); | |
4501 | spare->family_num = __cpu_to_le32(sum); | |
4502 | spare->orig_family_num = 0; | |
4503 | sum = __gen_imsm_checksum(spare); | |
4504 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 4505 | |
f796af5d | 4506 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
4507 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4508 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 4509 | return 1; |
d23fe947 DW |
4510 | } |
4511 | if (doclose) { | |
4512 | close(d->fd); | |
4513 | d->fd = -1; | |
4514 | } | |
4515 | } | |
4516 | ||
e74255d9 | 4517 | return 0; |
d23fe947 DW |
4518 | } |
4519 | ||
36988a3d | 4520 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 4521 | { |
36988a3d | 4522 | struct intel_super *super = st->sb; |
949c47a0 | 4523 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4524 | struct dl *d; |
4525 | __u32 generation; | |
4526 | __u32 sum; | |
d23fe947 | 4527 | int spares = 0; |
949c47a0 | 4528 | int i; |
a48ac0a8 | 4529 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 4530 | int num_disks = 0; |
146c6260 | 4531 | int clear_migration_record = 1; |
cdddbdbc | 4532 | |
c2c087e6 DW |
4533 | /* 'generation' is incremented everytime the metadata is written */ |
4534 | generation = __le32_to_cpu(mpb->generation_num); | |
4535 | generation++; | |
4536 | mpb->generation_num = __cpu_to_le32(generation); | |
4537 | ||
148acb7b DW |
4538 | /* fix up cases where previous mdadm releases failed to set |
4539 | * orig_family_num | |
4540 | */ | |
4541 | if (mpb->orig_family_num == 0) | |
4542 | mpb->orig_family_num = mpb->family_num; | |
4543 | ||
d23fe947 | 4544 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 4545 | if (d->index == -1) |
d23fe947 | 4546 | spares++; |
36988a3d | 4547 | else { |
d23fe947 | 4548 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4549 | num_disks++; |
4550 | } | |
d23fe947 | 4551 | } |
36988a3d | 4552 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 4553 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4554 | num_disks++; |
4555 | } | |
4556 | mpb->num_disks = num_disks; | |
4557 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 4558 | |
949c47a0 DW |
4559 | for (i = 0; i < mpb->num_raid_devs; i++) { |
4560 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
4561 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
4562 | if (dev && dev2) { | |
4563 | imsm_copy_dev(dev, dev2); | |
4564 | mpb_size += sizeof_imsm_dev(dev, 0); | |
4565 | } | |
146c6260 AK |
4566 | if (is_gen_migration(dev2)) |
4567 | clear_migration_record = 0; | |
949c47a0 | 4568 | } |
a48ac0a8 DW |
4569 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
4570 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 4571 | |
c2c087e6 | 4572 | /* recalculate checksum */ |
949c47a0 | 4573 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
4574 | mpb->check_sum = __cpu_to_le32(sum); |
4575 | ||
146c6260 AK |
4576 | if (clear_migration_record) |
4577 | memset(super->migr_rec_buf, 0, 512); | |
4578 | ||
d23fe947 | 4579 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 4580 | for (d = super->disks; d ; d = d->next) { |
d23fe947 DW |
4581 | if (d->index < 0) |
4582 | continue; | |
f796af5d | 4583 | if (store_imsm_mpb(d->fd, mpb)) |
c2c087e6 DW |
4584 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4585 | __func__, d->major, d->minor, strerror(errno)); | |
146c6260 AK |
4586 | if (clear_migration_record) { |
4587 | unsigned long long dsize; | |
4588 | ||
4589 | get_dev_size(d->fd, NULL, &dsize); | |
4590 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
9e2d750d N |
4591 | if (write(d->fd, super->migr_rec_buf, 512) != 512) |
4592 | perror("Write migr_rec failed"); | |
146c6260 AK |
4593 | } |
4594 | } | |
c2c087e6 DW |
4595 | if (doclose) { |
4596 | close(d->fd); | |
4597 | d->fd = -1; | |
4598 | } | |
4599 | } | |
4600 | ||
d23fe947 DW |
4601 | if (spares) |
4602 | return write_super_imsm_spares(super, doclose); | |
4603 | ||
e74255d9 | 4604 | return 0; |
c2c087e6 DW |
4605 | } |
4606 | ||
0e600426 | 4607 | |
9b1fb677 | 4608 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
4609 | { |
4610 | size_t len; | |
4611 | struct imsm_update_create_array *u; | |
4612 | struct intel_super *super = st->sb; | |
9b1fb677 | 4613 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
54c2c1ea DW |
4614 | struct imsm_map *map = get_imsm_map(dev, 0); |
4615 | struct disk_info *inf; | |
4616 | struct imsm_disk *disk; | |
4617 | int i; | |
43dad3d6 | 4618 | |
54c2c1ea DW |
4619 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
4620 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
4621 | u = malloc(len); |
4622 | if (!u) { | |
4623 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4624 | __func__); | |
4625 | return 1; | |
4626 | } | |
4627 | ||
4628 | u->type = update_create_array; | |
9b1fb677 | 4629 | u->dev_idx = dev_idx; |
43dad3d6 | 4630 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
4631 | inf = get_disk_info(u); |
4632 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 4633 | int idx = get_imsm_disk_idx(dev, i, -1); |
9b1fb677 | 4634 | |
54c2c1ea DW |
4635 | disk = get_imsm_disk(super, idx); |
4636 | serialcpy(inf[i].serial, disk->serial); | |
4637 | } | |
43dad3d6 DW |
4638 | append_metadata_update(st, u, len); |
4639 | ||
4640 | return 0; | |
4641 | } | |
4642 | ||
1a64be56 | 4643 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
4644 | { |
4645 | struct intel_super *super = st->sb; | |
4646 | size_t len; | |
1a64be56 | 4647 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 4648 | |
1a64be56 | 4649 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
4650 | return 0; |
4651 | ||
4652 | len = sizeof(*u); | |
4653 | u = malloc(len); | |
4654 | if (!u) { | |
4655 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4656 | __func__); | |
4657 | return 1; | |
4658 | } | |
4659 | ||
1a64be56 | 4660 | u->type = update_add_remove_disk; |
43dad3d6 DW |
4661 | append_metadata_update(st, u, len); |
4662 | ||
4663 | return 0; | |
4664 | } | |
4665 | ||
c2c087e6 DW |
4666 | static int write_init_super_imsm(struct supertype *st) |
4667 | { | |
9b1fb677 DW |
4668 | struct intel_super *super = st->sb; |
4669 | int current_vol = super->current_vol; | |
4670 | ||
4671 | /* we are done with current_vol reset it to point st at the container */ | |
4672 | super->current_vol = -1; | |
4673 | ||
8273f55e | 4674 | if (st->update_tail) { |
43dad3d6 DW |
4675 | /* queue the recently created array / added disk |
4676 | * as a metadata update */ | |
43dad3d6 | 4677 | int rv; |
8273f55e | 4678 | |
43dad3d6 | 4679 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 4680 | if (current_vol < 0) { |
1a64be56 LM |
4681 | /* in the mgmt (add/remove) disk case we are running |
4682 | * in mdmon context, so don't close fd's | |
43dad3d6 | 4683 | */ |
1a64be56 | 4684 | return mgmt_disk(st); |
43dad3d6 | 4685 | } else |
9b1fb677 | 4686 | rv = create_array(st, current_vol); |
8273f55e | 4687 | |
43dad3d6 | 4688 | return rv; |
d682f344 N |
4689 | } else { |
4690 | struct dl *d; | |
4691 | for (d = super->disks; d; d = d->next) | |
4692 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 4693 | return write_super_imsm(st, 1); |
d682f344 | 4694 | } |
cdddbdbc | 4695 | } |
0e600426 | 4696 | #endif |
cdddbdbc | 4697 | |
e683ca88 | 4698 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 4699 | { |
e683ca88 DW |
4700 | struct intel_super *super = st->sb; |
4701 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 4702 | |
e683ca88 | 4703 | if (!mpb) |
ad97895e DW |
4704 | return 1; |
4705 | ||
1799c9e8 | 4706 | #ifndef MDASSEMBLE |
e683ca88 | 4707 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
4708 | #else |
4709 | return 1; | |
4710 | #endif | |
cdddbdbc DW |
4711 | } |
4712 | ||
0e600426 N |
4713 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
4714 | { | |
4715 | return __le32_to_cpu(mpb->bbm_log_size); | |
4716 | } | |
4717 | ||
4718 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
4719 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
4720 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 4721 | unsigned long long size, char *dev, |
2c514b71 NB |
4722 | unsigned long long *freesize, |
4723 | int verbose) | |
cdddbdbc | 4724 | { |
c2c087e6 DW |
4725 | int fd; |
4726 | unsigned long long ldsize; | |
f2f5c343 LM |
4727 | struct intel_super *super=NULL; |
4728 | int rv = 0; | |
cdddbdbc | 4729 | |
c2c087e6 DW |
4730 | if (level != LEVEL_CONTAINER) |
4731 | return 0; | |
4732 | if (!dev) | |
4733 | return 1; | |
4734 | ||
4735 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4736 | if (fd < 0) { | |
2c514b71 NB |
4737 | if (verbose) |
4738 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
4739 | dev, strerror(errno)); | |
c2c087e6 DW |
4740 | return 0; |
4741 | } | |
4742 | if (!get_dev_size(fd, dev, &ldsize)) { | |
4743 | close(fd); | |
4744 | return 0; | |
4745 | } | |
f2f5c343 LM |
4746 | |
4747 | /* capabilities retrieve could be possible | |
4748 | * note that there is no fd for the disks in array. | |
4749 | */ | |
4750 | super = alloc_super(); | |
4751 | if (!super) { | |
4752 | fprintf(stderr, | |
4753 | Name ": malloc of %zu failed.\n", | |
4754 | sizeof(*super)); | |
4755 | close(fd); | |
4756 | return 0; | |
4757 | } | |
4758 | ||
d424212e | 4759 | rv = find_intel_hba_capability(fd, super, verbose ? dev : NULL); |
f2f5c343 LM |
4760 | if (rv != 0) { |
4761 | #if DEBUG | |
4762 | char str[256]; | |
4763 | fd2devname(fd, str); | |
4764 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
4765 | fd, str, super->orom, rv, raiddisks); | |
4766 | #endif | |
4767 | /* no orom/efi or non-intel hba of the disk */ | |
4768 | close(fd); | |
4769 | free_imsm(super); | |
4770 | return 0; | |
4771 | } | |
c2c087e6 | 4772 | close(fd); |
f2f5c343 LM |
4773 | if (super->orom && raiddisks > super->orom->tds) { |
4774 | if (verbose) | |
4775 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
4776 | " platform supported disks: %d\n", | |
4777 | raiddisks, super->orom->tds); | |
4778 | ||
4779 | free_imsm(super); | |
4780 | return 0; | |
4781 | } | |
c2c087e6 DW |
4782 | |
4783 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 4784 | free_imsm(super); |
c2c087e6 DW |
4785 | |
4786 | return 1; | |
cdddbdbc DW |
4787 | } |
4788 | ||
0dcecb2e DW |
4789 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
4790 | { | |
4791 | const unsigned long long base_start = e[*idx].start; | |
4792 | unsigned long long end = base_start + e[*idx].size; | |
4793 | int i; | |
4794 | ||
4795 | if (base_start == end) | |
4796 | return 0; | |
4797 | ||
4798 | *idx = *idx + 1; | |
4799 | for (i = *idx; i < num_extents; i++) { | |
4800 | /* extend overlapping extents */ | |
4801 | if (e[i].start >= base_start && | |
4802 | e[i].start <= end) { | |
4803 | if (e[i].size == 0) | |
4804 | return 0; | |
4805 | if (e[i].start + e[i].size > end) | |
4806 | end = e[i].start + e[i].size; | |
4807 | } else if (e[i].start > end) { | |
4808 | *idx = i; | |
4809 | break; | |
4810 | } | |
4811 | } | |
4812 | ||
4813 | return end - base_start; | |
4814 | } | |
4815 | ||
4816 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
4817 | { | |
4818 | /* build a composite disk with all known extents and generate a new | |
4819 | * 'maxsize' given the "all disks in an array must share a common start | |
4820 | * offset" constraint | |
4821 | */ | |
4822 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
4823 | struct dl *dl; | |
4824 | int i, j; | |
4825 | int start_extent; | |
4826 | unsigned long long pos; | |
b9d77223 | 4827 | unsigned long long start = 0; |
0dcecb2e DW |
4828 | unsigned long long maxsize; |
4829 | unsigned long reserve; | |
4830 | ||
4831 | if (!e) | |
a7dd165b | 4832 | return 0; |
0dcecb2e DW |
4833 | |
4834 | /* coalesce and sort all extents. also, check to see if we need to | |
4835 | * reserve space between member arrays | |
4836 | */ | |
4837 | j = 0; | |
4838 | for (dl = super->disks; dl; dl = dl->next) { | |
4839 | if (!dl->e) | |
4840 | continue; | |
4841 | for (i = 0; i < dl->extent_cnt; i++) | |
4842 | e[j++] = dl->e[i]; | |
4843 | } | |
4844 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
4845 | ||
4846 | /* merge extents */ | |
4847 | i = 0; | |
4848 | j = 0; | |
4849 | while (i < sum_extents) { | |
4850 | e[j].start = e[i].start; | |
4851 | e[j].size = find_size(e, &i, sum_extents); | |
4852 | j++; | |
4853 | if (e[j-1].size == 0) | |
4854 | break; | |
4855 | } | |
4856 | ||
4857 | pos = 0; | |
4858 | maxsize = 0; | |
4859 | start_extent = 0; | |
4860 | i = 0; | |
4861 | do { | |
4862 | unsigned long long esize; | |
4863 | ||
4864 | esize = e[i].start - pos; | |
4865 | if (esize >= maxsize) { | |
4866 | maxsize = esize; | |
4867 | start = pos; | |
4868 | start_extent = i; | |
4869 | } | |
4870 | pos = e[i].start + e[i].size; | |
4871 | i++; | |
4872 | } while (e[i-1].size); | |
4873 | free(e); | |
4874 | ||
a7dd165b DW |
4875 | if (maxsize == 0) |
4876 | return 0; | |
4877 | ||
4878 | /* FIXME assumes volume at offset 0 is the first volume in a | |
4879 | * container | |
4880 | */ | |
0dcecb2e DW |
4881 | if (start_extent > 0) |
4882 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
4883 | else | |
4884 | reserve = 0; | |
4885 | ||
4886 | if (maxsize < reserve) | |
a7dd165b | 4887 | return 0; |
0dcecb2e DW |
4888 | |
4889 | super->create_offset = ~((__u32) 0); | |
4890 | if (start + reserve > super->create_offset) | |
a7dd165b | 4891 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
4892 | super->create_offset = start + reserve; |
4893 | ||
4894 | return maxsize - reserve; | |
4895 | } | |
4896 | ||
88c32bb1 DW |
4897 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
4898 | { | |
4899 | if (level < 0 || level == 6 || level == 4) | |
4900 | return 0; | |
4901 | ||
4902 | /* if we have an orom prevent invalid raid levels */ | |
4903 | if (orom) | |
4904 | switch (level) { | |
4905 | case 0: return imsm_orom_has_raid0(orom); | |
4906 | case 1: | |
4907 | if (raiddisks > 2) | |
4908 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
4909 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
4910 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
4911 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
4912 | } |
4913 | else | |
4914 | return 1; /* not on an Intel RAID platform so anything goes */ | |
4915 | ||
4916 | return 0; | |
4917 | } | |
4918 | ||
cd9d1ac7 DW |
4919 | static int imsm_default_chunk(const struct imsm_orom *orom) |
4920 | { | |
4921 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
4922 | * 128 if no platform detected | |
4923 | */ | |
4924 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
4925 | ||
4926 | return min(512, (1 << fs)); | |
4927 | } | |
73408129 | 4928 | |
35f81cbb | 4929 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) |
73408129 LM |
4930 | /* |
4931 | * validate volume parameters with OROM/EFI capabilities | |
4932 | */ | |
6592ce37 DW |
4933 | static int |
4934 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
c21e737b | 4935 | int raiddisks, int *chunk, int verbose) |
6592ce37 | 4936 | { |
73408129 LM |
4937 | #if DEBUG |
4938 | verbose = 1; | |
4939 | #endif | |
4940 | /* validate container capabilities */ | |
4941 | if (super->orom && raiddisks > super->orom->tds) { | |
4942 | if (verbose) | |
4943 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
4944 | " platform supported disks: %d\n", | |
4945 | raiddisks, super->orom->tds); | |
4946 | return 0; | |
4947 | } | |
4948 | ||
4949 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
4950 | if (super->orom && (!is_raid_level_supported(super->orom, level, | |
4951 | raiddisks))) { | |
6592ce37 DW |
4952 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
4953 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
4954 | return 0; | |
4955 | } | |
cd9d1ac7 DW |
4956 | |
4957 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
4958 | *chunk = imsm_default_chunk(super->orom); | |
4959 | ||
4960 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
4961 | pr_vrb(": platform does not support a chunk size of: " | |
4962 | "%d\n", *chunk); | |
4963 | return 0; | |
6592ce37 | 4964 | } |
cd9d1ac7 | 4965 | |
6592ce37 DW |
4966 | if (layout != imsm_level_to_layout(level)) { |
4967 | if (level == 5) | |
4968 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
4969 | else if (level == 10) | |
4970 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
4971 | else | |
4972 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
4973 | layout, level); | |
4974 | return 0; | |
4975 | } | |
6592ce37 DW |
4976 | return 1; |
4977 | } | |
4978 | ||
c2c087e6 DW |
4979 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
4980 | * FIX ME add ahci details | |
4981 | */ | |
8b353278 | 4982 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 4983 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 4984 | unsigned long long size, char *dev, |
2c514b71 NB |
4985 | unsigned long long *freesize, |
4986 | int verbose) | |
cdddbdbc | 4987 | { |
c2c087e6 DW |
4988 | struct stat stb; |
4989 | struct intel_super *super = st->sb; | |
a20d2ba5 | 4990 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4991 | struct dl *dl; |
4992 | unsigned long long pos = 0; | |
4993 | unsigned long long maxsize; | |
4994 | struct extent *e; | |
4995 | int i; | |
cdddbdbc | 4996 | |
88c32bb1 DW |
4997 | /* We must have the container info already read in. */ |
4998 | if (!super) | |
c2c087e6 DW |
4999 | return 0; |
5000 | ||
d54559f0 LM |
5001 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, verbose)) { |
5002 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
5003 | "Cannot proceed with the action(s).\n"); | |
c2c087e6 | 5004 | return 0; |
d54559f0 | 5005 | } |
c2c087e6 DW |
5006 | if (!dev) { |
5007 | /* General test: make sure there is space for | |
2da8544a DW |
5008 | * 'raiddisks' device extents of size 'size' at a given |
5009 | * offset | |
c2c087e6 | 5010 | */ |
e46273eb | 5011 | unsigned long long minsize = size; |
b7528a20 | 5012 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5013 | int dcnt = 0; |
5014 | if (minsize == 0) | |
5015 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5016 | for (dl = super->disks; dl ; dl = dl->next) { | |
5017 | int found = 0; | |
5018 | ||
bf5a934a | 5019 | pos = 0; |
c2c087e6 DW |
5020 | i = 0; |
5021 | e = get_extents(super, dl); | |
5022 | if (!e) continue; | |
5023 | do { | |
5024 | unsigned long long esize; | |
5025 | esize = e[i].start - pos; | |
5026 | if (esize >= minsize) | |
5027 | found = 1; | |
b7528a20 | 5028 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5029 | start_offset = pos; |
5030 | break; | |
5031 | } else if (found && pos != start_offset) { | |
5032 | found = 0; | |
5033 | break; | |
5034 | } | |
c2c087e6 DW |
5035 | pos = e[i].start + e[i].size; |
5036 | i++; | |
5037 | } while (e[i-1].size); | |
5038 | if (found) | |
5039 | dcnt++; | |
5040 | free(e); | |
5041 | } | |
5042 | if (dcnt < raiddisks) { | |
2c514b71 NB |
5043 | if (verbose) |
5044 | fprintf(stderr, Name ": imsm: Not enough " | |
5045 | "devices with space for this array " | |
5046 | "(%d < %d)\n", | |
5047 | dcnt, raiddisks); | |
c2c087e6 DW |
5048 | return 0; |
5049 | } | |
5050 | return 1; | |
5051 | } | |
0dcecb2e | 5052 | |
c2c087e6 DW |
5053 | /* This device must be a member of the set */ |
5054 | if (stat(dev, &stb) < 0) | |
5055 | return 0; | |
5056 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
5057 | return 0; | |
5058 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
5059 | if (dl->major == (int)major(stb.st_rdev) && |
5060 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
5061 | break; |
5062 | } | |
5063 | if (!dl) { | |
2c514b71 NB |
5064 | if (verbose) |
5065 | fprintf(stderr, Name ": %s is not in the " | |
5066 | "same imsm set\n", dev); | |
c2c087e6 | 5067 | return 0; |
a20d2ba5 DW |
5068 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
5069 | /* If a volume is present then the current creation attempt | |
5070 | * cannot incorporate new spares because the orom may not | |
5071 | * understand this configuration (all member disks must be | |
5072 | * members of each array in the container). | |
5073 | */ | |
5074 | fprintf(stderr, Name ": %s is a spare and a volume" | |
5075 | " is already defined for this container\n", dev); | |
5076 | fprintf(stderr, Name ": The option-rom requires all member" | |
5077 | " disks to be a member of all volumes\n"); | |
5078 | return 0; | |
c2c087e6 | 5079 | } |
0dcecb2e DW |
5080 | |
5081 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
5082 | e = get_extents(super, dl); |
5083 | maxsize = 0; | |
5084 | i = 0; | |
0dcecb2e DW |
5085 | if (e) { |
5086 | do { | |
5087 | unsigned long long esize; | |
5088 | ||
5089 | esize = e[i].start - pos; | |
5090 | if (esize >= maxsize) | |
5091 | maxsize = esize; | |
5092 | pos = e[i].start + e[i].size; | |
5093 | i++; | |
5094 | } while (e[i-1].size); | |
5095 | dl->e = e; | |
5096 | dl->extent_cnt = i; | |
5097 | } else { | |
5098 | if (verbose) | |
5099 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
