]>
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 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
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 | \ | |
418f9b36 N |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE) |
85 | ||
86 | /* Define attributes that are unused but not harmful */ | |
87 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 88 | |
8e59f3d8 | 89 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 90 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 91 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 92 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
93 | |
94 | /* Disk configuration info. */ | |
95 | #define IMSM_MAX_DEVICES 255 | |
96 | struct imsm_disk { | |
97 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 98 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 99 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
100 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
101 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
102 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 103 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb | 104 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
105 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
106 | #define IMSM_DISK_FILLERS 3 | |
107 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
108 | }; |
109 | ||
3b451610 AK |
110 | /* map selector for map managment |
111 | */ | |
238c0a71 AK |
112 | #define MAP_0 0 |
113 | #define MAP_1 1 | |
114 | #define MAP_X -1 | |
3b451610 | 115 | |
cdddbdbc DW |
116 | /* RAID map configuration infos. */ |
117 | struct imsm_map { | |
5551b113 CA |
118 | __u32 pba_of_lba0_lo; /* start address of partition */ |
119 | __u32 blocks_per_member_lo;/* blocks per member */ | |
120 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
121 | __u16 blocks_per_strip; |
122 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
123 | #define IMSM_T_STATE_NORMAL 0 | |
124 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
125 | #define IMSM_T_STATE_DEGRADED 2 |
126 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
127 | __u8 raid_level; |
128 | #define IMSM_T_RAID0 0 | |
129 | #define IMSM_T_RAID1 1 | |
130 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
131 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
132 | __u8 num_domains; /* number of parity domains */ |
133 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 134 | __u8 ddf; |
5551b113 CA |
135 | __u32 pba_of_lba0_hi; |
136 | __u32 blocks_per_member_hi; | |
137 | __u32 num_data_stripes_hi; | |
138 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 139 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 140 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
141 | * top byte contains some flags |
142 | */ | |
cdddbdbc DW |
143 | } __attribute__ ((packed)); |
144 | ||
145 | struct imsm_vol { | |
f8f603f1 | 146 | __u32 curr_migr_unit; |
fe7ed8cb | 147 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 148 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
149 | #define MIGR_INIT 0 |
150 | #define MIGR_REBUILD 1 | |
151 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
152 | #define MIGR_GEN_MIGR 3 | |
153 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 154 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
155 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
156 | __u8 dirty; | |
fe7ed8cb DW |
157 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
158 | __u16 verify_errors; /* number of mismatches */ | |
159 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
160 | __u32 filler[4]; | |
cdddbdbc DW |
161 | struct imsm_map map[1]; |
162 | /* here comes another one if migr_state */ | |
163 | } __attribute__ ((packed)); | |
164 | ||
165 | struct imsm_dev { | |
fe7ed8cb | 166 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
167 | __u32 size_low; |
168 | __u32 size_high; | |
fe7ed8cb DW |
169 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
170 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
171 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
172 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
173 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
174 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
175 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
176 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
177 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
178 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
179 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
180 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
181 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
182 | __u32 status; /* Persistent RaidDev status */ |
183 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
184 | __u8 migr_priority; |
185 | __u8 num_sub_vols; | |
186 | __u8 tid; | |
187 | __u8 cng_master_disk; | |
188 | __u16 cache_policy; | |
189 | __u8 cng_state; | |
190 | __u8 cng_sub_state; | |
191 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
192 | __u32 filler[IMSM_DEV_FILLERS]; |
193 | struct imsm_vol vol; | |
194 | } __attribute__ ((packed)); | |
195 | ||
196 | struct imsm_super { | |
197 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
198 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
199 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
200 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
201 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
202 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
203 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
204 | __u8 num_disks; /* 0x38 Number of configured disks */ |
205 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
206 | __u8 error_log_pos; /* 0x3A */ |
207 | __u8 fill[1]; /* 0x3B */ | |
208 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
209 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
210 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
211 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
212 | #define IMSM_FILLERS 35 | |
213 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
214 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
215 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 216 | /* here comes BBM logs */ |
cdddbdbc DW |
217 | } __attribute__ ((packed)); |
218 | ||
604b746f JD |
219 | #define BBM_LOG_MAX_ENTRIES 254 |
220 | ||
221 | struct bbm_log_entry { | |
222 | __u64 defective_block_start; | |
223 | #define UNREADABLE 0xFFFFFFFF | |
224 | __u32 spare_block_offset; | |
225 | __u16 remapped_marked_count; | |
226 | __u16 disk_ordinal; | |
227 | } __attribute__ ((__packed__)); | |
228 | ||
229 | struct bbm_log { | |
230 | __u32 signature; /* 0xABADB10C */ | |
231 | __u32 entry_count; | |
232 | __u32 reserved_spare_block_count; /* 0 */ | |
233 | __u32 reserved; /* 0xFFFF */ | |
234 | __u64 first_spare_lba; | |
235 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
236 | } __attribute__ ((__packed__)); | |
237 | ||
238 | ||
cdddbdbc DW |
239 | #ifndef MDASSEMBLE |
240 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
241 | #endif | |
242 | ||
8e59f3d8 AK |
243 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
244 | ||
245 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
246 | ||
17a4eaf9 AK |
247 | #define MIGR_REC_BUF_SIZE 512 /* size of migr_record i/o buffer */ |
248 | #define MIGR_REC_POSITION 512 /* migr_record position offset on disk, | |
249 | * MIGR_REC_BUF_SIZE <= MIGR_REC_POSITION | |
250 | */ | |
251 | ||
252 | ||
8e59f3d8 AK |
253 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
254 | * be recovered using srcMap */ | |
255 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
256 | * already been migrated and must | |
257 | * be recovered from checkpoint area */ | |
258 | struct migr_record { | |
259 | __u32 rec_status; /* Status used to determine how to restart | |
260 | * migration in case it aborts | |
261 | * in some fashion */ | |
262 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
263 | __u32 family_num; /* Family number of MPB | |
264 | * containing the RaidDev | |
265 | * that is migrating */ | |
266 | __u32 ascending_migr; /* True if migrating in increasing | |
267 | * order of lbas */ | |
268 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
269 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
270 | * member disk | |
271 | * advances per unit-of-operation */ | |
272 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
273 | __u32 dest_1st_member_lba; /* First member lba on first | |
274 | * stripe of destination */ | |
275 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
276 | __u32 post_migr_vol_cap; /* Size of volume after | |
277 | * migration completes */ | |
278 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
279 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
280 | * migration ckpt record was read from | |
281 | * (for recovered migrations) */ | |
282 | } __attribute__ ((__packed__)); | |
283 | ||
ec50f7b6 LM |
284 | struct md_list { |
285 | /* usage marker: | |
286 | * 1: load metadata | |
287 | * 2: metadata does not match | |
288 | * 4: already checked | |
289 | */ | |
290 | int used; | |
291 | char *devname; | |
292 | int found; | |
293 | int container; | |
294 | dev_t st_rdev; | |
295 | struct md_list *next; | |
296 | }; | |
297 | ||
298 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) | |
299 | ||
1484e727 DW |
300 | static __u8 migr_type(struct imsm_dev *dev) |
301 | { | |
302 | if (dev->vol.migr_type == MIGR_VERIFY && | |
303 | dev->status & DEV_VERIFY_AND_FIX) | |
304 | return MIGR_REPAIR; | |
305 | else | |
306 | return dev->vol.migr_type; | |
307 | } | |
308 | ||
309 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
310 | { | |
311 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
312 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
313 | */ | |
314 | if (migr_type == MIGR_REPAIR) { | |
315 | dev->vol.migr_type = MIGR_VERIFY; | |
316 | dev->status |= DEV_VERIFY_AND_FIX; | |
317 | } else { | |
318 | dev->vol.migr_type = migr_type; | |
319 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
320 | } | |
321 | } | |
322 | ||
87eb16df | 323 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 324 | { |
654a3817 | 325 | return ROUND_UP(bytes, 512) / 512; |
87eb16df | 326 | } |
cdddbdbc | 327 | |
87eb16df DW |
328 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
329 | { | |
330 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
331 | } |
332 | ||
ba2de7ba DW |
333 | struct intel_dev { |
334 | struct imsm_dev *dev; | |
335 | struct intel_dev *next; | |
f21e18ca | 336 | unsigned index; |
ba2de7ba DW |
337 | }; |
338 | ||
88654014 LM |
339 | struct intel_hba { |
340 | enum sys_dev_type type; | |
341 | char *path; | |
342 | char *pci_id; | |
343 | struct intel_hba *next; | |
344 | }; | |
345 | ||
1a64be56 LM |
346 | enum action { |
347 | DISK_REMOVE = 1, | |
348 | DISK_ADD | |
349 | }; | |
cdddbdbc DW |
350 | /* internal representation of IMSM metadata */ |
351 | struct intel_super { | |
352 | union { | |
949c47a0 DW |
353 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
354 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 355 | }; |
8e59f3d8 AK |
356 | union { |
357 | void *migr_rec_buf; /* buffer for I/O operations */ | |
358 | struct migr_record *migr_rec; /* migration record */ | |
359 | }; | |
51d83f5d AK |
360 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
361 | array, it indicates that mdmon is allowed to clean migration | |
362 | record */ | |
949c47a0 | 363 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
364 | void *next_buf; /* for realloc'ing buf from the manager */ |
365 | size_t next_len; | |
c2c087e6 | 366 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 367 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 368 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 369 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 370 | struct intel_dev *devlist; |
cdddbdbc DW |
371 | struct dl { |
372 | struct dl *next; | |
373 | int index; | |
374 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
375 | int major, minor; | |
376 | char *devname; | |
b9f594fe | 377 | struct imsm_disk disk; |
cdddbdbc | 378 | int fd; |
0dcecb2e DW |
379 | int extent_cnt; |
380 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 381 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 382 | enum action action; |
ca0748fa | 383 | } *disks, *current_disk; |
1a64be56 LM |
384 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
385 | active */ | |
47ee5a45 | 386 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 387 | struct bbm_log *bbm_log; |
88654014 | 388 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 389 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
390 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
391 | }; | |
392 | ||
393 | struct intel_disk { | |
394 | struct imsm_disk disk; | |
395 | #define IMSM_UNKNOWN_OWNER (-1) | |
396 | int owner; | |
397 | struct intel_disk *next; | |
cdddbdbc DW |
398 | }; |
399 | ||
c2c087e6 DW |
400 | struct extent { |
401 | unsigned long long start, size; | |
402 | }; | |
403 | ||
694575e7 KW |
404 | /* definitions of reshape process types */ |
405 | enum imsm_reshape_type { | |
406 | CH_TAKEOVER, | |
b5347799 | 407 | CH_MIGRATION, |
7abc9871 | 408 | CH_ARRAY_SIZE, |
694575e7 KW |
409 | }; |
410 | ||
88758e9d DW |
411 | /* definition of messages passed to imsm_process_update */ |
412 | enum imsm_update_type { | |
413 | update_activate_spare, | |
8273f55e | 414 | update_create_array, |
33414a01 | 415 | update_kill_array, |
aa534678 | 416 | update_rename_array, |
1a64be56 | 417 | update_add_remove_disk, |
78b10e66 | 418 | update_reshape_container_disks, |
48c5303a | 419 | update_reshape_migration, |
2d40f3a1 AK |
420 | update_takeover, |
421 | update_general_migration_checkpoint, | |
f3871fdc | 422 | update_size_change, |
88758e9d DW |
423 | }; |
424 | ||
425 | struct imsm_update_activate_spare { | |
426 | enum imsm_update_type type; | |
d23fe947 | 427 | struct dl *dl; |
88758e9d DW |
428 | int slot; |
429 | int array; | |
430 | struct imsm_update_activate_spare *next; | |
431 | }; | |
432 | ||
78b10e66 N |
433 | struct geo_params { |
434 | int dev_id; | |
435 | char *dev_name; | |
436 | long long size; | |
437 | int level; | |
438 | int layout; | |
439 | int chunksize; | |
440 | int raid_disks; | |
441 | }; | |
442 | ||
bb025c2f KW |
443 | enum takeover_direction { |
444 | R10_TO_R0, | |
445 | R0_TO_R10 | |
446 | }; | |
447 | struct imsm_update_takeover { | |
448 | enum imsm_update_type type; | |
449 | int subarray; | |
450 | enum takeover_direction direction; | |
451 | }; | |
78b10e66 N |
452 | |
453 | struct imsm_update_reshape { | |
454 | enum imsm_update_type type; | |
455 | int old_raid_disks; | |
456 | int new_raid_disks; | |
48c5303a PC |
457 | |
458 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
459 | }; | |
460 | ||
461 | struct imsm_update_reshape_migration { | |
462 | enum imsm_update_type type; | |
463 | int old_raid_disks; | |
464 | int new_raid_disks; | |
465 | /* fields for array migration changes | |
466 | */ | |
467 | int subdev; | |
468 | int new_level; | |
469 | int new_layout; | |
4bba0439 | 470 | int new_chunksize; |
48c5303a | 471 | |
d195167d | 472 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
473 | }; |
474 | ||
f3871fdc AK |
475 | struct imsm_update_size_change { |
476 | enum imsm_update_type type; | |
477 | int subdev; | |
478 | long long new_size; | |
479 | }; | |
480 | ||
2d40f3a1 AK |
481 | struct imsm_update_general_migration_checkpoint { |
482 | enum imsm_update_type type; | |
483 | __u32 curr_migr_unit; | |
484 | }; | |
485 | ||
54c2c1ea DW |
486 | struct disk_info { |
487 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
488 | }; | |
489 | ||
8273f55e DW |
490 | struct imsm_update_create_array { |
491 | enum imsm_update_type type; | |
8273f55e | 492 | int dev_idx; |
6a3e913e | 493 | struct imsm_dev dev; |
8273f55e DW |
494 | }; |
495 | ||
33414a01 DW |
496 | struct imsm_update_kill_array { |
497 | enum imsm_update_type type; | |
498 | int dev_idx; | |
499 | }; | |
500 | ||
aa534678 DW |
501 | struct imsm_update_rename_array { |
502 | enum imsm_update_type type; | |
503 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
504 | int dev_idx; | |
505 | }; | |
506 | ||
1a64be56 | 507 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
508 | enum imsm_update_type type; |
509 | }; | |
510 | ||
88654014 LM |
511 | |
512 | static const char *_sys_dev_type[] = { | |
513 | [SYS_DEV_UNKNOWN] = "Unknown", | |
514 | [SYS_DEV_SAS] = "SAS", | |
515 | [SYS_DEV_SATA] = "SATA" | |
516 | }; | |
517 | ||
518 | const char *get_sys_dev_type(enum sys_dev_type type) | |
519 | { | |
520 | if (type >= SYS_DEV_MAX) | |
521 | type = SYS_DEV_UNKNOWN; | |
522 | ||
523 | return _sys_dev_type[type]; | |
524 | } | |
525 | ||
526 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
527 | { | |
528 | struct intel_hba *result = malloc(sizeof(*result)); | |
529 | if (result) { | |
530 | result->type = device->type; | |
531 | result->path = strdup(device->path); | |
532 | result->next = NULL; | |
533 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
534 | result->pci_id++; | |
535 | } | |
536 | return result; | |
537 | } | |
538 | ||
539 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
540 | { | |
541 | struct intel_hba *result=NULL; | |
542 | for (result = hba; result; result = result->next) { | |
543 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
544 | break; | |
545 | } | |
546 | return result; | |
547 | } | |
548 | ||
b4cf4cba | 549 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
550 | { |
551 | struct intel_hba *hba; | |
552 | ||
553 | /* check if disk attached to Intel HBA */ | |
554 | hba = find_intel_hba(super->hba, device); | |
555 | if (hba != NULL) | |
556 | return 1; | |
557 | /* Check if HBA is already attached to super */ | |
558 | if (super->hba == NULL) { | |
559 | super->hba = alloc_intel_hba(device); | |
560 | return 1; | |
561 | } | |
562 | ||
563 | hba = super->hba; | |
564 | /* Intel metadata allows for all disks attached to the same type HBA. | |
565 | * Do not sypport odf HBA types mixing | |
566 | */ | |
567 | if (device->type != hba->type) | |
568 | return 2; | |
569 | ||
570 | while (hba->next) | |
571 | hba = hba->next; | |
572 | ||
573 | hba->next = alloc_intel_hba(device); | |
574 | return 1; | |
575 | } | |
576 | ||
577 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
578 | { | |
579 | struct sys_dev *list, *elem, *prev; | |
580 | char *disk_path; | |
581 | ||
582 | if ((list = find_intel_devices()) == NULL) | |
583 | return 0; | |
584 | ||
585 | if (fd < 0) | |
586 | disk_path = (char *) devname; | |
587 | else | |
588 | disk_path = diskfd_to_devpath(fd); | |
589 | ||
590 | if (!disk_path) { | |
591 | free_sys_dev(&list); | |
592 | return 0; | |
593 | } | |
594 | ||
595 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
596 | if (path_attached_to_hba(disk_path, elem->path)) { | |
597 | if (prev == NULL) | |
598 | list = list->next; | |
599 | else | |
600 | prev->next = elem->next; | |
601 | elem->next = NULL; | |
602 | if (disk_path != devname) | |
603 | free(disk_path); | |
604 | free_sys_dev(&list); | |
605 | return elem; | |
606 | } | |
607 | } | |
608 | if (disk_path != devname) | |
609 | free(disk_path); | |
610 | free_sys_dev(&list); | |
611 | ||
612 | return NULL; | |
613 | } | |
614 | ||
615 | ||
d424212e N |
616 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
617 | char *devname); | |
f2f5c343 | 618 | |
cdddbdbc DW |
619 | static struct supertype *match_metadata_desc_imsm(char *arg) |
620 | { | |
621 | struct supertype *st; | |
622 | ||
623 | if (strcmp(arg, "imsm") != 0 && | |
624 | strcmp(arg, "default") != 0 | |
625 | ) | |
626 | return NULL; | |
627 | ||
628 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
629 | if (!st) |
630 | return NULL; | |
ef609477 | 631 | memset(st, 0, sizeof(*st)); |
d1d599ea | 632 | st->container_dev = NoMdDev; |
cdddbdbc DW |
633 | st->ss = &super_imsm; |
634 | st->max_devs = IMSM_MAX_DEVICES; | |
635 | st->minor_version = 0; | |
636 | st->sb = NULL; | |
637 | return st; | |
638 | } | |
639 | ||
0e600426 | 640 | #ifndef MDASSEMBLE |
cdddbdbc DW |
641 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
642 | { | |
643 | return &mpb->sig[MPB_SIG_LEN]; | |
644 | } | |
9e2d750d | 645 | #endif |
cdddbdbc | 646 | |
949c47a0 DW |
647 | /* retrieve a disk directly from the anchor when the anchor is known to be |
648 | * up-to-date, currently only at load time | |
649 | */ | |
650 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 651 | { |
949c47a0 | 652 | if (index >= mpb->num_disks) |
cdddbdbc DW |
653 | return NULL; |
654 | return &mpb->disk[index]; | |
655 | } | |
656 | ||
95d07a2c LM |
657 | /* retrieve the disk description based on a index of the disk |
658 | * in the sub-array | |
659 | */ | |
660 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 661 | { |
b9f594fe DW |
662 | struct dl *d; |
663 | ||
664 | for (d = super->disks; d; d = d->next) | |
665 | if (d->index == index) | |
95d07a2c LM |
666 | return d; |
667 | ||
668 | return NULL; | |
669 | } | |
670 | /* retrieve a disk from the parsed metadata */ | |
671 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
672 | { | |
673 | struct dl *dl; | |
674 | ||
675 | dl = get_imsm_dl_disk(super, index); | |
676 | if (dl) | |
677 | return &dl->disk; | |
678 | ||
b9f594fe | 679 | return NULL; |
949c47a0 DW |
680 | } |
681 | ||
682 | /* generate a checksum directly from the anchor when the anchor is known to be | |
683 | * up-to-date, currently only at load or write_super after coalescing | |
684 | */ | |
685 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
686 | { |
687 | __u32 end = mpb->mpb_size / sizeof(end); | |
688 | __u32 *p = (__u32 *) mpb; | |
689 | __u32 sum = 0; | |
690 | ||
97f734fd N |
691 | while (end--) { |
692 | sum += __le32_to_cpu(*p); | |
693 | p++; | |
694 | } | |
cdddbdbc DW |
695 | |
696 | return sum - __le32_to_cpu(mpb->check_sum); | |