5100 | dev); | |
5101 | return 0; | |
5102 | } | |
5103 | if (maxsize < size) { | |
5104 | if (verbose) | |
5105 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
5106 | dev, maxsize, size); | |
5107 | return 0; | |
5108 | } | |
5109 | ||
5110 | /* count total number of extents for merge */ | |
5111 | i = 0; | |
5112 | for (dl = super->disks; dl; dl = dl->next) | |
5113 | if (dl->e) | |
5114 | i += dl->extent_cnt; | |
5115 | ||
5116 | maxsize = merge_extents(super, i); | |
a7dd165b | 5117 | if (maxsize < size || maxsize == 0) { |
0dcecb2e DW |
5118 | if (verbose) |
5119 | fprintf(stderr, Name ": not enough space after merge (%llu < %llu)\n", | |
5120 | maxsize, size); | |
5121 | return 0; | |
0dcecb2e DW |
5122 | } |
5123 | ||
c2c087e6 DW |
5124 | *freesize = maxsize; |
5125 | ||
5126 | return 1; | |
cdddbdbc DW |
5127 | } |
5128 | ||
efb30e7f DW |
5129 | static int reserve_space(struct supertype *st, int raiddisks, |
5130 | unsigned long long size, int chunk, | |
5131 | unsigned long long *freesize) | |
5132 | { | |
5133 | struct intel_super *super = st->sb; | |
5134 | struct imsm_super *mpb = super->anchor; | |
5135 | struct dl *dl; | |
5136 | int i; | |
5137 | int extent_cnt; | |
5138 | struct extent *e; | |
5139 | unsigned long long maxsize; | |
5140 | unsigned long long minsize; | |
5141 | int cnt; | |
5142 | int used; | |
5143 | ||
5144 | /* find the largest common start free region of the possible disks */ | |
5145 | used = 0; | |
5146 | extent_cnt = 0; | |
5147 | cnt = 0; | |
5148 | for (dl = super->disks; dl; dl = dl->next) { | |
5149 | dl->raiddisk = -1; | |
5150 | ||
5151 | if (dl->index >= 0) | |
5152 | used++; | |
5153 | ||
5154 | /* don't activate new spares if we are orom constrained | |
5155 | * and there is already a volume active in the container | |
5156 | */ | |
5157 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
5158 | continue; | |
5159 | ||
5160 | e = get_extents(super, dl); | |
5161 | if (!e) | |
5162 | continue; | |
5163 | for (i = 1; e[i-1].size; i++) | |
5164 | ; | |
5165 | dl->e = e; | |
5166 | dl->extent_cnt = i; | |
5167 | extent_cnt += i; | |
5168 | cnt++; | |
5169 | } | |
5170 | ||
5171 | maxsize = merge_extents(super, extent_cnt); | |
5172 | minsize = size; | |
5173 | if (size == 0) | |
612e59d8 CA |
5174 | /* chunk is in K */ |
5175 | minsize = chunk * 2; | |
efb30e7f DW |
5176 | |
5177 | if (cnt < raiddisks || | |
5178 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
5179 | maxsize < minsize || |
5180 | maxsize == 0) { | |
efb30e7f DW |
5181 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
5182 | return 0; /* No enough free spaces large enough */ | |
5183 | } | |
5184 | ||
5185 | if (size == 0) { | |
5186 | size = maxsize; | |
5187 | if (chunk) { | |
612e59d8 CA |
5188 | size /= 2 * chunk; |
5189 | size *= 2 * chunk; | |
efb30e7f DW |
5190 | } |
5191 | } | |
5192 | ||
5193 | cnt = 0; | |
5194 | for (dl = super->disks; dl; dl = dl->next) | |
5195 | if (dl->e) | |
5196 | dl->raiddisk = cnt++; | |
5197 | ||
5198 | *freesize = size; | |
5199 | ||
5200 | return 1; | |
5201 | } | |
5202 | ||
bf5a934a | 5203 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 5204 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
5205 | char *dev, unsigned long long *freesize, |
5206 | int verbose) | |
5207 | { | |
5208 | int fd, cfd; | |
5209 | struct mdinfo *sra; | |
20cbe8d2 | 5210 | int is_member = 0; |
bf5a934a | 5211 | |
d54559f0 LM |
5212 | /* load capability |
5213 | * if given unused devices create a container | |
bf5a934a DW |
5214 | * if given given devices in a container create a member volume |
5215 | */ | |
5216 | if (level == LEVEL_CONTAINER) { | |
5217 | /* Must be a fresh device to add to a container */ | |
5218 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
5219 | raiddisks, |
5220 | chunk?*chunk:0, size, | |
bf5a934a DW |
5221 | dev, freesize, |
5222 | verbose); | |
5223 | } | |
5224 | ||
8592f29d N |
5225 | if (!dev) { |
5226 | if (st->sb && freesize) { | |
efb30e7f DW |
5227 | /* we are being asked to automatically layout a |
5228 | * new volume based on the current contents of | |
5229 | * the container. If the the parameters can be | |
5230 | * satisfied reserve_space will record the disks, | |
5231 | * start offset, and size of the volume to be | |
5232 | * created. add_to_super and getinfo_super | |
5233 | * detect when autolayout is in progress. | |
5234 | */ | |
6592ce37 DW |
5235 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
5236 | raiddisks, chunk, | |
5237 | verbose)) | |
5238 | return 0; | |
c21e737b CA |
5239 | return reserve_space(st, raiddisks, size, |
5240 | chunk?*chunk:0, freesize); | |
8592f29d N |
5241 | } |
5242 | return 1; | |
5243 | } | |
bf5a934a DW |
5244 | if (st->sb) { |
5245 | /* creating in a given container */ | |
5246 | return validate_geometry_imsm_volume(st, level, layout, | |
5247 | raiddisks, chunk, size, | |
5248 | dev, freesize, verbose); | |
5249 | } | |
5250 | ||
bf5a934a DW |
5251 | /* This device needs to be a device in an 'imsm' container */ |
5252 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5253 | if (fd >= 0) { | |
5254 | if (verbose) | |
5255 | fprintf(stderr, | |
5256 | Name ": Cannot create this array on device %s\n", | |
5257 | dev); | |
5258 | close(fd); | |
5259 | return 0; | |
5260 | } | |
5261 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
5262 | if (verbose) | |
5263 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
5264 | dev, strerror(errno)); | |
5265 | return 0; | |
5266 | } | |
5267 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
5268 | cfd = open_container(fd); | |
20cbe8d2 | 5269 | close(fd); |
bf5a934a | 5270 | if (cfd < 0) { |
bf5a934a DW |
5271 | if (verbose) |
5272 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
5273 | dev); | |
5274 | return 0; | |
5275 | } | |
5276 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 5277 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
5278 | strcmp(sra->text_version, "imsm") == 0) |
5279 | is_member = 1; | |
5280 | sysfs_free(sra); | |
5281 | if (is_member) { | |
bf5a934a DW |
5282 | /* This is a member of a imsm container. Load the container |
5283 | * and try to create a volume | |
5284 | */ | |
5285 | struct intel_super *super; | |
5286 | ||
e1902a7b | 5287 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL) == 0) { |
bf5a934a DW |
5288 | st->sb = super; |
5289 | st->container_dev = fd2devnum(cfd); | |
5290 | close(cfd); | |
5291 | return validate_geometry_imsm_volume(st, level, layout, | |
5292 | raiddisks, chunk, | |
5293 | size, dev, | |
5294 | freesize, verbose); | |
5295 | } | |
20cbe8d2 | 5296 | } |
bf5a934a | 5297 | |
20cbe8d2 AW |
5298 | if (verbose) |
5299 | fprintf(stderr, Name ": failed container membership check\n"); | |
5300 | ||
5301 | close(cfd); | |
5302 | return 0; | |
bf5a934a | 5303 | } |
0bd16cf2 | 5304 | |
30f58b22 | 5305 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
5306 | { |
5307 | struct intel_super *super = st->sb; | |
5308 | ||
30f58b22 DW |
5309 | if (level && *level == UnSet) |
5310 | *level = LEVEL_CONTAINER; | |
5311 | ||
5312 | if (level && layout && *layout == UnSet) | |
5313 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 5314 | |
cd9d1ac7 DW |
5315 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
5316 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
5317 | } |
5318 | ||
33414a01 DW |
5319 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
5320 | ||
5321 | static int kill_subarray_imsm(struct supertype *st) | |
5322 | { | |
5323 | /* remove the subarray currently referenced by ->current_vol */ | |
5324 | __u8 i; | |
5325 | struct intel_dev **dp; | |
5326 | struct intel_super *super = st->sb; | |
5327 | __u8 current_vol = super->current_vol; | |
5328 | struct imsm_super *mpb = super->anchor; | |
5329 | ||
5330 | if (super->current_vol < 0) | |
5331 | return 2; | |
5332 | super->current_vol = -1; /* invalidate subarray cursor */ | |
5333 | ||
5334 | /* block deletions that would change the uuid of active subarrays | |
5335 | * | |
5336 | * FIXME when immutable ids are available, but note that we'll | |
5337 | * also need to fixup the invalidated/active subarray indexes in | |
5338 | * mdstat | |
5339 | */ | |
5340 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5341 | char subarray[4]; | |
5342 | ||
5343 | if (i < current_vol) | |
5344 | continue; | |
5345 | sprintf(subarray, "%u", i); | |
5346 | if (is_subarray_active(subarray, st->devname)) { | |
5347 | fprintf(stderr, | |
5348 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
5349 | current_vol, i); | |
5350 | ||
5351 | return 2; | |
5352 | } | |
5353 | } | |
5354 | ||
5355 | if (st->update_tail) { | |
5356 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
5357 | ||
5358 | if (!u) | |
5359 | return 2; | |
5360 | u->type = update_kill_array; | |
5361 | u->dev_idx = current_vol; | |
5362 | append_metadata_update(st, u, sizeof(*u)); | |
5363 | ||
5364 | return 0; | |
5365 | } | |
5366 | ||
5367 | for (dp = &super->devlist; *dp;) | |
5368 | if ((*dp)->index == current_vol) { | |
5369 | *dp = (*dp)->next; | |
5370 | } else { | |
5371 | handle_missing(super, (*dp)->dev); | |
5372 | if ((*dp)->index > current_vol) | |
5373 | (*dp)->index--; | |
5374 | dp = &(*dp)->next; | |
5375 | } | |
5376 | ||
5377 | /* no more raid devices, all active components are now spares, | |
5378 | * but of course failed are still failed | |
5379 | */ | |
5380 | if (--mpb->num_raid_devs == 0) { | |
5381 | struct dl *d; | |
5382 | ||
5383 | for (d = super->disks; d; d = d->next) | |
5384 | if (d->index > -2) { | |
5385 | d->index = -1; | |
5386 | d->disk.status = SPARE_DISK; | |
5387 | } | |
5388 | } | |
5389 | ||
5390 | super->updates_pending++; | |
5391 | ||
5392 | return 0; | |
5393 | } | |
aa534678 | 5394 | |
a951a4f7 | 5395 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 5396 | char *update, struct mddev_ident *ident) |
aa534678 DW |
5397 | { |
5398 | /* update the subarray currently referenced by ->current_vol */ | |
5399 | struct intel_super *super = st->sb; | |
5400 | struct imsm_super *mpb = super->anchor; | |
5401 | ||
aa534678 DW |
5402 | if (strcmp(update, "name") == 0) { |
5403 | char *name = ident->name; | |
a951a4f7 N |
5404 | char *ep; |
5405 | int vol; | |
aa534678 | 5406 | |
a951a4f7 | 5407 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
5408 | fprintf(stderr, |
5409 | Name ": Unable to update name of active subarray\n"); | |
5410 | return 2; | |
5411 | } | |
5412 | ||
5413 | if (!check_name(super, name, 0)) | |
5414 | return 2; | |
5415 | ||
a951a4f7 N |
5416 | vol = strtoul(subarray, &ep, 10); |
5417 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
5418 | return 2; | |
5419 | ||
aa534678 DW |
5420 | if (st->update_tail) { |
5421 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
5422 | ||
5423 | if (!u) | |
5424 | return 2; | |
5425 | u->type = update_rename_array; | |
a951a4f7 | 5426 | u->dev_idx = vol; |
aa534678 DW |
5427 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
5428 | append_metadata_update(st, u, sizeof(*u)); | |
5429 | } else { | |
5430 | struct imsm_dev *dev; | |
5431 | int i; | |
5432 | ||
a951a4f7 | 5433 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
5434 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
5435 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5436 | dev = get_imsm_dev(super, i); | |
5437 | handle_missing(super, dev); | |
5438 | } | |
5439 | super->updates_pending++; | |
5440 | } | |
5441 | } else | |
5442 | return 2; | |
5443 | ||
5444 | return 0; | |
5445 | } | |
bf5a934a | 5446 | |
28bce06f AK |
5447 | static int is_gen_migration(struct imsm_dev *dev) |
5448 | { | |
7534230b AK |
5449 | if (dev == NULL) |
5450 | return 0; | |
5451 | ||
28bce06f AK |
5452 | if (!dev->vol.migr_state) |
5453 | return 0; | |
5454 | ||
5455 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
5456 | return 1; | |
5457 | ||
5458 | return 0; | |
5459 | } | |
71204a50 | 5460 | #endif /* MDASSEMBLE */ |
28bce06f | 5461 | |
1e5c6983 DW |
5462 | static int is_rebuilding(struct imsm_dev *dev) |
5463 | { | |
5464 | struct imsm_map *migr_map; | |
5465 | ||
5466 | if (!dev->vol.migr_state) | |
5467 | return 0; | |
5468 | ||
5469 | if (migr_type(dev) != MIGR_REBUILD) | |
5470 | return 0; | |
5471 | ||
5472 | migr_map = get_imsm_map(dev, 1); | |
5473 | ||
5474 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
5475 | return 1; | |
5476 | else | |
5477 | return 0; | |
5478 | } | |
5479 | ||
c47b0ff6 AK |
5480 | static void update_recovery_start(struct intel_super *super, |
5481 | struct imsm_dev *dev, | |
5482 | struct mdinfo *array) | |
1e5c6983 DW |
5483 | { |
5484 | struct mdinfo *rebuild = NULL; | |
5485 | struct mdinfo *d; | |
5486 | __u32 units; | |
5487 | ||
5488 | if (!is_rebuilding(dev)) | |
5489 | return; | |
5490 | ||
5491 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
5492 | for (d = array->devs; d; d = d->next) | |
5493 | if (d->recovery_start == 0) { | |
5494 | if (rebuild) | |
5495 | return; | |
5496 | rebuild = d; | |
5497 | } | |
5498 | ||
4363fd80 DW |
5499 | if (!rebuild) { |
5500 | /* (?) none of the disks are marked with | |
5501 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
5502 | * disk_ord_tbl was not correctly updated | |
5503 | */ | |
5504 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
5505 | return; | |
5506 | } | |
5507 | ||
1e5c6983 | 5508 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 5509 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
5510 | } |
5511 | ||
9e2d750d | 5512 | #ifndef MDASSEMBLE |
276d77db | 5513 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 5514 | #endif |
1e5c6983 | 5515 | |
00bbdbda | 5516 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 5517 | { |
4f5bc454 DW |
5518 | /* Given a container loaded by load_super_imsm_all, |
5519 | * extract information about all the arrays into | |
5520 | * an mdinfo tree. | |
00bbdbda | 5521 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
5522 | * |
5523 | * For each imsm_dev create an mdinfo, fill it in, | |
5524 | * then look for matching devices in super->disks | |
5525 | * and create appropriate device mdinfo. | |
5526 | */ | |
5527 | struct intel_super *super = st->sb; | |
949c47a0 | 5528 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 5529 | struct mdinfo *rest = NULL; |
00bbdbda | 5530 | unsigned int i; |
a06d022d | 5531 | int bbm_errors = 0; |
abef11a3 AK |
5532 | struct dl *d; |
5533 | int spare_disks = 0; | |
cdddbdbc | 5534 | |
19482bcc AK |
5535 | /* do not assemble arrays when not all attributes are supported */ |
5536 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
5537 | fprintf(stderr, Name ": IMSM metadata loading not allowed " | |
5538 | "due to attributes incompatibility.\n"); | |
5539 | return NULL; | |
5540 | } | |
5541 | ||
a06d022d KW |
5542 | /* check for bad blocks */ |
5543 | if (imsm_bbm_log_size(super->anchor)) | |
5544 | bbm_errors = 1; | |
604b746f | 5545 | |
abef11a3 AK |
5546 | /* count spare devices, not used in maps |
5547 | */ | |
5548 | for (d = super->disks; d; d = d->next) | |
5549 | if (d->index == -1) | |
5550 | spare_disks++; | |
5551 | ||
4f5bc454 | 5552 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
5553 | struct imsm_dev *dev; |
5554 | struct imsm_map *map; | |
86e3692b | 5555 | struct imsm_map *map2; |
4f5bc454 | 5556 | struct mdinfo *this; |
2db86302 | 5557 | int slot, chunk; |
00bbdbda N |
5558 | char *ep; |
5559 | ||
5560 | if (subarray && | |
5561 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
5562 | continue; | |
5563 | ||
5564 | dev = get_imsm_dev(super, i); | |
5565 | map = get_imsm_map(dev, 0); | |
86e3692b | 5566 | map2 = get_imsm_map(dev, 1); |
4f5bc454 | 5567 | |
1ce0101c DW |
5568 | /* do not publish arrays that are in the middle of an |
5569 | * unsupported migration | |
5570 | */ | |
5571 | if (dev->vol.migr_state && | |
28bce06f | 5572 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
5573 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
5574 | " unsupported migration in progress\n", | |
5575 | dev->volume); | |
5576 | continue; | |
5577 | } | |
2db86302 LM |
5578 | /* do not publish arrays that are not support by controller's |
5579 | * OROM/EFI | |
5580 | */ | |
1ce0101c | 5581 | |
2db86302 | 5582 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
7b0bbd0f | 5583 | #ifndef MDASSEMBLE |
2db86302 LM |
5584 | if (!validate_geometry_imsm_orom(super, |
5585 | get_imsm_raid_level(map), /* RAID level */ | |
5586 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
5587 | map->num_members, /* raid disks */ | |
5588 | &chunk, | |
5589 | 1 /* verbose */)) { | |
5590 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
5591 | "Cannot proceed with the action(s).\n"); | |
5592 | continue; | |
5593 | } | |
7b0bbd0f | 5594 | #endif /* MDASSEMBLE */ |
4f5bc454 | 5595 | this = malloc(sizeof(*this)); |
0fbd635c | 5596 | if (!this) { |
cf1be220 | 5597 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
5598 | sizeof(*this)); |
5599 | break; | |
5600 | } | |
4f5bc454 | 5601 | |
301406c9 | 5602 | super->current_vol = i; |
a5d85af7 | 5603 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 5604 | this->next = rest; |
4f5bc454 | 5605 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 5606 | unsigned long long recovery_start; |
4f5bc454 DW |
5607 | struct mdinfo *info_d; |
5608 | struct dl *d; | |
5609 | int idx; | |
9a1608e5 | 5610 | int skip; |
7eef0453 | 5611 | __u32 ord; |
4f5bc454 | 5612 | |
9a1608e5 | 5613 | skip = 0; |
98130f40 | 5614 | idx = get_imsm_disk_idx(dev, slot, 0); |
196b0d44 | 5615 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
4f5bc454 DW |
5616 | for (d = super->disks; d ; d = d->next) |
5617 | if (d->index == idx) | |
0fbd635c | 5618 | break; |
4f5bc454 | 5619 | |
1e5c6983 | 5620 | recovery_start = MaxSector; |
4f5bc454 | 5621 | if (d == NULL) |
9a1608e5 | 5622 | skip = 1; |
25ed7e59 | 5623 | if (d && is_failed(&d->disk)) |
9a1608e5 | 5624 | skip = 1; |
7eef0453 | 5625 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 5626 | recovery_start = 0; |
9a1608e5 DW |
5627 | |
5628 | /* | |
5629 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
5630 | * reset resync start to avoid a dirty-degraded |
5631 | * situation when performing the intial sync | |
9a1608e5 DW |
5632 | * |
5633 | * FIXME handle dirty degraded | |
5634 | */ | |
1e5c6983 | 5635 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 5636 | this->resync_start = MaxSector; |
9a1608e5 DW |
5637 | if (skip) |
5638 | continue; | |
4f5bc454 | 5639 | |
1e5c6983 | 5640 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
5641 | if (!info_d) { |
5642 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 5643 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
5644 | info_d = this->devs; |
5645 | while (info_d) { | |
5646 | struct mdinfo *d = info_d->next; | |
5647 | ||
5648 | free(info_d); | |
5649 | info_d = d; | |
5650 | } | |
9a1608e5 DW |
5651 | free(this); |
5652 | this = rest; | |
5653 | break; | |
5654 | } | |
4f5bc454 DW |
5655 | info_d->next = this->devs; |
5656 | this->devs = info_d; | |
5657 | ||
4f5bc454 DW |
5658 | info_d->disk.number = d->index; |
5659 | info_d->disk.major = d->major; | |
5660 | info_d->disk.minor = d->minor; | |
5661 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 5662 | info_d->recovery_start = recovery_start; |
86e3692b AK |
5663 | if (map2) { |
5664 | if (slot < map2->num_members) | |
5665 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5666 | else |
5667 | this->array.spare_disks++; | |
86e3692b AK |
5668 | } else { |
5669 | if (slot < map->num_members) | |
5670 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5671 | else |
5672 | this->array.spare_disks++; | |
86e3692b | 5673 | } |
1e5c6983 DW |
5674 | if (info_d->recovery_start == MaxSector) |
5675 | this->array.working_disks++; | |
4f5bc454 DW |
5676 | |
5677 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5678 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
5679 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
4f5bc454 | 5680 | } |
1e5c6983 | 5681 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 5682 | update_recovery_start(super, dev, this); |
abef11a3 | 5683 | this->array.spare_disks += spare_disks; |
276d77db | 5684 | |
9e2d750d | 5685 | #ifndef MDASSEMBLE |
276d77db AK |
5686 | /* check for reshape */ |
5687 | if (this->reshape_active == 1) | |
5688 | recover_backup_imsm(st, this); | |
9e2d750d | 5689 | #endif |
9a1608e5 | 5690 | rest = this; |
4f5bc454 DW |
5691 | } |
5692 | ||
a06d022d KW |
5693 | /* if array has bad blocks, set suitable bit in array status */ |
5694 | if (bbm_errors) | |
5695 | rest->array.state |= (1<<MD_SB_BBM_ERRORS); | |
5696 | ||
4f5bc454 | 5697 | return rest; |
cdddbdbc DW |
5698 | } |
5699 | ||
845dea95 | 5700 | |
fb49eef2 | 5701 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed) |
c2a1e7da | 5702 | { |
a965f303 | 5703 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
5704 | |
5705 | if (!failed) | |
3393c6af DW |
5706 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
5707 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
5708 | |
5709 | switch (get_imsm_raid_level(map)) { | |
5710 | case 0: | |
5711 | return IMSM_T_STATE_FAILED; | |
5712 | break; | |
5713 | case 1: | |
5714 | if (failed < map->num_members) | |
5715 | return IMSM_T_STATE_DEGRADED; | |
5716 | else | |
5717 | return IMSM_T_STATE_FAILED; | |
5718 | break; | |
5719 | case 10: | |
5720 | { | |
5721 | /** | |
c92a2527 DW |
5722 | * check to see if any mirrors have failed, otherwise we |
5723 | * are degraded. Even numbered slots are mirrored on | |
5724 | * slot+1 | |
c2a1e7da | 5725 | */ |
c2a1e7da | 5726 | int i; |
d9b420a5 N |
5727 | /* gcc -Os complains that this is unused */ |
5728 | int insync = insync; | |
c2a1e7da DW |
5729 | |
5730 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 5731 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
c92a2527 DW |
5732 | int idx = ord_to_idx(ord); |
5733 | struct imsm_disk *disk; | |
c2a1e7da | 5734 | |
c92a2527 DW |
5735 | /* reset the potential in-sync count on even-numbered |
5736 | * slots. num_copies is always 2 for imsm raid10 | |
5737 | */ | |
5738 | if ((i & 1) == 0) | |
5739 | insync = 2; | |
c2a1e7da | 5740 | |
c92a2527 | 5741 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 5742 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 5743 | insync--; |
c2a1e7da | 5744 | |
c92a2527 DW |
5745 | /* no in-sync disks left in this mirror the |
5746 | * array has failed | |
5747 | */ | |
5748 | if (insync == 0) | |
5749 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
5750 | } |
5751 | ||
5752 | return IMSM_T_STATE_DEGRADED; | |
5753 | } | |
5754 | case 5: | |
5755 | if (failed < 2) | |
5756 | return IMSM_T_STATE_DEGRADED; | |
5757 | else | |
5758 | return IMSM_T_STATE_FAILED; | |
5759 | break; | |
5760 | default: | |
5761 | break; | |
5762 | } | |
5763 | ||
5764 | return map->map_state; | |
5765 | } | |
5766 | ||
ff077194 | 5767 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev) |
c2a1e7da DW |
5768 | { |
5769 | int i; | |
5770 | int failed = 0; | |
5771 | struct imsm_disk *disk; | |
ff077194 | 5772 | struct imsm_map *map = get_imsm_map(dev, 0); |
0556e1a2 DW |
5773 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
5774 | __u32 ord; | |
5775 | int idx; | |
c2a1e7da | 5776 | |
0556e1a2 DW |
5777 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
5778 | * disks that are being rebuilt. New failures are recorded to | |
5779 | * map[0]. So we look through all the disks we started with and | |
5780 | * see if any failures are still present, or if any new ones | |
5781 | * have arrived | |
5782 | * | |
5783 | * FIXME add support for online capacity expansion and | |
5784 | * raid-level-migration | |
5785 | */ | |
5786 | for (i = 0; i < prev->num_members; i++) { | |
5787 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); | |
5788 | ord |= __le32_to_cpu(map->disk_ord_tbl[i]); | |
5789 | idx = ord_to_idx(ord); | |
c2a1e7da | 5790 | |
949c47a0 | 5791 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 5792 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
fcb84475 | 5793 | failed++; |
c2a1e7da DW |
5794 | } |
5795 | ||
5796 | return failed; | |
845dea95 NB |
5797 | } |
5798 | ||
97b4d0e9 DW |
5799 | #ifndef MDASSEMBLE |
5800 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
5801 | char *inst) | |
5802 | { | |
5803 | struct intel_super *super = c->sb; | |
5804 | struct imsm_super *mpb = super->anchor; | |
5805 | ||
5806 | if (atoi(inst) >= mpb->num_raid_devs) { | |
5807 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
5808 | __func__, atoi(inst)); | |
5809 | return -ENODEV; | |
5810 | } | |
5811 | ||
5812 | dprintf("imsm: open_new %s\n", inst); | |
5813 | a->info.container_member = atoi(inst); | |
5814 | return 0; | |
5815 | } | |
5816 | ||
0c046afd DW |
5817 | static int is_resyncing(struct imsm_dev *dev) |
5818 | { | |
5819 | struct imsm_map *migr_map; | |
5820 | ||
5821 | if (!dev->vol.migr_state) | |
5822 | return 0; | |
5823 | ||
1484e727 DW |
5824 | if (migr_type(dev) == MIGR_INIT || |
5825 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
5826 | return 1; |
5827 | ||
4c9bc37b AK |
5828 | if (migr_type(dev) == MIGR_GEN_MIGR) |
5829 | return 0; | |
5830 | ||
0c046afd DW |
5831 | migr_map = get_imsm_map(dev, 1); |
5832 | ||
4c9bc37b AK |
5833 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
5834 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
5835 | return 1; |
5836 | else | |
5837 | return 0; | |
5838 | } | |
5839 | ||
0556e1a2 DW |
5840 | /* return true if we recorded new information */ |
5841 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 5842 | { |
0556e1a2 DW |
5843 | __u32 ord; |
5844 | int slot; | |
5845 | struct imsm_map *map; | |
5846 | ||
5847 | /* new failures are always set in map[0] */ | |
5848 | map = get_imsm_map(dev, 0); | |
5849 | ||
5850 | slot = get_imsm_disk_slot(map, idx); | |
5851 | if (slot < 0) | |
5852 | return 0; | |
5853 | ||
5854 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 5855 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
5856 | return 0; |
5857 | ||
f2f27e63 | 5858 | disk->status |= FAILED_DISK; |
0556e1a2 | 5859 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
f21e18ca | 5860 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
5861 | map->failed_disk_num = slot; |
5862 | return 1; | |
5863 | } | |
5864 | ||
5865 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
5866 | { | |
5867 | mark_failure(dev, disk, idx); | |
5868 | ||
5869 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
5870 | return; | |
5871 | ||
47ee5a45 DW |
5872 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
5873 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
5874 | } | |
5875 | ||
33414a01 DW |
5876 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
5877 | { | |
5878 | __u8 map_state; | |
5879 | struct dl *dl; | |
5880 | int failed; | |
5881 | ||
5882 | if (!super->missing) | |
5883 | return; | |
5884 | failed = imsm_count_failed(super, dev); | |
5885 | map_state = imsm_check_degraded(super, dev, failed); | |
5886 | ||
5887 | dprintf("imsm: mark missing\n"); | |
5888 | end_migration(dev, map_state); | |
5889 | for (dl = super->missing; dl; dl = dl->next) | |
5890 | mark_missing(dev, &dl->disk, dl->index); | |
5891 | super->updates_pending++; | |
5892 | } | |
5893 | ||
70bdf0dc AK |
5894 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev) |
5895 | { | |
5896 | int used_disks = imsm_num_data_members(dev, 0); | |
5897 | unsigned long long array_blocks; | |
5898 | struct imsm_map *map; | |
5899 | ||
5900 | if (used_disks == 0) { | |
5901 | /* when problems occures | |
5902 | * return current array_blocks value | |
5903 | */ | |
5904 | array_blocks = __le32_to_cpu(dev->size_high); | |
5905 | array_blocks = array_blocks << 32; | |
5906 | array_blocks += __le32_to_cpu(dev->size_low); | |
5907 | ||
5908 | return array_blocks; | |
5909 | } | |
5910 | ||
5911 | /* set array size in metadata | |
5912 | */ | |
5913 | map = get_imsm_map(dev, 0); | |
5914 | array_blocks = map->blocks_per_member * used_disks; | |
5915 | ||
5916 | /* round array size down to closest MB | |
5917 | */ | |
5918 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
5919 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
5920 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
5921 | ||
5922 | return array_blocks; | |
5923 | } | |
5924 | ||
28bce06f AK |
5925 | static void imsm_set_disk(struct active_array *a, int n, int state); |
5926 | ||
0e2d1a4e AK |
5927 | static void imsm_progress_container_reshape(struct intel_super *super) |
5928 | { | |
5929 | /* if no device has a migr_state, but some device has a | |
5930 | * different number of members than the previous device, start | |
5931 | * changing the number of devices in this device to match | |
5932 | * previous. | |
5933 | */ | |
5934 | struct imsm_super *mpb = super->anchor; | |
5935 | int prev_disks = -1; | |
5936 | int i; | |
1dfaa380 | 5937 | int copy_map_size; |
0e2d1a4e AK |
5938 | |
5939 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5940 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5941 | struct imsm_map *map = get_imsm_map(dev, 0); | |
5942 | struct imsm_map *map2; | |
5943 | int prev_num_members; | |
0e2d1a4e AK |
5944 | |
5945 | if (dev->vol.migr_state) | |
5946 | return; | |
5947 | ||
5948 | if (prev_disks == -1) | |
5949 | prev_disks = map->num_members; | |
5950 | if (prev_disks == map->num_members) | |
5951 | continue; | |
5952 | ||
5953 | /* OK, this array needs to enter reshape mode. | |
5954 | * i.e it needs a migr_state | |
5955 | */ | |
5956 | ||
1dfaa380 | 5957 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
5958 | prev_num_members = map->num_members; |
5959 | map->num_members = prev_disks; | |
5960 | dev->vol.migr_state = 1; | |
5961 | dev->vol.curr_migr_unit = 0; | |
5962 | dev->vol.migr_type = MIGR_GEN_MIGR; | |
5963 | for (i = prev_num_members; | |
5964 | i < map->num_members; i++) | |
5965 | set_imsm_ord_tbl_ent(map, i, i); | |
5966 | map2 = get_imsm_map(dev, 1); | |
5967 | /* Copy the current map */ | |
1dfaa380 | 5968 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
5969 | map2->num_members = prev_num_members; |
5970 | ||
70bdf0dc | 5971 | imsm_set_array_size(dev); |
0e2d1a4e AK |
5972 | super->updates_pending++; |
5973 | } | |
5974 | } | |
5975 | ||
aad6f216 | 5976 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
5977 | * states are handled in imsm_set_disk() with one exception, when a |
5978 | * resync is stopped due to a new failure this routine will set the | |
5979 | * 'degraded' state for the array. | |
5980 | */ | |
01f157d7 | 5981 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
5982 | { |
5983 | int inst = a->info.container_member; | |
5984 | struct intel_super *super = a->container->sb; | |
949c47a0 | 5985 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 5986 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd DW |
5987 | int failed = imsm_count_failed(super, dev); |
5988 | __u8 map_state = imsm_check_degraded(super, dev, failed); | |
1e5c6983 | 5989 | __u32 blocks_per_unit; |
a862209d | 5990 | |
1af97990 AK |
5991 | if (dev->vol.migr_state && |
5992 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
5993 | /* array state change is blocked due to reshape action | |
aad6f216 N |
5994 | * We might need to |
5995 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
5996 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
5997 | * - update curr_migr_unit | |
1af97990 | 5998 | */ |
aad6f216 N |
5999 | if (a->curr_action == reshape) { |
6000 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 6001 | goto mark_checkpoint; |
aad6f216 N |
6002 | } else { |
6003 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
6004 | /* for some reason we aborted the reshape. | |
b66e591b AK |
6005 | * |
6006 | * disable automatic metadata rollback | |
6007 | * user action is required to recover process | |
aad6f216 | 6008 | */ |
b66e591b | 6009 | if (0) { |
aad6f216 N |
6010 | struct imsm_map *map2 = get_imsm_map(dev, 1); |
6011 | dev->vol.migr_state = 0; | |
6012 | dev->vol.migr_type = 0; | |
6013 | dev->vol.curr_migr_unit = 0; | |
6014 | memcpy(map, map2, sizeof_imsm_map(map2)); | |
6015 | super->updates_pending++; | |
b66e591b | 6016 | } |
aad6f216 N |
6017 | } |
6018 | if (a->last_checkpoint >= a->info.component_size) { | |
6019 | unsigned long long array_blocks; | |
6020 | int used_disks; | |
e154ced3 | 6021 | struct mdinfo *mdi; |
aad6f216 | 6022 | |
9653001d | 6023 | used_disks = imsm_num_data_members(dev, 0); |
d55adef9 AK |
6024 | if (used_disks > 0) { |
6025 | array_blocks = | |
6026 | map->blocks_per_member * | |
6027 | used_disks; | |
6028 | /* round array size down to closest MB | |
6029 | */ | |
6030 | array_blocks = (array_blocks | |
6031 | >> SECT_PER_MB_SHIFT) | |
6032 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
6033 | a->info.custom_array_size = array_blocks; |
6034 | /* encourage manager to update array | |
6035 | * size | |
6036 | */ | |
e154ced3 | 6037 | |
d55adef9 | 6038 | a->check_reshape = 1; |
633b5610 | 6039 | } |
e154ced3 AK |
6040 | /* finalize online capacity expansion/reshape */ |
6041 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
6042 | imsm_set_disk(a, | |
6043 | mdi->disk.raid_disk, | |
6044 | mdi->curr_state); | |
6045 | ||
0e2d1a4e | 6046 | imsm_progress_container_reshape(super); |
e154ced3 | 6047 | } |
aad6f216 | 6048 | } |
1af97990 AK |
6049 | } |
6050 | ||
47ee5a45 | 6051 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
6052 | if (consistent == 2) |
6053 | handle_missing(super, dev); | |
1e5c6983 | 6054 | |
0c046afd | 6055 | if (consistent == 2 && |
b7941fd6 | 6056 | (!is_resync_complete(&a->info) || |
0c046afd DW |
6057 | map_state != IMSM_T_STATE_NORMAL || |
6058 | dev->vol.migr_state)) | |
01f157d7 | 6059 | consistent = 0; |
272906ef | 6060 | |
b7941fd6 | 6061 | if (is_resync_complete(&a->info)) { |
0c046afd | 6062 | /* complete intialization / resync, |
0556e1a2 DW |
6063 | * recovery and interrupted recovery is completed in |
6064 | * ->set_disk | |
0c046afd DW |
6065 | */ |
6066 | if (is_resyncing(dev)) { | |
6067 | dprintf("imsm: mark resync done\n"); | |
f8f603f1 | 6068 | end_migration(dev, map_state); |
115c3803 | 6069 | super->updates_pending++; |
484240d8 | 6070 | a->last_checkpoint = 0; |
115c3803 | 6071 | } |
0c046afd DW |
6072 | } else if (!is_resyncing(dev) && !failed) { |
6073 | /* mark the start of the init process if nothing is failed */ | |
b7941fd6 | 6074 | dprintf("imsm: mark resync start\n"); |
1484e727 | 6075 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 6076 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 6077 | else |
8e59f3d8 | 6078 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 6079 | super->updates_pending++; |
115c3803 | 6080 | } |
a862209d | 6081 | |
633b5610 | 6082 | mark_checkpoint: |
5b83bacf AK |
6083 | /* skip checkpointing for general migration, |
6084 | * it is controlled in mdadm | |
6085 | */ | |
6086 | if (is_gen_migration(dev)) | |
6087 | goto skip_mark_checkpoint; | |
6088 | ||
1e5c6983 | 6089 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 6090 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 6091 | if (blocks_per_unit) { |
1e5c6983 DW |
6092 | __u32 units32; |
6093 | __u64 units; | |
6094 | ||
4f0a7acc | 6095 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
6096 | units32 = units; |
6097 | ||
6098 | /* check that we did not overflow 32-bits, and that | |
6099 | * curr_migr_unit needs updating | |
6100 | */ | |
6101 | if (units32 == units && | |
bfd80a56 | 6102 | units32 != 0 && |
1e5c6983 DW |
6103 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
6104 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
6105 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
6106 | super->updates_pending++; | |
6107 | } | |
6108 | } | |
f8f603f1 | 6109 | |
5b83bacf | 6110 | skip_mark_checkpoint: |
3393c6af | 6111 | /* mark dirty / clean */ |
0c046afd | 6112 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 6113 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
6114 | if (consistent) |
6115 | dev->vol.dirty = 0; | |
6116 | else | |
6117 | dev->vol.dirty = 1; | |
a862209d DW |
6118 | super->updates_pending++; |
6119 | } | |
28bce06f | 6120 | |
01f157d7 | 6121 | return consistent; |
a862209d DW |
6122 | } |
6123 | ||
8d45d196 | 6124 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 6125 | { |
8d45d196 DW |
6126 | int inst = a->info.container_member; |
6127 | struct intel_super *super = a->container->sb; | |
949c47a0 | 6128 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 6129 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 | 6130 | struct imsm_disk *disk; |
0c046afd | 6131 | int failed; |
b10b37b8 | 6132 | __u32 ord; |
0c046afd | 6133 | __u8 map_state; |
8d45d196 DW |
6134 | |
6135 | if (n > map->num_members) | |
6136 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
6137 | n, map->num_members - 1); | |
6138 | ||
6139 | if (n < 0) | |
6140 | return; | |
6141 | ||
4e6e574a | 6142 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 6143 | |
98130f40 | 6144 | ord = get_imsm_ord_tbl_ent(dev, n, -1); |
b10b37b8 | 6145 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 6146 | |
5802a811 | 6147 | /* check for new failures */ |
0556e1a2 DW |
6148 | if (state & DS_FAULTY) { |
6149 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
6150 | super->updates_pending++; | |
8d45d196 | 6151 | } |
47ee5a45 | 6152 | |
19859edc | 6153 | /* check if in_sync */ |
0556e1a2 | 6154 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
b10b37b8 DW |
6155 | struct imsm_map *migr_map = get_imsm_map(dev, 1); |
6156 | ||
6157 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
6158 | super->updates_pending++; |
6159 | } | |
8d45d196 | 6160 | |
0c046afd DW |
6161 | failed = imsm_count_failed(super, dev); |
6162 | map_state = imsm_check_degraded(super, dev, failed); | |
5802a811 | 6163 | |
0c046afd DW |
6164 | /* check if recovery complete, newly degraded, or failed */ |
6165 | if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) { | |
f8f603f1 | 6166 | end_migration(dev, map_state); |
0556e1a2 DW |
6167 | map = get_imsm_map(dev, 0); |
6168 | map->failed_disk_num = ~0; | |
0c046afd | 6169 | super->updates_pending++; |
484240d8 | 6170 | a->last_checkpoint = 0; |
0c046afd DW |
6171 | } else if (map_state == IMSM_T_STATE_DEGRADED && |
6172 | map->map_state != map_state && | |
6173 | !dev->vol.migr_state) { | |
6174 | dprintf("imsm: mark degraded\n"); | |
6175 | map->map_state = map_state; | |
6176 | super->updates_pending++; | |
484240d8 | 6177 | a->last_checkpoint = 0; |
0c046afd DW |
6178 | } else if (map_state == IMSM_T_STATE_FAILED && |
6179 | map->map_state != map_state) { | |
6180 | dprintf("imsm: mark failed\n"); | |
f8f603f1 | 6181 | end_migration(dev, map_state); |
0c046afd | 6182 | super->updates_pending++; |
484240d8 | 6183 | a->last_checkpoint = 0; |
28bce06f AK |
6184 | } else if (is_gen_migration(dev)) { |
6185 | dprintf("imsm: Detected General Migration in state: "); | |
6186 | if (map_state == IMSM_T_STATE_NORMAL) { | |
6187 | end_migration(dev, map_state); | |
6188 | map = get_imsm_map(dev, 0); | |
6189 | map->failed_disk_num = ~0; | |
6190 | dprintf("normal\n"); | |
6191 | } else { | |
6192 | if (map_state == IMSM_T_STATE_DEGRADED) { | |
6193 | printf("degraded\n"); | |
6194 | end_migration(dev, map_state); | |
6195 | } else { | |
6196 | dprintf("failed\n"); | |
6197 | } | |
6198 | map->map_state = map_state; | |
6199 | } | |
6200 | super->updates_pending++; | |
5802a811 | 6201 | } |
845dea95 NB |
6202 | } |
6203 | ||
f796af5d | 6204 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 6205 | { |
f796af5d | 6206 | void *buf = mpb; |
c2a1e7da DW |
6207 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
6208 | unsigned long long dsize; | |
6209 | unsigned long long sectors; | |
6210 | ||
6211 | get_dev_size(fd, NULL, &dsize); | |
6212 | ||
272f648f DW |
6213 | if (mpb_size > 512) { |
6214 | /* -1 to account for anchor */ | |
6215 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 6216 | |
272f648f DW |
6217 | /* write the extended mpb to the sectors preceeding the anchor */ |
6218 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
6219 | return 1; | |
c2a1e7da | 6220 | |
f21e18ca N |
6221 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
6222 | != 512 * sectors) | |
272f648f DW |
6223 | return 1; |
6224 | } | |
c2a1e7da | 6225 | |
272f648f DW |
6226 | /* first block is stored on second to last sector of the disk */ |
6227 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
6228 | return 1; |
6229 | ||
f796af5d | 6230 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
6231 | return 1; |
6232 | ||
c2a1e7da DW |
6233 | return 0; |
6234 | } | |
6235 | ||
2e735d19 | 6236 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 6237 | { |
2e735d19 | 6238 | struct intel_super *super = container->sb; |
c2a1e7da | 6239 | |
1a64be56 | 6240 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
6241 | if (!super->updates_pending) |
6242 | return; | |
6243 | ||
36988a3d | 6244 | write_super_imsm(container, 0); |
c2a1e7da DW |
6245 | |
6246 | super->updates_pending = 0; | |
845dea95 NB |
6247 | } |
6248 | ||
272906ef DW |
6249 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
6250 | { | |
6251 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 6252 | int i = get_imsm_disk_idx(dev, idx, -1); |
272906ef DW |
6253 | struct dl *dl; |
6254 | ||
6255 | for (dl = super->disks; dl; dl = dl->next) | |
6256 | if (dl->index == i) | |
6257 | break; | |
6258 | ||
25ed7e59 | 6259 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
6260 | dl = NULL; |
6261 | ||
6262 | if (dl) | |
6263 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
6264 | ||
6265 | return dl; | |
6266 | } | |
6267 | ||
a20d2ba5 | 6268 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
6269 | struct active_array *a, int activate_new, |
6270 | struct mdinfo *additional_test_list) | |
272906ef DW |
6271 | { |
6272 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 6273 | int idx = get_imsm_disk_idx(dev, slot, -1); |