697 | } | |
698 | ||
a965f303 DW |
699 | static size_t sizeof_imsm_map(struct imsm_map *map) |
700 | { | |
701 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
702 | } | |
703 | ||
704 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 705 | { |
5e7b0330 AK |
706 | /* A device can have 2 maps if it is in the middle of a migration. |
707 | * If second_map is: | |
238c0a71 AK |
708 | * MAP_0 - we return the first map |
709 | * MAP_1 - we return the second map if it exists, else NULL | |
710 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 711 | */ |
a965f303 | 712 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 713 | struct imsm_map *map2 = NULL; |
a965f303 | 714 | |
9535fc47 AK |
715 | if (dev->vol.migr_state) |
716 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 717 | |
9535fc47 | 718 | switch (second_map) { |
3b451610 | 719 | case MAP_0: |
9535fc47 | 720 | break; |
3b451610 | 721 | case MAP_1: |
9535fc47 AK |
722 | map = map2; |
723 | break; | |
238c0a71 | 724 | case MAP_X: |
9535fc47 AK |
725 | if (map2) |
726 | map = map2; | |
727 | break; | |
9535fc47 AK |
728 | default: |
729 | map = NULL; | |
730 | } | |
731 | return map; | |
5e7b0330 | 732 | |
a965f303 | 733 | } |
cdddbdbc | 734 | |
3393c6af DW |
735 | /* return the size of the device. |
736 | * migr_state increases the returned size if map[0] were to be duplicated | |
737 | */ | |
738 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
739 | { |
740 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 741 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
742 | |
743 | /* migrating means an additional map */ | |
a965f303 | 744 | if (dev->vol.migr_state) |
238c0a71 | 745 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 746 | else if (migr_state) |
238c0a71 | 747 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
748 | |
749 | return size; | |
750 | } | |
751 | ||
54c2c1ea DW |
752 | #ifndef MDASSEMBLE |
753 | /* retrieve disk serial number list from a metadata update */ | |
754 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
755 | { | |
756 | void *u = update; | |
757 | struct disk_info *inf; | |
758 | ||
759 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
760 | sizeof_imsm_dev(&update->dev, 0); | |
761 | ||
762 | return inf; | |
763 | } | |
764 | #endif | |
765 | ||
949c47a0 | 766 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
767 | { |
768 | int offset; | |
769 | int i; | |
770 | void *_mpb = mpb; | |
771 | ||
949c47a0 | 772 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
773 | return NULL; |
774 | ||
775 | /* devices start after all disks */ | |
776 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
777 | ||
778 | for (i = 0; i <= index; i++) | |
779 | if (i == index) | |
780 | return _mpb + offset; | |
781 | else | |
3393c6af | 782 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
783 | |
784 | return NULL; | |
785 | } | |
786 | ||
949c47a0 DW |
787 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
788 | { | |
ba2de7ba DW |
789 | struct intel_dev *dv; |
790 | ||
949c47a0 DW |
791 | if (index >= super->anchor->num_raid_devs) |
792 | return NULL; | |
ba2de7ba DW |
793 | for (dv = super->devlist; dv; dv = dv->next) |
794 | if (dv->index == index) | |
795 | return dv->dev; | |
796 | return NULL; | |
949c47a0 DW |
797 | } |
798 | ||
98130f40 AK |
799 | /* |
800 | * for second_map: | |
238c0a71 AK |
801 | * == MAP_0 get first map |
802 | * == MAP_1 get second map | |
803 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
804 | */ |
805 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
806 | int slot, | |
807 | int second_map) | |
7eef0453 DW |
808 | { |
809 | struct imsm_map *map; | |
810 | ||
5e7b0330 | 811 | map = get_imsm_map(dev, second_map); |
7eef0453 | 812 | |
ff077194 DW |
813 | /* top byte identifies disk under rebuild */ |
814 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
815 | } | |
816 | ||
817 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 818 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 819 | { |
98130f40 | 820 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
821 | |
822 | return ord_to_idx(ord); | |
7eef0453 DW |
823 | } |
824 | ||
be73972f DW |
825 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
826 | { | |
827 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
828 | } | |
829 | ||
f21e18ca | 830 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
831 | { |
832 | int slot; | |
833 | __u32 ord; | |
834 | ||
835 | for (slot = 0; slot < map->num_members; slot++) { | |
836 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
837 | if (ord_to_idx(ord) == idx) | |
838 | return slot; | |
839 | } | |
840 | ||
841 | return -1; | |
842 | } | |
843 | ||
cdddbdbc DW |
844 | static int get_imsm_raid_level(struct imsm_map *map) |
845 | { | |
846 | if (map->raid_level == 1) { | |
847 | if (map->num_members == 2) | |
848 | return 1; | |
849 | else | |
850 | return 10; | |
851 | } | |
852 | ||
853 | return map->raid_level; | |
854 | } | |
855 | ||
c2c087e6 DW |
856 | static int cmp_extent(const void *av, const void *bv) |
857 | { | |
858 | const struct extent *a = av; | |
859 | const struct extent *b = bv; | |
860 | if (a->start < b->start) | |
861 | return -1; | |
862 | if (a->start > b->start) | |
863 | return 1; | |
864 | return 0; | |
865 | } | |
866 | ||
0dcecb2e | 867 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 868 | { |
c2c087e6 | 869 | int memberships = 0; |
620b1713 | 870 | int i; |
c2c087e6 | 871 | |
949c47a0 DW |
872 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
873 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 874 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 875 | |
620b1713 DW |
876 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
877 | memberships++; | |
c2c087e6 | 878 | } |
0dcecb2e DW |
879 | |
880 | return memberships; | |
881 | } | |
882 | ||
b81221b7 CA |
883 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
884 | ||
5551b113 CA |
885 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
886 | { | |
887 | if (lo == 0 || hi == 0) | |
888 | return 1; | |
889 | *lo = __le32_to_cpu((unsigned)n); | |
890 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
891 | return 0; | |
892 | } | |
893 | ||
894 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
895 | { | |
896 | return (unsigned long long)__le32_to_cpu(lo) | | |
897 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
898 | } | |
899 | ||
900 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
901 | { | |
902 | if (disk == NULL) | |
903 | return 0; | |
904 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
905 | } | |
906 | ||
907 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
908 | { | |
909 | if (map == NULL) | |
910 | return 0; | |
911 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
912 | } | |
913 | ||
914 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
915 | { | |
916 | if (map == NULL) | |
917 | return 0; | |
918 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
919 | } | |
920 | ||
921 | static unsigned long long num_data_stripes(struct imsm_map *map) | |
922 | { | |
923 | if (map == NULL) | |
924 | return 0; | |
925 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
926 | } | |
927 | ||
928 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
929 | { | |
930 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
931 | } | |
932 | ||
933 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
934 | { | |
935 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
936 | } | |
937 | ||
938 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
939 | { | |
940 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
941 | } | |
942 | ||
943 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
944 | { | |
945 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
946 | } | |
947 | ||
0dcecb2e DW |
948 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
949 | { | |
950 | /* find a list of used extents on the given physical device */ | |
951 | struct extent *rv, *e; | |
620b1713 | 952 | int i; |
0dcecb2e | 953 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
954 | __u32 reservation; |
955 | ||
956 | /* trim the reserved area for spares, so they can join any array | |
957 | * regardless of whether the OROM has assigned sectors from the | |
958 | * IMSM_RESERVED_SECTORS region | |
959 | */ | |
960 | if (dl->index == -1) | |
b81221b7 | 961 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
962 | else |
963 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 964 | |
c2c087e6 DW |
965 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
966 | if (!rv) | |
967 | return NULL; | |
968 | e = rv; | |
969 | ||
949c47a0 DW |
970 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
971 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 972 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 973 | |
620b1713 | 974 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
975 | e->start = pba_of_lba0(map); |
976 | e->size = blocks_per_member(map); | |
620b1713 | 977 | e++; |
c2c087e6 DW |
978 | } |
979 | } | |
980 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
981 | ||
14e8215b DW |
982 | /* determine the start of the metadata |
983 | * when no raid devices are defined use the default | |
984 | * ...otherwise allow the metadata to truncate the value | |
985 | * as is the case with older versions of imsm | |
986 | */ | |
987 | if (memberships) { | |
988 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 989 | unsigned long long remainder; |
14e8215b | 990 | |
5551b113 | 991 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
992 | /* round down to 1k block to satisfy precision of the kernel |
993 | * 'size' interface | |
994 | */ | |
995 | remainder &= ~1UL; | |
996 | /* make sure remainder is still sane */ | |
f21e18ca | 997 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 998 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
999 | if (reservation > remainder) |
1000 | reservation = remainder; | |
1001 | } | |
5551b113 | 1002 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1003 | e->size = 0; |
1004 | return rv; | |
1005 | } | |
1006 | ||
14e8215b DW |
1007 | /* try to determine how much space is reserved for metadata from |
1008 | * the last get_extents() entry, otherwise fallback to the | |
1009 | * default | |
1010 | */ | |
1011 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1012 | { | |
1013 | struct extent *e; | |
1014 | int i; | |
1015 | __u32 rv; | |
1016 | ||
1017 | /* for spares just return a minimal reservation which will grow | |
1018 | * once the spare is picked up by an array | |
1019 | */ | |
1020 | if (dl->index == -1) | |
1021 | return MPB_SECTOR_CNT; | |
1022 | ||
1023 | e = get_extents(super, dl); | |
1024 | if (!e) | |
1025 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1026 | ||
1027 | /* scroll to last entry */ | |
1028 | for (i = 0; e[i].size; i++) | |
1029 | continue; | |
1030 | ||
5551b113 | 1031 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1032 | |
1033 | free(e); | |
1034 | ||
1035 | return rv; | |
1036 | } | |
1037 | ||
25ed7e59 DW |
1038 | static int is_spare(struct imsm_disk *disk) |
1039 | { | |
1040 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1041 | } | |
1042 | ||
1043 | static int is_configured(struct imsm_disk *disk) | |
1044 | { | |
1045 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1046 | } | |
1047 | ||
1048 | static int is_failed(struct imsm_disk *disk) | |
1049 | { | |
1050 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1051 | } | |
1052 | ||
b81221b7 CA |
1053 | /* try to determine how much space is reserved for metadata from |
1054 | * the last get_extents() entry on the smallest active disk, | |
1055 | * otherwise fallback to the default | |
1056 | */ | |
1057 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1058 | { | |
1059 | struct extent *e; | |
1060 | int i; | |
5551b113 CA |
1061 | unsigned long long min_active; |
1062 | __u32 remainder; | |
b81221b7 CA |
1063 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1064 | struct dl *dl, *dl_min = NULL; | |
1065 | ||
1066 | if (!super) | |
1067 | return rv; | |
1068 | ||
1069 | min_active = 0; | |
1070 | for (dl = super->disks; dl; dl = dl->next) { | |
1071 | if (dl->index < 0) | |
1072 | continue; | |
5551b113 CA |
1073 | unsigned long long blocks = total_blocks(&dl->disk); |
1074 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1075 | dl_min = dl; |
5551b113 | 1076 | min_active = blocks; |
b81221b7 CA |
1077 | } |
1078 | } | |
1079 | if (!dl_min) | |
1080 | return rv; | |
1081 | ||
1082 | /* find last lba used by subarrays on the smallest active disk */ | |
1083 | e = get_extents(super, dl_min); | |
1084 | if (!e) | |
1085 | return rv; | |
1086 | for (i = 0; e[i].size; i++) | |
1087 | continue; | |
1088 | ||
1089 | remainder = min_active - e[i].start; | |
1090 | free(e); | |
1091 | ||
1092 | /* to give priority to recovery we should not require full | |
1093 | IMSM_RESERVED_SECTORS from the spare */ | |
1094 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1095 | ||
1096 | /* if real reservation is smaller use that value */ | |
1097 | return (remainder < rv) ? remainder : rv; | |
1098 | } | |
1099 | ||
80e7f8c3 AC |
1100 | /* Return minimum size of a spare that can be used in this array*/ |
1101 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1102 | { | |
1103 | struct intel_super *super = st->sb; | |
1104 | struct dl *dl; | |
1105 | struct extent *e; | |
1106 | int i; | |
1107 | unsigned long long rv = 0; | |
1108 | ||
1109 | if (!super) | |
1110 | return rv; | |
1111 | /* find first active disk in array */ | |
1112 | dl = super->disks; | |
1113 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1114 | dl = dl->next; | |
1115 | if (!dl) | |
1116 | return rv; | |
1117 | /* find last lba used by subarrays */ | |
1118 | e = get_extents(super, dl); | |
1119 | if (!e) | |
1120 | return rv; | |
1121 | for (i = 0; e[i].size; i++) | |
1122 | continue; | |
1123 | if (i > 0) | |
1124 | rv = e[i-1].start + e[i-1].size; | |
1125 | free(e); | |
b81221b7 | 1126 | |
80e7f8c3 | 1127 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1128 | rv = rv + imsm_min_reserved_sectors(super); |
1129 | ||
80e7f8c3 AC |
1130 | return rv * 512; |
1131 | } | |
1132 | ||
d1e02575 AK |
1133 | static int is_gen_migration(struct imsm_dev *dev); |
1134 | ||
1799c9e8 | 1135 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1136 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1137 | struct imsm_dev *dev); | |
1e5c6983 | 1138 | |
c47b0ff6 AK |
1139 | static void print_imsm_dev(struct intel_super *super, |
1140 | struct imsm_dev *dev, | |
1141 | char *uuid, | |
1142 | int disk_idx) | |
cdddbdbc DW |
1143 | { |
1144 | __u64 sz; | |
0d80bb2f | 1145 | int slot, i; |
238c0a71 AK |
1146 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1147 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1148 | __u32 ord; |
cdddbdbc DW |
1149 | |
1150 | printf("\n"); | |
1e7bc0ed | 1151 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1152 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1153 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1154 | if (map2) | |
1155 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1156 | printf("\n"); | |
1157 | printf(" Members : %d", map->num_members); | |
1158 | if (map2) | |
1159 | printf(" <-- %d", map2->num_members); | |
1160 | printf("\n"); | |
0d80bb2f DW |
1161 | printf(" Slots : ["); |
1162 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1163 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1164 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1165 | } | |
dd8bcb3b AK |
1166 | printf("]"); |
1167 | if (map2) { | |
1168 | printf(" <-- ["); | |
1169 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1170 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1171 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1172 | } | |
1173 | printf("]"); | |
1174 | } | |
1175 | printf("\n"); | |
7095bccb AK |
1176 | printf(" Failed disk : "); |
1177 | if (map->failed_disk_num == 0xff) | |
1178 | printf("none"); | |
1179 | else | |
1180 | printf("%i", map->failed_disk_num); | |
1181 | printf("\n"); | |
620b1713 DW |
1182 | slot = get_imsm_disk_slot(map, disk_idx); |
1183 | if (slot >= 0) { | |
238c0a71 | 1184 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1185 | printf(" This Slot : %d%s\n", slot, |
1186 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1187 | } else | |
cdddbdbc DW |
1188 | printf(" This Slot : ?\n"); |
1189 | sz = __le32_to_cpu(dev->size_high); | |
1190 | sz <<= 32; | |
1191 | sz += __le32_to_cpu(dev->size_low); | |
1192 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1193 | human_size(sz * 512)); | |
5551b113 | 1194 | sz = blocks_per_member(map); |
cdddbdbc DW |
1195 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, |
1196 | human_size(sz * 512)); | |
5551b113 CA |
1197 | printf(" Sector Offset : %llu\n", |
1198 | pba_of_lba0(map)); | |
1199 | printf(" Num Stripes : %llu\n", | |
1200 | num_data_stripes(map)); | |
dd8bcb3b | 1201 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1202 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1203 | if (map2) |
1204 | printf(" <-- %u KiB", | |
1205 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1206 | printf("\n"); | |
cdddbdbc | 1207 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1208 | printf(" Migrate State : "); |
1484e727 DW |
1209 | if (dev->vol.migr_state) { |
1210 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1211 | printf("initialize\n"); |
1484e727 | 1212 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1213 | printf("rebuild\n"); |
1484e727 | 1214 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1215 | printf("check\n"); |
1484e727 | 1216 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1217 | printf("general migration\n"); |
1484e727 | 1218 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1219 | printf("state change\n"); |
1484e727 | 1220 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1221 | printf("repair\n"); |
1484e727 | 1222 | else |
8655a7b1 DW |
1223 | printf("<unknown:%d>\n", migr_type(dev)); |
1224 | } else | |
1225 | printf("idle\n"); | |
3393c6af DW |
1226 | printf(" Map State : %s", map_state_str[map->map_state]); |
1227 | if (dev->vol.migr_state) { | |
238c0a71 | 1228 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1229 | |
b10b37b8 | 1230 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1231 | printf("\n Checkpoint : %u ", |
1232 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
3136abe5 | 1233 | if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0))) |
464d40e8 LD |
1234 | printf("(N/A)"); |
1235 | else | |
1236 | printf("(%llu)", (unsigned long long) | |
1237 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1238 | } |
1239 | printf("\n"); | |
cdddbdbc | 1240 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1241 | } |
1242 | ||
0ec1f4e8 | 1243 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1244 | { |
1f24f035 | 1245 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1246 | __u64 sz; |
1247 | ||
0ec1f4e8 | 1248 | if (index < -1 || !disk) |
e9d82038 DW |
1249 | return; |
1250 | ||
cdddbdbc | 1251 | printf("\n"); |
1f24f035 | 1252 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1253 | if (index >= 0) |
1254 | printf(" Disk%02d Serial : %s\n", index, str); | |
1255 | else | |
1256 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1257 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1258 | is_configured(disk) ? " active" : "", | |
1259 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1260 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1261 | sz = total_blocks(disk) - reserved; |
cdddbdbc DW |
1262 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1263 | human_size(sz * 512)); | |
1264 | } | |
1265 | ||
520e69e2 AK |
1266 | void examine_migr_rec_imsm(struct intel_super *super) |
1267 | { | |
1268 | struct migr_record *migr_rec = super->migr_rec; | |
1269 | struct imsm_super *mpb = super->anchor; | |
1270 | int i; | |
1271 | ||
1272 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1273 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1274 | struct imsm_map *map; |
b4ab44d8 | 1275 | int slot = -1; |
3136abe5 | 1276 | |
520e69e2 AK |
1277 | if (is_gen_migration(dev) == 0) |
1278 | continue; | |
1279 | ||
1280 | printf("\nMigration Record Information:"); | |
3136abe5 | 1281 | |
44bfe6df AK |
1282 | /* first map under migration */ |
1283 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1284 | if (map) |
1285 | slot = get_imsm_disk_slot(map, super->disks->index); | |
1286 | if ((map == NULL) || (slot > 1) || (slot < 0)) { | |
520e69e2 AK |
1287 | printf(" Empty\n "); |
1288 | printf("Examine one of first two disks in array\n"); | |
1289 | break; | |
1290 | } | |
1291 | printf("\n Status : "); | |
1292 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1293 | printf("Normal\n"); | |
1294 | else | |
1295 | printf("Contains Data\n"); | |
1296 | printf(" Current Unit : %u\n", | |
1297 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1298 | printf(" Family : %u\n", | |
1299 | __le32_to_cpu(migr_rec->family_num)); | |
1300 | printf(" Ascending : %u\n", | |
1301 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1302 | printf(" Blocks Per Unit : %u\n", | |
1303 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1304 | printf(" Dest. Depth Per Unit : %u\n", | |
1305 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1306 | printf(" Checkpoint Area pba : %u\n", | |
1307 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1308 | printf(" First member lba : %u\n", | |
1309 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1310 | printf(" Total Number of Units : %u\n", | |
1311 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1312 | printf(" Size of volume : %u\n", | |
1313 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1314 | printf(" Expansion space for LBA64 : %u\n", | |
1315 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1316 | printf(" Record was read from : %u\n", | |