a20d2ba5 DW |
6274 | struct imsm_super *mpb = super->anchor; |
6275 | struct imsm_map *map; | |
272906ef DW |
6276 | unsigned long long pos; |
6277 | struct mdinfo *d; | |
6278 | struct extent *ex; | |
a20d2ba5 | 6279 | int i, j; |
272906ef | 6280 | int found; |
569cc43f DW |
6281 | __u32 array_start = 0; |
6282 | __u32 array_end = 0; | |
272906ef | 6283 | struct dl *dl; |
6c932028 | 6284 | struct mdinfo *test_list; |
272906ef DW |
6285 | |
6286 | for (dl = super->disks; dl; dl = dl->next) { | |
6287 | /* If in this array, skip */ | |
6288 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
6289 | if (d->state_fd >= 0 && |
6290 | d->disk.major == dl->major && | |
272906ef | 6291 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
6292 | dprintf("%x:%x already in array\n", |
6293 | dl->major, dl->minor); | |
272906ef DW |
6294 | break; |
6295 | } | |
6296 | if (d) | |
6297 | continue; | |
6c932028 AK |
6298 | test_list = additional_test_list; |
6299 | while (test_list) { | |
6300 | if (test_list->disk.major == dl->major && | |
6301 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
6302 | dprintf("%x:%x already in additional test list\n", |
6303 | dl->major, dl->minor); | |
6304 | break; | |
6305 | } | |
6c932028 | 6306 | test_list = test_list->next; |
8ba77d32 | 6307 | } |
6c932028 | 6308 | if (test_list) |
8ba77d32 | 6309 | continue; |
272906ef | 6310 | |
e553d2a4 | 6311 | /* skip in use or failed drives */ |
25ed7e59 | 6312 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
6313 | dl->index == -2) { |
6314 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 6315 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
6316 | continue; |
6317 | } | |
6318 | ||
a20d2ba5 DW |
6319 | /* skip pure spares when we are looking for partially |
6320 | * assimilated drives | |
6321 | */ | |
6322 | if (dl->index == -1 && !activate_new) | |
6323 | continue; | |
6324 | ||
272906ef | 6325 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 6326 | * It needs to be able to cover all member volumes |
272906ef DW |
6327 | */ |
6328 | ex = get_extents(super, dl); | |
6329 | if (!ex) { | |
6330 | dprintf("cannot get extents\n"); | |
6331 | continue; | |
6332 | } | |
a20d2ba5 DW |
6333 | for (i = 0; i < mpb->num_raid_devs; i++) { |
6334 | dev = get_imsm_dev(super, i); | |
6335 | map = get_imsm_map(dev, 0); | |
272906ef | 6336 | |
a20d2ba5 DW |
6337 | /* check if this disk is already a member of |
6338 | * this array | |
272906ef | 6339 | */ |
620b1713 | 6340 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
6341 | continue; |
6342 | ||
6343 | found = 0; | |
6344 | j = 0; | |
6345 | pos = 0; | |
6346 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
329c8278 DW |
6347 | array_end = array_start + |
6348 | __le32_to_cpu(map->blocks_per_member) - 1; | |
a20d2ba5 DW |
6349 | |
6350 | do { | |
6351 | /* check that we can start at pba_of_lba0 with | |
6352 | * blocks_per_member of space | |
6353 | */ | |
329c8278 | 6354 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
6355 | found = 1; |
6356 | break; | |
6357 | } | |
6358 | pos = ex[j].start + ex[j].size; | |
6359 | j++; | |
6360 | } while (ex[j-1].size); | |
6361 | ||
6362 | if (!found) | |
272906ef | 6363 | break; |
a20d2ba5 | 6364 | } |
272906ef DW |
6365 | |
6366 | free(ex); | |
a20d2ba5 | 6367 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
6368 | dprintf("%x:%x does not have %u to %u available\n", |
6369 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
6370 | /* No room */ |
6371 | continue; | |
a20d2ba5 DW |
6372 | } |
6373 | return dl; | |
272906ef DW |
6374 | } |
6375 | ||
6376 | return dl; | |
6377 | } | |
6378 | ||
95d07a2c LM |
6379 | |
6380 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
6381 | { | |
6382 | struct imsm_dev *dev2; | |
6383 | struct imsm_map *map; | |
6384 | struct dl *idisk; | |
6385 | int slot; | |
6386 | int idx; | |
6387 | __u8 state; | |
6388 | ||
6389 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
6390 | if (dev2) { | |
6391 | state = imsm_check_degraded(cont->sb, dev2, failed); | |
6392 | if (state == IMSM_T_STATE_FAILED) { | |
6393 | map = get_imsm_map(dev2, 0); | |
6394 | if (!map) | |
6395 | return 1; | |
6396 | for (slot = 0; slot < map->num_members; slot++) { | |
6397 | /* | |
6398 | * Check if failed disks are deleted from intel | |
6399 | * disk list or are marked to be deleted | |
6400 | */ | |
98130f40 | 6401 | idx = get_imsm_disk_idx(dev2, slot, -1); |
95d07a2c LM |
6402 | idisk = get_imsm_dl_disk(cont->sb, idx); |
6403 | /* | |
6404 | * Do not rebuild the array if failed disks | |
6405 | * from failed sub-array are not removed from | |
6406 | * container. | |
6407 | */ | |
6408 | if (idisk && | |
6409 | is_failed(&idisk->disk) && | |
6410 | (idisk->action != DISK_REMOVE)) | |
6411 | return 0; | |
6412 | } | |
6413 | } | |
6414 | } | |
6415 | return 1; | |
6416 | } | |
6417 | ||
88758e9d DW |
6418 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
6419 | struct metadata_update **updates) | |
6420 | { | |
6421 | /** | |
d23fe947 DW |
6422 | * Find a device with unused free space and use it to replace a |
6423 | * failed/vacant region in an array. We replace failed regions one a | |
6424 | * array at a time. The result is that a new spare disk will be added | |
6425 | * to the first failed array and after the monitor has finished | |
6426 | * propagating failures the remainder will be consumed. | |
88758e9d | 6427 | * |
d23fe947 DW |
6428 | * FIXME add a capability for mdmon to request spares from another |
6429 | * container. | |
88758e9d DW |
6430 | */ |
6431 | ||
6432 | struct intel_super *super = a->container->sb; | |
88758e9d | 6433 | int inst = a->info.container_member; |
949c47a0 | 6434 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 6435 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
6436 | int failed = a->info.array.raid_disks; |
6437 | struct mdinfo *rv = NULL; | |
6438 | struct mdinfo *d; | |
6439 | struct mdinfo *di; | |
6440 | struct metadata_update *mu; | |
6441 | struct dl *dl; | |
6442 | struct imsm_update_activate_spare *u; | |
6443 | int num_spares = 0; | |
6444 | int i; | |
95d07a2c | 6445 | int allowed; |
88758e9d DW |
6446 | |
6447 | for (d = a->info.devs ; d ; d = d->next) { | |
6448 | if ((d->curr_state & DS_FAULTY) && | |
6449 | d->state_fd >= 0) | |
6450 | /* wait for Removal to happen */ | |
6451 | return NULL; | |
6452 | if (d->state_fd >= 0) | |
6453 | failed--; | |
6454 | } | |
6455 | ||
6456 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
6457 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 AK |
6458 | |
6459 | if (dev->vol.migr_state && | |
6460 | dev->vol.migr_type == MIGR_GEN_MIGR) | |
6461 | /* No repair during migration */ | |
6462 | return NULL; | |
6463 | ||
89c67882 AK |
6464 | if (a->info.array.level == 4) |
6465 | /* No repair for takeovered array | |
6466 | * imsm doesn't support raid4 | |
6467 | */ | |
6468 | return NULL; | |
6469 | ||
fb49eef2 | 6470 | if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
6471 | return NULL; |
6472 | ||
95d07a2c LM |
6473 | /* |
6474 | * If there are any failed disks check state of the other volume. | |
6475 | * Block rebuild if the another one is failed until failed disks | |
6476 | * are removed from container. | |
6477 | */ | |
6478 | if (failed) { | |
6479 | dprintf("found failed disks in %s, check if there another" | |
6480 | "failed sub-array.\n", | |
6481 | dev->volume); | |
6482 | /* check if states of the other volumes allow for rebuild */ | |
6483 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
6484 | if (i != inst) { | |
6485 | allowed = imsm_rebuild_allowed(a->container, | |
6486 | i, failed); | |
6487 | if (!allowed) | |
6488 | return NULL; | |
6489 | } | |
6490 | } | |
6491 | } | |
6492 | ||
88758e9d | 6493 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
6494 | for (i = 0; i < a->info.array.raid_disks; i++) { |
6495 | for (d = a->info.devs ; d ; d = d->next) | |
6496 | if (d->disk.raid_disk == i) | |
6497 | break; | |
6498 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
6499 | if (d && (d->state_fd >= 0)) | |
6500 | continue; | |
6501 | ||
272906ef | 6502 | /* |
a20d2ba5 DW |
6503 | * OK, this device needs recovery. Try to re-add the |
6504 | * previous occupant of this slot, if this fails see if | |
6505 | * we can continue the assimilation of a spare that was | |
6506 | * partially assimilated, finally try to activate a new | |
6507 | * spare. | |
272906ef DW |
6508 | */ |
6509 | dl = imsm_readd(super, i, a); | |
6510 | if (!dl) | |
8ba77d32 | 6511 | dl = imsm_add_spare(super, i, a, 0, NULL); |
a20d2ba5 | 6512 | if (!dl) |
8ba77d32 | 6513 | dl = imsm_add_spare(super, i, a, 1, NULL); |
272906ef DW |
6514 | if (!dl) |
6515 | continue; | |
6516 | ||
6517 | /* found a usable disk with enough space */ | |
6518 | di = malloc(sizeof(*di)); | |
79244939 DW |
6519 | if (!di) |
6520 | continue; | |
272906ef DW |
6521 | memset(di, 0, sizeof(*di)); |
6522 | ||
6523 | /* dl->index will be -1 in the case we are activating a | |
6524 | * pristine spare. imsm_process_update() will create a | |
6525 | * new index in this case. Once a disk is found to be | |
6526 | * failed in all member arrays it is kicked from the | |
6527 | * metadata | |
6528 | */ | |
6529 | di->disk.number = dl->index; | |
d23fe947 | 6530 | |
272906ef DW |
6531 | /* (ab)use di->devs to store a pointer to the device |
6532 | * we chose | |
6533 | */ | |
6534 | di->devs = (struct mdinfo *) dl; | |
6535 | ||
6536 | di->disk.raid_disk = i; | |
6537 | di->disk.major = dl->major; | |
6538 | di->disk.minor = dl->minor; | |
6539 | di->disk.state = 0; | |
d23534e4 | 6540 | di->recovery_start = 0; |
272906ef DW |
6541 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); |
6542 | di->component_size = a->info.component_size; | |
6543 | di->container_member = inst; | |
148acb7b | 6544 | super->random = random32(); |
272906ef DW |
6545 | di->next = rv; |
6546 | rv = di; | |
6547 | num_spares++; | |
6548 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
6549 | i, di->data_offset); | |
88758e9d | 6550 | |
272906ef | 6551 | break; |
88758e9d DW |
6552 | } |
6553 | ||
6554 | if (!rv) | |
6555 | /* No spares found */ | |
6556 | return rv; | |
6557 | /* Now 'rv' has a list of devices to return. | |
6558 | * Create a metadata_update record to update the | |
6559 | * disk_ord_tbl for the array | |
6560 | */ | |
6561 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
6562 | if (mu) { |
6563 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
6564 | if (mu->buf == NULL) { | |
6565 | free(mu); | |
6566 | mu = NULL; | |
6567 | } | |
6568 | } | |
6569 | if (!mu) { | |
6570 | while (rv) { | |
6571 | struct mdinfo *n = rv->next; | |
6572 | ||
6573 | free(rv); | |
6574 | rv = n; | |
6575 | } | |
6576 | return NULL; | |
6577 | } | |
6578 | ||
88758e9d | 6579 | mu->space = NULL; |
cb23f1f4 | 6580 | mu->space_list = NULL; |
88758e9d DW |
6581 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
6582 | mu->next = *updates; | |
6583 | u = (struct imsm_update_activate_spare *) mu->buf; | |
6584 | ||
6585 | for (di = rv ; di ; di = di->next) { | |
6586 | u->type = update_activate_spare; | |
d23fe947 DW |
6587 | u->dl = (struct dl *) di->devs; |
6588 | di->devs = NULL; | |
88758e9d DW |
6589 | u->slot = di->disk.raid_disk; |
6590 | u->array = inst; | |
6591 | u->next = u + 1; | |
6592 | u++; | |
6593 | } | |
6594 | (u-1)->next = NULL; | |
6595 | *updates = mu; | |
6596 | ||
6597 | return rv; | |
6598 | } | |
6599 | ||
54c2c1ea | 6600 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 6601 | { |
54c2c1ea DW |
6602 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
6603 | struct imsm_map *map = get_imsm_map(dev, 0); | |
6604 | struct imsm_map *new_map = get_imsm_map(&u->dev, 0); | |
6605 | struct disk_info *inf = get_disk_info(u); | |
6606 | struct imsm_disk *disk; | |
8273f55e DW |
6607 | int i; |
6608 | int j; | |
8273f55e | 6609 | |
54c2c1ea | 6610 | for (i = 0; i < map->num_members; i++) { |
98130f40 | 6611 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, -1)); |
54c2c1ea DW |
6612 | for (j = 0; j < new_map->num_members; j++) |
6613 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
6614 | return 1; |
6615 | } | |
6616 | ||
6617 | return 0; | |
6618 | } | |
6619 | ||
1a64be56 LM |
6620 | |
6621 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
6622 | { | |
6623 | struct dl *dl = NULL; | |
6624 | for (dl = super->disks; dl; dl = dl->next) | |
6625 | if ((dl->major == major) && (dl->minor == minor)) | |
6626 | return dl; | |
6627 | return NULL; | |
6628 | } | |
6629 | ||
6630 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
6631 | { | |
6632 | struct dl *prev = NULL; | |
6633 | struct dl *dl; | |
6634 | ||
6635 | prev = NULL; | |
6636 | for (dl = super->disks; dl; dl = dl->next) { | |
6637 | if ((dl->major == major) && (dl->minor == minor)) { | |
6638 | /* remove */ | |
6639 | if (prev) | |
6640 | prev->next = dl->next; | |
6641 | else | |
6642 | super->disks = dl->next; | |
6643 | dl->next = NULL; | |
6644 | __free_imsm_disk(dl); | |
6645 | dprintf("%s: removed %x:%x\n", | |
6646 | __func__, major, minor); | |
6647 | break; | |
6648 | } | |
6649 | prev = dl; | |
6650 | } | |
6651 | return 0; | |
6652 | } | |
6653 | ||
f21e18ca | 6654 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 6655 | |
1a64be56 LM |
6656 | static int add_remove_disk_update(struct intel_super *super) |
6657 | { | |
6658 | int check_degraded = 0; | |
6659 | struct dl *disk = NULL; | |
6660 | /* add/remove some spares to/from the metadata/contrainer */ | |
6661 | while (super->disk_mgmt_list) { | |
6662 | struct dl *disk_cfg; | |
6663 | ||
6664 | disk_cfg = super->disk_mgmt_list; | |
6665 | super->disk_mgmt_list = disk_cfg->next; | |
6666 | disk_cfg->next = NULL; | |
6667 | ||
6668 | if (disk_cfg->action == DISK_ADD) { | |
6669 | disk_cfg->next = super->disks; | |
6670 | super->disks = disk_cfg; | |
6671 | check_degraded = 1; | |
6672 | dprintf("%s: added %x:%x\n", | |
6673 | __func__, disk_cfg->major, | |
6674 | disk_cfg->minor); | |
6675 | } else if (disk_cfg->action == DISK_REMOVE) { | |
6676 | dprintf("Disk remove action processed: %x.%x\n", | |
6677 | disk_cfg->major, disk_cfg->minor); | |
6678 | disk = get_disk_super(super, | |
6679 | disk_cfg->major, | |
6680 | disk_cfg->minor); | |
6681 | if (disk) { | |
6682 | /* store action status */ | |
6683 | disk->action = DISK_REMOVE; | |
6684 | /* remove spare disks only */ | |
6685 | if (disk->index == -1) { | |
6686 | remove_disk_super(super, | |
6687 | disk_cfg->major, | |
6688 | disk_cfg->minor); | |
6689 | } | |
6690 | } | |
6691 | /* release allocate disk structure */ | |
6692 | __free_imsm_disk(disk_cfg); | |
6693 | } | |
6694 | } | |
6695 | return check_degraded; | |
6696 | } | |
6697 | ||
a29911da PC |
6698 | |
6699 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
6700 | struct intel_super *super, | |
6701 | void ***space_list) | |
6702 | { | |
6703 | struct intel_dev *id; | |
6704 | void **tofree = NULL; | |
6705 | int ret_val = 0; | |
6706 | ||
6707 | dprintf("apply_reshape_migration_update()\n"); | |
6708 | if ((u->subdev < 0) || | |
6709 | (u->subdev > 1)) { | |
6710 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
6711 | return ret_val; | |
6712 | } | |
6713 | if ((space_list == NULL) || (*space_list == NULL)) { | |
6714 | dprintf("imsm: Error: Memory is not allocated\n"); | |
6715 | return ret_val; | |
6716 | } | |
6717 | ||
6718 | for (id = super->devlist ; id; id = id->next) { | |
6719 | if (id->index == (unsigned)u->subdev) { | |
6720 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
6721 | struct imsm_map *map; | |
6722 | struct imsm_dev *new_dev = | |
6723 | (struct imsm_dev *)*space_list; | |
6724 | struct imsm_map *migr_map = get_imsm_map(dev, 1); | |
6725 | int to_state; | |
6726 | struct dl *new_disk; | |
6727 | ||
6728 | if (new_dev == NULL) | |
6729 | return ret_val; | |
6730 | *space_list = **space_list; | |
6731 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
6732 | map = get_imsm_map(new_dev, 0); | |
6733 | if (migr_map) { | |
6734 | dprintf("imsm: Error: migration in progress"); | |
6735 | return ret_val; | |
6736 | } | |
6737 | ||
6738 | to_state = map->map_state; | |
6739 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
6740 | map->num_members++; | |
6741 | /* this should not happen */ | |
6742 | if (u->new_disks[0] < 0) { | |
6743 | map->failed_disk_num = | |
6744 | map->num_members - 1; | |
6745 | to_state = IMSM_T_STATE_DEGRADED; | |
6746 | } else | |
6747 | to_state = IMSM_T_STATE_NORMAL; | |
6748 | } | |
8e59f3d8 | 6749 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
6750 | if (u->new_level > -1) |
6751 | map->raid_level = u->new_level; | |
6752 | migr_map = get_imsm_map(new_dev, 1); | |
6753 | if ((u->new_level == 5) && | |
6754 | (migr_map->raid_level == 0)) { | |
6755 | int ord = map->num_members - 1; | |
6756 | migr_map->num_members--; | |
6757 | if (u->new_disks[0] < 0) | |
6758 | ord |= IMSM_ORD_REBUILD; | |
6759 | set_imsm_ord_tbl_ent(map, | |
6760 | map->num_members - 1, | |
6761 | ord); | |
6762 | } | |
6763 | id->dev = new_dev; | |
6764 | tofree = (void **)dev; | |
6765 | ||
4bba0439 PC |
6766 | /* update chunk size |
6767 | */ | |
6768 | if (u->new_chunksize > 0) | |
6769 | map->blocks_per_strip = | |
6770 | __cpu_to_le16(u->new_chunksize * 2); | |
6771 | ||
a29911da PC |
6772 | /* add disk |
6773 | */ | |
6774 | if ((u->new_level != 5) || | |
6775 | (migr_map->raid_level != 0) || | |
6776 | (migr_map->raid_level == map->raid_level)) | |
6777 | goto skip_disk_add; | |
6778 | ||
6779 | if (u->new_disks[0] >= 0) { | |
6780 | /* use passes spare | |
6781 | */ | |
6782 | new_disk = get_disk_super(super, | |
6783 | major(u->new_disks[0]), | |
6784 | minor(u->new_disks[0])); | |
6785 | dprintf("imsm: new disk for reshape is: %i:%i " | |
6786 | "(%p, index = %i)\n", | |
6787 | major(u->new_disks[0]), | |
6788 | minor(u->new_disks[0]), | |
6789 | new_disk, new_disk->index); | |
6790 | if (new_disk == NULL) | |
6791 | goto error_disk_add; | |
6792 | ||
6793 | new_disk->index = map->num_members - 1; | |
6794 | /* slot to fill in autolayout | |
6795 | */ | |
6796 | new_disk->raiddisk = new_disk->index; | |
6797 | new_disk->disk.status |= CONFIGURED_DISK; | |
6798 | new_disk->disk.status &= ~SPARE_DISK; | |
6799 | } else | |
6800 | goto error_disk_add; | |
6801 | ||
6802 | skip_disk_add: | |
6803 | *tofree = *space_list; | |
6804 | /* calculate new size | |
6805 | */ | |
6806 | imsm_set_array_size(new_dev); | |
6807 | ||
6808 | ret_val = 1; | |
6809 | } | |
6810 | } | |
6811 | ||
6812 | if (tofree) | |
6813 | *space_list = tofree; | |
6814 | return ret_val; | |
6815 | ||
6816 | error_disk_add: | |
6817 | dprintf("Error: imsm: Cannot find disk.\n"); | |
6818 | return ret_val; | |
6819 | } | |
6820 | ||
6821 | ||
2e5dc010 N |
6822 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
6823 | struct intel_super *super, | |
6824 | void ***space_list) | |
6825 | { | |
6826 | struct dl *new_disk; | |
6827 | struct intel_dev *id; | |
6828 | int i; | |
6829 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 6830 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
6831 | void **tofree = NULL; |
6832 | int devices_to_reshape = 1; | |
6833 | struct imsm_super *mpb = super->anchor; | |
6834 | int ret_val = 0; | |
d098291a | 6835 | unsigned int dev_id; |
2e5dc010 | 6836 | |
ed7333bd | 6837 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
6838 | |
6839 | /* enable spares to use in array */ | |
6840 | for (i = 0; i < delta_disks; i++) { | |
6841 | new_disk = get_disk_super(super, | |
6842 | major(u->new_disks[i]), | |
6843 | minor(u->new_disks[i])); | |
ed7333bd AK |
6844 | dprintf("imsm: new disk for reshape is: %i:%i " |
6845 | "(%p, index = %i)\n", | |
2e5dc010 N |
6846 | major(u->new_disks[i]), minor(u->new_disks[i]), |
6847 | new_disk, new_disk->index); | |
6848 | if ((new_disk == NULL) || | |
6849 | ((new_disk->index >= 0) && | |
6850 | (new_disk->index < u->old_raid_disks))) | |
6851 | goto update_reshape_exit; | |
ee4beede | 6852 | new_disk->index = disk_count++; |
2e5dc010 N |
6853 | /* slot to fill in autolayout |
6854 | */ | |
6855 | new_disk->raiddisk = new_disk->index; | |
6856 | new_disk->disk.status |= | |
6857 | CONFIGURED_DISK; | |
6858 | new_disk->disk.status &= ~SPARE_DISK; | |
6859 | } | |
6860 | ||
ed7333bd AK |
6861 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
6862 | mpb->num_raid_devs); | |
2e5dc010 N |
6863 | /* manage changes in volume |
6864 | */ | |
d098291a | 6865 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
6866 | void **sp = *space_list; |
6867 | struct imsm_dev *newdev; | |
6868 | struct imsm_map *newmap, *oldmap; | |
6869 | ||
d098291a AK |
6870 | for (id = super->devlist ; id; id = id->next) { |
6871 | if (id->index == dev_id) | |
6872 | break; | |
6873 | } | |
6874 | if (id == NULL) | |
6875 | break; | |
2e5dc010 N |
6876 | if (!sp) |
6877 | continue; | |
6878 | *space_list = *sp; | |
6879 | newdev = (void*)sp; | |
6880 | /* Copy the dev, but not (all of) the map */ | |
6881 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
6882 | oldmap = get_imsm_map(id->dev, 0); | |
6883 | newmap = get_imsm_map(newdev, 0); | |
6884 | /* Copy the current map */ | |
6885 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
6886 | /* update one device only | |
6887 | */ | |
6888 | if (devices_to_reshape) { | |
ed7333bd AK |
6889 | dprintf("imsm: modifying subdev: %i\n", |
6890 | id->index); | |
2e5dc010 N |
6891 | devices_to_reshape--; |
6892 | newdev->vol.migr_state = 1; | |
6893 | newdev->vol.curr_migr_unit = 0; | |
6894 | newdev->vol.migr_type = MIGR_GEN_MIGR; | |
6895 | newmap->num_members = u->new_raid_disks; | |
6896 | for (i = 0; i < delta_disks; i++) { | |
6897 | set_imsm_ord_tbl_ent(newmap, | |
6898 | u->old_raid_disks + i, | |
6899 | u->old_raid_disks + i); | |
6900 | } | |
6901 | /* New map is correct, now need to save old map | |
6902 | */ | |
6903 | newmap = get_imsm_map(newdev, 1); | |
6904 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
6905 | ||
70bdf0dc | 6906 | imsm_set_array_size(newdev); |
2e5dc010 N |
6907 | } |
6908 | ||
6909 | sp = (void **)id->dev; | |
6910 | id->dev = newdev; | |
6911 | *sp = tofree; | |
6912 | tofree = sp; | |
8e59f3d8 AK |
6913 | |
6914 | /* Clear migration record */ | |
6915 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 6916 | } |
819bc634 AK |
6917 | if (tofree) |
6918 | *space_list = tofree; | |
2e5dc010 N |
6919 | ret_val = 1; |
6920 | ||
6921 | update_reshape_exit: | |
6922 | ||
6923 | return ret_val; | |
6924 | } | |
6925 | ||
bb025c2f | 6926 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
6927 | struct intel_super *super, |
6928 | void ***space_list) | |
bb025c2f KW |
6929 | { |
6930 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
6931 | struct intel_dev *dv; |
6932 | struct imsm_dev *dev_new; | |
bb025c2f KW |
6933 | struct imsm_map *map; |
6934 | struct dl *dm, *du; | |
8ca6df95 | 6935 | int i; |
bb025c2f KW |
6936 | |
6937 | for (dv = super->devlist; dv; dv = dv->next) | |
6938 | if (dv->index == (unsigned int)u->subarray) { | |
6939 | dev = dv->dev; | |
6940 | break; | |
6941 | } | |
6942 | ||
6943 | if (dev == NULL) | |
6944 | return 0; | |
6945 | ||
6946 | map = get_imsm_map(dev, 0); | |
6947 | ||
6948 | if (u->direction == R10_TO_R0) { | |
43d5ec18 KW |
6949 | /* Number of failed disks must be half of initial disk number */ |
6950 | if (imsm_count_failed(super, dev) != (map->num_members / 2)) | |
6951 | return 0; | |
6952 | ||
bb025c2f KW |
6953 | /* iterate through devices to mark removed disks as spare */ |
6954 | for (dm = super->disks; dm; dm = dm->next) { | |
6955 | if (dm->disk.status & FAILED_DISK) { | |
6956 | int idx = dm->index; | |
6957 | /* update indexes on the disk list */ | |
6958 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
6959 | the index values will end up being correct.... NB */ | |
6960 | for (du = super->disks; du; du = du->next) | |
6961 | if (du->index > idx) | |
6962 | du->index--; | |
6963 | /* mark as spare disk */ | |
6964 | dm->disk.status = SPARE_DISK; | |
6965 | dm->index = -1; | |
6966 | } | |
6967 | } | |
bb025c2f KW |
6968 | /* update map */ |
6969 | map->num_members = map->num_members / 2; | |
6970 | map->map_state = IMSM_T_STATE_NORMAL; | |
6971 | map->num_domains = 1; | |
6972 | map->raid_level = 0; | |
6973 | map->failed_disk_num = -1; | |
6974 | } | |
6975 | ||
8ca6df95 KW |
6976 | if (u->direction == R0_TO_R10) { |
6977 | void **space; | |
6978 | /* update slots in current disk list */ | |
6979 | for (dm = super->disks; dm; dm = dm->next) { | |
6980 | if (dm->index >= 0) | |
6981 | dm->index *= 2; | |
6982 | } | |
6983 | /* create new *missing* disks */ | |
6984 | for (i = 0; i < map->num_members; i++) { | |
6985 | space = *space_list; | |
6986 | if (!space) | |
6987 | continue; | |
6988 | *space_list = *space; | |
6989 | du = (void *)space; | |
6990 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
6991 | du->fd = -1; |
6992 | du->minor = 0; | |
6993 | du->major = 0; | |
6994 | du->index = (i * 2) + 1; | |
6995 | sprintf((char *)du->disk.serial, | |
6996 | " MISSING_%d", du->index); | |
6997 | sprintf((char *)du->serial, | |
6998 | "MISSING_%d", du->index); | |
6999 | du->next = super->missing; | |
7000 | super->missing = du; | |
7001 | } | |
7002 | /* create new dev and map */ | |
7003 | space = *space_list; | |
7004 | if (!space) | |
7005 | return 0; | |
7006 | *space_list = *space; | |
7007 | dev_new = (void *)space; | |
7008 | memcpy(dev_new, dev, sizeof(*dev)); | |
7009 | /* update new map */ | |
7010 | map = get_imsm_map(dev_new, 0); | |
8ca6df95 | 7011 | map->num_members = map->num_members * 2; |
1a2487c2 | 7012 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
7013 | map->num_domains = 2; |
7014 | map->raid_level = 1; | |
7015 | /* replace dev<->dev_new */ | |
7016 | dv->dev = dev_new; | |
7017 | } | |
bb025c2f KW |
7018 | /* update disk order table */ |
7019 | for (du = super->disks; du; du = du->next) | |
7020 | if (du->index >= 0) | |
7021 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 7022 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
7023 | if (du->index >= 0) { |
7024 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 7025 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 7026 | } |
bb025c2f KW |
7027 | |
7028 | return 1; | |
7029 | } | |
7030 | ||
e8319a19 DW |
7031 | static void imsm_process_update(struct supertype *st, |
7032 | struct metadata_update *update) | |
7033 | { | |
7034 | /** | |
7035 | * crack open the metadata_update envelope to find the update record | |
7036 | * update can be one of: | |
d195167d AK |
7037 | * update_reshape_container_disks - all the arrays in the container |
7038 | * are being reshaped to have more devices. We need to mark | |
7039 | * the arrays for general migration and convert selected spares | |
7040 | * into active devices. | |
7041 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
7042 | * device in an array, update the disk_ord_tbl. If this disk is |
7043 | * present in all member arrays then also clear the SPARE_DISK | |
7044 | * flag | |
d195167d AK |
7045 | * update_create_array |
7046 | * update_kill_array | |
7047 | * update_rename_array | |
7048 | * update_add_remove_disk | |
e8319a19 DW |
7049 | */ |
7050 | struct intel_super *super = st->sb; | |
4d7b1503 | 7051 | struct imsm_super *mpb; |
e8319a19 DW |
7052 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
7053 | ||
4d7b1503 DW |
7054 | /* update requires a larger buf but the allocation failed */ |
7055 | if (super->next_len && !super->next_buf) { | |
7056 | super->next_len = 0; | |
7057 | return; | |
7058 | } | |
7059 | ||
7060 | if (super->next_buf) { | |
7061 | memcpy(super->next_buf, super->buf, super->len); | |
7062 | free(super->buf); | |
7063 | super->len = super->next_len; | |
7064 | super->buf = super->next_buf; | |
7065 | ||
7066 | super->next_len = 0; | |
7067 | super->next_buf = NULL; | |
7068 | } | |
7069 | ||
7070 | mpb = super->anchor; | |
7071 | ||
e8319a19 | 7072 | switch (type) { |
0ec5d470 AK |
7073 | case update_general_migration_checkpoint: { |
7074 | struct intel_dev *id; | |
7075 | struct imsm_update_general_migration_checkpoint *u = | |
7076 | (void *)update->buf; | |
7077 | ||
7078 | dprintf("imsm: process_update() " | |
7079 | "for update_general_migration_checkpoint called\n"); | |
7080 | ||
7081 | /* find device under general migration */ | |
7082 | for (id = super->devlist ; id; id = id->next) { | |
7083 | if (is_gen_migration(id->dev)) { | |
7084 | id->dev->vol.curr_migr_unit = | |
7085 | __cpu_to_le32(u->curr_migr_unit); | |
7086 | super->updates_pending++; | |
7087 | } | |
7088 | } | |
7089 | break; | |
7090 | } | |
bb025c2f KW |
7091 | case update_takeover: { |
7092 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
7093 | if (apply_takeover_update(u, super, &update->space_list)) { |
7094 | imsm_update_version_info(super); | |
bb025c2f | 7095 | super->updates_pending++; |
1a2487c2 | 7096 | } |
bb025c2f KW |
7097 | break; |
7098 | } | |
7099 | ||
78b10e66 | 7100 | case update_reshape_container_disks: { |
d195167d | 7101 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
7102 | if (apply_reshape_container_disks_update( |
7103 | u, super, &update->space_list)) | |
7104 | super->updates_pending++; | |
78b10e66 N |
7105 | break; |
7106 | } | |
48c5303a | 7107 | case update_reshape_migration: { |
a29911da PC |
7108 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
7109 | if (apply_reshape_migration_update( | |
7110 | u, super, &update->space_list)) | |
7111 | super->updates_pending++; | |
48c5303a PC |
7112 | break; |
7113 | } | |
e8319a19 DW |
7114 | case update_activate_spare: { |
7115 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 7116 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 7117 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd | 7118 | struct imsm_map *migr_map; |
e8319a19 DW |
7119 | struct active_array *a; |
7120 | struct imsm_disk *disk; | |
0c046afd | 7121 | __u8 to_state; |
e8319a19 | 7122 | struct dl *dl; |
e8319a19 | 7123 | unsigned int found; |
0c046afd | 7124 | int failed; |
98130f40 | 7125 | int victim = get_imsm_disk_idx(dev, u->slot, -1); |
e8319a19 DW |
7126 | int i; |
7127 | ||
7128 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 7129 | if (dl == u->dl) |
e8319a19 DW |
7130 | break; |
7131 | ||
7132 | if (!dl) { | |
7133 | fprintf(stderr, "error: imsm_activate_spare passed " | |
1f24f035 DW |
7134 | "an unknown disk (index: %d)\n", |
7135 | u->dl->index); | |
e8319a19 DW |
7136 | return; |
7137 | } | |
7138 | ||
7139 | super->updates_pending++; | |
0c046afd DW |
7140 | /* count failures (excluding rebuilds and the victim) |
7141 | * to determine map[0] state | |
7142 | */ | |
7143 | failed = 0; | |
7144 | for (i = 0; i < map->num_members; i++) { | |
7145 | if (i == u->slot) | |
7146 | continue; | |
98130f40 AK |
7147 | disk = get_imsm_disk(super, |
7148 | get_imsm_disk_idx(dev, i, -1)); | |
25ed7e59 | 7149 | if (!disk || is_failed(disk)) |
0c046afd DW |
7150 | failed++; |
7151 | } | |
7152 | ||
d23fe947 DW |
7153 | /* adding a pristine spare, assign a new index */ |
7154 | if (dl->index < 0) { | |
7155 | dl->index = super->anchor->num_disks; | |
7156 | super->anchor->num_disks++; | |
7157 | } | |
d23fe947 | 7158 | disk = &dl->disk; |
f2f27e63 DW |
7159 | disk->status |= CONFIGURED_DISK; |
7160 | disk->status &= ~SPARE_DISK; | |
e8319a19 | 7161 | |
0c046afd DW |
7162 | /* mark rebuild */ |
7163 | to_state = imsm_check_degraded(super, dev, failed); | |
7164 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8e59f3d8 | 7165 | migrate(dev, super, to_state, MIGR_REBUILD); |
0c046afd DW |
7166 | migr_map = get_imsm_map(dev, 1); |
7167 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); | |
7168 | set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD); | |
7169 | ||
148acb7b DW |
7170 | /* update the family_num to mark a new container |
7171 | * generation, being careful to record the existing | |
7172 | * family_num in orig_family_num to clean up after | |
7173 | * earlier mdadm versions that neglected to set it. | |
7174 | */ | |
7175 | if (mpb->orig_family_num == 0) | |
7176 | mpb->orig_family_num = mpb->family_num; | |
7177 | mpb->family_num += super->random; | |
7178 | ||
e8319a19 DW |
7179 | /* count arrays using the victim in the metadata */ |
7180 | found = 0; | |
7181 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 7182 | dev = get_imsm_dev(super, a->info.container_member); |
620b1713 DW |
7183 | map = get_imsm_map(dev, 0); |
7184 | ||
7185 | if (get_imsm_disk_slot(map, victim) >= 0) | |
7186 | found++; | |
e8319a19 DW |
7187 | } |
7188 | ||
24565c9a | 7189 | /* delete the victim if it is no longer being |
e8319a19 DW |
7190 | * utilized anywhere |
7191 | */ | |
e8319a19 | 7192 | if (!found) { |
ae6aad82 | 7193 | struct dl **dlp; |
24565c9a | 7194 | |
47ee5a45 DW |
7195 | /* We know that 'manager' isn't touching anything, |
7196 | * so it is safe to delete | |
7197 | */ | |
24565c9a | 7198 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) |
ae6aad82 DW |
7199 | if ((*dlp)->index == victim) |
7200 | break; | |
47ee5a45 DW |
7201 | |
7202 | /* victim may be on the missing list */ | |
7203 | if (!*dlp) | |
7204 | for (dlp = &super->missing; *dlp; dlp = &(*dlp)->next) | |
7205 | if ((*dlp)->index == victim) | |
7206 | break; | |
24565c9a | 7207 | imsm_delete(super, dlp, victim); |
e8319a19 | 7208 | } |
8273f55e DW |
7209 | break; |
7210 | } | |
7211 | case update_create_array: { | |
7212 | /* someone wants to create a new array, we need to be aware of | |
7213 | * a few races/collisions: | |
7214 | * 1/ 'Create' called by two separate instances of mdadm | |
7215 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
7216 | * devices that have since been assimilated via | |
7217 | * activate_spare. | |
7218 | * In the event this update can not be carried out mdadm will | |
7219 | * (FIX ME) notice that its update did not take hold. | |
7220 | */ | |
7221 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 7222 | struct intel_dev *dv; |
8273f55e DW |
7223 | struct imsm_dev *dev; |
7224 | struct imsm_map *map, *new_map; | |
7225 | unsigned long long start, end; | |
7226 | unsigned long long new_start, new_end; | |
7227 | int i; | |
54c2c1ea DW |
7228 | struct disk_info *inf; |
7229 | struct dl *dl; | |
8273f55e DW |
7230 | |
7231 | /* handle racing creates: first come first serve */ | |
7232 | if (u->dev_idx < mpb->num_raid_devs) { | |
7233 | dprintf("%s: subarray %d already defined\n", | |
7234 | __func__, u->dev_idx); | |
ba2de7ba | 7235 | goto create_error; |
8273f55e DW |
7236 | } |
7237 | ||
7238 | /* check update is next in sequence */ | |
7239 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
7240 | dprintf("%s: can not create array %d expected index %d\n", |
7241 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 7242 | goto create_error; |
8273f55e DW |
7243 | } |
7244 | ||
a965f303 | 7245 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
7246 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
7247 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
54c2c1ea | 7248 | inf = get_disk_info(u); |
8273f55e DW |
7249 | |
7250 | /* handle activate_spare versus create race: | |
7251 | * check to make sure that overlapping arrays do not include | |
7252 | * overalpping disks | |
7253 | */ | |
7254 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 7255 | dev = get_imsm_dev(super, i); |
a965f303 | 7256 | map = get_imsm_map(dev, 0); |
8273f55e DW |
7257 | start = __le32_to_cpu(map->pba_of_lba0); |
7258 | end = start + __le32_to_cpu(map->blocks_per_member); | |
7259 | if ((new_start >= start && new_start <= end) || | |
7260 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
7261 | /* overlap */; |
7262 | else | |
7263 | continue; | |
7264 | ||
7265 | if (disks_overlap(super, i, u)) { | |
8273f55e | 7266 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 7267 | goto create_error; |
8273f55e DW |
7268 | } |
7269 | } | |
8273f55e | 7270 | |
949c47a0 DW |
7271 | /* check that prepare update was successful */ |
7272 | if (!update->space) { | |
7273 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 7274 | goto create_error; |
949c47a0 DW |
7275 | } |
7276 | ||
54c2c1ea DW |
7277 | /* check that all disks are still active before committing |
7278 | * changes. FIXME: could we instead handle this by creating a | |
7279 | * degraded array? That's probably not what the user expects, | |
7280 | * so better to drop this update on the floor. | |
7281 | */ | |
7282 | for (i = 0; i < new_map->num_members; i++) { | |
7283 | dl = serial_to_dl(inf[i].serial, super); | |
7284 | if (!dl) { | |
7285 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 7286 | goto create_error; |
54c2c1ea | 7287 | } |
949c47a0 DW |
7288 | } |
7289 | ||
8273f55e | 7290 | super->updates_pending++; |
54c2c1ea DW |
7291 | |
7292 | /* convert spares to members and fixup ord_tbl */ | |
7293 | for (i = 0; i < new_map->num_members; i++) { | |
7294 | dl = serial_to_dl(inf[i].serial, super); | |
7295 | if (dl->index == -1) { | |
7296 | dl->index = mpb->num_disks; | |
7297 | mpb->num_disks++; | |
7298 | dl->disk.status |= CONFIGURED_DISK; | |
7299 | dl->disk.status &= ~SPARE_DISK; | |
7300 | } | |
7301 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
7302 | } | |
7303 | ||
ba2de7ba DW |
7304 | dv = update->space; |
7305 | dev = dv->dev; | |
949c47a0 DW |
7306 | update->space = NULL; |
7307 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
7308 | dv->index = u->dev_idx; |
7309 | dv->next = super->devlist; | |
7310 | super->devlist = dv; | |
8273f55e | 7311 | mpb->num_raid_devs++; |
8273f55e | 7312 | |
4d1313e9 | 7313 | imsm_update_version_info(super); |
8273f55e | 7314 | break; |
ba2de7ba DW |
7315 | create_error: |
7316 | /* mdmon knows how to release update->space, but not | |
7317 | * ((struct intel_dev *) update->space)->dev | |
7318 | */ | |
7319 | if (update->space) { | |
7320 | dv = update->space; | |
7321 | free(dv->dev); | |
7322 | } | |
8273f55e | 7323 | break; |
e8319a19 | 7324 | } |
33414a01 DW |
7325 | case update_kill_array: { |
7326 | struct imsm_update_kill_array *u = (void *) update->buf; | |
7327 | int victim = u->dev_idx; | |
7328 | struct active_array *a; | |
7329 | struct intel_dev **dp; | |
7330 | struct imsm_dev *dev; | |
7331 | ||
7332 | /* sanity check that we are not affecting the uuid of | |
7333 | * active arrays, or deleting an active array | |
7334 | * | |
7335 | * FIXME when immutable ids are available, but note that | |
7336 | * we'll also need to fixup the invalidated/active | |
7337 | * subarray indexes in mdstat | |
7338 | */ | |
7339 | for (a = st->arrays; a; a = a->next) | |
7340 | if (a->info.container_member >= victim) | |
7341 | break; | |
7342 | /* by definition if mdmon is running at least one array | |
7343 | * is active in the container, so checking | |
7344 | * mpb->num_raid_devs is just extra paranoia | |
7345 | */ | |
7346 | dev = get_imsm_dev(super, victim); | |
7347 | if (a || !dev || mpb->num_raid_devs == 1) { | |
7348 | dprintf("failed to delete subarray-%d\n", victim); | |
7349 | break; | |
7350 | } | |
7351 | ||
7352 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 7353 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
7354 | *dp = (*dp)->next; |
7355 | } else { | |
f21e18ca | 7356 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
7357 | (*dp)->index--; |
7358 | dp = &(*dp)->next; | |
7359 | } | |
7360 | mpb->num_raid_devs--; | |
7361 | super->updates_pending++; | |
7362 | break; | |
7363 | } | |
aa534678 DW |
7364 | case update_rename_array: { |
7365 | struct imsm_update_rename_array *u = (void *) update->buf; | |
7366 | char name[MAX_RAID_SERIAL_LEN+1]; | |
7367 | int target = u->dev_idx; | |
7368 | struct active_array *a; | |
7369 | struct imsm_dev *dev; | |
7370 | ||
7371 | /* sanity check that we are not affecting the uuid of | |
7372 | * an active array | |
7373 | */ | |
7374 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
7375 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
7376 | for (a = st->arrays; a; a = a->next) | |
7377 | if (a->info.container_member == target) | |
7378 | break; | |
7379 | dev = get_imsm_dev(super, u->dev_idx); | |
7380 | if (a || !dev || !check_name(super, name, 1)) { | |
7381 | dprintf("failed to rename subarray-%d\n", target); | |
7382 | break; | |
7383 | } | |
7384 | ||
cdbe98cd | 7385 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
7386 | super->updates_pending++; |
7387 | break; | |
7388 | } | |
1a64be56 | 7389 | case update_add_remove_disk: { |
43dad3d6 | 7390 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
7391 | * being added, check teh status of add_remove_disk |
7392 | * if discs has been added. | |
7393 | */ | |
7394 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 7395 | struct active_array *a; |
072b727f DW |
7396 | |
7397 | super->updates_pending++; | |
1a64be56 | 7398 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
7399 | a->check_degraded = 1; |
7400 | } | |
43dad3d6 | 7401 | break; |
e8319a19 | 7402 | } |
1a64be56 LM |
7403 | default: |
7404 | fprintf(stderr, "error: unsuported process update type:" | |
7405 | "(type: %d)\n", type); | |
7406 | } | |
e8319a19 | 7407 | } |
88758e9d | 7408 | |
bc0b9d34 PC |
7409 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
7410 | ||
8273f55e DW |
7411 | static void imsm_prepare_update(struct supertype *st, |
7412 | struct metadata_update *update) | |
7413 | { | |
949c47a0 | 7414 | /** |
4d7b1503 DW |
7415 | * Allocate space to hold new disk entries, raid-device entries or a new |
7416 | * mpb if necessary. The manager synchronously waits for updates to | |
7417 | * complete in the monitor, so new mpb buffers allocated here can be | |
7418 | * integrated by the monitor thread without worrying about live pointers | |
7419 | * in the manager thread. | |
8273f55e | 7420 | */ |
949c47a0 | 7421 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
7422 | struct intel_super *super = st->sb; |
7423 | struct imsm_super *mpb = super->anchor; | |