1317 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1318 | ||
1319 | break; | |
1320 | } | |
1321 | } | |
9e2d750d | 1322 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1323 | /******************************************************************************* |
1324 | * function: imsm_check_attributes | |
1325 | * Description: Function checks if features represented by attributes flags | |
1326 | * are supported by mdadm. | |
1327 | * Parameters: | |
1328 | * attributes - Attributes read from metadata | |
1329 | * Returns: | |
1330 | * 0 - passed attributes contains unsupported features flags | |
1331 | * 1 - all features are supported | |
1332 | ******************************************************************************/ | |
1333 | static int imsm_check_attributes(__u32 attributes) | |
1334 | { | |
1335 | int ret_val = 1; | |
418f9b36 N |
1336 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1337 | ||
1338 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1339 | |
1340 | not_supported &= attributes; | |
1341 | if (not_supported) { | |
418f9b36 N |
1342 | fprintf(stderr, Name "(IMSM): Unsupported attributes : %x\n", |
1343 | (unsigned)__le32_to_cpu(not_supported)); | |
19482bcc AK |
1344 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1345 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1346 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1347 | } | |
1348 | if (not_supported & MPB_ATTRIB_2TB) { | |
1349 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1350 | not_supported ^= MPB_ATTRIB_2TB; | |
1351 | } | |
1352 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1353 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1354 | not_supported ^= MPB_ATTRIB_RAID0; | |
1355 | } | |
1356 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1357 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1358 | not_supported ^= MPB_ATTRIB_RAID1; | |
1359 | } | |
1360 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1361 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1362 | not_supported ^= MPB_ATTRIB_RAID10; | |
1363 | } | |
1364 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1365 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1366 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1367 | } | |
1368 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1369 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1370 | not_supported ^= MPB_ATTRIB_RAID5; | |
1371 | } | |
1372 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1373 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1374 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1375 | } | |
1376 | if (not_supported & MPB_ATTRIB_BBM) { | |
1377 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1378 | not_supported ^= MPB_ATTRIB_BBM; | |
1379 | } | |
1380 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1381 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1382 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1383 | } | |
1384 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1385 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1386 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1387 | } | |
1388 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1389 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1390 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1391 | } | |
1392 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1393 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1394 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1395 | } | |
1396 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1397 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1398 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1399 | } | |
1400 | ||
1401 | if (not_supported) | |
1402 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1403 | ||
1404 | ret_val = 0; | |
1405 | } | |
1406 | ||
1407 | return ret_val; | |
1408 | } | |
1409 | ||
9e2d750d | 1410 | #ifndef MDASSEMBLE |
a5d85af7 | 1411 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1412 | |
cdddbdbc DW |
1413 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1414 | { | |
1415 | struct intel_super *super = st->sb; | |
949c47a0 | 1416 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1417 | char str[MAX_SIGNATURE_LENGTH]; |
1418 | int i; | |
27fd6274 DW |
1419 | struct mdinfo info; |
1420 | char nbuf[64]; | |
cdddbdbc | 1421 | __u32 sum; |
14e8215b | 1422 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1423 | struct dl *dl; |
27fd6274 | 1424 | |
cdddbdbc DW |
1425 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1426 | printf(" Magic : %s\n", str); | |
1427 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1428 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1429 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1430 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1431 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1432 | printf(" Attributes : "); |
1433 | if (imsm_check_attributes(mpb->attributes)) | |
1434 | printf("All supported\n"); | |
1435 | else | |
1436 | printf("not supported\n"); | |
a5d85af7 | 1437 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1438 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1439 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1440 | sum = __le32_to_cpu(mpb->check_sum); |
1441 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1442 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1443 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1444 | printf(" Disks : %d\n", mpb->num_disks); |
1445 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1446 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1447 | if (super->bbm_log) { |
1448 | struct bbm_log *log = super->bbm_log; | |
1449 | ||
1450 | printf("\n"); | |
1451 | printf("Bad Block Management Log:\n"); | |
1452 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1453 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1454 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1455 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1456 | printf(" First Spare : %llx\n", |
1457 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1458 | } |
44470971 DW |
1459 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1460 | struct mdinfo info; | |
1461 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1462 | ||
1463 | super->current_vol = i; | |
a5d85af7 | 1464 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1465 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1466 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1467 | } |
cdddbdbc DW |
1468 | for (i = 0; i < mpb->num_disks; i++) { |
1469 | if (i == super->disks->index) | |
1470 | continue; | |
0ec1f4e8 | 1471 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1472 | } |
94827db3 | 1473 | |
0ec1f4e8 DW |
1474 | for (dl = super->disks; dl; dl = dl->next) |
1475 | if (dl->index == -1) | |
1476 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1477 | |
1478 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1479 | } |
1480 | ||
061f2c6a | 1481 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1482 | { |
27fd6274 | 1483 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1484 | struct mdinfo info; |
1485 | char nbuf[64]; | |
1e7bc0ed | 1486 | struct intel_super *super = st->sb; |
1e7bc0ed | 1487 | |
0d5a423f DW |
1488 | if (!super->anchor->num_raid_devs) { |
1489 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1490 | return; |
0d5a423f | 1491 | } |
ff54de6e | 1492 | |
a5d85af7 | 1493 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1494 | fname_from_uuid(st, &info, nbuf, ':'); |
1495 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1496 | } | |
1497 | ||
1498 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1499 | { | |
1500 | /* We just write a generic IMSM ARRAY entry */ | |
1501 | struct mdinfo info; | |
1502 | char nbuf[64]; | |
1503 | char nbuf1[64]; | |
1504 | struct intel_super *super = st->sb; | |
1505 | int i; | |
1506 | ||
1507 | if (!super->anchor->num_raid_devs) | |
1508 | return; | |
1509 | ||
a5d85af7 | 1510 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1511 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1512 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1513 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1514 | ||
1515 | super->current_vol = i; | |
a5d85af7 | 1516 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1517 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1518 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1519 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1520 | } |
cdddbdbc DW |
1521 | } |
1522 | ||
9d84c8ea DW |
1523 | static void export_examine_super_imsm(struct supertype *st) |
1524 | { | |
1525 | struct intel_super *super = st->sb; | |
1526 | struct imsm_super *mpb = super->anchor; | |
1527 | struct mdinfo info; | |
1528 | char nbuf[64]; | |
1529 | ||
a5d85af7 | 1530 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1531 | fname_from_uuid(st, &info, nbuf, ':'); |
1532 | printf("MD_METADATA=imsm\n"); | |
1533 | printf("MD_LEVEL=container\n"); | |
1534 | printf("MD_UUID=%s\n", nbuf+5); | |
1535 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1536 | } | |
1537 | ||
cdddbdbc DW |
1538 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1539 | { | |
3ebe00a1 DW |
1540 | struct mdinfo info; |
1541 | char nbuf[64]; | |
1542 | ||
a5d85af7 | 1543 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1544 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1545 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1546 | } |
1547 | ||
1548 | static void brief_detail_super_imsm(struct supertype *st) | |
1549 | { | |
ff54de6e N |
1550 | struct mdinfo info; |
1551 | char nbuf[64]; | |
a5d85af7 | 1552 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1553 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1554 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1555 | } |
d665cc31 DW |
1556 | |
1557 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1558 | static void fd2devname(int fd, char *name); | |
1559 | ||
120dc887 | 1560 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1561 | { |
120dc887 LM |
1562 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1563 | * controller, as well as non-connected ports | |
d665cc31 DW |
1564 | */ |
1565 | int hba_len = strlen(hba_path) + 1; | |
1566 | struct dirent *ent; | |
1567 | DIR *dir; | |
1568 | char *path = NULL; | |
1569 | int err = 0; | |
1570 | unsigned long port_mask = (1 << port_count) - 1; | |
1571 | ||
f21e18ca | 1572 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1573 | if (verbose) |
1574 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1575 | return 2; | |
1576 | } | |
1577 | ||
1578 | /* scroll through /sys/dev/block looking for devices attached to | |
1579 | * this hba | |
1580 | */ | |
1581 | dir = opendir("/sys/dev/block"); | |
1582 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1583 | int fd; | |
1584 | char model[64]; | |
1585 | char vendor[64]; | |
1586 | char buf[1024]; | |
1587 | int major, minor; | |
1588 | char *device; | |
1589 | char *c; | |
1590 | int port; | |
1591 | int type; | |
1592 | ||
1593 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1594 | continue; | |
1595 | path = devt_to_devpath(makedev(major, minor)); | |
1596 | if (!path) | |
1597 | continue; | |
1598 | if (!path_attached_to_hba(path, hba_path)) { | |
1599 | free(path); | |
1600 | path = NULL; | |
1601 | continue; | |
1602 | } | |
1603 | ||
1604 | /* retrieve the scsi device type */ | |
1605 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1606 | if (verbose) | |
1607 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1608 | err = 2; | |
1609 | break; | |
1610 | } | |
1611 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1612 | if (load_sys(device, buf) != 0) { | |
1613 | if (verbose) | |
1614 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1615 | path); | |
1616 | err = 2; | |
1617 | free(device); | |
1618 | break; | |
1619 | } | |
1620 | type = strtoul(buf, NULL, 10); | |
1621 | ||
1622 | /* if it's not a disk print the vendor and model */ | |
1623 | if (!(type == 0 || type == 7 || type == 14)) { | |
1624 | vendor[0] = '\0'; | |
1625 | model[0] = '\0'; | |
1626 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1627 | if (load_sys(device, buf) == 0) { | |
1628 | strncpy(vendor, buf, sizeof(vendor)); | |
1629 | vendor[sizeof(vendor) - 1] = '\0'; | |
1630 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1631 | while (isspace(*c) || *c == '\0') | |
1632 | *c-- = '\0'; | |
1633 | ||
1634 | } | |
1635 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1636 | if (load_sys(device, buf) == 0) { | |
1637 | strncpy(model, buf, sizeof(model)); | |
1638 | model[sizeof(model) - 1] = '\0'; | |
1639 | c = (char *) &model[sizeof(model) - 1]; | |
1640 | while (isspace(*c) || *c == '\0') | |
1641 | *c-- = '\0'; | |
1642 | } | |
1643 | ||
1644 | if (vendor[0] && model[0]) | |
1645 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1646 | else | |
1647 | switch (type) { /* numbers from hald/linux/device.c */ | |
1648 | case 1: sprintf(buf, "tape"); break; | |
1649 | case 2: sprintf(buf, "printer"); break; | |
1650 | case 3: sprintf(buf, "processor"); break; | |
1651 | case 4: | |
1652 | case 5: sprintf(buf, "cdrom"); break; | |
1653 | case 6: sprintf(buf, "scanner"); break; | |
1654 | case 8: sprintf(buf, "media_changer"); break; | |
1655 | case 9: sprintf(buf, "comm"); break; | |
1656 | case 12: sprintf(buf, "raid"); break; | |
1657 | default: sprintf(buf, "unknown"); | |
1658 | } | |
1659 | } else | |
1660 | buf[0] = '\0'; | |
1661 | free(device); | |
1662 | ||
1663 | /* chop device path to 'host%d' and calculate the port number */ | |
1664 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1665 | if (!c) { |
1666 | if (verbose) | |
1667 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1668 | err = 2; | |
1669 | break; | |
1670 | } | |
d665cc31 DW |
1671 | *c = '\0'; |
1672 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1673 | port -= host_base; | |
1674 | else { | |
1675 | if (verbose) { | |
1676 | *c = '/'; /* repair the full string */ | |
1677 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1678 | path); | |
1679 | } | |
1680 | err = 2; | |
1681 | break; | |
1682 | } | |
1683 | ||
1684 | /* mark this port as used */ | |
1685 | port_mask &= ~(1 << port); | |
1686 | ||
1687 | /* print out the device information */ | |
1688 | if (buf[0]) { | |
1689 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1690 | continue; | |
1691 | } | |
1692 | ||
1693 | fd = dev_open(ent->d_name, O_RDONLY); | |
1694 | if (fd < 0) | |
1695 | printf(" Port%d : - disk info unavailable -\n", port); | |
1696 | else { | |
1697 | fd2devname(fd, buf); | |
1698 | printf(" Port%d : %s", port, buf); | |
1699 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1700 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1701 | else |
664d5325 | 1702 | printf(" ()\n"); |
4dab422a | 1703 | close(fd); |
d665cc31 | 1704 | } |
d665cc31 DW |
1705 | free(path); |
1706 | path = NULL; | |
1707 | } | |
1708 | if (path) | |
1709 | free(path); | |
1710 | if (dir) | |
1711 | closedir(dir); | |
1712 | if (err == 0) { | |
1713 | int i; | |
1714 | ||
1715 | for (i = 0; i < port_count; i++) | |
1716 | if (port_mask & (1 << i)) | |
1717 | printf(" Port%d : - no device attached -\n", i); | |
1718 | } | |
1719 | ||
1720 | return err; | |
1721 | } | |
1722 | ||
120dc887 LM |
1723 | static void print_found_intel_controllers(struct sys_dev *elem) |
1724 | { | |
1725 | for (; elem; elem = elem->next) { | |
1726 | fprintf(stderr, Name ": found Intel(R) "); | |
1727 | if (elem->type == SYS_DEV_SATA) | |
1728 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1729 | else if (elem->type == SYS_DEV_SAS) |
1730 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1731 | fprintf(stderr, "RAID controller"); |
1732 | if (elem->pci_id) | |
1733 | fprintf(stderr, " at %s", elem->pci_id); | |
1734 | fprintf(stderr, ".\n"); | |
1735 | } | |
1736 | fflush(stderr); | |
1737 | } | |
1738 | ||
120dc887 LM |
1739 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1740 | { | |
1741 | struct dirent *ent; | |
1742 | DIR *dir; | |
1743 | int host_base = -1; | |
1744 | ||
1745 | *port_count = 0; | |
1746 | if ((dir = opendir(hba_path)) == NULL) | |
1747 | return -1; | |
1748 | ||
1749 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1750 | int host; | |
1751 | ||
1752 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1753 | continue; | |
1754 | if (*port_count == 0) | |
1755 | host_base = host; | |
1756 | else if (host < host_base) | |
1757 | host_base = host; | |
1758 | ||
1759 | if (host + 1 > *port_count + host_base) | |
1760 | *port_count = host + 1 - host_base; | |
1761 | } | |
1762 | closedir(dir); | |
1763 | return host_base; | |
1764 | } | |
1765 | ||
a891a3c2 LM |
1766 | static void print_imsm_capability(const struct imsm_orom *orom) |
1767 | { | |
1768 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1769 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1770 | orom->hotfix_ver, orom->build); | |
1771 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1772 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1773 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1774 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1775 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1776 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1777 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1778 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1779 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1780 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1781 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1782 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1783 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1784 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1785 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1786 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1787 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1788 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1789 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1790 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1791 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1792 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1793 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
1794 | printf(" 2TB volumes :%s supported\n", |
1795 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
1796 | printf(" 2TB disks :%s supported\n", | |
1797 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 LM |
1798 | printf(" Max Disks : %d\n", orom->tds); |
1799 | printf(" Max Volumes : %d per array, %d per controller\n", | |
1800 | orom->vpa, orom->vphba); | |
a891a3c2 LM |
1801 | return; |
1802 | } | |
1803 | ||
5615172f | 1804 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1805 | { |
1806 | /* There are two components to imsm platform support, the ahci SATA | |
1807 | * controller and the option-rom. To find the SATA controller we | |
1808 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1809 | * controller with the Intel vendor id is present. This approach | |
1810 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1811 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1812 | * detect platform raid capabilities. The option-rom resides in a | |
1813 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1814 | * platform capabilities. If raid support is disabled in the BIOS the | |
1815 | * option-rom capability structure will not be available. | |
1816 | */ | |
1817 | const struct imsm_orom *orom; | |
1818 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1819 | int host_base = 0; |
1820 | int port_count = 0; | |
120dc887 | 1821 | int result=0; |
d665cc31 | 1822 | |
5615172f | 1823 | if (enumerate_only) { |
a891a3c2 | 1824 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1825 | return 0; |
a891a3c2 LM |
1826 | list = find_intel_devices(); |
1827 | if (!list) | |
1828 | return 2; | |
1829 | for (hba = list; hba; hba = hba->next) { | |
1830 | orom = find_imsm_capability(hba->type); | |
1831 | if (!orom) { | |
1832 | result = 2; | |
1833 | break; | |
1834 | } | |
1835 | } | |
1836 | free_sys_dev(&list); | |
1837 | return result; | |
5615172f DW |
1838 | } |
1839 | ||
155cbb4c LM |
1840 | list = find_intel_devices(); |
1841 | if (!list) { | |
d665cc31 | 1842 | if (verbose) |
155cbb4c LM |
1843 | fprintf(stderr, Name ": no active Intel(R) RAID " |
1844 | "controller found.\n"); | |
d665cc31 DW |
1845 | free_sys_dev(&list); |
1846 | return 2; | |
1847 | } else if (verbose) | |
155cbb4c | 1848 | print_found_intel_controllers(list); |
d665cc31 | 1849 | |
a891a3c2 LM |
1850 | for (hba = list; hba; hba = hba->next) { |
1851 | orom = find_imsm_capability(hba->type); | |
1852 | if (!orom) | |
1853 | fprintf(stderr, Name ": imsm capabilities not found for controller: %s (type %s)\n", | |
1854 | hba->path, get_sys_dev_type(hba->type)); | |
1855 | else | |
1856 | print_imsm_capability(orom); | |
d665cc31 DW |
1857 | } |
1858 | ||
120dc887 LM |
1859 | for (hba = list; hba; hba = hba->next) { |
1860 | printf(" I/O Controller : %s (%s)\n", | |
1861 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1862 | |
120dc887 LM |
1863 | if (hba->type == SYS_DEV_SATA) { |
1864 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1865 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1866 | if (verbose) | |
1867 | fprintf(stderr, Name ": failed to enumerate " | |
1868 | "ports on SATA controller at %s.", hba->pci_id); | |
1869 | result |= 2; | |
1870 | } | |
1871 | } | |
d665cc31 | 1872 | } |
155cbb4c | 1873 | |
120dc887 LM |
1874 | free_sys_dev(&list); |
1875 | return result; | |
d665cc31 | 1876 | } |
cdddbdbc DW |
1877 | #endif |
1878 | ||
1879 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1880 | { | |
5115ca67 DW |
1881 | /* the imsm metadata format does not specify any host |