7424 | size_t buf_len; | |
7425 | size_t len = 0; | |
949c47a0 DW |
7426 | |
7427 | switch (type) { | |
0ec5d470 AK |
7428 | case update_general_migration_checkpoint: |
7429 | dprintf("imsm: prepare_update() " | |
7430 | "for update_general_migration_checkpoint called\n"); | |
7431 | break; | |
abedf5fc KW |
7432 | case update_takeover: { |
7433 | struct imsm_update_takeover *u = (void *)update->buf; | |
7434 | if (u->direction == R0_TO_R10) { | |
7435 | void **tail = (void **)&update->space_list; | |
7436 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
7437 | struct imsm_map *map = get_imsm_map(dev, 0); | |
7438 | int num_members = map->num_members; | |
7439 | void *space; | |
7440 | int size, i; | |
7441 | int err = 0; | |
7442 | /* allocate memory for added disks */ | |
7443 | for (i = 0; i < num_members; i++) { | |
7444 | size = sizeof(struct dl); | |
7445 | space = malloc(size); | |
7446 | if (!space) { | |
7447 | err++; | |
7448 | break; | |
7449 | } | |
7450 | *tail = space; | |
7451 | tail = space; | |
7452 | *tail = NULL; | |
7453 | } | |
7454 | /* allocate memory for new device */ | |
7455 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
7456 | (num_members * sizeof(__u32)); | |
7457 | space = malloc(size); | |
7458 | if (!space) | |
7459 | err++; | |
7460 | else { | |
7461 | *tail = space; | |
7462 | tail = space; | |
7463 | *tail = NULL; | |
7464 | } | |
7465 | if (!err) { | |
7466 | len = disks_to_mpb_size(num_members * 2); | |
7467 | } else { | |
7468 | /* if allocation didn't success, free buffer */ | |
7469 | while (update->space_list) { | |
7470 | void **sp = update->space_list; | |
7471 | update->space_list = *sp; | |
7472 | free(sp); | |
7473 | } | |
7474 | } | |
7475 | } | |
7476 | ||
7477 | break; | |
7478 | } | |
78b10e66 | 7479 | case update_reshape_container_disks: { |
d195167d AK |
7480 | /* Every raid device in the container is about to |
7481 | * gain some more devices, and we will enter a | |
7482 | * reconfiguration. | |
7483 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
7484 | * will now hold 2 of them. | |
7485 | * Thus we need new 'struct imsm_dev' allocations sized | |
7486 | * as sizeof_imsm_dev but with more devices in both maps. | |
7487 | */ | |
7488 | struct imsm_update_reshape *u = (void *)update->buf; | |
7489 | struct intel_dev *dl; | |
7490 | void **space_tail = (void**)&update->space_list; | |
7491 | ||
7492 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7493 | ||
7494 | for (dl = super->devlist; dl; dl = dl->next) { | |
7495 | int size = sizeof_imsm_dev(dl->dev, 1); | |
7496 | void *s; | |
d677e0b8 AK |
7497 | if (u->new_raid_disks > u->old_raid_disks) |
7498 | size += sizeof(__u32)*2* | |
7499 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
7500 | s = malloc(size); |
7501 | if (!s) | |
7502 | break; | |
7503 | *space_tail = s; | |
7504 | space_tail = s; | |
7505 | *space_tail = NULL; | |
7506 | } | |
7507 | ||
7508 | len = disks_to_mpb_size(u->new_raid_disks); | |
7509 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
7510 | break; |
7511 | } | |
48c5303a | 7512 | case update_reshape_migration: { |
bc0b9d34 PC |
7513 | /* for migration level 0->5 we need to add disks |
7514 | * so the same as for container operation we will copy | |
7515 | * device to the bigger location. | |
7516 | * in memory prepared device and new disk area are prepared | |
7517 | * for usage in process update | |
7518 | */ | |
7519 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
7520 | struct intel_dev *id; | |
7521 | void **space_tail = (void **)&update->space_list; | |
7522 | int size; | |
7523 | void *s; | |
7524 | int current_level = -1; | |
7525 | ||
7526 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7527 | ||
7528 | /* add space for bigger array in update | |
7529 | */ | |
7530 | for (id = super->devlist; id; id = id->next) { | |
7531 | if (id->index == (unsigned)u->subdev) { | |
7532 | size = sizeof_imsm_dev(id->dev, 1); | |
7533 | if (u->new_raid_disks > u->old_raid_disks) | |
7534 | size += sizeof(__u32)*2* | |
7535 | (u->new_raid_disks - u->old_raid_disks); | |
7536 | s = malloc(size); | |
7537 | if (!s) | |
7538 | break; | |
7539 | *space_tail = s; | |
7540 | space_tail = s; | |
7541 | *space_tail = NULL; | |
7542 | break; | |
7543 | } | |
7544 | } | |
7545 | if (update->space_list == NULL) | |
7546 | break; | |
7547 | ||
7548 | /* add space for disk in update | |
7549 | */ | |
7550 | size = sizeof(struct dl); | |
7551 | s = malloc(size); | |
7552 | if (!s) { | |
7553 | free(update->space_list); | |
7554 | update->space_list = NULL; | |
7555 | break; | |
7556 | } | |
7557 | *space_tail = s; | |
7558 | space_tail = s; | |
7559 | *space_tail = NULL; | |
7560 | ||
7561 | /* add spare device to update | |
7562 | */ | |
7563 | for (id = super->devlist ; id; id = id->next) | |
7564 | if (id->index == (unsigned)u->subdev) { | |
7565 | struct imsm_dev *dev; | |
7566 | struct imsm_map *map; | |
7567 | ||
7568 | dev = get_imsm_dev(super, u->subdev); | |
7569 | map = get_imsm_map(dev, 0); | |
7570 | current_level = map->raid_level; | |
7571 | break; | |
7572 | } | |
7573 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
7574 | struct mdinfo *spares; | |
7575 | ||
7576 | spares = get_spares_for_grow(st); | |
7577 | if (spares) { | |
7578 | struct dl *dl; | |
7579 | struct mdinfo *dev; | |
7580 | ||
7581 | dev = spares->devs; | |
7582 | if (dev) { | |
7583 | u->new_disks[0] = | |
7584 | makedev(dev->disk.major, | |
7585 | dev->disk.minor); | |
7586 | dl = get_disk_super(super, | |
7587 | dev->disk.major, | |
7588 | dev->disk.minor); | |
7589 | dl->index = u->old_raid_disks; | |
7590 | dev = dev->next; | |
7591 | } | |
7592 | sysfs_free(spares); | |
7593 | } | |
7594 | } | |
7595 | len = disks_to_mpb_size(u->new_raid_disks); | |
7596 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
7597 | break; |
7598 | } | |
949c47a0 DW |
7599 | case update_create_array: { |
7600 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 7601 | struct intel_dev *dv; |
54c2c1ea DW |
7602 | struct imsm_dev *dev = &u->dev; |
7603 | struct imsm_map *map = get_imsm_map(dev, 0); | |
7604 | struct dl *dl; | |
7605 | struct disk_info *inf; | |
7606 | int i; | |
7607 | int activate = 0; | |
949c47a0 | 7608 | |
54c2c1ea DW |
7609 | inf = get_disk_info(u); |
7610 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
7611 | /* allocate a new super->devlist entry */ |
7612 | dv = malloc(sizeof(*dv)); | |
7613 | if (dv) { | |
7614 | dv->dev = malloc(len); | |
7615 | if (dv->dev) | |
7616 | update->space = dv; | |
7617 | else { | |
7618 | free(dv); | |
7619 | update->space = NULL; | |
7620 | } | |
7621 | } | |
949c47a0 | 7622 | |
54c2c1ea DW |
7623 | /* count how many spares will be converted to members */ |
7624 | for (i = 0; i < map->num_members; i++) { | |
7625 | dl = serial_to_dl(inf[i].serial, super); | |
7626 | if (!dl) { | |
7627 | /* hmm maybe it failed?, nothing we can do about | |
7628 | * it here | |
7629 | */ | |
7630 | continue; | |
7631 | } | |
7632 | if (count_memberships(dl, super) == 0) | |
7633 | activate++; | |
7634 | } | |
7635 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
7636 | break; |
7637 | default: | |
7638 | break; | |
7639 | } | |
7640 | } | |
8273f55e | 7641 | |
4d7b1503 DW |
7642 | /* check if we need a larger metadata buffer */ |
7643 | if (super->next_buf) | |
7644 | buf_len = super->next_len; | |
7645 | else | |
7646 | buf_len = super->len; | |
7647 | ||
7648 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
7649 | /* ok we need a larger buf than what is currently allocated | |
7650 | * if this allocation fails process_update will notice that | |
7651 | * ->next_len is set and ->next_buf is NULL | |
7652 | */ | |
7653 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
7654 | if (super->next_buf) | |
7655 | free(super->next_buf); | |
7656 | ||
7657 | super->next_len = buf_len; | |
1f45a8ad DW |
7658 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
7659 | memset(super->next_buf, 0, buf_len); | |
7660 | else | |
4d7b1503 DW |
7661 | super->next_buf = NULL; |
7662 | } | |
8273f55e DW |
7663 | } |
7664 | ||
ae6aad82 | 7665 | /* must be called while manager is quiesced */ |
f21e18ca | 7666 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
7667 | { |
7668 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
7669 | struct dl *iter; |
7670 | struct imsm_dev *dev; | |
7671 | struct imsm_map *map; | |
24565c9a DW |
7672 | int i, j, num_members; |
7673 | __u32 ord; | |
ae6aad82 | 7674 | |
24565c9a DW |
7675 | dprintf("%s: deleting device[%d] from imsm_super\n", |
7676 | __func__, index); | |
ae6aad82 DW |
7677 | |
7678 | /* shift all indexes down one */ | |
7679 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 7680 | if (iter->index > (int)index) |
ae6aad82 | 7681 | iter->index--; |
47ee5a45 | 7682 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 7683 | if (iter->index > (int)index) |
47ee5a45 | 7684 | iter->index--; |
ae6aad82 DW |
7685 | |
7686 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7687 | dev = get_imsm_dev(super, i); | |
7688 | map = get_imsm_map(dev, 0); | |
24565c9a DW |
7689 | num_members = map->num_members; |
7690 | for (j = 0; j < num_members; j++) { | |
7691 | /* update ord entries being careful not to propagate | |
7692 | * ord-flags to the first map | |
7693 | */ | |
98130f40 | 7694 | ord = get_imsm_ord_tbl_ent(dev, j, -1); |
ae6aad82 | 7695 | |
24565c9a DW |
7696 | if (ord_to_idx(ord) <= index) |
7697 | continue; | |
ae6aad82 | 7698 | |
24565c9a DW |
7699 | map = get_imsm_map(dev, 0); |
7700 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); | |
7701 | map = get_imsm_map(dev, 1); | |
7702 | if (map) | |
7703 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
7704 | } |
7705 | } | |
7706 | ||
7707 | mpb->num_disks--; | |
7708 | super->updates_pending++; | |
24565c9a DW |
7709 | if (*dlp) { |
7710 | struct dl *dl = *dlp; | |
7711 | ||
7712 | *dlp = (*dlp)->next; | |
7713 | __free_imsm_disk(dl); | |
7714 | } | |
ae6aad82 | 7715 | } |
9e2d750d | 7716 | #endif /* MDASSEMBLE */ |
687629c2 AK |
7717 | /******************************************************************************* |
7718 | * Function: open_backup_targets | |
7719 | * Description: Function opens file descriptors for all devices given in | |
7720 | * info->devs | |
7721 | * Parameters: | |
7722 | * info : general array info | |
7723 | * raid_disks : number of disks | |
7724 | * raid_fds : table of device's file descriptors | |
7725 | * Returns: | |
7726 | * 0 : success | |
7727 | * -1 : fail | |
7728 | ******************************************************************************/ | |
7729 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds) | |
7730 | { | |
7731 | struct mdinfo *sd; | |
7732 | ||
7733 | for (sd = info->devs ; sd ; sd = sd->next) { | |
7734 | char *dn; | |
7735 | ||
7736 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
7737 | dprintf("disk is faulty!!\n"); | |
7738 | continue; | |
7739 | } | |
7740 | ||
7741 | if ((sd->disk.raid_disk >= raid_disks) || | |
7742 | (sd->disk.raid_disk < 0)) | |
7743 | continue; | |
7744 | ||
7745 | dn = map_dev(sd->disk.major, | |
7746 | sd->disk.minor, 1); | |
7747 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
7748 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
7749 | fprintf(stderr, "cannot open component\n"); | |
7750 | return -1; | |
7751 | } | |
7752 | } | |
7753 | return 0; | |
7754 | } | |
7755 | ||
9e2d750d | 7756 | #ifndef MDASSEMBLE |
687629c2 AK |
7757 | /******************************************************************************* |
7758 | * Function: init_migr_record_imsm | |
7759 | * Description: Function inits imsm migration record | |
7760 | * Parameters: | |
7761 | * super : imsm internal array info | |
7762 | * dev : device under migration | |
7763 | * info : general array info to find the smallest device | |
7764 | * Returns: | |
7765 | * none | |
7766 | ******************************************************************************/ | |
7767 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
7768 | struct mdinfo *info) | |
7769 | { | |
7770 | struct intel_super *super = st->sb; | |
7771 | struct migr_record *migr_rec = super->migr_rec; | |
7772 | int new_data_disks; | |
7773 | unsigned long long dsize, dev_sectors; | |
7774 | long long unsigned min_dev_sectors = -1LLU; | |
7775 | struct mdinfo *sd; | |
7776 | char nm[30]; | |
7777 | int fd; | |
7778 | struct imsm_map *map_dest = get_imsm_map(dev, 0); | |
7779 | struct imsm_map *map_src = get_imsm_map(dev, 1); | |
7780 | unsigned long long num_migr_units; | |
3ef4403c | 7781 | unsigned long long array_blocks; |
687629c2 AK |
7782 | |
7783 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
7784 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
7785 | ||
7786 | /* only ascending reshape supported now */ | |
7787 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
7788 | ||
7789 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
7790 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
7791 | migr_rec->dest_depth_per_unit *= map_dest->blocks_per_strip; | |
7792 | new_data_disks = imsm_num_data_members(dev, 0); | |
7793 | migr_rec->blocks_per_unit = | |
7794 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
7795 | migr_rec->dest_depth_per_unit = | |
7796 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 7797 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
7798 | num_migr_units = |
7799 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
7800 | ||
7801 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
7802 | num_migr_units++; | |
7803 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
7804 | ||
7805 | migr_rec->post_migr_vol_cap = dev->size_low; | |
7806 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
7807 | ||
7808 | ||
7809 | /* Find the smallest dev */ | |
7810 | for (sd = info->devs ; sd ; sd = sd->next) { | |
7811 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
7812 | fd = dev_open(nm, O_RDONLY); | |
7813 | if (fd < 0) | |
7814 | continue; | |
7815 | get_dev_size(fd, NULL, &dsize); | |
7816 | dev_sectors = dsize / 512; | |
7817 | if (dev_sectors < min_dev_sectors) | |
7818 | min_dev_sectors = dev_sectors; | |
7819 | close(fd); | |
7820 | } | |
7821 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
7822 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
7823 | ||
7824 | write_imsm_migr_rec(st); | |
7825 | ||
7826 | return; | |
7827 | } | |
7828 | ||
7829 | /******************************************************************************* | |
7830 | * Function: save_backup_imsm | |
7831 | * Description: Function saves critical data stripes to Migration Copy Area | |
7832 | * and updates the current migration unit status. | |
7833 | * Use restore_stripes() to form a destination stripe, | |
7834 | * and to write it to the Copy Area. | |
7835 | * Parameters: | |
7836 | * st : supertype information | |
aea93171 | 7837 | * dev : imsm device that backup is saved for |
687629c2 AK |
7838 | * info : general array info |
7839 | * buf : input buffer | |
687629c2 AK |
7840 | * length : length of data to backup (blocks_per_unit) |
7841 | * Returns: | |
7842 | * 0 : success | |
7843 | *, -1 : fail | |
7844 | ******************************************************************************/ | |
7845 | int save_backup_imsm(struct supertype *st, | |
7846 | struct imsm_dev *dev, | |
7847 | struct mdinfo *info, | |
7848 | void *buf, | |
687629c2 AK |
7849 | int length) |
7850 | { | |
7851 | int rv = -1; | |
7852 | struct intel_super *super = st->sb; | |
7853 | unsigned long long *target_offsets = NULL; | |
7854 | int *targets = NULL; | |
7855 | int i; | |
7856 | struct imsm_map *map_dest = get_imsm_map(dev, 0); | |
7857 | int new_disks = map_dest->num_members; | |
ab724b98 AK |
7858 | int dest_layout = 0; |
7859 | int dest_chunk; | |
d1877f69 AK |
7860 | unsigned long long start; |
7861 | int data_disks = imsm_num_data_members(dev, 0); | |
687629c2 AK |
7862 | |
7863 | targets = malloc(new_disks * sizeof(int)); | |
7864 | if (!targets) | |
7865 | goto abort; | |
7866 | ||
7e45b550 AK |
7867 | for (i = 0; i < new_disks; i++) |
7868 | targets[i] = -1; | |
7869 | ||
687629c2 AK |
7870 | target_offsets = malloc(new_disks * sizeof(unsigned long long)); |
7871 | if (!target_offsets) | |
7872 | goto abort; | |
7873 | ||
d1877f69 | 7874 | start = info->reshape_progress * 512; |
687629c2 | 7875 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
7876 | target_offsets[i] = (unsigned long long) |
7877 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
7878 | /* move back copy area adderss, it will be moved forward |
7879 | * in restore_stripes() using start input variable | |
7880 | */ | |
7881 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
7882 | } |
7883 | ||
7884 | if (open_backup_targets(info, new_disks, targets)) | |
7885 | goto abort; | |
7886 | ||
68eb8bc6 | 7887 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
7888 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
7889 | ||
687629c2 AK |
7890 | if (restore_stripes(targets, /* list of dest devices */ |
7891 | target_offsets, /* migration record offsets */ | |
7892 | new_disks, | |
ab724b98 AK |
7893 | dest_chunk, |
7894 | map_dest->raid_level, | |
7895 | dest_layout, | |
7896 | -1, /* source backup file descriptor */ | |
7897 | 0, /* input buf offset | |
7898 | * always 0 buf is already offseted */ | |
d1877f69 | 7899 | start, |
687629c2 AK |
7900 | length, |
7901 | buf) != 0) { | |
7902 | fprintf(stderr, Name ": Error restoring stripes\n"); | |
7903 | goto abort; | |
7904 | } | |
7905 | ||
7906 | rv = 0; | |
7907 | ||
7908 | abort: | |
7909 | if (targets) { | |
7910 | for (i = 0; i < new_disks; i++) | |
7911 | if (targets[i] >= 0) | |
7912 | close(targets[i]); | |
7913 | free(targets); | |
7914 | } | |
7915 | free(target_offsets); | |
7916 | ||
7917 | return rv; | |
7918 | } | |
7919 | ||
7920 | /******************************************************************************* | |
7921 | * Function: save_checkpoint_imsm | |
7922 | * Description: Function called for current unit status update | |
7923 | * in the migration record. It writes it to disk. | |
7924 | * Parameters: | |
7925 | * super : imsm internal array info | |
7926 | * info : general array info | |
7927 | * Returns: | |
7928 | * 0: success | |
7929 | * 1: failure | |
0228d92c AK |
7930 | * 2: failure, means no valid migration record |
7931 | * / no general migration in progress / | |
687629c2 AK |
7932 | ******************************************************************************/ |
7933 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
7934 | { | |
7935 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
7936 | unsigned long long blocks_per_unit; |
7937 | unsigned long long curr_migr_unit; | |
7938 | ||
2e062e82 AK |
7939 | if (load_imsm_migr_rec(super, info) != 0) { |
7940 | dprintf("imsm: ERROR: Cannot read migration record " | |
7941 | "for checkpoint save.\n"); | |
7942 | return 1; | |
7943 | } | |
7944 | ||
f8b72ef5 AK |
7945 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
7946 | if (blocks_per_unit == 0) { | |
0228d92c AK |
7947 | dprintf("imsm: no migration in progress.\n"); |
7948 | return 2; | |
687629c2 | 7949 | } |
f8b72ef5 AK |
7950 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
7951 | /* check if array is alligned to copy area | |
7952 | * if it is not alligned, add one to current migration unit value | |
7953 | * this can happend on array reshape finish only | |
7954 | */ | |
7955 | if (info->reshape_progress % blocks_per_unit) | |
7956 | curr_migr_unit++; | |
687629c2 AK |
7957 | |
7958 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 7959 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
7960 | super->migr_rec->rec_status = __cpu_to_le32(state); |
7961 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
7962 | __cpu_to_le32(curr_migr_unit * |
7963 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
7964 | if (write_imsm_migr_rec(st) < 0) { |
7965 | dprintf("imsm: Cannot write migration record " | |
7966 | "outside backup area\n"); | |
7967 | return 1; | |
7968 | } | |
7969 | ||
7970 | return 0; | |
7971 | } | |
7972 | ||
276d77db AK |
7973 | /******************************************************************************* |
7974 | * Function: recover_backup_imsm | |
7975 | * Description: Function recovers critical data from the Migration Copy Area | |
7976 | * while assembling an array. | |
7977 | * Parameters: | |
7978 | * super : imsm internal array info | |
7979 | * info : general array info | |
7980 | * Returns: | |
7981 | * 0 : success (or there is no data to recover) | |
7982 | * 1 : fail | |
7983 | ******************************************************************************/ | |
7984 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
7985 | { | |
7986 | struct intel_super *super = st->sb; | |
7987 | struct migr_record *migr_rec = super->migr_rec; | |
7988 | struct imsm_map *map_dest = NULL; | |
7989 | struct intel_dev *id = NULL; | |
7990 | unsigned long long read_offset; | |