1882 | * identification information. We return -1 since we can never | |
1883 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1884 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1885 | * exclude member disks that do not belong, and we rely on |
1886 | * mdadm.conf to specify the arrays that should be assembled. | |
1887 | * Auto-assembly may still pick up "foreign" arrays. | |
1888 | */ | |
cdddbdbc | 1889 | |
9362c1c8 | 1890 | return -1; |
cdddbdbc DW |
1891 | } |
1892 | ||
1893 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1894 | { | |
51006d85 N |
1895 | /* The uuid returned here is used for: |
1896 | * uuid to put into bitmap file (Create, Grow) | |
1897 | * uuid for backup header when saving critical section (Grow) | |
1898 | * comparing uuids when re-adding a device into an array | |
1899 | * In these cases the uuid required is that of the data-array, | |
1900 | * not the device-set. | |
1901 | * uuid to recognise same set when adding a missing device back | |
1902 | * to an array. This is a uuid for the device-set. | |
1903 | * | |
1904 | * For each of these we can make do with a truncated | |
1905 | * or hashed uuid rather than the original, as long as | |
1906 | * everyone agrees. | |
1907 | * In each case the uuid required is that of the data-array, | |
1908 | * not the device-set. | |
43dad3d6 | 1909 | */ |
51006d85 N |
1910 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1911 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1912 | * - the orig_family_num of the container |
51006d85 N |
1913 | * - the index number of the volume |
1914 | * - the 'serial' number of the volume. | |
1915 | * Hopefully these are all constant. | |
1916 | */ | |
1917 | struct intel_super *super = st->sb; | |
43dad3d6 | 1918 | |
51006d85 N |
1919 | char buf[20]; |
1920 | struct sha1_ctx ctx; | |
1921 | struct imsm_dev *dev = NULL; | |
148acb7b | 1922 | __u32 family_num; |
51006d85 | 1923 | |
148acb7b DW |
1924 | /* some mdadm versions failed to set ->orig_family_num, in which |
1925 | * case fall back to ->family_num. orig_family_num will be | |
1926 | * fixed up with the first metadata update. | |
1927 | */ | |
1928 | family_num = super->anchor->orig_family_num; | |
1929 | if (family_num == 0) | |
1930 | family_num = super->anchor->family_num; | |
51006d85 | 1931 | sha1_init_ctx(&ctx); |
92bd8f8d | 1932 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1933 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1934 | if (super->current_vol >= 0) |
1935 | dev = get_imsm_dev(super, super->current_vol); | |
1936 | if (dev) { | |
1937 | __u32 vol = super->current_vol; | |
1938 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1939 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1940 | } | |
1941 | sha1_finish_ctx(&ctx, buf); | |
1942 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1943 | } |
1944 | ||
0d481d37 | 1945 | #if 0 |
4f5bc454 DW |
1946 | static void |
1947 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1948 | { |
cdddbdbc DW |
1949 | __u8 *v = get_imsm_version(mpb); |
1950 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1951 | char major[] = { 0, 0, 0 }; | |
1952 | char minor[] = { 0 ,0, 0 }; | |
1953 | char patch[] = { 0, 0, 0 }; | |
1954 | char *ver_parse[] = { major, minor, patch }; | |
1955 | int i, j; | |
1956 | ||
1957 | i = j = 0; | |
1958 | while (*v != '\0' && v < end) { | |
1959 | if (*v != '.' && j < 2) | |
1960 | ver_parse[i][j++] = *v; | |
1961 | else { | |
1962 | i++; | |
1963 | j = 0; | |
1964 | } | |
1965 | v++; | |
1966 | } | |
1967 | ||
4f5bc454 DW |
1968 | *m = strtol(minor, NULL, 0); |
1969 | *p = strtol(patch, NULL, 0); | |
1970 | } | |
0d481d37 | 1971 | #endif |
4f5bc454 | 1972 | |
1e5c6983 DW |
1973 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1974 | { | |
1975 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 1976 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1977 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1978 | ||
1979 | switch (get_imsm_raid_level(map)) { | |
1980 | case 5: | |
1981 | case 10: | |
1982 | return chunk; | |
1983 | default: | |
1984 | return 128*1024 >> 9; | |
1985 | } | |
1986 | } | |
1987 | ||
1988 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1989 | { | |
1990 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1991 | * this is different than migr_strip_size_resync(), but it's good | |
1992 | * to be compatible | |
1993 | */ | |
238c0a71 | 1994 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
1995 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1996 | ||
1997 | switch (get_imsm_raid_level(map)) { | |
1998 | case 1: | |
1999 | case 10: | |
2000 | if (map->num_members % map->num_domains == 0) | |
2001 | return 128*1024 >> 9; | |
2002 | else | |
2003 | return chunk; | |
2004 | case 5: | |
2005 | return max((__u32) 64*1024 >> 9, chunk); | |
2006 | default: | |
2007 | return 128*1024 >> 9; | |
2008 | } | |
2009 | } | |
2010 | ||
2011 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2012 | { | |
238c0a71 AK |
2013 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2014 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2015 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2016 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2017 | ||
2018 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2019 | } | |
2020 | ||
2021 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2022 | { | |
238c0a71 | 2023 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2024 | int level = get_imsm_raid_level(lo); |
2025 | ||
2026 | if (level == 1 || level == 10) { | |
238c0a71 | 2027 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2028 | |
2029 | return hi->num_domains; | |
2030 | } else | |
2031 | return num_stripes_per_unit_resync(dev); | |
2032 | } | |
2033 | ||
98130f40 | 2034 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2035 | { |
2036 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2037 | * counter-intuitively have the same number of data disks | |
2038 | */ | |
98130f40 | 2039 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2040 | |
2041 | switch (get_imsm_raid_level(map)) { | |
2042 | case 0: | |
36fd8ccc AK |
2043 | return map->num_members; |
2044 | break; | |
1e5c6983 DW |
2045 | case 1: |
2046 | case 10: | |
36fd8ccc | 2047 | return map->num_members/2; |
1e5c6983 DW |
2048 | case 5: |
2049 | return map->num_members - 1; | |
2050 | default: | |
2051 | dprintf("%s: unsupported raid level\n", __func__); | |
2052 | return 0; | |
2053 | } | |
2054 | } | |
2055 | ||
2056 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2057 | { | |
238c0a71 | 2058 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2059 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2060 | ||
2061 | switch(get_imsm_raid_level(map)) { | |
2062 | case 1: | |
2063 | case 10: | |
2064 | return chunk * map->num_domains; | |
2065 | case 5: | |
2066 | return chunk * map->num_members; | |
2067 | default: | |
2068 | return chunk; | |
2069 | } | |
2070 | } | |
2071 | ||
2072 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2073 | { | |
238c0a71 | 2074 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2075 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2076 | __u32 strip = block / chunk; | |
2077 | ||
2078 | switch (get_imsm_raid_level(map)) { | |
2079 | case 1: | |
2080 | case 10: { | |
2081 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2082 | __u32 vol_stripe = vol_strip / map->num_members; | |
2083 | ||
2084 | return vol_stripe * chunk + block % chunk; | |
2085 | } case 5: { | |
2086 | __u32 stripe = strip / (map->num_members - 1); | |
2087 | ||
2088 | return stripe * chunk + block % chunk; | |
2089 | } | |
2090 | default: | |
2091 | return 0; | |
2092 | } | |
2093 | } | |
2094 | ||
c47b0ff6 AK |
2095 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2096 | struct imsm_dev *dev) | |
1e5c6983 DW |
2097 | { |
2098 | /* calculate the conversion factor between per member 'blocks' | |
2099 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2100 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2101 | */ | |
2102 | if (!dev->vol.migr_state) | |
2103 | return 0; | |
2104 | ||
2105 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2106 | case MIGR_GEN_MIGR: { |
2107 | struct migr_record *migr_rec = super->migr_rec; | |
2108 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2109 | } | |
1e5c6983 DW |
2110 | case MIGR_VERIFY: |
2111 | case MIGR_REPAIR: | |
2112 | case MIGR_INIT: { | |
238c0a71 | 2113 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2114 | __u32 stripes_per_unit; |
2115 | __u32 blocks_per_unit; | |
2116 | __u32 parity_depth; | |
2117 | __u32 migr_chunk; | |
2118 | __u32 block_map; | |
2119 | __u32 block_rel; | |
2120 | __u32 segment; | |
2121 | __u32 stripe; | |
2122 | __u8 disks; | |
2123 | ||
2124 | /* yes, this is really the translation of migr_units to | |
2125 | * per-member blocks in the 'resync' case | |
2126 | */ | |
2127 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2128 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2129 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2130 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2131 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2132 | segment = blocks_per_unit / stripe; |
2133 | block_rel = blocks_per_unit - segment * stripe; | |
2134 | parity_depth = parity_segment_depth(dev); | |
2135 | block_map = map_migr_block(dev, block_rel); | |
2136 | return block_map + parity_depth * segment; | |
2137 | } | |
2138 | case MIGR_REBUILD: { | |
2139 | __u32 stripes_per_unit; | |
2140 | __u32 migr_chunk; | |
2141 | ||
2142 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2143 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2144 | return migr_chunk * stripes_per_unit; | |
2145 | } | |
1e5c6983 DW |
2146 | case MIGR_STATE_CHANGE: |
2147 | default: | |
2148 | return 0; | |
2149 | } | |
2150 | } | |
2151 | ||
c2c087e6 DW |
2152 | static int imsm_level_to_layout(int level) |
2153 | { | |
2154 | switch (level) { | |
2155 | case 0: | |
2156 | case 1: | |
2157 | return 0; | |
2158 | case 5: | |
2159 | case 6: | |
a380c027 | 2160 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2161 | case 10: |
c92a2527 | 2162 | return 0x102; |
c2c087e6 | 2163 | } |
a18a888e | 2164 | return UnSet; |
c2c087e6 DW |
2165 | } |
2166 | ||
8e59f3d8 AK |
2167 | /******************************************************************************* |
2168 | * Function: read_imsm_migr_rec | |
2169 | * Description: Function reads imsm migration record from last sector of disk | |
2170 | * Parameters: | |
2171 | * fd : disk descriptor | |
2172 | * super : metadata info | |
2173 | * Returns: | |
2174 | * 0 : success, | |
2175 | * -1 : fail | |
2176 | ******************************************************************************/ | |
2177 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2178 | { | |
2179 | int ret_val = -1; | |
2180 | unsigned long long dsize; | |
2181 | ||
2182 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2183 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
8e59f3d8 AK |
2184 | fprintf(stderr, |
2185 | Name ": Cannot seek to anchor block: %s\n", | |
2186 | strerror(errno)); | |
2187 | goto out; | |
2188 | } | |
17a4eaf9 AK |
2189 | if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2190 | MIGR_REC_BUF_SIZE) { | |
8e59f3d8 AK |
2191 | fprintf(stderr, |
2192 | Name ": Cannot read migr record block: %s\n", | |
2193 | strerror(errno)); | |
2194 | goto out; | |
2195 | } | |
2196 | ret_val = 0; | |
2197 | ||
2198 | out: | |
2199 | return ret_val; | |
2200 | } | |
2201 | ||
3136abe5 AK |
2202 | static struct imsm_dev *imsm_get_device_during_migration( |
2203 | struct intel_super *super) | |
2204 | { | |
2205 | ||
2206 | struct intel_dev *dv; | |
2207 | ||
2208 | for (dv = super->devlist; dv; dv = dv->next) { | |
2209 | if (is_gen_migration(dv->dev)) | |
2210 | return dv->dev; | |
2211 | } | |
2212 | return NULL; | |
2213 | } | |
2214 | ||
8e59f3d8 AK |
2215 | /******************************************************************************* |
2216 | * Function: load_imsm_migr_rec | |
2217 | * Description: Function reads imsm migration record (it is stored at the last | |
2218 | * sector of disk) | |
2219 | * Parameters: | |
2220 | * super : imsm internal array info | |
2221 | * info : general array info | |
2222 | * Returns: | |
2223 | * 0 : success | |
2224 | * -1 : fail | |
4c965cc9 | 2225 | * -2 : no migration in progress |
8e59f3d8 AK |
2226 | ******************************************************************************/ |
2227 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2228 | { | |
2229 | struct mdinfo *sd; | |
2230 | struct dl *dl = NULL; | |
2231 | char nm[30]; | |
2232 | int retval = -1; | |
2233 | int fd = -1; | |
3136abe5 AK |
2234 | struct imsm_dev *dev; |
2235 | struct imsm_map *map = NULL; | |
b4ab44d8 | 2236 | int slot = -1; |
3136abe5 AK |
2237 | |
2238 | /* find map under migration */ | |
2239 | dev = imsm_get_device_during_migration(super); | |
2240 | /* nothing to load,no migration in progress? | |
2241 | */ | |
2242 | if (dev == NULL) | |
4c965cc9 | 2243 | return -2; |
44bfe6df | 2244 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 AK |
2245 | |
2246 | if (info) { | |
2247 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3136abe5 AK |
2248 | /* skip spare and failed disks |
2249 | */ | |
2250 | if (sd->disk.raid_disk < 0) | |
2251 | continue; | |
8e59f3d8 | 2252 | /* read only from one of the first two slots */ |
3136abe5 AK |
2253 | if (map) |
2254 | slot = get_imsm_disk_slot(map, | |
2255 | sd->disk.raid_disk); | |
2256 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 | 2257 | continue; |
3136abe5 | 2258 | |
8e59f3d8 AK |
2259 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2260 | fd = dev_open(nm, O_RDONLY); | |
2261 | if (fd >= 0) | |
2262 | break; | |
2263 | } | |
2264 | } | |
2265 | if (fd < 0) { | |
2266 | for (dl = super->disks; dl; dl = dl->next) { | |
3136abe5 AK |
2267 | /* skip spare and failed disks |
2268 | */ | |
2269 | if (dl->index < 0) | |
2270 | continue; | |
8e59f3d8 | 2271 | /* read only from one of the first two slots */ |
3136abe5 AK |
2272 | if (map) |
2273 | slot = get_imsm_disk_slot(map, dl->index); | |
2274 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 AK |
2275 | continue; |
2276 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2277 | fd = dev_open(nm, O_RDONLY); | |
2278 | if (fd >= 0) | |
2279 | break; | |
2280 | } | |
2281 | } | |
2282 | if (fd < 0) | |
2283 | goto out; | |
2284 | retval = read_imsm_migr_rec(fd, super); | |
2285 | ||
2286 | out: | |
2287 | if (fd >= 0) | |
2288 | close(fd); | |
2289 | return retval; | |
2290 | } | |
2291 | ||
9e2d750d | 2292 | #ifndef MDASSEMBLE |
c17608ea AK |
2293 | /******************************************************************************* |
2294 | * function: imsm_create_metadata_checkpoint_update | |
2295 | * Description: It creates update for checkpoint change. | |
2296 | * Parameters: | |
2297 | * super : imsm internal array info | |
2298 | * u : pointer to prepared update | |
2299 | * Returns: | |
2300 | * Uptate length. | |
2301 | * If length is equal to 0, input pointer u contains no update | |
2302 | ******************************************************************************/ | |
2303 | static int imsm_create_metadata_checkpoint_update( | |
2304 | struct intel_super *super, | |
2305 | struct imsm_update_general_migration_checkpoint **u) | |
2306 | { | |
2307 | ||
2308 | int update_memory_size = 0; | |
2309 | ||
2310 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2311 | ||
2312 | if (u == NULL) | |
2313 | return 0; | |
2314 | *u = NULL; | |
2315 | ||
2316 | /* size of all update data without anchor */ | |
2317 | update_memory_size = | |
2318 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2319 | ||
2320 | *u = calloc(1, update_memory_size); | |
2321 | if (*u == NULL) { | |
2322 | dprintf("error: cannot get memory for " | |
2323 | "imsm_create_metadata_checkpoint_update update\n"); | |
2324 | return 0; | |
2325 | } | |
2326 | (*u)->type = update_general_migration_checkpoint; | |
2327 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2328 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2329 | (*u)->curr_migr_unit); | |
2330 | ||
2331 | return update_memory_size; | |
2332 | } | |
2333 | ||
2334 | ||
2335 | static void imsm_update_metadata_locally(struct supertype *st, | |
2336 | void *buf, int len); | |
2337 | ||
687629c2 AK |
2338 | /******************************************************************************* |
2339 | * Function: write_imsm_migr_rec | |
2340 | * Description: Function writes imsm migration record | |
2341 | * (at the last sector of disk) | |
2342 | * Parameters: | |
2343 | * super : imsm internal array info | |
2344 | * Returns: | |
2345 | * 0 : success | |
2346 | * -1 : if fail | |
2347 | ******************************************************************************/ | |
2348 | static int write_imsm_migr_rec(struct supertype *st) | |
2349 | { | |
2350 | struct intel_super *super = st->sb; | |
2351 | unsigned long long dsize; | |
2352 | char nm[30]; | |
2353 | int fd = -1; | |
2354 | int retval = -1; | |
2355 | struct dl *sd; | |
c17608ea AK |
2356 | int len; |
2357 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 AK |
2358 | struct imsm_dev *dev; |
2359 | struct imsm_map *map = NULL; | |
2360 | ||
2361 | /* find map under migration */ | |
2362 | dev = imsm_get_device_during_migration(super); | |
2363 | /* if no migration, write buffer anyway to clear migr_record | |
2364 | * on disk based on first available device | |
2365 | */ | |
2366 | if (dev == NULL) | |
2367 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2368 | super->current_vol); | |
2369 | ||
44bfe6df | 2370 | map = get_imsm_map(dev, MAP_0); |
687629c2 AK |
2371 | |
2372 | for (sd = super->disks ; sd ; sd = sd->next) { | |
b4ab44d8 | 2373 | int slot = -1; |
3136abe5 AK |
2374 | |
2375 | /* skip failed and spare devices */ | |
2376 | if (sd->index < 0) | |
2377 | continue; | |
687629c2 | 2378 | /* write to 2 first slots only */ |
3136abe5 AK |
2379 | if (map) |
2380 | slot = get_imsm_disk_slot(map, sd->index); | |
2381 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
687629c2 | 2382 | continue; |
3136abe5 | 2383 | |
687629c2 AK |
2384 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
2385 | fd = dev_open(nm, O_RDWR); | |
2386 | if (fd < 0) | |
2387 | continue; | |
2388 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2389 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
687629c2 AK |
2390 | fprintf(stderr, |
2391 | Name ": Cannot seek to anchor block: %s\n", | |
2392 | strerror(errno)); | |
2393 | goto out; | |
2394 | } | |
17a4eaf9 AK |
2395 | if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2396 | MIGR_REC_BUF_SIZE) { | |
687629c2 AK |
2397 | fprintf(stderr, |
2398 | Name ": Cannot write migr record block: %s\n", | |
2399 | strerror(errno)); | |
2400 | goto out; | |
2401 | } | |
2402 | close(fd); | |
2403 | fd = -1; | |
2404 | } | |
c17608ea AK |
2405 | /* update checkpoint information in metadata */ |
2406 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2407 | ||
2408 | if (len <= 0) { | |
2409 | dprintf("imsm: Cannot prepare update\n"); | |
2410 | goto out; | |
2411 | } | |
2412 | /* update metadata locally */ | |
2413 | imsm_update_metadata_locally(st, u, len); | |
2414 | /* and possibly remotely */ | |
2415 | if (st->update_tail) { | |
2416 | append_metadata_update(st, u, len); | |
2417 | /* during reshape we do all work inside metadata handler | |
2418 | * manage_reshape(), so metadata update has to be triggered | |
2419 | * insida it | |
2420 | */ | |
2421 | flush_metadata_updates(st); | |
2422 | st->update_tail = &st->updates; | |
2423 | } else | |
2424 | free(u); | |
687629c2 AK |
2425 | |
2426 | retval = 0; | |
2427 | out: | |
2428 | if (fd >= 0) | |
2429 | close(fd); | |
2430 | return retval; | |
2431 | } | |
9e2d750d | 2432 | #endif /* MDASSEMBLE */ |
687629c2 | 2433 | |
e2962bfc AK |
2434 | /* spare/missing disks activations are not allowe when |
2435 | * array/container performs reshape operation, because | |
2436 | * all arrays in container works on the same disks set | |
2437 | */ | |
2438 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2439 | { | |
2440 | int rv = 0; | |
2441 | struct intel_dev *i_dev; | |
2442 | struct imsm_dev *dev; | |
2443 | ||
2444 | /* check whole container | |
2445 | */ | |
2446 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2447 | dev = i_dev->dev; | |
3ad25638 | 2448 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2449 | /* No repair during any migration in container |
2450 | */ | |
2451 | rv = 1; | |
2452 | break; | |
2453 | } | |
2454 | } | |
2455 | return rv; | |
2456 | } | |
2457 | ||
a5d85af7 | 2458 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2459 | { |
2460 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2461 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2462 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2463 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2464 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2465 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2466 | struct dl *dl; |
e207da2f | 2467 | char *devname; |
139dae11 | 2468 | unsigned int component_size_alligment; |
a5d85af7 | 2469 | int map_disks = info->array.raid_disks; |
bf5a934a | 2470 | |
95eeceeb | 2471 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2472 | if (prev_map) |
2473 | map_to_analyse = prev_map; | |
2474 | ||
ca0748fa | 2475 | dl = super->current_disk; |
9894ec0d | 2476 | |
bf5a934a | 2477 | info->container_member = super->current_vol; |
cd0430a1 | 2478 | info->array.raid_disks = map->num_members; |
b335e593 | 2479 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2480 | info->array.layout = imsm_level_to_layout(info->array.level); |
2481 | info->array.md_minor = -1; | |
2482 | info->array.ctime = 0; | |
2483 | info->array.utime = 0; | |
b335e593 AK |
2484 | info->array.chunk_size = |
2485 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2486 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2487 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2488 | info->custom_array_size <<= 32; | |
2489 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2490 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2491 | ||
3f510843 | 2492 | if (is_gen_migration(dev)) { |
3f83228a | 2493 | info->reshape_active = 1; |
b335e593 AK |
2494 | info->new_level = get_imsm_raid_level(map); |
2495 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2496 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2497 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2498 | if (info->delta_disks) { |
2499 | /* this needs to be applied to every array | |
2500 | * in the container. | |
2501 | */ | |
81219e70 | 2502 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2503 | } |
3f83228a N |
2504 | /* We shape information that we give to md might have to be |
2505 | * modify to cope with md's requirement for reshaping arrays. | |
2506 | * For example, when reshaping a RAID0, md requires it to be | |
2507 | * presented as a degraded RAID4. | |
2508 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2509 | * the array as already being RAID5, but the 'before' layout | |
2510 | * is a RAID4-like layout. | |
2511 | */ | |
2512 | switch (info->array.level) { | |
2513 | case 0: | |
2514 | switch(info->new_level) { | |
2515 | case 0: | |
2516 | /* conversion is happening as RAID4 */ | |
2517 | info->array.level = 4; | |
2518 | info->array.raid_disks += 1; | |
2519 | break; | |
2520 | case 5: | |
2521 | /* conversion is happening as RAID5 */ | |
2522 | info->array.level = 5; | |
2523 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2524 | info->delta_disks -= 1; |
2525 | break; | |
2526 | default: | |
2527 | /* FIXME error message */ | |
2528 | info->array.level = UnSet; | |
2529 | break; | |
2530 | } | |
2531 | break; | |
2532 | } | |
b335e593 AK |
2533 | } else { |
2534 | info->new_level = UnSet; | |
2535 | info->new_layout = UnSet; | |
2536 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2537 | info->delta_disks = 0; |
b335e593 | 2538 | } |
ca0748fa | 2539 | |
efb30e7f DW |
2540 | if (dl) { |
2541 | info->disk.major = dl->major; | |
2542 | info->disk.minor = dl->minor; | |
ca0748fa | 2543 | info->disk.number = dl->index; |
656b6b5a N |
2544 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2545 | dl->index); | |
efb30e7f | 2546 | } |
bf5a934a | 2547 | |
5551b113 CA |
2548 | info->data_offset = pba_of_lba0(map_to_analyse); |
2549 | info->component_size = blocks_per_member(map_to_analyse); | |
139dae11 AK |
2550 | |
2551 | /* check component size aligment | |
2552 | */ | |
2553 | component_size_alligment = | |
2554 | info->component_size % (info->array.chunk_size/512); | |
2555 | ||
2556 | if (component_size_alligment && | |
2557 | (info->array.level != 1) && (info->array.level != UnSet)) { | |
2558 | dprintf("imsm: reported component size alligned from %llu ", | |
2559 | info->component_size); | |
2560 | info->component_size -= component_size_alligment; | |
2561 | dprintf("to %llu (%i).\n", | |
2562 | info->component_size, component_size_alligment); | |
2563 | } | |
2564 | ||
301406c9 | 2565 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2566 | info->recovery_start = MaxSector; |
bf5a934a | 2567 | |
d2e6d5d6 | 2568 | info->reshape_progress = 0; |
b6796ce1 | 2569 | info->resync_start = MaxSector; |
b9172665 AK |
2570 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2571 | dev->vol.dirty) && | |
2572 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2573 | info->resync_start = 0; |
b6796ce1 AK |
2574 | } |
2575 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2576 | switch (migr_type(dev)) { |
2577 | case MIGR_REPAIR: | |
2578 | case MIGR_INIT: { | |
c47b0ff6 AK |
2579 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2580 | dev); | |
1e5c6983 DW |
2581 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2582 | ||
2583 | info->resync_start = blocks_per_unit * units; | |
2584 | break; | |
2585 | } | |
d2e6d5d6 | 2586 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2587 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2588 | dev); | |
2589 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2590 | unsigned long long array_blocks; |
2591 | int used_disks; | |
d2e6d5d6 | 2592 | |
befb629b AK |
2593 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2594 | (units < | |
2595 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2596 | (super->migr_rec->rec_status == | |
2597 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2598 | units++; | |
2599 | ||
d2e6d5d6 | 2600 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2601 | |
d2e6d5d6 AK |
2602 | dprintf("IMSM: General Migration checkpoint : %llu " |
2603 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2604 | (unsigned long long)units, |
2605 | (unsigned long long)blocks_per_unit, | |
2606 | info->reshape_progress); | |
75156c46 | 2607 | |
238c0a71 | 2608 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 2609 | if (used_disks > 0) { |
5551b113 | 2610 | array_blocks = blocks_per_member(map) * |
75156c46 AK |
2611 | used_disks; |
2612 | /* round array size down to closest MB | |
2613 | */ | |
2614 | info->custom_array_size = (array_blocks | |
2615 | >> SECT_PER_MB_SHIFT) | |
2616 | << SECT_PER_MB_SHIFT; | |
2617 | } | |
d2e6d5d6 | 2618 | } |
1e5c6983 DW |
2619 | case MIGR_VERIFY: |
2620 | /* we could emulate the checkpointing of | |
2621 | * 'sync_action=check' migrations, but for now | |
2622 | * we just immediately complete them | |
2623 | */ | |
2624 | case MIGR_REBUILD: | |
2625 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2626 | case MIGR_STATE_CHANGE: |
2627 | /* FIXME handle other migrations */ | |
2628 | default: | |
2629 | /* we are not dirty, so... */ | |
2630 | info->resync_start = MaxSector; | |
2631 | } | |
b6796ce1 | 2632 | } |
301406c9 DW |
2633 | |
2634 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2635 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2636 | |
f35f2525 N |
2637 | info->array.major_version = -1; |
2638 | info->array.minor_version = -2; | |
e207da2f AW |
2639 | devname = devnum2devname(st->container_dev); |
2640 | *info->text_version = '\0'; | |
2641 | if (devname) | |
2642 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2643 | free(devname); | |
a67dd8cc | 2644 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2645 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2646 | |
2647 | if (dmap) { | |
2648 | int i, j; | |
2649 | for (i=0; i<map_disks; i++) { | |
2650 | dmap[i] = 0; | |
2651 | if (i < info->array.raid_disks) { | |
2652 | struct imsm_disk *dsk; | |
238c0a71 | 2653 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2654 | dsk = get_imsm_disk(super, j); |
2655 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2656 | dmap[i] = 1; | |
2657 | } | |
2658 | } | |
2659 | } | |
81ac8b4d | 2660 | } |
bf5a934a | 2661 | |
3b451610 AK |
2662 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2663 | int failed, int look_in_map); | |
2664 | ||
2665 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2666 | int look_in_map); | |
2667 | ||
b4ab44d8 AK |
2668 | |
2669 | #ifndef MDASSEMBLE | |
3b451610 AK |
2670 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
2671 | { | |
2672 | if (is_gen_migration(dev)) { | |
2673 | int failed; | |
2674 | __u8 map_state; | |
2675 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2676 | ||
2677 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2678 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2679 | if (map2->map_state != map_state) { |
2680 | map2->map_state = map_state; | |
2681 | super->updates_pending++; | |
2682 | } | |
2683 | } | |
2684 | } | |
b4ab44d8 | 2685 | #endif |
97b4d0e9 DW |
2686 | |
2687 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2688 | { | |
2689 | struct dl *d; | |
2690 | ||
2691 | for (d = super->missing; d; d = d->next) | |
2692 | if (d->index == index) | |
2693 | return &d->disk; | |
2694 | return NULL; | |
2695 | } | |
2696 | ||
a5d85af7 | 2697 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2698 | { |
2699 | struct intel_super *super = st->sb; | |
4f5bc454 | 2700 | struct imsm_disk *disk; |
a5d85af7 | 2701 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2702 | int max_enough = -1; |
2703 | int i; | |
2704 | struct imsm_super *mpb; | |
4f5bc454 | 2705 | |
bf5a934a | 2706 | if (super->current_vol >= 0) { |
a5d85af7 | 2707 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2708 | return; |
2709 | } | |
95eeceeb | 2710 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2711 | |
2712 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2713 | * spares | |
2714 | */ | |
2715 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2716 | info->array.level = LEVEL_CONTAINER; |
2717 | info->array.layout = 0; | |
2718 | info->array.md_minor = -1; | |
c2c087e6 | 2719 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2720 | info->array.utime = 0; |
2721 | info->array.chunk_size = 0; | |
2722 | ||
2723 | info->disk.major = 0; | |
2724 | info->disk.minor = 0; | |
cdddbdbc | 2725 | info->disk.raid_disk = -1; |
c2c087e6 | 2726 | info->reshape_active = 0; |
f35f2525 N |
2727 | info->array.major_version = -1; |
2728 | info->array.minor_version = -2; | |
c2c087e6 | 2729 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2730 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2731 | info->disk.number = -1; |
2732 | info->disk.state = 0; | |
c5afc314 | 2733 | info->name[0] = 0; |
921d9e16 | 2734 | info->recovery_start = MaxSector; |
3ad25638 | 2735 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2736 | |
97b4d0e9 | 2737 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2738 | mpb = super->anchor; |
97b4d0e9 | 2739 | |
ab3cb6b3 N |
2740 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2741 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2742 | int failed, enough, j, missing = 0; | |
2743 | struct imsm_map *map; | |
2744 | __u8 state; | |
97b4d0e9 | 2745 | |
3b451610 AK |
2746 | failed = imsm_count_failed(super, dev, MAP_0); |
2747 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2748 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2749 | |
2750 | /* any newly missing disks? | |
2751 | * (catches single-degraded vs double-degraded) | |
2752 | */ | |
2753 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2754 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2755 | __u32 idx = ord_to_idx(ord); |
2756 | ||
2757 | if (!(ord & IMSM_ORD_REBUILD) && | |
2758 | get_imsm_missing(super, idx)) { | |
2759 | missing = 1; | |
2760 | break; | |
2761 | } | |
97b4d0e9 | 2762 | } |
ab3cb6b3 N |
2763 | |
2764 | if (state == IMSM_T_STATE_FAILED) | |
2765 | enough = -1; | |
2766 | else if (state == IMSM_T_STATE_DEGRADED && | |
2767 | (state != map->map_state || missing)) | |
2768 | enough = 0; | |
2769 | else /* we're normal, or already degraded */ | |
2770 | enough = 1; | |
d2bde6d3 AK |
2771 | if (is_gen_migration(dev) && missing) { |
2772 | /* during general migration we need all disks | |
2773 | * that process is running on. | |
2774 | * No new missing disk is allowed. | |
2775 | */ | |
2776 | max_enough = -1; | |
2777 | enough = -1; | |
2778 | /* no more checks necessary | |
2779 | */ | |
2780 | break; | |
2781 | } | |
ab3cb6b3 N |
2782 | /* in the missing/failed disk case check to see |
2783 | * if at least one array is runnable | |
2784 | */ | |
2785 | max_enough = max(max_enough, enough); | |
2786 | } | |
2787 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2788 | info->container_enough = max_enough; | |
97b4d0e9 | 2789 | |
4a04ec6c | 2790 | if (super->disks) { |
14e8215b DW |
2791 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2792 | ||
b9f594fe | 2793 | disk = &super->disks->disk; |
5551b113 | 2794 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 2795 | info->component_size = reserved; |
25ed7e59 | 2796 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2797 | /* we don't change info->disk.raid_disk here because |
2798 | * this state will be finalized in mdmon after we have | |
2799 | * found the 'most fresh' version of the metadata | |
2800 | */ | |
25ed7e59 DW |
2801 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2802 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2803 | } |
a575e2a7 DW |
2804 | |
2805 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2806 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2807 | */ | |
2808 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2809 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2810 | else |
2811 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2812 | |
2813 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2814 | if (map) { | |
2815 | int i; | |
2816 | for (i = 0; i < map_disks; i++) | |
2817 | map[i] = 1; | |
2818 | } | |
2819 | ||
cdddbdbc DW |
2820 | } |
2821 | ||
5c4cd5da AC |
2822 | /* allocates memory and fills disk in mdinfo structure |
2823 | * for each disk in array */ | |
2824 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2825 | { | |
2826 | struct mdinfo *mddev = NULL; | |
2827 | struct intel_super *super = st->sb; | |
2828 | struct imsm_disk *disk; | |
2829 | int count = 0; | |
2830 | struct dl *dl; | |
2831 | if (!super || !super->disks) | |
2832 | return NULL; | |
2833 | dl = super->disks; | |
2834 | mddev = malloc(sizeof(*mddev)); | |
2835 | if (!mddev) { | |
2836 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2837 | return NULL; | |
2838 | } | |
2839 | memset(mddev, 0, sizeof(*mddev)); | |
2840 | while (dl) { | |
2841 | struct mdinfo *tmp; | |
2842 | disk = &dl->disk; | |
2843 | tmp = malloc(sizeof(*tmp)); | |
2844 | if (!tmp) { | |
2845 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2846 | if (mddev) | |
2847 | sysfs_free(mddev); | |
2848 | return NULL; | |
2849 | } | |
2850 | memset(tmp, 0, sizeof(*tmp)); | |
2851 | if (mddev->devs) | |
2852 | tmp->next = mddev->devs; | |
2853 | mddev->devs = tmp; | |
2854 | tmp->disk.number = count++; | |
2855 | tmp->disk.major = dl->major; | |
2856 | tmp->disk.minor = dl->minor; | |
2857 | tmp->disk.state = is_configured(disk) ? | |
2858 | (1 << MD_DISK_ACTIVE) : 0; | |
2859 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2860 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2861 | tmp->disk.raid_disk = -1; | |
2862 | dl = dl->next; | |
2863 | } | |
2864 | return mddev; | |
2865 | } | |
2866 | ||
cdddbdbc DW |
2867 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2868 | char *update, char *devname, int verbose, | |
2869 | int uuid_set, char *homehost) | |
2870 | { | |
f352c545 DW |
2871 | /* For 'assemble' and 'force' we need to return non-zero if any |
2872 | * change was made. For others, the return value is ignored. | |
2873 | * Update options are: | |
2874 | * force-one : This device looks a bit old but needs to be included, | |
2875 | * update age info appropriately. | |
2876 | * assemble: clear any 'faulty' flag to allow this device to | |
2877 | * be assembled. | |
2878 | * force-array: Array is degraded but being forced, mark it clean | |
2879 | * if that will be needed to assemble it. | |
2880 | * | |
2881 | * newdev: not used ???? | |
2882 | * grow: Array has gained a new device - this is currently for | |
2883 | * linear only | |
2884 | * resync: mark as dirty so a resync will happen. | |
2885 | * name: update the name - preserving the homehost | |
6e46bf34 | 2886 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2887 | * |
2888 | * Following are not relevant for this imsm: | |
2889 | * sparc2.2 : update from old dodgey metadata | |
2890 | * super-minor: change the preferred_minor number | |
2891 | * summaries: update redundant counters. | |
f352c545 DW |
2892 | * homehost: update the recorded homehost |
2893 | * _reshape_progress: record new reshape_progress position. | |
2894 | */ | |
6e46bf34 DW |
2895 | int rv = 1; |
2896 | struct intel_super *super = st->sb; | |
2897 | struct imsm_super *mpb; | |
f352c545 | 2898 | |
6e46bf34 DW |
2899 | /* we can only update container info */ |
2900 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2901 | return 1; | |
2902 | ||
2903 | mpb = super->anchor; | |
2904 | ||
81a5b4f5 N |
2905 | if (strcmp(update, "uuid") == 0) { |
2906 | /* We take this to mean that the family_num should be updated. | |
2907 | * However that is much smaller than the uuid so we cannot really | |
2908 | * allow an explicit uuid to be given. And it is hard to reliably | |
2909 | * know if one was. | |
2910 | * So if !uuid_set we know the current uuid is random and just used | |
2911 | * the first 'int' and copy it to the other 3 positions. | |
2912 | * Otherwise we require the 4 'int's to be the same as would be the | |
2913 | * case if we are using a random uuid. So an explicit uuid will be | |
2914 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 2915 | */ |
81a5b4f5 N |
2916 | if (!uuid_set) { |
2917 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 2918 | rv = 0; |
81a5b4f5 N |
2919 | } else { |
2920 | if (info->uuid[0] != info->uuid[1] || | |
2921 | info->uuid[1] != info->uuid[2] || | |
2922 | info->uuid[2] != info->uuid[3]) | |
2923 | rv = -1; | |
2924 | else | |
2925 | rv = 0; | |
6e46bf34 | 2926 | } |
81a5b4f5 N |
2927 | if (rv == 0) |
2928 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
2929 | } else if (strcmp(update, "assemble") == 0) |
2930 | rv = 0; | |
2931 | else | |
1e2b2765 | 2932 | rv = -1; |
f352c545 | 2933 | |
6e46bf34 DW |
2934 | /* successful update? recompute checksum */ |
2935 | if (rv == 0) | |
2936 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2937 | |
2938 | return rv; | |
cdddbdbc DW |
2939 | } |
2940 | ||
c2c087e6 | 2941 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2942 | { |
c2c087e6 | 2943 | size_t size; |
cdddbdbc | 2944 | |
c2c087e6 DW |
2945 | size = sizeof(struct imsm_super); |
2946 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2947 | size += 2 * sizeof(struct imsm_dev); | |
2948 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2949 | size += (4 - 2) * sizeof(struct imsm_map); | |
2950 | /* 4 possible disk_ord_tbl's */ | |
2951 | size += 4 * (disks - 1) * sizeof(__u32); | |
2952 | ||
2953 | return size; | |
2954 | } | |
2955 | ||
2956 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
2957 | { | |
2958 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
2959 | return 0; | |
2960 | ||
2961 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
2962 | } |
2963 | ||
ba2de7ba DW |
2964 | static void free_devlist(struct intel_super *super) |
2965 | { | |
2966 | struct intel_dev *dv; | |
2967 | ||
2968 | while (super->devlist) { | |
2969 | dv = super->devlist->next; | |
2970 | free(super->devlist->dev); | |
2971 | free(super->devlist); | |
2972 | super->devlist = dv; | |
2973 | } | |
2974 | } | |
2975 | ||
2976 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
2977 | { | |
2978 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
2979 | } | |
2980 | ||
cdddbdbc DW |
2981 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
2982 | { | |
2983 | /* | |
2984 | * return: | |
2985 | * 0 same, or first was empty, and second was copied | |
2986 | * 1 second had wrong number | |
2987 | * 2 wrong uuid | |
2988 | * 3 wrong other info | |
2989 | */ | |
2990 | struct intel_super *first = st->sb; | |
2991 | struct intel_super *sec = tst->sb; | |
2992 | ||
2993 | if (!first) { | |
2994 | st->sb = tst->sb; | |
2995 | tst->sb = NULL; | |
2996 | return 0; | |
2997 | } | |
8603ea6f LM |
2998 | /* in platform dependent environment test if the disks |
2999 | * use the same Intel hba | |
3000 | */ | |
3001 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
3002 | if (!first->hba || !sec->hba || |
3003 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
3004 | fprintf(stderr, |
3005 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
3006 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
3007 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
3008 | return 3; |
3009 | } | |
3010 | } | |
cdddbdbc | 3011 | |
d23fe947 DW |
3012 | /* if an anchor does not have num_raid_devs set then it is a free |
3013 | * floating spare | |
3014 | */ | |
3015 | if (first->anchor->num_raid_devs > 0 && | |
3016 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3017 | /* Determine if these disks might ever have been |
3018 | * related. Further disambiguation can only take place | |
3019 | * in load_super_imsm_all | |
3020 | */ | |
3021 | __u32 first_family = first->anchor->orig_family_num; | |
3022 | __u32 sec_family = sec->anchor->orig_family_num; | |
3023 | ||
f796af5d DW |
3024 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3025 | MAX_SIGNATURE_LENGTH) != 0) | |
3026 | return 3; | |
3027 | ||
a2b97981 DW |
3028 | if (first_family == 0) |
3029 | first_family = first->anchor->family_num; | |
3030 | if (sec_family == 0) | |
3031 | sec_family = sec->anchor->family_num; | |
3032 | ||
3033 | if (first_family != sec_family) | |
d23fe947 | 3034 | return 3; |
f796af5d | 3035 | |
d23fe947 | 3036 | } |
cdddbdbc | 3037 | |
f796af5d | 3038 | |
3e372e5a DW |
3039 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3040 | * family number | |
3041 | */ | |
3042 | if (first->anchor->num_raid_devs == 0 && | |
3043 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3044 | int i; |
ba2de7ba DW |
3045 | struct intel_dev *dv; |
3046 | struct imsm_dev *dev; | |
78d30f94 DW |
3047 | |
3048 | /* we need to copy raid device info from sec if an allocation | |
3049 | * fails here we don't associate the spare | |
3050 | */ | |
3051 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
3052 | dv = malloc(sizeof(*dv)); |
3053 | if (!dv) | |
3054 | break; | |
3055 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
3056 | if (!dev) { | |
3057 | free(dv); | |
3058 | break; | |
78d30f94 | 3059 | } |
ba2de7ba DW |
3060 | dv->dev = dev; |
3061 | dv->index = i; | |
3062 | dv->next = first->devlist; | |
3063 | first->devlist = dv; | |