7991 | unsigned long long write_offset; | |
7992 | unsigned unit_len; | |
7993 | int *targets = NULL; | |
7994 | int new_disks, i, err; | |
7995 | char *buf = NULL; | |
7996 | int retval = 1; | |
7997 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
7998 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 7999 | char buffer[20]; |
6c3560c0 AK |
8000 | int skipped_disks = 0; |
8001 | int max_degradation; | |
276d77db AK |
8002 | |
8003 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
8004 | if (err < 1) | |
8005 | return 1; | |
8006 | ||
8007 | /* recover data only during assemblation */ | |
8008 | if (strncmp(buffer, "inactive", 8) != 0) | |
8009 | return 0; | |
8010 | /* no data to recover */ | |
8011 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
8012 | return 0; | |
8013 | if (curr_migr_unit >= num_migr_units) | |
8014 | return 1; | |
8015 | ||
8016 | /* find device during reshape */ | |
8017 | for (id = super->devlist; id; id = id->next) | |
8018 | if (is_gen_migration(id->dev)) | |
8019 | break; | |
8020 | if (id == NULL) | |
8021 | return 1; | |
8022 | ||
8023 | map_dest = get_imsm_map(id->dev, 0); | |
8024 | new_disks = map_dest->num_members; | |
6c3560c0 | 8025 | max_degradation = new_disks - imsm_num_data_members(id->dev, 0); |
276d77db AK |
8026 | |
8027 | read_offset = (unsigned long long) | |
8028 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
8029 | ||
8030 | write_offset = ((unsigned long long) | |
8031 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
75b69ea4 | 8032 | __le32_to_cpu(map_dest->pba_of_lba0)) * 512; |
276d77db AK |
8033 | |
8034 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
8035 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
8036 | goto abort; | |
8037 | targets = malloc(new_disks * sizeof(int)); | |
8038 | if (!targets) | |
8039 | goto abort; | |
8040 | ||
8041 | open_backup_targets(info, new_disks, targets); | |
8042 | ||
8043 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
8044 | if (targets[i] < 0) { |
8045 | skipped_disks++; | |
8046 | continue; | |
8047 | } | |
276d77db AK |
8048 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
8049 | fprintf(stderr, | |
8050 | Name ": Cannot seek to block: %s\n", | |
8051 | strerror(errno)); | |
8052 | goto abort; | |
8053 | } | |
9ec11d1a | 8054 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
8055 | fprintf(stderr, |
8056 | Name ": Cannot read copy area block: %s\n", | |
8057 | strerror(errno)); | |
8058 | goto abort; | |
8059 | } | |
8060 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
8061 | fprintf(stderr, | |
8062 | Name ": Cannot seek to block: %s\n", | |
8063 | strerror(errno)); | |
8064 | goto abort; | |
8065 | } | |
9ec11d1a | 8066 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
8067 | fprintf(stderr, |
8068 | Name ": Cannot restore block: %s\n", | |
8069 | strerror(errno)); | |
8070 | goto abort; | |
8071 | } | |
8072 | } | |
8073 | ||
6c3560c0 AK |
8074 | if (skipped_disks > max_degradation) { |
8075 | fprintf(stderr, | |
8076 | Name ": Cannot restore data from backup." | |
8077 | " Too many failed disks\n"); | |
8078 | goto abort; | |
8079 | } | |
8080 | ||
befb629b AK |
8081 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
8082 | /* ignore error == 2, this can mean end of reshape here | |
8083 | */ | |
8084 | dprintf("imsm: Cannot write checkpoint to " | |
8085 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
8086 | } else | |
276d77db | 8087 | retval = 0; |
276d77db AK |
8088 | |
8089 | abort: | |
8090 | if (targets) { | |
8091 | for (i = 0; i < new_disks; i++) | |
8092 | if (targets[i]) | |
8093 | close(targets[i]); | |
8094 | free(targets); | |
8095 | } | |
8096 | free(buf); | |
8097 | return retval; | |
8098 | } | |
8099 | ||
2cda7640 ML |
8100 | static char disk_by_path[] = "/dev/disk/by-path/"; |
8101 | ||
8102 | static const char *imsm_get_disk_controller_domain(const char *path) | |
8103 | { | |
2cda7640 | 8104 | char disk_path[PATH_MAX]; |
96234762 LM |
8105 | char *drv=NULL; |
8106 | struct stat st; | |
2cda7640 | 8107 | |
96234762 LM |
8108 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
8109 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
8110 | if (stat(disk_path, &st) == 0) { | |
8111 | struct sys_dev* hba; | |
8112 | char *path=NULL; | |
8113 | ||
8114 | path = devt_to_devpath(st.st_rdev); | |
8115 | if (path == NULL) | |
8116 | return "unknown"; | |
8117 | hba = find_disk_attached_hba(-1, path); | |
8118 | if (hba && hba->type == SYS_DEV_SAS) | |
8119 | drv = "isci"; | |
8120 | else if (hba && hba->type == SYS_DEV_SATA) | |
8121 | drv = "ahci"; | |
8122 | else | |
8123 | drv = "unknown"; | |
8124 | dprintf("path: %s hba: %s attached: %s\n", | |
8125 | path, (hba) ? hba->path : "NULL", drv); | |
8126 | free(path); | |
8127 | if (hba) | |
8128 | free_sys_dev(&hba); | |
2cda7640 | 8129 | } |
96234762 | 8130 | return drv; |
2cda7640 ML |
8131 | } |
8132 | ||
78b10e66 N |
8133 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
8134 | { | |
8135 | char subdev_name[20]; | |
8136 | struct mdstat_ent *mdstat; | |
8137 | ||
8138 | sprintf(subdev_name, "%d", subdev); | |
8139 | mdstat = mdstat_by_subdev(subdev_name, container); | |
8140 | if (!mdstat) | |
8141 | return -1; | |
8142 | ||
8143 | *minor = mdstat->devnum; | |
8144 | free_mdstat(mdstat); | |
8145 | return 0; | |
8146 | } | |
8147 | ||
8148 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
8149 | struct geo_params *geo, | |
8150 | int *old_raid_disks) | |
8151 | { | |
694575e7 KW |
8152 | /* currently we only support increasing the number of devices |
8153 | * for a container. This increases the number of device for each | |
8154 | * member array. They must all be RAID0 or RAID5. | |
8155 | */ | |
78b10e66 N |
8156 | int ret_val = 0; |
8157 | struct mdinfo *info, *member; | |
8158 | int devices_that_can_grow = 0; | |
8159 | ||
8160 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
8161 | "st->devnum = (%i)\n", | |
8162 | st->devnum); | |
8163 | ||
8164 | if (geo->size != -1 || | |
8165 | geo->level != UnSet || | |
8166 | geo->layout != UnSet || | |
8167 | geo->chunksize != 0 || | |
8168 | geo->raid_disks == UnSet) { | |
8169 | dprintf("imsm: Container operation is allowed for " | |
8170 | "raid disks number change only.\n"); | |
8171 | return ret_val; | |
8172 | } | |
8173 | ||
8174 | info = container_content_imsm(st, NULL); | |
8175 | for (member = info; member; member = member->next) { | |
8176 | int result; | |
8177 | int minor; | |
8178 | ||
8179 | dprintf("imsm: checking device_num: %i\n", | |
8180 | member->container_member); | |
8181 | ||
d7d205bd | 8182 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
8183 | /* we work on container for Online Capacity Expansion |
8184 | * only so raid_disks has to grow | |
8185 | */ | |
8186 | dprintf("imsm: for container operation raid disks " | |
8187 | "increase is required\n"); | |
8188 | break; | |
8189 | } | |
8190 | ||
8191 | if ((info->array.level != 0) && | |
8192 | (info->array.level != 5)) { | |
8193 | /* we cannot use this container with other raid level | |
8194 | */ | |
690aae1a | 8195 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
8196 | " raid level (%i) detected\n", |
8197 | info->array.level); | |
8198 | break; | |
8199 | } else { | |
8200 | /* check for platform support | |
8201 | * for this raid level configuration | |
8202 | */ | |
8203 | struct intel_super *super = st->sb; | |
8204 | if (!is_raid_level_supported(super->orom, | |
8205 | member->array.level, | |
8206 | geo->raid_disks)) { | |
690aae1a | 8207 | dprintf("platform does not support raid%d with" |
78b10e66 N |
8208 | " %d disk%s\n", |
8209 | info->array.level, | |
8210 | geo->raid_disks, | |
8211 | geo->raid_disks > 1 ? "s" : ""); | |
8212 | break; | |
8213 | } | |
2a4a08e7 AK |
8214 | /* check if component size is aligned to chunk size |
8215 | */ | |
8216 | if (info->component_size % | |
8217 | (info->array.chunk_size/512)) { | |
8218 | dprintf("Component size is not aligned to " | |
8219 | "chunk size\n"); | |
8220 | break; | |
8221 | } | |
78b10e66 N |
8222 | } |
8223 | ||
8224 | if (*old_raid_disks && | |
8225 | info->array.raid_disks != *old_raid_disks) | |
8226 | break; | |
8227 | *old_raid_disks = info->array.raid_disks; | |
8228 | ||
8229 | /* All raid5 and raid0 volumes in container | |
8230 | * have to be ready for Online Capacity Expansion | |
8231 | * so they need to be assembled. We have already | |
8232 | * checked that no recovery etc is happening. | |
8233 | */ | |
8234 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
8235 | st->container_dev, | |
8236 | &minor); | |
8237 | if (result < 0) { | |
8238 | dprintf("imsm: cannot find array\n"); | |
8239 | break; | |
8240 | } | |
8241 | devices_that_can_grow++; | |
8242 | } | |
8243 | sysfs_free(info); | |
8244 | if (!member && devices_that_can_grow) | |
8245 | ret_val = 1; | |
8246 | ||
8247 | if (ret_val) | |
8248 | dprintf("\tContainer operation allowed\n"); | |
8249 | else | |
8250 | dprintf("\tError: %i\n", ret_val); | |
8251 | ||
8252 | return ret_val; | |
8253 | } | |
8254 | ||
8255 | /* Function: get_spares_for_grow | |
8256 | * Description: Allocates memory and creates list of spare devices | |
8257 | * avaliable in container. Checks if spare drive size is acceptable. | |
8258 | * Parameters: Pointer to the supertype structure | |
8259 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
8260 | * NULL if fail | |
8261 | */ | |
8262 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
8263 | { | |
78b10e66 | 8264 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 8265 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
8266 | } |
8267 | ||
8268 | /****************************************************************************** | |
8269 | * function: imsm_create_metadata_update_for_reshape | |
8270 | * Function creates update for whole IMSM container. | |
8271 | * | |
8272 | ******************************************************************************/ | |
8273 | static int imsm_create_metadata_update_for_reshape( | |
8274 | struct supertype *st, | |
8275 | struct geo_params *geo, | |
8276 | int old_raid_disks, | |
8277 | struct imsm_update_reshape **updatep) | |
8278 | { | |
8279 | struct intel_super *super = st->sb; | |
8280 | struct imsm_super *mpb = super->anchor; | |
8281 | int update_memory_size = 0; | |
8282 | struct imsm_update_reshape *u = NULL; | |
8283 | struct mdinfo *spares = NULL; | |
8284 | int i; | |
8285 | int delta_disks = 0; | |
bbd24d86 | 8286 | struct mdinfo *dev; |
78b10e66 N |
8287 | |
8288 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
8289 | geo->raid_disks); | |
8290 | ||
8291 | delta_disks = geo->raid_disks - old_raid_disks; | |
8292 | ||
8293 | /* size of all update data without anchor */ | |
8294 | update_memory_size = sizeof(struct imsm_update_reshape); | |
8295 | ||
8296 | /* now add space for spare disks that we need to add. */ | |
8297 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
8298 | ||
8299 | u = calloc(1, update_memory_size); | |
8300 | if (u == NULL) { | |
8301 | dprintf("error: " | |
8302 | "cannot get memory for imsm_update_reshape update\n"); | |
8303 | return 0; | |
8304 | } | |
8305 | u->type = update_reshape_container_disks; | |
8306 | u->old_raid_disks = old_raid_disks; | |
8307 | u->new_raid_disks = geo->raid_disks; | |
8308 | ||
8309 | /* now get spare disks list | |
8310 | */ | |
8311 | spares = get_spares_for_grow(st); | |
8312 | ||
8313 | if (spares == NULL | |
8314 | || delta_disks > spares->array.spare_disks) { | |
e14e5960 KW |
8315 | fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices " |
8316 | "for %s.\n", geo->dev_name); | |
e4c72d1d | 8317 | i = -1; |
78b10e66 N |
8318 | goto abort; |
8319 | } | |
8320 | ||
8321 | /* we have got spares | |
8322 | * update disk list in imsm_disk list table in anchor | |
8323 | */ | |
8324 | dprintf("imsm: %i spares are available.\n\n", | |
8325 | spares->array.spare_disks); | |
8326 | ||
bbd24d86 | 8327 | dev = spares->devs; |
78b10e66 | 8328 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
8329 | struct dl *dl; |
8330 | ||
bbd24d86 AK |
8331 | if (dev == NULL) |
8332 | break; | |
78b10e66 N |
8333 | u->new_disks[i] = makedev(dev->disk.major, |
8334 | dev->disk.minor); | |
8335 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
8336 | dl->index = mpb->num_disks; |
8337 | mpb->num_disks++; | |
bbd24d86 | 8338 | dev = dev->next; |
78b10e66 | 8339 | } |
78b10e66 N |
8340 | |
8341 | abort: | |
8342 | /* free spares | |
8343 | */ | |
8344 | sysfs_free(spares); | |
8345 | ||
d677e0b8 | 8346 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 8347 | if (i == delta_disks) { |
d677e0b8 | 8348 | dprintf(" OK\n"); |
78b10e66 N |
8349 | *updatep = u; |
8350 | return update_memory_size; | |
8351 | } | |
8352 | free(u); | |
d677e0b8 | 8353 | dprintf(" Error\n"); |
78b10e66 N |
8354 | |
8355 | return 0; | |
8356 | } | |
8357 | ||
48c5303a PC |
8358 | /****************************************************************************** |
8359 | * function: imsm_create_metadata_update_for_migration() | |
8360 | * Creates update for IMSM array. | |
8361 | * | |
8362 | ******************************************************************************/ | |
8363 | static int imsm_create_metadata_update_for_migration( | |
8364 | struct supertype *st, | |
8365 | struct geo_params *geo, | |
8366 | struct imsm_update_reshape_migration **updatep) | |
8367 | { | |
8368 | struct intel_super *super = st->sb; | |
8369 | int update_memory_size = 0; | |
8370 | struct imsm_update_reshape_migration *u = NULL; | |
8371 | struct imsm_dev *dev; | |
8372 | int previous_level = -1; | |
8373 | ||
8374 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
8375 | " New Level = %i\n", geo->level); | |
8376 | ||
8377 | /* size of all update data without anchor */ | |
8378 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
8379 | ||
8380 | u = calloc(1, update_memory_size); | |
8381 | if (u == NULL) { | |
8382 | dprintf("error: cannot get memory for " | |
8383 | "imsm_create_metadata_update_for_migration\n"); | |
8384 | return 0; | |
8385 | } | |
8386 | u->type = update_reshape_migration; | |
8387 | u->subdev = super->current_vol; | |
8388 | u->new_level = geo->level; | |
8389 | u->new_layout = geo->layout; | |
8390 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
8391 | u->new_disks[0] = -1; | |
4bba0439 | 8392 | u->new_chunksize = -1; |
48c5303a PC |
8393 | |
8394 | dev = get_imsm_dev(super, u->subdev); | |
8395 | if (dev) { | |
8396 | struct imsm_map *map; | |
8397 | ||
8398 | map = get_imsm_map(dev, 0); | |
4bba0439 PC |
8399 | if (map) { |
8400 | int current_chunk_size = | |
8401 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
8402 | ||
8403 | if (geo->chunksize != current_chunk_size) { | |
8404 | u->new_chunksize = geo->chunksize / 1024; | |
8405 | dprintf("imsm: " | |
8406 | "chunk size change from %i to %i\n", | |
8407 | current_chunk_size, u->new_chunksize); | |
8408 | } | |
48c5303a | 8409 | previous_level = map->raid_level; |
4bba0439 | 8410 | } |
48c5303a PC |
8411 | } |
8412 | if ((geo->level == 5) && (previous_level == 0)) { | |
8413 | struct mdinfo *spares = NULL; | |
8414 | ||
8415 | u->new_raid_disks++; | |
8416 | spares = get_spares_for_grow(st); | |
8417 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
8418 | free(u); | |
8419 | sysfs_free(spares); | |
8420 | update_memory_size = 0; | |
8421 | dprintf("error: cannot get spare device " | |
8422 | "for requested migration"); | |
8423 | return 0; | |
8424 | } | |
8425 | sysfs_free(spares); | |
8426 | } | |
8427 | dprintf("imsm: reshape update preparation : OK\n"); | |
8428 | *updatep = u; | |
8429 | ||
8430 | return update_memory_size; | |
8431 | } | |
8432 | ||
8dd70bce AK |
8433 | static void imsm_update_metadata_locally(struct supertype *st, |
8434 | void *buf, int len) | |
8435 | { | |
8436 | struct metadata_update mu; | |
8437 | ||
8438 | mu.buf = buf; | |
8439 | mu.len = len; | |
8440 | mu.space = NULL; | |
8441 | mu.space_list = NULL; | |
8442 | mu.next = NULL; | |
8443 | imsm_prepare_update(st, &mu); | |
8444 | imsm_process_update(st, &mu); | |
8445 | ||
8446 | while (mu.space_list) { | |
8447 | void **space = mu.space_list; | |
8448 | mu.space_list = *space; | |
8449 | free(space); | |
8450 | } | |
8451 | } | |
78b10e66 | 8452 | |
471bceb6 | 8453 | /*************************************************************************** |
694575e7 | 8454 | * Function: imsm_analyze_change |
471bceb6 KW |
8455 | * Description: Function analyze change for single volume |
8456 | * and validate if transition is supported | |
694575e7 KW |
8457 | * Parameters: Geometry parameters, supertype structure |
8458 | * Returns: Operation type code on success, -1 if fail | |
471bceb6 KW |
8459 | ****************************************************************************/ |
8460 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
8461 | struct geo_params *geo) | |
694575e7 | 8462 | { |
471bceb6 KW |
8463 | struct mdinfo info; |
8464 | int change = -1; | |
8465 | int check_devs = 0; | |
c21e737b | 8466 | int chunk; |
471bceb6 KW |
8467 | |
8468 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
8469 | if ((geo->level != info.array.level) && |
8470 | (geo->level >= 0) && | |
8471 | (geo->level != UnSet)) { | |
8472 | switch (info.array.level) { | |
8473 | case 0: | |
8474 | if (geo->level == 5) { | |
b5347799 | 8475 | change = CH_MIGRATION; |
e13ce846 AK |
8476 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
8477 | fprintf(stderr, | |
8478 | Name " Error. Requested Layout " | |
8479 | "not supported (left-asymmetric layout " | |
8480 | "is supported only)!\n"); | |
8481 | change = -1; | |
8482 | goto analyse_change_exit; | |
8483 | } | |
471bceb6 KW |
8484 | check_devs = 1; |
8485 | } | |
8486 | if (geo->level == 10) { | |
8487 | change = CH_TAKEOVER; | |
8488 | check_devs = 1; | |
8489 | } | |
dfe77a9e KW |
8490 | break; |
8491 | case 1: | |
8492 | if (geo->level == 0) { | |
8493 | change = CH_TAKEOVER; | |
8494 | check_devs = 1; | |
8495 | } | |
471bceb6 | 8496 | break; |
471bceb6 KW |
8497 | case 10: |
8498 | if (geo->level == 0) { | |
8499 | change = CH_TAKEOVER; | |
8500 | check_devs = 1; | |
8501 | } | |
8502 | break; | |
8503 | } | |
8504 | if (change == -1) { | |
8505 | fprintf(stderr, | |
8506 | Name " Error. Level Migration from %d to %d " | |
8507 | "not supported!\n", | |
8508 | info.array.level, geo->level); | |
8509 | goto analyse_change_exit; | |
8510 | } | |
8511 | } else | |
8512 | geo->level = info.array.level; | |
8513 | ||
8514 | if ((geo->layout != info.array.layout) | |
8515 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 8516 | change = CH_MIGRATION; |
471bceb6 KW |
8517 | if ((info.array.layout == 0) |
8518 | && (info.array.level == 5) | |
8519 | && (geo->layout == 5)) { | |
8520 | /* reshape 5 -> 4 */ | |
8521 | } else if ((info.array.layout == 5) | |
8522 | && (info.array.level == 5) | |
8523 | && (geo->layout == 0)) { | |
8524 | /* reshape 4 -> 5 */ | |
8525 | geo->layout = 0; | |
8526 | geo->level = 5; | |
8527 | } else { | |
8528 | fprintf(stderr, | |
8529 | Name " Error. Layout Migration from %d to %d " | |
8530 | "not supported!\n", | |
8531 | info.array.layout, geo->layout); | |
8532 | change = -1; | |
8533 | goto analyse_change_exit; | |
8534 | } | |
8535 | } else | |
8536 | geo->layout = info.array.layout; | |
8537 | ||
8538 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
8539 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 8540 | change = CH_MIGRATION; |
471bceb6 KW |
8541 | else |
8542 | geo->chunksize = info.array.chunk_size; | |
8543 | ||
c21e737b | 8544 | chunk = geo->chunksize / 1024; |
471bceb6 KW |
8545 | if (!validate_geometry_imsm(st, |
8546 | geo->level, | |
8547 | geo->layout, | |
8548 | geo->raid_disks, | |
c21e737b | 8549 | &chunk, |
471bceb6 KW |
8550 | geo->size, |
8551 | 0, 0, 1)) | |
8552 | change = -1; | |
8553 | ||
8554 | if (check_devs) { | |
8555 | struct intel_super *super = st->sb; | |
8556 | struct imsm_super *mpb = super->anchor; | |
8557 | ||
8558 | if (mpb->num_raid_devs > 1) { | |
8559 | fprintf(stderr, | |
8560 | Name " Error. Cannot perform operation on %s" | |
8561 | "- for this operation it MUST be single " | |
8562 | "array in container\n", | |
8563 | geo->dev_name); | |
8564 | change = -1; | |
8565 | } | |
8566 | } | |
8567 | ||
8568 | analyse_change_exit: | |
8569 | ||
8570 | return change; | |
694575e7 KW |
8571 | } |
8572 | ||
bb025c2f KW |
8573 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
8574 | { | |
8575 | struct intel_super *super = st->sb; | |
8576 | struct imsm_update_takeover *u; | |
8577 | ||
8578 | u = malloc(sizeof(struct imsm_update_takeover)); | |
8579 | if (u == NULL) | |
8580 | return 1; | |
8581 | ||
8582 | u->type = update_takeover; | |
8583 | u->subarray = super->current_vol; | |
8584 | ||
8585 | /* 10->0 transition */ | |
8586 | if (geo->level == 0) | |
8587 | u->direction = R10_TO_R0; | |
8588 | ||
0529c688 KW |
8589 | /* 0->10 transition */ |
8590 | if (geo->level == 10) | |