78d30f94 | 3064 | } |
709743c5 | 3065 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3066 | /* allocation failure */ |
3067 | free_devlist(first); | |
3068 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
3069 | return 3; | |
78d30f94 | 3070 | } |
3e372e5a | 3071 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3072 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3073 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3074 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3075 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3076 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3077 | } |
3078 | ||
cdddbdbc DW |
3079 | return 0; |
3080 | } | |
3081 | ||
0030e8d6 DW |
3082 | static void fd2devname(int fd, char *name) |
3083 | { | |
3084 | struct stat st; | |
3085 | char path[256]; | |
33a6535d | 3086 | char dname[PATH_MAX]; |
0030e8d6 DW |
3087 | char *nm; |
3088 | int rv; | |
3089 | ||
3090 | name[0] = '\0'; | |
3091 | if (fstat(fd, &st) != 0) | |
3092 | return; | |
3093 | sprintf(path, "/sys/dev/block/%d:%d", | |
3094 | major(st.st_rdev), minor(st.st_rdev)); | |
3095 | ||
9cf014ec | 3096 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3097 | if (rv <= 0) |
3098 | return; | |
9587c373 | 3099 | |
0030e8d6 DW |
3100 | dname[rv] = '\0'; |
3101 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3102 | if (nm) { |
3103 | nm++; | |
3104 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3105 | } | |
0030e8d6 DW |
3106 | } |
3107 | ||
cdddbdbc DW |
3108 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3109 | ||
3110 | static int imsm_read_serial(int fd, char *devname, | |
3111 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3112 | { | |
3113 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
3114 | int rv; |
3115 | int rsp_len; | |
1f24f035 | 3116 | int len; |
316e2bf4 DW |
3117 | char *dest; |
3118 | char *src; | |
3119 | char *rsp_buf; | |
3120 | int i; | |
cdddbdbc DW |
3121 | |
3122 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 3123 | |
f9ba0ff1 DW |
3124 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
3125 | ||
40ebbb9c | 3126 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3127 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3128 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3129 | return 0; |
3130 | } | |
3131 | ||
cdddbdbc DW |
3132 | if (rv != 0) { |
3133 | if (devname) | |
3134 | fprintf(stderr, | |
3135 | Name ": Failed to retrieve serial for %s\n", | |
3136 | devname); | |
3137 | return rv; | |
3138 | } | |
3139 | ||
3140 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
3141 | if (!rsp_len) { |
3142 | if (devname) | |
3143 | fprintf(stderr, | |
3144 | Name ": Failed to retrieve serial for %s\n", | |
3145 | devname); | |
3146 | return 2; | |
3147 | } | |
1f24f035 | 3148 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 3149 | |
316e2bf4 DW |
3150 | /* trim all whitespace and non-printable characters and convert |
3151 | * ':' to ';' | |
3152 | */ | |
3153 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
3154 | src = &rsp_buf[i]; | |
3155 | if (*src > 0x20) { | |
3156 | /* ':' is reserved for use in placeholder serial | |
3157 | * numbers for missing disks | |
3158 | */ | |
3159 | if (*src == ':') | |
3160 | *dest++ = ';'; | |
3161 | else | |
3162 | *dest++ = *src; | |
3163 | } | |
3164 | } | |
3165 | len = dest - rsp_buf; | |
3166 | dest = rsp_buf; | |
3167 | ||
3168 | /* truncate leading characters */ | |
3169 | if (len > MAX_RAID_SERIAL_LEN) { | |
3170 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3171 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3172 | } |
5c3db629 | 3173 | |
5c3db629 | 3174 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3175 | memcpy(serial, dest, len); |
cdddbdbc DW |
3176 | |
3177 | return 0; | |
3178 | } | |
3179 | ||
1f24f035 DW |
3180 | static int serialcmp(__u8 *s1, __u8 *s2) |
3181 | { | |
3182 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3183 | } | |
3184 | ||
3185 | static void serialcpy(__u8 *dest, __u8 *src) | |
3186 | { | |
3187 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3188 | } | |
3189 | ||
54c2c1ea DW |
3190 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3191 | { | |
3192 | struct dl *dl; | |
3193 | ||
3194 | for (dl = super->disks; dl; dl = dl->next) | |
3195 | if (serialcmp(dl->serial, serial) == 0) | |
3196 | break; | |
3197 | ||
3198 | return dl; | |
3199 | } | |
3200 | ||
a2b97981 DW |
3201 | static struct imsm_disk * |
3202 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3203 | { | |
3204 | int i; | |
3205 | ||
3206 | for (i = 0; i < mpb->num_disks; i++) { | |
3207 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3208 | ||
3209 | if (serialcmp(disk->serial, serial) == 0) { | |
3210 | if (idx) | |
3211 | *idx = i; | |
3212 | return disk; | |
3213 | } | |
3214 | } | |
3215 | ||
3216 | return NULL; | |
3217 | } | |
3218 | ||
cdddbdbc DW |
3219 | static int |
3220 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3221 | { | |
a2b97981 | 3222 | struct imsm_disk *disk; |
cdddbdbc DW |
3223 | struct dl *dl; |
3224 | struct stat stb; | |
cdddbdbc | 3225 | int rv; |
a2b97981 | 3226 | char name[40]; |
d23fe947 DW |
3227 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3228 | ||
3229 | rv = imsm_read_serial(fd, devname, serial); | |
3230 | ||
3231 | if (rv != 0) | |
3232 | return 2; | |
3233 | ||
a2b97981 | 3234 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 3235 | if (!dl) { |
cdddbdbc DW |
3236 | if (devname) |
3237 | fprintf(stderr, | |
3238 | Name ": failed to allocate disk buffer for %s\n", | |
3239 | devname); | |
3240 | return 2; | |
3241 | } | |
cdddbdbc | 3242 | |
a2b97981 DW |
3243 | fstat(fd, &stb); |
3244 | dl->major = major(stb.st_rdev); | |
3245 | dl->minor = minor(stb.st_rdev); | |
3246 | dl->next = super->disks; | |
3247 | dl->fd = keep_fd ? fd : -1; | |
3248 | assert(super->disks == NULL); | |
3249 | super->disks = dl; | |
3250 | serialcpy(dl->serial, serial); | |
3251 | dl->index = -2; | |
3252 | dl->e = NULL; | |
3253 | fd2devname(fd, name); | |
3254 | if (devname) | |
3255 | dl->devname = strdup(devname); | |
3256 | else | |
3257 | dl->devname = strdup(name); | |
cdddbdbc | 3258 | |
d23fe947 | 3259 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3260 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3261 | if (disk) { | |
3262 | dl->disk = *disk; | |
3263 | /* only set index on disks that are a member of a | |
3264 | * populated contianer, i.e. one with raid_devs | |
3265 | */ | |
3266 | if (is_failed(&dl->disk)) | |
3f6efecc | 3267 | dl->index = -2; |
a2b97981 DW |
3268 | else if (is_spare(&dl->disk)) |
3269 | dl->index = -1; | |
3f6efecc DW |
3270 | } |
3271 | ||
949c47a0 DW |
3272 | return 0; |
3273 | } | |
3274 | ||
0e600426 | 3275 | #ifndef MDASSEMBLE |
0c046afd DW |
3276 | /* When migrating map0 contains the 'destination' state while map1 |
3277 | * contains the current state. When not migrating map0 contains the | |
3278 | * current state. This routine assumes that map[0].map_state is set to | |
3279 | * the current array state before being called. | |
3280 | * | |
3281 | * Migration is indicated by one of the following states | |
3282 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3283 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3284 | * map1state=unitialized) |
1484e727 | 3285 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3286 | * map1state=normal) |
e3bba0e0 | 3287 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3288 | * map1state=degraded) |
8e59f3d8 AK |
3289 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3290 | * map1state=normal) | |
0c046afd | 3291 | */ |
8e59f3d8 AK |
3292 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3293 | __u8 to_state, int migr_type) | |
3393c6af | 3294 | { |
0c046afd | 3295 | struct imsm_map *dest; |
238c0a71 | 3296 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3297 | |
0c046afd | 3298 | dev->vol.migr_state = 1; |
1484e727 | 3299 | set_migr_type(dev, migr_type); |
f8f603f1 | 3300 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3301 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3302 | |
0556e1a2 | 3303 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3304 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3305 | if ((migr_type == MIGR_REBUILD) || |
3306 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3307 | __u32 ord; |
3308 | int i; | |
3309 | ||
3310 | for (i = 0; i < src->num_members; i++) { | |
3311 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3312 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3313 | } | |
3314 | } | |
3315 | ||
8e59f3d8 AK |
3316 | if (migr_type == MIGR_GEN_MIGR) |
3317 | /* Clear migration record */ | |
3318 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3319 | ||
0c046afd | 3320 | src->map_state = to_state; |
949c47a0 | 3321 | } |
f8f603f1 | 3322 | |
809da78e AK |
3323 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3324 | __u8 map_state) | |
f8f603f1 | 3325 | { |
238c0a71 AK |
3326 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3327 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3328 | MAP_0 : MAP_1); | |
28bce06f | 3329 | int i, j; |
0556e1a2 DW |
3330 | |
3331 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3332 | * completed in the last migration. | |
3333 | * | |
28bce06f | 3334 | * FIXME add support for raid-level-migration |
0556e1a2 | 3335 | */ |
809da78e AK |
3336 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3337 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3338 | /* when final map state is other than expected | |
3339 | * merge maps (not for migration) | |
3340 | */ | |
3341 | int failed; | |
3342 | ||
3343 | for (i = 0; i < prev->num_members; i++) | |
3344 | for (j = 0; j < map->num_members; j++) | |
3345 | /* during online capacity expansion | |
3346 | * disks position can be changed | |
3347 | * if takeover is used | |
3348 | */ | |
3349 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3350 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3351 | map->disk_ord_tbl[j] |= | |
3352 | prev->disk_ord_tbl[i]; | |
3353 | break; | |
3354 | } | |
3355 | failed = imsm_count_failed(super, dev, MAP_0); | |
3356 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3357 | } | |
f8f603f1 DW |
3358 | |
3359 | dev->vol.migr_state = 0; | |
ea672ee1 | 3360 | set_migr_type(dev, 0); |
f8f603f1 DW |
3361 | dev->vol.curr_migr_unit = 0; |
3362 | map->map_state = map_state; | |
3363 | } | |
0e600426 | 3364 | #endif |
949c47a0 DW |
3365 | |
3366 | static int parse_raid_devices(struct intel_super *super) | |
3367 | { | |
3368 | int i; | |
3369 | struct imsm_dev *dev_new; | |
4d7b1503 | 3370 | size_t len, len_migr; |
401d313b | 3371 | size_t max_len = 0; |
4d7b1503 DW |
3372 | size_t space_needed = 0; |
3373 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3374 | |
3375 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3376 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3377 | struct intel_dev *dv; |
949c47a0 | 3378 | |
4d7b1503 DW |
3379 | len = sizeof_imsm_dev(dev_iter, 0); |
3380 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3381 | if (len_migr > len) | |
3382 | space_needed += len_migr - len; | |
ca9de185 | 3383 | |
ba2de7ba DW |
3384 | dv = malloc(sizeof(*dv)); |
3385 | if (!dv) | |
3386 | return 1; | |
401d313b AK |
3387 | if (max_len < len_migr) |
3388 | max_len = len_migr; | |
3389 | if (max_len > len_migr) | |
3390 | space_needed += max_len - len_migr; | |
3391 | dev_new = malloc(max_len); | |
ba2de7ba DW |
3392 | if (!dev_new) { |
3393 | free(dv); | |
949c47a0 | 3394 | return 1; |
ba2de7ba | 3395 | } |
949c47a0 | 3396 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3397 | dv->dev = dev_new; |
3398 | dv->index = i; | |
3399 | dv->next = super->devlist; | |
3400 | super->devlist = dv; | |
949c47a0 | 3401 | } |
cdddbdbc | 3402 | |
4d7b1503 DW |
3403 | /* ensure that super->buf is large enough when all raid devices |
3404 | * are migrating | |
3405 | */ | |
3406 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3407 | void *buf; | |
3408 | ||
3409 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3410 | if (posix_memalign(&buf, 512, len) != 0) | |
3411 | return 1; | |
3412 | ||
1f45a8ad DW |
3413 | memcpy(buf, super->buf, super->len); |
3414 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3415 | free(super->buf); |
3416 | super->buf = buf; | |
3417 | super->len = len; | |
3418 | } | |
ca9de185 | 3419 | |
cdddbdbc DW |
3420 | return 0; |
3421 | } | |
3422 | ||
604b746f JD |
3423 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3424 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3425 | { | |
3426 | void *ptr = NULL; | |
3427 | ||
3428 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3429 | ptr = mpb; | |
3430 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3431 | } | |
3432 | ||
3433 | return ptr; | |
3434 | } | |
3435 | ||
e2f41b2c AK |
3436 | /******************************************************************************* |
3437 | * Function: check_mpb_migr_compatibility | |
3438 | * Description: Function checks for unsupported migration features: | |
3439 | * - migration optimization area (pba_of_lba0) | |
3440 | * - descending reshape (ascending_migr) | |
3441 | * Parameters: | |
3442 | * super : imsm metadata information | |
3443 | * Returns: | |
3444 | * 0 : migration is compatible | |
3445 | * -1 : migration is not compatible | |
3446 | ******************************************************************************/ | |
3447 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3448 | { | |
3449 | struct imsm_map *map0, *map1; | |
3450 | struct migr_record *migr_rec = super->migr_rec; | |
3451 | int i; | |
3452 | ||
3453 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3454 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3455 | ||
3456 | if (dev_iter && | |
3457 | dev_iter->vol.migr_state == 1 && | |
3458 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3459 | /* This device is migrating */ | |
238c0a71 AK |
3460 | map0 = get_imsm_map(dev_iter, MAP_0); |
3461 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 3462 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
3463 | /* migration optimization area was used */ |
3464 | return -1; | |
3465 | if (migr_rec->ascending_migr == 0 | |
3466 | && migr_rec->dest_depth_per_unit > 0) | |
3467 | /* descending reshape not supported yet */ | |
3468 | return -1; | |
3469 | } | |
3470 | } | |
3471 | return 0; | |
3472 | } | |
3473 | ||
d23fe947 | 3474 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3475 | |
cdddbdbc | 3476 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3477 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3478 | */ |
3479 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3480 | { | |
3481 | unsigned long long dsize; | |
cdddbdbc DW |
3482 | unsigned long long sectors; |
3483 | struct stat; | |
6416d527 | 3484 | struct imsm_super *anchor; |
cdddbdbc DW |
3485 | __u32 check_sum; |
3486 | ||
cdddbdbc | 3487 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3488 | if (dsize < 1024) { |
3489 | if (devname) | |
3490 | fprintf(stderr, | |
3491 | Name ": %s: device to small for imsm\n", | |
3492 | devname); | |
3493 | return 1; | |
3494 | } | |
cdddbdbc DW |
3495 | |
3496 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3497 | if (devname) | |
2e062e82 AK |
3498 | fprintf(stderr, Name |
3499 | ": Cannot seek to anchor block on %s: %s\n", | |
cdddbdbc DW |
3500 | devname, strerror(errno)); |
3501 | return 1; | |
3502 | } | |
3503 | ||
949c47a0 | 3504 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
3505 | if (devname) |
3506 | fprintf(stderr, | |
3507 | Name ": Failed to allocate imsm anchor buffer" | |
3508 | " on %s\n", devname); | |
3509 | return 1; | |
3510 | } | |
949c47a0 | 3511 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
3512 | if (devname) |
3513 | fprintf(stderr, | |
3514 | Name ": Cannot read anchor block on %s: %s\n", | |
3515 | devname, strerror(errno)); | |
6416d527 | 3516 | free(anchor); |
cdddbdbc DW |
3517 | return 1; |
3518 | } | |
3519 | ||
6416d527 | 3520 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
3521 | if (devname) |
3522 | fprintf(stderr, | |
3523 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 3524 | free(anchor); |
cdddbdbc DW |
3525 | return 2; |
3526 | } | |
3527 | ||
d23fe947 | 3528 | __free_imsm(super, 0); |
f2f5c343 LM |
3529 | /* reload capability and hba */ |
3530 | ||
3531 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3532 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3533 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3534 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
3535 | if (devname) |
3536 | fprintf(stderr, | |
3537 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 3538 | super->len); |
6416d527 | 3539 | free(anchor); |
cdddbdbc DW |
3540 | return 2; |
3541 | } | |
949c47a0 | 3542 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3543 | |
6416d527 NB |
3544 | sectors = mpb_sectors(anchor) - 1; |
3545 | free(anchor); | |
8e59f3d8 | 3546 | |
17a4eaf9 | 3547 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
8e59f3d8 AK |
3548 | fprintf(stderr, Name |
3549 | ": %s could not allocate migr_rec buffer\n", __func__); | |
3550 | free(super->buf); | |
3551 | return 2; | |
3552 | } | |
51d83f5d | 3553 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 3554 | |
949c47a0 | 3555 | if (!sectors) { |
ecf45690 DW |
3556 | check_sum = __gen_imsm_checksum(super->anchor); |
3557 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3558 | if (devname) | |
3559 | fprintf(stderr, | |
3560 | Name ": IMSM checksum %x != %x on %s\n", | |
3561 | check_sum, | |
3562 | __le32_to_cpu(super->anchor->check_sum), | |
3563 | devname); | |
3564 | return 2; | |
3565 | } | |
3566 | ||
a2b97981 | 3567 | return 0; |
949c47a0 | 3568 | } |
cdddbdbc DW |
3569 | |
3570 | /* read the extended mpb */ | |
3571 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3572 | if (devname) | |
3573 | fprintf(stderr, | |
3574 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
3575 | devname, strerror(errno)); | |
3576 | return 1; | |
3577 | } | |
3578 | ||
f21e18ca | 3579 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
3580 | if (devname) |
3581 | fprintf(stderr, | |
3582 | Name ": Cannot read extended mpb on %s: %s\n", | |
3583 | devname, strerror(errno)); | |
3584 | return 2; | |
3585 | } | |
3586 | ||
949c47a0 DW |
3587 | check_sum = __gen_imsm_checksum(super->anchor); |
3588 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
3589 | if (devname) |
3590 | fprintf(stderr, | |
3591 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 3592 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 3593 | devname); |
db575f3b | 3594 | return 3; |
cdddbdbc DW |
3595 | } |
3596 | ||
604b746f JD |
3597 | /* FIXME the BBM log is disk specific so we cannot use this global |
3598 | * buffer for all disks. Ok for now since we only look at the global | |
3599 | * bbm_log_size parameter to gate assembly | |
3600 | */ | |
3601 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3602 | ||
a2b97981 DW |
3603 | return 0; |
3604 | } | |
3605 | ||
8e59f3d8 AK |
3606 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3607 | ||
97f81ee2 CA |
3608 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
3609 | static void clear_hi(struct intel_super *super) | |
3610 | { | |
3611 | struct imsm_super *mpb = super->anchor; | |
3612 | int i, n; | |
3613 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
3614 | return; | |
3615 | for (i = 0; i < mpb->num_disks; ++i) { | |
3616 | struct imsm_disk *disk = &mpb->disk[i]; | |
3617 | disk->total_blocks_hi = 0; | |
3618 | } | |
3619 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
3620 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3621 | if (!dev) | |
3622 | return; | |
3623 | for (n = 0; n < 2; ++n) { | |
3624 | struct imsm_map *map = get_imsm_map(dev, n); | |
3625 | if (!map) | |
3626 | continue; | |
3627 | map->pba_of_lba0_hi = 0; | |
3628 | map->blocks_per_member_hi = 0; | |
3629 | map->num_data_stripes_hi = 0; | |
3630 | } | |
3631 | } | |
3632 | } | |
3633 | ||
a2b97981 DW |
3634 | static int |
3635 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3636 | { | |
3637 | int err; | |
3638 | ||
3639 | err = load_imsm_mpb(fd, super, devname); | |
3640 | if (err) | |
3641 | return err; | |
3642 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3643 | if (err) | |
3644 | return err; | |
3645 | err = parse_raid_devices(super); | |
97f81ee2 | 3646 | clear_hi(super); |
a2b97981 | 3647 | return err; |
cdddbdbc DW |
3648 | } |
3649 | ||
ae6aad82 DW |
3650 | static void __free_imsm_disk(struct dl *d) |
3651 | { | |
3652 | if (d->fd >= 0) | |
3653 | close(d->fd); | |
3654 | if (d->devname) | |
3655 | free(d->devname); | |
0dcecb2e DW |
3656 | if (d->e) |
3657 | free(d->e); | |
ae6aad82 DW |
3658 | free(d); |
3659 | ||
3660 | } | |
1a64be56 | 3661 | |
cdddbdbc DW |
3662 | static void free_imsm_disks(struct intel_super *super) |
3663 | { | |
47ee5a45 | 3664 | struct dl *d; |
cdddbdbc | 3665 | |
47ee5a45 DW |
3666 | while (super->disks) { |
3667 | d = super->disks; | |
cdddbdbc | 3668 | super->disks = d->next; |
ae6aad82 | 3669 | __free_imsm_disk(d); |
cdddbdbc | 3670 | } |
cb82edca AK |
3671 | while (super->disk_mgmt_list) { |
3672 | d = super->disk_mgmt_list; | |
3673 | super->disk_mgmt_list = d->next; | |
3674 | __free_imsm_disk(d); | |
3675 | } | |
47ee5a45 DW |
3676 | while (super->missing) { |
3677 | d = super->missing; | |
3678 | super->missing = d->next; | |
3679 | __free_imsm_disk(d); | |
3680 | } | |
3681 | ||
cdddbdbc DW |
3682 | } |
3683 | ||
9ca2c81c | 3684 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3685 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3686 | { |
88654014 LM |
3687 | struct intel_hba *elem, *next; |
3688 | ||
9ca2c81c | 3689 | if (super->buf) { |
949c47a0 | 3690 | free(super->buf); |
9ca2c81c DW |
3691 | super->buf = NULL; |
3692 | } | |
f2f5c343 LM |
3693 | /* unlink capability description */ |
3694 | super->orom = NULL; | |
8e59f3d8 AK |
3695 | if (super->migr_rec_buf) { |
3696 | free(super->migr_rec_buf); | |
3697 | super->migr_rec_buf = NULL; | |
3698 | } | |
d23fe947 DW |
3699 | if (free_disks) |
3700 | free_imsm_disks(super); | |
ba2de7ba | 3701 | free_devlist(super); |
88654014 LM |
3702 | elem = super->hba; |
3703 | while (elem) { | |
3704 | if (elem->path) | |
3705 | free((void *)elem->path); | |
3706 | next = elem->next; | |
3707 | free(elem); | |
3708 | elem = next; | |
88c32bb1 | 3709 | } |
88654014 | 3710 | super->hba = NULL; |
cdddbdbc DW |
3711 | } |
3712 | ||
9ca2c81c DW |
3713 | static void free_imsm(struct intel_super *super) |
3714 | { | |
d23fe947 | 3715 | __free_imsm(super, 1); |
9ca2c81c DW |
3716 | free(super); |
3717 | } | |
cdddbdbc DW |
3718 | |
3719 | static void free_super_imsm(struct supertype *st) | |
3720 | { | |
3721 | struct intel_super *super = st->sb; | |
3722 | ||
3723 | if (!super) | |
3724 | return; | |
3725 | ||
3726 | free_imsm(super); | |
3727 | st->sb = NULL; | |
3728 | } | |
3729 | ||
49133e57 | 3730 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
3731 | { |
3732 | struct intel_super *super = malloc(sizeof(*super)); | |
3733 | ||
3734 | if (super) { | |
3735 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 3736 | super->current_vol = -1; |
5551b113 | 3737 | super->create_offset = ~((unsigned long long) 0); |
c2c087e6 | 3738 | } |
c2c087e6 DW |
3739 | return super; |
3740 | } | |
3741 | ||
f0f5a016 LM |
3742 | /* |
3743 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3744 | */ | |
d424212e | 3745 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3746 | { |
3747 | struct sys_dev *hba_name; | |
3748 | int rv = 0; | |
3749 | ||
3750 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3751 | super->orom = NULL; |
f0f5a016 LM |
3752 | super->hba = NULL; |
3753 | return 0; | |
3754 | } | |
3755 | hba_name = find_disk_attached_hba(fd, NULL); | |
3756 | if (!hba_name) { | |
d424212e | 3757 | if (devname) |
f0f5a016 LM |
3758 | fprintf(stderr, |
3759 | Name ": %s is not attached to Intel(R) RAID controller.\n", | |
d424212e | 3760 | devname); |
f0f5a016 LM |
3761 | return 1; |
3762 | } | |
3763 | rv = attach_hba_to_super(super, hba_name); | |
3764 | if (rv == 2) { | |
d424212e N |
3765 | if (devname) { |
3766 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3767 | |
f0f5a016 LM |
3768 | fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID " |
3769 | "controller (%s),\n" | |
3770 | " but the container is assigned to Intel(R) " | |
3771 | "%s RAID controller (", | |
d424212e | 3772 | devname, |
f0f5a016 LM |
3773 | hba_name->path, |
3774 | hba_name->pci_id ? : "Err!", | |
3775 | get_sys_dev_type(hba_name->type)); | |
3776 | ||
f0f5a016 LM |
3777 | while (hba) { |
3778 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3779 | if (hba->next) | |
3780 | fprintf(stderr, ", "); | |
3781 | hba = hba->next; | |
3782 | } | |
3783 | ||
3784 | fprintf(stderr, ").\n" | |
3785 | " Mixing devices attached to different controllers " | |
3786 | "is not allowed.\n"); | |
3787 | } | |
3788 | free_sys_dev(&hba_name); | |
3789 | return 2; | |
3790 | } | |
f2f5c343 | 3791 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3792 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3793 | if (!super->orom) |
3794 | return 3; | |
f0f5a016 LM |
3795 | return 0; |
3796 | } | |
3797 | ||
47ee5a45 DW |
3798 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3799 | * disks that have disappeared from the system. This routine relies on | |
3800 | * the mpb being uptodate, which it is at load time. | |
3801 | */ | |
3802 | static int find_missing(struct intel_super *super) | |
3803 | { | |
3804 | int i; | |
3805 | struct imsm_super *mpb = super->anchor; | |
3806 | struct dl *dl; | |
3807 | struct imsm_disk *disk; | |
47ee5a45 DW |
3808 | |
3809 | for (i = 0; i < mpb->num_disks; i++) { | |
3810 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3811 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3812 | if (dl) |
3813 | continue; | |
47ee5a45 DW |
3814 | |
3815 | dl = malloc(sizeof(*dl)); | |
3816 | if (!dl) | |
3817 | return 1; | |
3818 | dl->major = 0; | |
3819 | dl->minor = 0; | |
3820 | dl->fd = -1; | |
3821 | dl->devname = strdup("missing"); | |
3822 | dl->index = i; | |
3823 | serialcpy(dl->serial, disk->serial); | |
3824 | dl->disk = *disk; | |
689c9bf3 | 3825 | dl->e = NULL; |
47ee5a45 DW |
3826 | dl->next = super->missing; |
3827 | super->missing = dl; | |
3828 | } | |
3829 | ||
3830 | return 0; | |
3831 | } | |
3832 | ||
3960e579 | 3833 | #ifndef MDASSEMBLE |
a2b97981 DW |
3834 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3835 | { | |
3836 | struct intel_disk *idisk = disk_list; | |
3837 | ||
3838 | while (idisk) { | |
3839 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3840 | break; | |
3841 | idisk = idisk->next; | |
3842 | } | |
3843 | ||
3844 | return idisk; | |
3845 | } | |
3846 | ||
3847 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3848 | struct intel_super *super, | |
3849 | struct intel_disk **disk_list) | |
3850 | { | |
3851 | struct imsm_disk *d = &super->disks->disk; | |
3852 | struct imsm_super *mpb = super->anchor; | |
3853 | int i, j; | |
3854 | ||
3855 | for (i = 0; i < tbl_size; i++) { | |
3856 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3857 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3858 | ||
3859 | if (tbl_mpb->family_num == mpb->family_num) { | |
3860 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3861 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3862 | __func__, super->disks->major, | |
3863 | super->disks->minor, | |
3864 | table[i]->disks->major, | |
3865 | table[i]->disks->minor); | |
3866 | break; | |
3867 | } | |
3868 | ||
3869 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3870 | is_configured(d) == is_configured(tbl_d)) && | |
3871 | tbl_mpb->generation_num < mpb->generation_num) { | |
3872 | /* current version of the mpb is a | |
3873 | * better candidate than the one in | |
3874 | * super_table, but copy over "cross | |
3875 | * generational" status | |
3876 | */ | |
3877 | struct intel_disk *idisk; | |
3878 | ||
3879 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3880 | __func__, super->disks->major, | |
3881 | super->disks->minor, | |
3882 | table[i]->disks->major, | |
3883 | table[i]->disks->minor); | |
3884 | ||
3885 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3886 | if (idisk && is_failed(&idisk->disk)) | |
3887 | tbl_d->status |= FAILED_DISK; | |
3888 | break; | |
3889 | } else { | |
3890 | struct intel_disk *idisk; | |
3891 | struct imsm_disk *disk; | |
3892 | ||
3893 | /* tbl_mpb is more up to date, but copy | |
3894 | * over cross generational status before | |
3895 | * returning | |
3896 | */ | |
3897 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3898 | if (disk && is_failed(disk)) | |
3899 | d->status |= FAILED_DISK; | |
3900 | ||
3901 | idisk = disk_list_get(d->serial, *disk_list); | |
3902 | if (idisk) { | |
3903 | idisk->owner = i; | |
3904 | if (disk && is_configured(disk)) | |
3905 | idisk->disk.status |= CONFIGURED_DISK; | |
3906 | } | |
3907 | ||
3908 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3909 | __func__, super->disks->major, | |
3910 | super->disks->minor, | |
3911 | table[i]->disks->major, | |
3912 | table[i]->disks->minor); | |
3913 | ||
3914 | return tbl_size; | |
3915 | } | |
3916 | } | |
3917 | } | |
3918 | ||
3919 | if (i >= tbl_size) | |
3920 | table[tbl_size++] = super; | |
3921 | else | |
3922 | table[i] = super; | |
3923 | ||
3924 | /* update/extend the merged list of imsm_disk records */ | |
3925 | for (j = 0; j < mpb->num_disks; j++) { | |
3926 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3927 | struct intel_disk *idisk; | |
3928 | ||
3929 | idisk = disk_list_get(disk->serial, *disk_list); | |
3930 | if (idisk) { | |
3931 | idisk->disk.status |= disk->status; | |
3932 | if (is_configured(&idisk->disk) || | |
3933 | is_failed(&idisk->disk)) | |
3934 | idisk->disk.status &= ~(SPARE_DISK); | |
3935 | } else { | |
3936 | idisk = calloc(1, sizeof(*idisk)); | |
3937 | if (!idisk) | |
3938 | return -1; | |
3939 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
3940 | idisk->disk = *disk; | |
3941 | idisk->next = *disk_list; | |
3942 | *disk_list = idisk; | |
3943 | } | |
3944 | ||
3945 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3946 | idisk->owner = i; | |
3947 | } | |
3948 | ||
3949 | return tbl_size; | |
3950 | } | |
3951 | ||
3952 | static struct intel_super * | |
3953 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3954 | const int owner) | |
3955 | { | |
3956 | struct imsm_super *mpb = super->anchor; | |
3957 | int ok_count = 0; | |
3958 | int i; | |
3959 | ||
3960 | for (i = 0; i < mpb->num_disks; i++) { | |
3961 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3962 | struct intel_disk *idisk; | |
3963 | ||
3964 | idisk = disk_list_get(disk->serial, disk_list); | |
3965 | if (idisk) { | |
3966 | if (idisk->owner == owner || | |
3967 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3968 | ok_count++; | |
3969 | else | |
3970 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3971 | __func__, disk->serial, idisk->owner, | |
3972 | owner); | |
3973 | } else { | |
3974 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
3975 | __func__, __le32_to_cpu(mpb->family_num), i, | |
3976 | disk->serial); | |
3977 | break; | |
3978 | } | |
3979 | } | |
3980 | ||
3981 | if (ok_count == mpb->num_disks) | |
3982 | return super; | |
3983 | return NULL; | |
3984 | } | |
3985 | ||
3986 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
3987 | { | |
3988 | struct intel_super *s; | |
3989 | ||
3990 | for (s = super_list; s; s = s->next) { | |
3991 | if (family_num != s->anchor->family_num) | |
3992 | continue; | |
3993 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
3994 | __le32_to_cpu(family_num), s->disks->devname); | |
3995 | } | |
3996 | } | |
3997 | ||
3998 | static struct intel_super * | |
3999 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4000 | { | |
4001 | struct intel_super *super_table[len]; | |
4002 | struct intel_disk *disk_list = NULL; | |
4003 | struct intel_super *champion, *spare; | |
4004 | struct intel_super *s, **del; | |
4005 | int tbl_size = 0; | |
4006 | int conflict; | |
4007 | int i; | |
4008 | ||
4009 | memset(super_table, 0, sizeof(super_table)); | |
4010 | for (s = *super_list; s; s = s->next) | |
4011 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4012 | ||
4013 | for (i = 0; i < tbl_size; i++) { | |
4014 | struct imsm_disk *d; | |
4015 | struct intel_disk *idisk; | |
4016 | struct imsm_super *mpb = super_table[i]->anchor; | |
4017 | ||
4018 | s = super_table[i]; | |
4019 | d = &s->disks->disk; | |
4020 | ||
4021 | /* 'd' must appear in merged disk list for its | |
4022 | * configuration to be valid | |
4023 | */ | |
4024 | idisk = disk_list_get(d->serial, disk_list); | |
4025 | if (idisk && idisk->owner == i) | |
4026 | s = validate_members(s, disk_list, i); | |
4027 | else | |
4028 | s = NULL; | |
4029 | ||
4030 | if (!s) | |
4031 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
4032 | __func__, mpb->family_num, | |
4033 | super_table[i]->disks->major, | |
4034 | super_table[i]->disks->minor); | |
4035 | super_table[i] = s; | |
4036 | } | |
4037 | ||
4038 | /* This is where the mdadm implementation differs from the Windows | |
4039 | * driver which has no strict concept of a container. We can only | |
4040 | * assemble one family from a container, so when returning a prodigal | |
4041 | * array member to this system the code will not be able to disambiguate | |
4042 | * the container contents that should be assembled ("foreign" versus | |
4043 | * "local"). It requires user intervention to set the orig_family_num | |
4044 | * to a new value to establish a new container. The Windows driver in | |
4045 | * this situation fixes up the volume name in place and manages the | |
4046 | * foreign array as an independent entity. | |
4047 | */ | |
4048 | s = NULL; | |
4049 | spare = NULL; | |
4050 | conflict = 0; | |
4051 | for (i = 0; i < tbl_size; i++) { | |
4052 | struct intel_super *tbl_ent = super_table[i]; | |
4053 | int is_spare = 0; | |
4054 | ||
4055 | if (!tbl_ent) | |
4056 | continue; | |
4057 | ||
4058 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4059 | spare = tbl_ent; | |
4060 | is_spare = 1; | |
4061 | } | |
4062 | ||
4063 | if (s && !is_spare) { | |
4064 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4065 | conflict++; | |
4066 | } else if (!s && !is_spare) | |
4067 | s = tbl_ent; | |
4068 | } | |
4069 | ||
4070 | if (!s) | |
4071 | s = spare; | |
4072 | if (!s) { | |
4073 | champion = NULL; | |
4074 | goto out; | |
4075 | } | |
4076 | champion = s; | |
4077 | ||
4078 | if (conflict) | |
4079 | fprintf(stderr, "Chose family %#x on '%s', " | |
4080 | "assemble conflicts to new container with '--update=uuid'\n", | |
4081 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
4082 | ||
4083 | /* collect all dl's onto 'champion', and update them to | |
4084 | * champion's version of the status | |
4085 | */ | |
4086 | for (s = *super_list; s; s = s->next) { | |
4087 | struct imsm_super *mpb = champion->anchor; | |
4088 | struct dl *dl = s->disks; | |
4089 | ||
4090 | if (s == champion) | |
4091 | continue; | |
4092 | ||
4093 | for (i = 0; i < mpb->num_disks; i++) { | |
4094 | struct imsm_disk *disk; | |
4095 | ||
4096 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4097 | if (disk) { | |
4098 | dl->disk = *disk; | |
4099 | /* only set index on disks that are a member of | |
4100 | * a populated contianer, i.e. one with | |
4101 | * raid_devs | |
4102 | */ | |
4103 | if (is_failed(&dl->disk)) | |
4104 | dl->index = -2; | |
4105 | else if (is_spare(&dl->disk)) | |
4106 | dl->index = -1; | |
4107 | break; | |
4108 | } | |
4109 | } | |
4110 | ||
4111 | if (i >= mpb->num_disks) { | |
4112 | struct intel_disk *idisk; | |
4113 | ||
4114 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4115 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4116 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4117 | dl->index = -1; | |
4118 | else { | |
4119 | dl->index = -2; | |
4120 | continue; | |
4121 | } | |
4122 | } | |
4123 | ||
4124 | dl->next = champion->disks; | |
4125 | champion->disks = dl; | |
4126 | s->disks = NULL; | |
4127 | } | |
4128 | ||
4129 | /* delete 'champion' from super_list */ | |
4130 | for (del = super_list; *del; ) { | |
4131 | if (*del == champion) { | |
4132 | *del = (*del)->next; | |
4133 | break; | |
4134 | } else | |
4135 | del = &(*del)->next; | |
4136 | } | |
4137 | champion->next = NULL; | |
4138 | ||
4139 | out: | |
4140 | while (disk_list) { | |
4141 | struct intel_disk *idisk = disk_list; | |
4142 | ||
4143 | disk_list = disk_list->next; | |
4144 | free(idisk); | |
4145 | } | |
4146 | ||
4147 | return champion; | |
4148 | } | |
4149 | ||
9587c373 LM |
4150 | |
4151 | static int | |
4152 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
9587c373 LM |
4153 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, |
4154 | int major, int minor, int keep_fd); | |
ec50f7b6 LM |
4155 | static int |
4156 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4157 | int *max, int keep_fd); | |
4158 | ||
9587c373 | 4159 | |
cdddbdbc | 4160 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4161 | char *devname, struct md_list *devlist, |
4162 | int keep_fd) | |
cdddbdbc | 4163 | { |
a2b97981 DW |
4164 | struct intel_super *super_list = NULL; |
4165 | struct intel_super *super = NULL; | |
a2b97981 | 4166 | int err = 0; |
9587c373 | 4167 | int i = 0; |
dab4a513 | 4168 | |
9587c373 LM |
4169 | if (fd >= 0) |
4170 | /* 'fd' is an opened container */ | |
4171 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4172 | else | |
ec50f7b6 LM |
4173 | /* get super block from devlist devices */ |
4174 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4175 | if (err) |
1602d52c | 4176 | goto error; |
a2b97981 DW |
4177 | /* all mpbs enter, maybe one leaves */ |
4178 | super = imsm_thunderdome(&super_list, i); | |
4179 | if (!super) { | |
4180 | err = 1; | |
4181 | goto error; | |
cdddbdbc DW |
4182 | } |
4183 | ||
47ee5a45 DW |
4184 | if (find_missing(super) != 0) { |
4185 | free_imsm(super); | |
a2b97981 DW |
4186 | err = 2; |
4187 | goto error; | |
47ee5a45 | 4188 | } |
8e59f3d8 AK |
4189 | |
4190 | /* load migration record */ | |
4191 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4192 | if (err == -1) { |
4193 | /* migration is in progress, | |
4194 | * but migr_rec cannot be loaded, | |
4195 | */ | |
8e59f3d8 AK |
4196 | err = 4; |
4197 | goto error; | |
4198 | } | |
e2f41b2c AK |
4199 | |
4200 | /* Check migration compatibility */ | |
4c965cc9 | 4201 | if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) { |
e2f41b2c AK |
4202 | fprintf(stderr, Name ": Unsupported migration detected"); |
4203 | if (devname) | |
4204 | fprintf(stderr, " on %s\n", devname); | |
4205 | else | |
4206 | fprintf(stderr, " (IMSM).\n"); | |
4207 | ||
4208 | err = 5; | |
4209 | goto error; | |
4210 | } | |
4211 | ||
a2b97981 DW |
4212 | err = 0; |
4213 | ||
4214 | error: | |
4215 | while (super_list) { | |
4216 | struct intel_super *s = super_list; | |
4217 | ||
4218 | super_list = super_list->next; | |
4219 | free_imsm(s); | |
4220 | } | |
9587c373 | 4221 | |
a2b97981 DW |
4222 | |
4223 | if (err) | |
4224 | return err; | |
f7e7067b | 4225 | |
cdddbdbc | 4226 | *sbp = super; |
9587c373 LM |
4227 | if (fd >= 0) |
4228 | st->container_dev = fd2devnum(fd); | |
4229 | else | |
4230 | st->container_dev = NoMdDev; | |
a2b97981 | 4231 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4232 | st->ss = &super_imsm; |
cdddbdbc DW |
4233 | st->minor_version = 0; |
4234 | st->max_devs = IMSM_MAX_DEVICES; | |
4235 | } | |
cdddbdbc DW |
4236 | return 0; |
4237 | } | |
2b959fbf | 4238 | |
9587c373 | 4239 | |
ec50f7b6 LM |
4240 | static int |
4241 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4242 | int *max, int keep_fd) | |
4243 | { | |
4244 | struct md_list *tmpdev; | |
4245 | int err = 0; | |
4246 | int i = 0; | |
9587c373 | 4247 | |
ec50f7b6 LM |
4248 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4249 | if (tmpdev->used != 1) | |
4250 | continue; | |
4251 | if (tmpdev->container == 1) { | |
ca9de185 | 4252 | int lmax = 0; |
ec50f7b6 LM |
4253 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4254 | if (fd < 0) { | |
4255 | fprintf(stderr, Name ": cannot open device %s: %s\n", | |
4256 | tmpdev->devname, strerror(errno)); | |
4257 | err = 8; | |
4258 | goto error; | |
4259 | } | |
4260 | err = get_sra_super_block(fd, super_list, | |
4261 | tmpdev->devname, &lmax, | |
4262 | keep_fd); | |
4263 | i += lmax; | |
4264 | close(fd); | |
4265 | if (err) { | |
4266 | err = 7; | |
4267 | goto error; | |
4268 | } | |
4269 | } else { | |
4270 | int major = major(tmpdev->st_rdev); | |
4271 | int minor = minor(tmpdev->st_rdev); | |
4272 | err = get_super_block(super_list, | |
4273 | -1, | |
4274 | tmpdev->devname, | |
4275 | major, minor, | |
4276 | keep_fd); | |
4277 | i++; | |
4278 | if (err) { | |
4279 | err = 6; | |
4280 | goto error; | |
4281 | } | |
4282 | } | |
4283 | } | |
4284 | error: | |
4285 | *max = i; | |
4286 | return err; | |
4287 | } | |
9587c373 LM |
4288 | |
4289 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, | |
4290 | int major, int minor, int keep_fd) | |
4291 | { | |
4292 | struct intel_super*s = NULL; | |
4293 | char nm[32]; | |
4294 | int dfd = -1; | |
4295 | int rv; | |
4296 | int err = 0; | |
4297 | int retry; | |
4298 | ||
4299 | s = alloc_super(); | |
4300 | if (!s) { | |
4301 | err = 1; | |
4302 | goto error; | |
4303 | } | |
4304 | ||
4305 | sprintf(nm, "%d:%d", major, minor); | |
4306 | dfd = dev_open(nm, O_RDWR); | |
4307 | if (dfd < 0) { | |
4308 | err = 2; | |
4309 | goto error; | |
4310 | } | |
4311 | ||
4312 | rv = find_intel_hba_capability(dfd, s, devname); | |
4313 | /* no orom/efi or non-intel hba of the disk */ | |
4314 | if (rv != 0) { | |
4315 | err = 4; | |
4316 | goto error; | |
4317 | } | |
4318 | ||
4319 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4320 | ||
4321 | /* retry the load if we might have raced against mdmon */ | |
4322 | if (err == 3 && (devnum != -1) && mdmon_running(devnum)) | |
4323 | for (retry = 0; retry < 3; retry++) { | |
4324 | usleep(3000); | |
4325 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4326 | if (err != 3) | |
4327 | break; | |
4328 | } | |
4329 | error: | |
4330 | if (!err) { | |
4331 | s->next = *super_list; | |
4332 | *super_list = s; | |
4333 | } else { | |
4334 | if (s) | |
4335 | free(s); | |
4336 | if (dfd) | |
4337 | close(dfd); | |
4338 | } | |
4339 | if ((dfd >= 0) && (!keep_fd)) | |
4340 | close(dfd); | |
4341 | return err; | |
4342 | ||
4343 | } | |
4344 | ||
4345 | static int | |
4346 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4347 | { | |
4348 | struct mdinfo *sra; | |
4349 | int devnum; | |
4350 | struct mdinfo *sd; | |
4351 | int err = 0; | |
4352 | int i = 0; | |
4353 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
4354 | if (!sra) | |
4355 | return 1; | |
4356 | ||
4357 | if (sra->array.major_version != -1 || | |
4358 | sra->array.minor_version != -2 || | |
4359 | strcmp(sra->text_version, "imsm") != 0) { | |
4360 | err = 1; | |
4361 | goto error; | |
4362 | } | |
4363 | /* load all mpbs */ | |
4364 | devnum = fd2devnum(fd); | |
4365 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
4366 | if (get_super_block(super_list, devnum, devname, | |
4367 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { | |
4368 | err = 7; | |
4369 | goto error; | |
4370 | } | |
4371 | } | |
4372 | error: | |
4373 | sysfs_free(sra); | |
4374 | *max = i; | |
4375 | return err; | |
4376 | } | |
4377 | ||
2b959fbf N |
4378 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4379 | { | |
ec50f7b6 | 4380 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4381 | } |
cdddbdbc DW |
4382 | #endif |
4383 | ||
4384 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4385 | { | |