8591 | u->direction = R0_TO_R10; | |
8592 | ||
bb025c2f KW |
8593 | /* update metadata locally */ |
8594 | imsm_update_metadata_locally(st, u, | |
8595 | sizeof(struct imsm_update_takeover)); | |
8596 | /* and possibly remotely */ | |
8597 | if (st->update_tail) | |
8598 | append_metadata_update(st, u, | |
8599 | sizeof(struct imsm_update_takeover)); | |
8600 | else | |
8601 | free(u); | |
8602 | ||
8603 | return 0; | |
8604 | } | |
8605 | ||
78b10e66 N |
8606 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
8607 | int layout, int chunksize, int raid_disks, | |
41784c88 AK |
8608 | int delta_disks, char *backup, char *dev, |
8609 | int verbose) | |
78b10e66 | 8610 | { |
78b10e66 N |
8611 | int ret_val = 1; |
8612 | struct geo_params geo; | |
8613 | ||
8614 | dprintf("imsm: reshape_super called.\n"); | |
8615 | ||
71204a50 | 8616 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
8617 | |
8618 | geo.dev_name = dev; | |
694575e7 | 8619 | geo.dev_id = st->devnum; |
78b10e66 N |
8620 | geo.size = size; |
8621 | geo.level = level; | |
8622 | geo.layout = layout; | |
8623 | geo.chunksize = chunksize; | |
8624 | geo.raid_disks = raid_disks; | |
41784c88 AK |
8625 | if (delta_disks != UnSet) |
8626 | geo.raid_disks += delta_disks; | |
78b10e66 N |
8627 | |
8628 | dprintf("\tfor level : %i\n", geo.level); | |
8629 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
8630 | ||
8631 | if (experimental() == 0) | |
8632 | return ret_val; | |
8633 | ||
78b10e66 | 8634 | if (st->container_dev == st->devnum) { |
694575e7 KW |
8635 | /* On container level we can only increase number of devices. */ |
8636 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 8637 | int old_raid_disks = 0; |
6dc0be30 | 8638 | |
78b10e66 N |
8639 | if (imsm_reshape_is_allowed_on_container( |
8640 | st, &geo, &old_raid_disks)) { | |
8641 | struct imsm_update_reshape *u = NULL; | |
8642 | int len; | |
8643 | ||
8644 | len = imsm_create_metadata_update_for_reshape( | |
8645 | st, &geo, old_raid_disks, &u); | |
8646 | ||
ed08d51c AK |
8647 | if (len <= 0) { |
8648 | dprintf("imsm: Cannot prepare update\n"); | |
8649 | goto exit_imsm_reshape_super; | |
8650 | } | |
8651 | ||
8dd70bce AK |
8652 | ret_val = 0; |
8653 | /* update metadata locally */ | |
8654 | imsm_update_metadata_locally(st, u, len); | |
8655 | /* and possibly remotely */ | |
8656 | if (st->update_tail) | |
8657 | append_metadata_update(st, u, len); | |
8658 | else | |
ed08d51c | 8659 | free(u); |
8dd70bce | 8660 | |
694575e7 | 8661 | } else { |
e7ff7e40 AK |
8662 | fprintf(stderr, Name ": (imsm) Operation " |
8663 | "is not allowed on this container\n"); | |
694575e7 KW |
8664 | } |
8665 | } else { | |
8666 | /* On volume level we support following operations | |
471bceb6 KW |
8667 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
8668 | * - chunk size migration | |
8669 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
8670 | */ | |
8671 | struct intel_super *super = st->sb; | |
8672 | struct intel_dev *dev = super->devlist; | |
8673 | int change, devnum; | |
694575e7 | 8674 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
8675 | /* find requested device */ |
8676 | while (dev) { | |
19986c72 MB |
8677 | if (imsm_find_array_minor_by_subdev( |
8678 | dev->index, st->container_dev, &devnum) == 0 | |
8679 | && devnum == geo.dev_id) | |
471bceb6 KW |
8680 | break; |
8681 | dev = dev->next; | |
8682 | } | |
8683 | if (dev == NULL) { | |
8684 | fprintf(stderr, Name " Cannot find %s (%i) subarray\n", | |
8685 | geo.dev_name, geo.dev_id); | |
8686 | goto exit_imsm_reshape_super; | |
8687 | } | |
8688 | super->current_vol = dev->index; | |
694575e7 KW |
8689 | change = imsm_analyze_change(st, &geo); |
8690 | switch (change) { | |
471bceb6 | 8691 | case CH_TAKEOVER: |
bb025c2f | 8692 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 8693 | break; |
48c5303a PC |
8694 | case CH_MIGRATION: { |
8695 | struct imsm_update_reshape_migration *u = NULL; | |
8696 | int len = | |
8697 | imsm_create_metadata_update_for_migration( | |
8698 | st, &geo, &u); | |
8699 | if (len < 1) { | |
8700 | dprintf("imsm: " | |
8701 | "Cannot prepare update\n"); | |
8702 | break; | |
8703 | } | |
471bceb6 | 8704 | ret_val = 0; |
48c5303a PC |
8705 | /* update metadata locally */ |
8706 | imsm_update_metadata_locally(st, u, len); | |
8707 | /* and possibly remotely */ | |
8708 | if (st->update_tail) | |
8709 | append_metadata_update(st, u, len); | |
8710 | else | |
8711 | free(u); | |
8712 | } | |
8713 | break; | |
471bceb6 KW |
8714 | default: |
8715 | ret_val = 1; | |
694575e7 | 8716 | } |
694575e7 | 8717 | } |
78b10e66 | 8718 | |
ed08d51c | 8719 | exit_imsm_reshape_super: |
78b10e66 N |
8720 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
8721 | return ret_val; | |
8722 | } | |
2cda7640 | 8723 | |
eee67a47 AK |
8724 | /******************************************************************************* |
8725 | * Function: wait_for_reshape_imsm | |
8726 | * Description: Function writes new sync_max value and waits until | |
8727 | * reshape process reach new position | |
8728 | * Parameters: | |
8729 | * sra : general array info | |
eee67a47 AK |
8730 | * ndata : number of disks in new array's layout |
8731 | * Returns: | |
8732 | * 0 : success, | |
8733 | * 1 : there is no reshape in progress, | |
8734 | * -1 : fail | |
8735 | ******************************************************************************/ | |
ae9f01f8 | 8736 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
8737 | { |
8738 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
8739 | unsigned long long completed; | |
ae9f01f8 AK |
8740 | /* to_complete : new sync_max position */ |
8741 | unsigned long long to_complete = sra->reshape_progress; | |
8742 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 8743 | |
ae9f01f8 AK |
8744 | if (fd < 0) { |
8745 | dprintf("imsm: wait_for_reshape_imsm() " | |
8746 | "cannot open reshape_position\n"); | |
eee67a47 | 8747 | return 1; |
ae9f01f8 | 8748 | } |
eee67a47 | 8749 | |
ae9f01f8 AK |
8750 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
8751 | dprintf("imsm: wait_for_reshape_imsm() " | |
8752 | "cannot read reshape_position (no reshape in progres)\n"); | |
8753 | close(fd); | |
8754 | return 0; | |
8755 | } | |
eee67a47 | 8756 | |
ae9f01f8 AK |
8757 | if (completed > to_complete) { |
8758 | dprintf("imsm: wait_for_reshape_imsm() " | |
8759 | "wrong next position to set %llu (%llu)\n", | |
8760 | to_complete, completed); | |
8761 | close(fd); | |
8762 | return -1; | |
8763 | } | |
8764 | dprintf("Position set: %llu\n", position_to_set); | |
8765 | if (sysfs_set_num(sra, NULL, "sync_max", | |
8766 | position_to_set) != 0) { | |
8767 | dprintf("imsm: wait_for_reshape_imsm() " | |
8768 | "cannot set reshape position to %llu\n", | |
8769 | position_to_set); | |
8770 | close(fd); | |
8771 | return -1; | |
eee67a47 AK |
8772 | } |
8773 | ||
eee67a47 AK |
8774 | do { |
8775 | char action[20]; | |
8776 | fd_set rfds; | |
8777 | FD_ZERO(&rfds); | |
8778 | FD_SET(fd, &rfds); | |
a47e44fb AK |
8779 | select(fd+1, &rfds, NULL, NULL, NULL); |
8780 | if (sysfs_get_str(sra, NULL, "sync_action", | |
8781 | action, 20) > 0 && | |
8782 | strncmp(action, "reshape", 7) != 0) | |
8783 | break; | |
eee67a47 | 8784 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
8785 | dprintf("imsm: wait_for_reshape_imsm() " |
8786 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
8787 | close(fd); |
8788 | return 1; | |
8789 | } | |
eee67a47 AK |
8790 | } while (completed < to_complete); |
8791 | close(fd); | |
8792 | return 0; | |
8793 | ||
8794 | } | |
8795 | ||
b915c95f AK |
8796 | /******************************************************************************* |
8797 | * Function: check_degradation_change | |
8798 | * Description: Check that array hasn't become failed. | |
8799 | * Parameters: | |
8800 | * info : for sysfs access | |
8801 | * sources : source disks descriptors | |
8802 | * degraded: previous degradation level | |
8803 | * Returns: | |
8804 | * degradation level | |
8805 | ******************************************************************************/ | |
8806 | int check_degradation_change(struct mdinfo *info, | |
8807 | int *sources, | |
8808 | int degraded) | |
8809 | { | |
8810 | unsigned long long new_degraded; | |
8811 | sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
8812 | if (new_degraded != (unsigned long long)degraded) { | |
8813 | /* check each device to ensure it is still working */ | |
8814 | struct mdinfo *sd; | |
8815 | new_degraded = 0; | |
8816 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8817 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
8818 | continue; | |
8819 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
8820 | char sbuf[20]; | |
8821 | if (sysfs_get_str(info, | |
8822 | sd, "state", sbuf, 20) < 0 || | |
8823 | strstr(sbuf, "faulty") || | |
8824 | strstr(sbuf, "in_sync") == NULL) { | |
8825 | /* this device is dead */ | |
8826 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
8827 | if (sd->disk.raid_disk >= 0 && | |
8828 | sources[sd->disk.raid_disk] >= 0) { | |
8829 | close(sources[ | |
8830 | sd->disk.raid_disk]); | |
8831 | sources[sd->disk.raid_disk] = | |
8832 | -1; | |
8833 | } | |
8834 | new_degraded++; | |
8835 | } | |
8836 | } | |
8837 | } | |
8838 | } | |
8839 | ||
8840 | return new_degraded; | |
8841 | } | |
8842 | ||
10f22854 AK |
8843 | /******************************************************************************* |
8844 | * Function: imsm_manage_reshape | |
8845 | * Description: Function finds array under reshape and it manages reshape | |
8846 | * process. It creates stripes backups (if required) and sets | |
8847 | * checheckpoits. | |
8848 | * Parameters: | |
8849 | * afd : Backup handle (nattive) - not used | |
8850 | * sra : general array info | |
8851 | * reshape : reshape parameters - not used | |
8852 | * st : supertype structure | |
8853 | * blocks : size of critical section [blocks] | |
8854 | * fds : table of source device descriptor | |
8855 | * offsets : start of array (offest per devices) | |
8856 | * dests : not used | |
8857 | * destfd : table of destination device descriptor | |
8858 | * destoffsets : table of destination offsets (per device) | |
8859 | * Returns: | |
8860 | * 1 : success, reshape is done | |
8861 | * 0 : fail | |
8862 | ******************************************************************************/ | |
999b4972 N |
8863 | static int imsm_manage_reshape( |
8864 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 8865 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
8866 | int *fds, unsigned long long *offsets, |
8867 | int dests, int *destfd, unsigned long long *destoffsets) | |
8868 | { | |
10f22854 AK |
8869 | int ret_val = 0; |
8870 | struct intel_super *super = st->sb; | |
8871 | struct intel_dev *dv = NULL; | |
8872 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 8873 | struct imsm_map *map_src; |
10f22854 AK |
8874 | int migr_vol_qan = 0; |
8875 | int ndata, odata; /* [bytes] */ | |
8876 | int chunk; /* [bytes] */ | |
8877 | struct migr_record *migr_rec; | |
8878 | char *buf = NULL; | |
8879 | unsigned int buf_size; /* [bytes] */ | |
8880 | unsigned long long max_position; /* array size [bytes] */ | |
8881 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
8882 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
8883 | unsigned long long start_src; /* [bytes] */ |
8884 | unsigned long long start; /* [bytes] */ | |
8885 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 8886 | int degraded = 0; |
ab724b98 | 8887 | int source_layout = 0; |
10f22854 | 8888 | |
1ab242d8 | 8889 | if (!fds || !offsets || !sra) |
10f22854 AK |
8890 | goto abort; |
8891 | ||
8892 | /* Find volume during the reshape */ | |
8893 | for (dv = super->devlist; dv; dv = dv->next) { | |
8894 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
8895 | && dv->dev->vol.migr_state == 1) { | |
8896 | dev = dv->dev; | |
8897 | migr_vol_qan++; | |
8898 | } | |
8899 | } | |
8900 | /* Only one volume can migrate at the same time */ | |
8901 | if (migr_vol_qan != 1) { | |
8902 | fprintf(stderr, Name " : %s", migr_vol_qan ? | |
8903 | "Number of migrating volumes greater than 1\n" : | |
8904 | "There is no volume during migrationg\n"); | |
8905 | goto abort; | |
8906 | } | |
8907 | ||
8908 | map_src = get_imsm_map(dev, 1); | |
8909 | if (map_src == NULL) | |
8910 | goto abort; | |
10f22854 AK |
8911 | |
8912 | ndata = imsm_num_data_members(dev, 0); | |
8913 | odata = imsm_num_data_members(dev, 1); | |
8914 | ||
7b1ab482 | 8915 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
8916 | old_data_stripe_length = odata * chunk; |
8917 | ||
8918 | migr_rec = super->migr_rec; | |
8919 | ||
10f22854 AK |
8920 | /* initialize migration record for start condition */ |
8921 | if (sra->reshape_progress == 0) | |
8922 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
8923 | else { |
8924 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
8925 | dprintf("imsm: cannot restart migration when data " | |
8926 | "are present in copy area.\n"); | |
8927 | goto abort; | |
8928 | } | |
8929 | } | |
10f22854 AK |
8930 | |
8931 | /* size for data */ | |
8932 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
8933 | /* extend buffer size for parity disk */ | |
8934 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
8935 | /* add space for stripe aligment */ | |
8936 | buf_size += old_data_stripe_length; | |
8937 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
8938 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
8939 | goto abort; | |
8940 | } | |
8941 | ||
3ef4403c | 8942 | max_position = sra->component_size * ndata; |
68eb8bc6 | 8943 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
8944 | |
8945 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
8946 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
8947 | /* current reshape position [blocks] */ | |
8948 | unsigned long long current_position = | |
8949 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
8950 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
8951 | unsigned long long border; | |
8952 | ||
b915c95f AK |
8953 | /* Check that array hasn't become failed. |
8954 | */ | |
8955 | degraded = check_degradation_change(sra, fds, degraded); | |
8956 | if (degraded > 1) { | |
8957 | dprintf("imsm: Abort reshape due to degradation" | |
8958 | " level (%i)\n", degraded); | |
8959 | goto abort; | |
8960 | } | |
8961 | ||
10f22854 AK |
8962 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
8963 | ||
8964 | if ((current_position + next_step) > max_position) | |
8965 | next_step = max_position - current_position; | |
8966 | ||
92144abf | 8967 | start = current_position * 512; |
10f22854 AK |
8968 | |
8969 | /* allign reading start to old geometry */ | |
8970 | start_buf_shift = start % old_data_stripe_length; | |
8971 | start_src = start - start_buf_shift; | |
8972 | ||
8973 | border = (start_src / odata) - (start / ndata); | |
8974 | border /= 512; | |
8975 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
8976 | /* save critical stripes to buf | |
8977 | * start - start address of current unit | |
8978 | * to backup [bytes] | |
8979 | * start_src - start address of current unit | |
8980 | * to backup alligned to source array | |
8981 | * [bytes] | |
8982 | */ | |
8983 | unsigned long long next_step_filler = 0; | |
8984 | unsigned long long copy_length = next_step * 512; | |
8985 | ||
8986 | /* allign copy area length to stripe in old geometry */ | |
8987 | next_step_filler = ((copy_length + start_buf_shift) | |
8988 | % old_data_stripe_length); | |
8989 | if (next_step_filler) | |
8990 | next_step_filler = (old_data_stripe_length | |
8991 | - next_step_filler); | |
8992 | dprintf("save_stripes() parameters: start = %llu," | |
8993 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
8994 | "\tstart_in_buf_shift = %llu," | |
8995 | "\tnext_step_filler = %llu\n", | |
8996 | start, start_src, copy_length, | |
8997 | start_buf_shift, next_step_filler); | |
8998 | ||
8999 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
9000 | chunk, map_src->raid_level, |
9001 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
9002 | copy_length + |
9003 | next_step_filler + start_buf_shift, | |
9004 | buf)) { | |
9005 | dprintf("imsm: Cannot save stripes" | |
9006 | " to buffer\n"); | |
9007 | goto abort; | |
9008 | } | |
9009 | /* Convert data to destination format and store it | |
9010 | * in backup general migration area | |
9011 | */ | |
9012 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 9013 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
9014 | dprintf("imsm: Cannot save stripes to " |
9015 | "target devices\n"); | |
9016 | goto abort; | |
9017 | } | |
9018 | if (save_checkpoint_imsm(st, sra, | |
9019 | UNIT_SRC_IN_CP_AREA)) { | |
9020 | dprintf("imsm: Cannot write checkpoint to " | |
9021 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
9022 | goto abort; | |
9023 | } | |
8016a6d4 AK |
9024 | } else { |
9025 | /* set next step to use whole border area */ | |
9026 | border /= next_step; | |
9027 | if (border > 1) | |
9028 | next_step *= border; | |
10f22854 AK |
9029 | } |
9030 | /* When data backed up, checkpoint stored, | |
9031 | * kick the kernel to reshape unit of data | |
9032 | */ | |
9033 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
9034 | /* limit next step to array max position */ |
9035 | if (next_step > max_position) | |
9036 | next_step = max_position; | |
10f22854 AK |
9037 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
9038 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 9039 | sra->reshape_progress = next_step; |
10f22854 AK |
9040 | |
9041 | /* wait until reshape finish */ | |
ae9f01f8 | 9042 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
9043 | dprintf("wait_for_reshape_imsm returned error!\n"); |
9044 | goto abort; | |
9045 | } | |
10f22854 | 9046 | |
0228d92c AK |
9047 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
9048 | /* ignore error == 2, this can mean end of reshape here | |
9049 | */ | |
10f22854 AK |
9050 | dprintf("imsm: Cannot write checkpoint to " |
9051 | "migration record (UNIT_SRC_NORMAL)\n"); | |
9052 | goto abort; | |
9053 | } | |
9054 | ||
9055 | } | |
9056 | ||
9057 | /* return '1' if done */ | |
9058 | ret_val = 1; | |
9059 | abort: | |
9060 | free(buf); | |
9061 | abort_reshape(sra); | |
9062 | ||
9063 | return ret_val; | |
999b4972 | 9064 | } |
71204a50 | 9065 | #endif /* MDASSEMBLE */ |
999b4972 | 9066 | |
cdddbdbc DW |
9067 | struct superswitch super_imsm = { |
9068 | #ifndef MDASSEMBLE | |
9069 | .examine_super = examine_super_imsm, | |
9070 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 9071 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 9072 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
9073 | .detail_super = detail_super_imsm, |
9074 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 9075 | .write_init_super = write_init_super_imsm, |
0e600426 N |
9076 | .validate_geometry = validate_geometry_imsm, |
9077 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 9078 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 9079 | .detail_platform = detail_platform_imsm, |
33414a01 | 9080 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 9081 | .update_subarray = update_subarray_imsm, |
2b959fbf | 9082 | .load_container = load_container_imsm, |
71204a50 N |
9083 | .default_geometry = default_geometry_imsm, |
9084 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
9085 | .reshape_super = imsm_reshape_super, | |
9086 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 9087 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
9088 | #endif |
9089 | .match_home = match_home_imsm, | |
9090 | .uuid_from_super= uuid_from_super_imsm, | |
9091 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 9092 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
9093 | .update_super = update_super_imsm, |
9094 | ||
9095 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 9096 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
9097 | |
9098 | .compare_super = compare_super_imsm, | |
9099 | ||
9100 | .load_super = load_super_imsm, | |
bf5a934a | 9101 | .init_super = init_super_imsm, |
e683ca88 | 9102 | .store_super = store_super_imsm, |
cdddbdbc DW |
9103 | .free_super = free_super_imsm, |
9104 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 9105 | .container_content = container_content_imsm, |
cdddbdbc | 9106 | |
276d77db | 9107 | |
cdddbdbc | 9108 | .external = 1, |
4cce4069 | 9109 | .name = "imsm", |
845dea95 | 9110 | |
0e600426 | 9111 | #ifndef MDASSEMBLE |
845dea95 NB |
9112 | /* for mdmon */ |
9113 | .open_new = imsm_open_new, | |
ed9d66aa | 9114 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
9115 | .set_disk = imsm_set_disk, |
9116 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 9117 | .activate_spare = imsm_activate_spare, |
e8319a19 | 9118 | .process_update = imsm_process_update, |
8273f55e | 9119 | .prepare_update = imsm_prepare_update, |
0e600426 | 9120 | #endif /* MDASSEMBLE */ |
cdddbdbc | 9121 | }; |