4386 | struct intel_super *super; | |
4387 | int rv; | |
4388 | ||
691c6ee1 N |
4389 | if (test_partition(fd)) |
4390 | /* IMSM not allowed on partitions */ | |
4391 | return 1; | |
4392 | ||
37424f13 DW |
4393 | free_super_imsm(st); |
4394 | ||
49133e57 | 4395 | super = alloc_super(); |
cdddbdbc DW |
4396 | if (!super) { |
4397 | fprintf(stderr, | |
4398 | Name ": malloc of %zu failed.\n", | |
4399 | sizeof(*super)); | |
4400 | return 1; | |
4401 | } | |
ea2bc72b LM |
4402 | /* Load hba and capabilities if they exist. |
4403 | * But do not preclude loading metadata in case capabilities or hba are | |
4404 | * non-compliant and ignore_hw_compat is set. | |
4405 | */ | |
d424212e | 4406 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4407 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4408 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 LM |
4409 | if (devname) |
4410 | fprintf(stderr, | |
4411 | Name ": No OROM/EFI properties for %s\n", devname); | |
4412 | free_imsm(super); | |
4413 | return 2; | |
4414 | } | |
a2b97981 | 4415 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
4416 | |
4417 | if (rv) { | |
4418 | if (devname) | |
4419 | fprintf(stderr, | |
4420 | Name ": Failed to load all information " | |
4421 | "sections on %s\n", devname); | |
4422 | free_imsm(super); | |
4423 | return rv; | |
4424 | } | |
4425 | ||
4426 | st->sb = super; | |
4427 | if (st->ss == NULL) { | |
4428 | st->ss = &super_imsm; | |
4429 | st->minor_version = 0; | |
4430 | st->max_devs = IMSM_MAX_DEVICES; | |
4431 | } | |
8e59f3d8 AK |
4432 | |
4433 | /* load migration record */ | |
2e062e82 AK |
4434 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4435 | /* Check for unsupported migration features */ | |
4436 | if (check_mpb_migr_compatibility(super) != 0) { | |
4437 | fprintf(stderr, | |
4438 | Name ": Unsupported migration detected"); | |
4439 | if (devname) | |
4440 | fprintf(stderr, " on %s\n", devname); | |
4441 | else | |
4442 | fprintf(stderr, " (IMSM).\n"); | |
4443 | return 3; | |
4444 | } | |
e2f41b2c AK |
4445 | } |
4446 | ||
cdddbdbc DW |
4447 | return 0; |
4448 | } | |
4449 | ||
ef6ffade DW |
4450 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4451 | { | |
4452 | if (info->level == 1) | |
4453 | return 128; | |
4454 | return info->chunk_size >> 9; | |
4455 | } | |
4456 | ||
5551b113 CA |
4457 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
4458 | unsigned long long size) | |
fcfd9599 | 4459 | { |
4025c288 | 4460 | if (info->level == 1) |
5551b113 | 4461 | return size * 2; |
4025c288 | 4462 | else |
5551b113 | 4463 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
4464 | } |
4465 | ||
4d1313e9 DW |
4466 | static void imsm_update_version_info(struct intel_super *super) |
4467 | { | |
4468 | /* update the version and attributes */ | |
4469 | struct imsm_super *mpb = super->anchor; | |
4470 | char *version; | |
4471 | struct imsm_dev *dev; | |
4472 | struct imsm_map *map; | |
4473 | int i; | |
4474 | ||
4475 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4476 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4477 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4478 | if (__le32_to_cpu(dev->size_high) > 0) |
4479 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4480 | ||
4481 | /* FIXME detect when an array spans a port multiplier */ | |
4482 | #if 0 | |
4483 | mpb->attributes |= MPB_ATTRIB_PM; | |
4484 | #endif | |
4485 | ||
4486 | if (mpb->num_raid_devs > 1 || | |
4487 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4488 | version = MPB_VERSION_ATTRIBS; | |
4489 | switch (get_imsm_raid_level(map)) { | |
4490 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4491 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4492 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4493 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4494 | } | |
4495 | } else { | |
4496 | if (map->num_members >= 5) | |
4497 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4498 | else if (dev->status == DEV_CLONE_N_GO) | |
4499 | version = MPB_VERSION_CNG; | |
4500 | else if (get_imsm_raid_level(map) == 5) | |
4501 | version = MPB_VERSION_RAID5; | |
4502 | else if (map->num_members >= 3) | |
4503 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4504 | else if (get_imsm_raid_level(map) == 1) | |
4505 | version = MPB_VERSION_RAID1; | |
4506 | else | |
4507 | version = MPB_VERSION_RAID0; | |
4508 | } | |
4509 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4510 | } | |
4511 | } | |
4512 | ||
aa534678 DW |
4513 | static int check_name(struct intel_super *super, char *name, int quiet) |
4514 | { | |
4515 | struct imsm_super *mpb = super->anchor; | |
4516 | char *reason = NULL; | |
4517 | int i; | |
4518 | ||
4519 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4520 | reason = "must be 16 characters or less"; | |
4521 | ||
4522 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4523 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4524 | ||
4525 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4526 | reason = "already exists"; | |
4527 | break; | |
4528 | } | |
4529 | } | |
4530 | ||
4531 | if (reason && !quiet) | |
4532 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
4533 | ||
4534 | return !reason; | |
4535 | } | |
4536 | ||
8b353278 DW |
4537 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4538 | unsigned long long size, char *name, | |
4539 | char *homehost, int *uuid) | |
cdddbdbc | 4540 | { |
c2c087e6 DW |
4541 | /* We are creating a volume inside a pre-existing container. |
4542 | * so st->sb is already set. | |
4543 | */ | |
4544 | struct intel_super *super = st->sb; | |
949c47a0 | 4545 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4546 | struct intel_dev *dv; |
c2c087e6 DW |
4547 | struct imsm_dev *dev; |
4548 | struct imsm_vol *vol; | |
4549 | struct imsm_map *map; | |
4550 | int idx = mpb->num_raid_devs; | |
4551 | int i; | |
4552 | unsigned long long array_blocks; | |
2c092cad | 4553 | size_t size_old, size_new; |
5551b113 | 4554 | unsigned long long num_data_stripes; |
cdddbdbc | 4555 | |
88c32bb1 | 4556 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 4557 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 4558 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4559 | return 0; |
4560 | } | |
4561 | ||
2c092cad DW |
4562 | /* ensure the mpb is large enough for the new data */ |
4563 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4564 | size_new = disks_to_mpb_size(info->nr_disks); | |
4565 | if (size_new > size_old) { | |
4566 | void *mpb_new; | |
4567 | size_t size_round = ROUND_UP(size_new, 512); | |
4568 | ||
4569 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
4570 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
4571 | return 0; | |
4572 | } | |
17a4eaf9 AK |
4573 | if (posix_memalign(&super->migr_rec_buf, 512, |
4574 | MIGR_REC_BUF_SIZE) != 0) { | |
8e59f3d8 AK |
4575 | fprintf(stderr, Name |
4576 | ": %s could not allocate migr_rec buffer\n", | |
4577 | __func__); | |
4578 | free(super->buf); | |
4579 | free(super); | |
ea944c8f | 4580 | free(mpb_new); |
8e59f3d8 AK |
4581 | return 0; |
4582 | } | |
2c092cad DW |
4583 | memcpy(mpb_new, mpb, size_old); |
4584 | free(mpb); | |
4585 | mpb = mpb_new; | |
949c47a0 | 4586 | super->anchor = mpb_new; |
2c092cad DW |
4587 | mpb->mpb_size = __cpu_to_le32(size_new); |
4588 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4589 | } | |
bf5a934a | 4590 | super->current_vol = idx; |
3960e579 DW |
4591 | |
4592 | /* handle 'failed_disks' by either: | |
4593 | * a) create dummy disk entries in the table if this the first | |
4594 | * volume in the array. We add them here as this is the only | |
4595 | * opportunity to add them. add_to_super_imsm_volume() | |
4596 | * handles the non-failed disks and continues incrementing | |
4597 | * mpb->num_disks. | |
4598 | * b) validate that 'failed_disks' matches the current number | |
4599 | * of missing disks if the container is populated | |
d23fe947 | 4600 | */ |
3960e579 | 4601 | if (super->current_vol == 0) { |
d23fe947 | 4602 | mpb->num_disks = 0; |
3960e579 DW |
4603 | for (i = 0; i < info->failed_disks; i++) { |
4604 | struct imsm_disk *disk; | |
4605 | ||
4606 | mpb->num_disks++; | |
4607 | disk = __get_imsm_disk(mpb, i); | |
4608 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4609 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4610 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4611 | "missing:%d", i); | |
4612 | } | |
4613 | find_missing(super); | |
4614 | } else { | |
4615 | int missing = 0; | |
4616 | struct dl *d; | |
4617 | ||
4618 | for (d = super->missing; d; d = d->next) | |
4619 | missing++; | |
4620 | if (info->failed_disks > missing) { | |
4621 | fprintf(stderr, Name": unable to add 'missing' disk to container\n"); | |
4622 | return 0; | |
4623 | } | |
4624 | } | |
5a038140 | 4625 | |
aa534678 DW |
4626 | if (!check_name(super, name, 0)) |
4627 | return 0; | |
ba2de7ba DW |
4628 | dv = malloc(sizeof(*dv)); |
4629 | if (!dv) { | |
4630 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
4631 | return 0; | |
4632 | } | |
1a2487c2 | 4633 | dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
949c47a0 | 4634 | if (!dev) { |
ba2de7ba | 4635 | free(dv); |
949c47a0 DW |
4636 | fprintf(stderr, Name": could not allocate raid device\n"); |
4637 | return 0; | |
4638 | } | |
1a2487c2 | 4639 | |
c2c087e6 | 4640 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 4641 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 4642 | info->layout, info->chunk_size, |
5551b113 | 4643 | size * 2); |
979d38be DW |
4644 | /* round array size down to closest MB */ |
4645 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4646 | ||
c2c087e6 DW |
4647 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4648 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4649 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4650 | vol = &dev->vol; |
4651 | vol->migr_state = 0; | |
1484e727 | 4652 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4653 | vol->dirty = !info->state; |
f8f603f1 | 4654 | vol->curr_migr_unit = 0; |
238c0a71 | 4655 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
4656 | set_pba_of_lba0(map, super->create_offset); |
4657 | set_blocks_per_member(map, info_to_blocks_per_member(info, size)); | |
ef6ffade | 4658 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4659 | map->failed_disk_num = ~0; |
bf4442ab AK |
4660 | if (info->level > 0) |
4661 | map->map_state = IMSM_T_STATE_UNINITIALIZED; | |
4662 | else | |
4663 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4664 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4665 | map->ddf = 1; |
ef6ffade DW |
4666 | |
4667 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4668 | free(dev); |
4669 | free(dv); | |
ef6ffade DW |
4670 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
4671 | "in a raid1 volume\n"); | |
4672 | return 0; | |
4673 | } | |
81062a36 DW |
4674 | |
4675 | map->raid_level = info->level; | |
4d1313e9 | 4676 | if (info->level == 10) { |
c2c087e6 | 4677 | map->raid_level = 1; |
4d1313e9 | 4678 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4679 | } else if (info->level == 1) |
4680 | map->num_domains = info->raid_disks; | |
4681 | else | |
ff596308 | 4682 | map->num_domains = 1; |
81062a36 | 4683 | |
5551b113 CA |
4684 | /* info->size is only int so use the 'size' parameter instead */ |
4685 | num_data_stripes = (size * 2) / info_to_blocks_per_strip(info); | |
4686 | num_data_stripes /= map->num_domains; | |
4687 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 4688 | |
c2c087e6 DW |
4689 | map->num_members = info->raid_disks; |
4690 | for (i = 0; i < map->num_members; i++) { | |
4691 | /* initialized in add_to_super */ | |
4eb26970 | 4692 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4693 | } |
949c47a0 | 4694 | mpb->num_raid_devs++; |
ba2de7ba DW |
4695 | |
4696 | dv->dev = dev; | |
4697 | dv->index = super->current_vol; | |
4698 | dv->next = super->devlist; | |
4699 | super->devlist = dv; | |
c2c087e6 | 4700 | |
4d1313e9 DW |
4701 | imsm_update_version_info(super); |
4702 | ||
c2c087e6 | 4703 | return 1; |
cdddbdbc DW |
4704 | } |
4705 | ||
bf5a934a DW |
4706 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4707 | unsigned long long size, char *name, | |
4708 | char *homehost, int *uuid) | |
4709 | { | |
4710 | /* This is primarily called by Create when creating a new array. | |
4711 | * We will then get add_to_super called for each component, and then | |
4712 | * write_init_super called to write it out to each device. | |
4713 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4714 | * array. | |
4715 | * To create on a pre-existing array a different method will be called. | |
4716 | * This one is just for fresh drives. | |
4717 | */ | |
4718 | struct intel_super *super; | |
4719 | struct imsm_super *mpb; | |
4720 | size_t mpb_size; | |
4d1313e9 | 4721 | char *version; |
bf5a934a | 4722 | |
bf5a934a | 4723 | if (st->sb) |
e683ca88 DW |
4724 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
4725 | ||
4726 | if (info) | |
4727 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4728 | else | |
4729 | mpb_size = 512; | |
bf5a934a | 4730 | |
49133e57 | 4731 | super = alloc_super(); |
e683ca88 | 4732 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4733 | free(super); |
e683ca88 DW |
4734 | super = NULL; |
4735 | } | |
4736 | if (!super) { | |
4737 | fprintf(stderr, Name | |
4738 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
4739 | return 0; |
4740 | } | |
17a4eaf9 | 4741 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
8e59f3d8 AK |
4742 | fprintf(stderr, Name |
4743 | ": %s could not allocate migr_rec buffer\n", __func__); | |
4744 | free(super->buf); | |
4745 | free(super); | |
4746 | return 0; | |
4747 | } | |
e683ca88 | 4748 | memset(super->buf, 0, mpb_size); |
ef649044 | 4749 | mpb = super->buf; |
e683ca88 DW |
4750 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4751 | st->sb = super; | |
4752 | ||
4753 | if (info == NULL) { | |
4754 | /* zeroing superblock */ | |
4755 | return 0; | |
4756 | } | |
bf5a934a | 4757 | |
4d1313e9 DW |
4758 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4759 | ||
4760 | version = (char *) mpb->sig; | |
4761 | strcpy(version, MPB_SIGNATURE); | |
4762 | version += strlen(MPB_SIGNATURE); | |
4763 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4764 | |
bf5a934a DW |
4765 | return 1; |
4766 | } | |
4767 | ||
0e600426 | 4768 | #ifndef MDASSEMBLE |
f20c3968 | 4769 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4770 | int fd, char *devname) |
4771 | { | |
4772 | struct intel_super *super = st->sb; | |
d23fe947 | 4773 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4774 | struct imsm_disk *_disk; |
bf5a934a DW |
4775 | struct imsm_dev *dev; |
4776 | struct imsm_map *map; | |
3960e579 | 4777 | struct dl *dl, *df; |
4eb26970 | 4778 | int slot; |
bf5a934a | 4779 | |
949c47a0 | 4780 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4781 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4782 | |
208933a7 N |
4783 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
4784 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
4785 | devname); | |
4786 | return 1; | |
4787 | } | |
4788 | ||
efb30e7f DW |
4789 | if (fd == -1) { |
4790 | /* we're doing autolayout so grab the pre-marked (in | |
4791 | * validate_geometry) raid_disk | |
4792 | */ | |
4793 | for (dl = super->disks; dl; dl = dl->next) | |
4794 | if (dl->raiddisk == dk->raid_disk) | |
4795 | break; | |
4796 | } else { | |
4797 | for (dl = super->disks; dl ; dl = dl->next) | |
4798 | if (dl->major == dk->major && | |
4799 | dl->minor == dk->minor) | |
4800 | break; | |
4801 | } | |
d23fe947 | 4802 | |
208933a7 N |
4803 | if (!dl) { |
4804 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 4805 | return 1; |
208933a7 | 4806 | } |
bf5a934a | 4807 | |
d23fe947 DW |
4808 | /* add a pristine spare to the metadata */ |
4809 | if (dl->index < 0) { | |
4810 | dl->index = super->anchor->num_disks; | |
4811 | super->anchor->num_disks++; | |
4812 | } | |
4eb26970 DW |
4813 | /* Check the device has not already been added */ |
4814 | slot = get_imsm_disk_slot(map, dl->index); | |
4815 | if (slot >= 0 && | |
238c0a71 | 4816 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
4817 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
4818 | devname); | |
4819 | return 1; | |
4820 | } | |
656b6b5a | 4821 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4822 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4823 | |
3960e579 DW |
4824 | /* update size of 'missing' disks to be at least as large as the |
4825 | * largest acitve member (we only have dummy missing disks when | |
4826 | * creating the first volume) | |
4827 | */ | |
4828 | if (super->current_vol == 0) { | |
4829 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
4830 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
4831 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
4832 | _disk = __get_imsm_disk(mpb, df->index); |
4833 | *_disk = df->disk; | |
4834 | } | |
4835 | } | |
4836 | ||
4837 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
4838 | for (df = super->missing; df; df = df->next) | |
4839 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 4840 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
4841 | |
4842 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
4843 | continue; | |
4844 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 4845 | if (is_gen_migration(dev)) { |
238c0a71 AK |
4846 | struct imsm_map *map2 = get_imsm_map(dev, |
4847 | MAP_1); | |
0a108d63 AK |
4848 | int slot2 = get_imsm_disk_slot(map2, df->index); |
4849 | if ((slot2 < map2->num_members) && | |
4850 | (slot2 >= 0)) { | |
1ace8403 | 4851 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
4852 | slot2, |
4853 | MAP_1); | |
1ace8403 AK |
4854 | if ((unsigned)df->index == |
4855 | ord_to_idx(ord2)) | |
4856 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 4857 | slot2, |
1ace8403 AK |
4858 | df->index | |
4859 | IMSM_ORD_REBUILD); | |
4860 | } | |
4861 | } | |
3960e579 DW |
4862 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
4863 | break; | |
4864 | } | |
4865 | ||
d23fe947 DW |
4866 | /* if we are creating the first raid device update the family number */ |
4867 | if (super->current_vol == 0) { | |
4868 | __u32 sum; | |
4869 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 4870 | |
3960e579 | 4871 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a AW |
4872 | if (!_dev || !_disk) { |
4873 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
4874 | return 1; | |
4875 | } | |
d23fe947 DW |
4876 | *_dev = *dev; |
4877 | *_disk = dl->disk; | |
148acb7b DW |
4878 | sum = random32(); |
4879 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4880 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4881 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4882 | } |
ca0748fa | 4883 | super->current_disk = dl; |
f20c3968 | 4884 | return 0; |
bf5a934a DW |
4885 | } |
4886 | ||
a8619d23 AK |
4887 | /* mark_spare() |
4888 | * Function marks disk as spare and restores disk serial | |
4889 | * in case it was previously marked as failed by takeover operation | |
4890 | * reruns: | |
4891 | * -1 : critical error | |
4892 | * 0 : disk is marked as spare but serial is not set | |
4893 | * 1 : success | |
4894 | */ | |
4895 | int mark_spare(struct dl *disk) | |
4896 | { | |
4897 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
4898 | int ret_val = -1; | |
4899 | ||
4900 | if (!disk) | |
4901 | return ret_val; | |
4902 | ||
4903 | ret_val = 0; | |
4904 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
4905 | /* Restore disk serial number, because takeover marks disk | |
4906 | * as failed and adds to serial ':0' before it becomes | |
4907 | * a spare disk. | |
4908 | */ | |
4909 | serialcpy(disk->serial, serial); | |
4910 | serialcpy(disk->disk.serial, serial); | |
4911 | ret_val = 1; | |
4912 | } | |
4913 | disk->disk.status = SPARE_DISK; | |
4914 | disk->index = -1; | |
4915 | ||
4916 | return ret_val; | |
4917 | } | |
88654014 | 4918 | |
f20c3968 | 4919 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4920 | int fd, char *devname) |
cdddbdbc | 4921 | { |
c2c087e6 | 4922 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4923 | struct dl *dd; |
4924 | unsigned long long size; | |
f2f27e63 | 4925 | __u32 id; |
c2c087e6 DW |
4926 | int rv; |
4927 | struct stat stb; | |
4928 | ||
88654014 LM |
4929 | /* If we are on an RAID enabled platform check that the disk is |
4930 | * attached to the raid controller. | |
4931 | * We do not need to test disks attachment for container based additions, | |
4932 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4933 | */ |
d424212e | 4934 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4935 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4936 | if (rv != 0) { |
4937 | dprintf("capability: %p fd: %d ret: %d\n", | |
4938 | super->orom, fd, rv); | |
4939 | return 1; | |
88c32bb1 DW |
4940 | } |
4941 | ||
f20c3968 DW |
4942 | if (super->current_vol >= 0) |
4943 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4944 | |
c2c087e6 DW |
4945 | fstat(fd, &stb); |
4946 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 4947 | if (!dd) { |
c2c087e6 DW |
4948 | fprintf(stderr, |
4949 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 4950 | return 1; |
c2c087e6 DW |
4951 | } |
4952 | memset(dd, 0, sizeof(*dd)); | |
4953 | dd->major = major(stb.st_rdev); | |
4954 | dd->minor = minor(stb.st_rdev); | |
c2c087e6 | 4955 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 4956 | dd->fd = fd; |
689c9bf3 | 4957 | dd->e = NULL; |
1a64be56 | 4958 | dd->action = DISK_ADD; |
c2c087e6 | 4959 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4960 | if (rv) { |
c2c087e6 | 4961 | fprintf(stderr, |
0030e8d6 | 4962 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4963 | free(dd); |
0030e8d6 | 4964 | abort(); |
c2c087e6 DW |
4965 | } |
4966 | ||
c2c087e6 DW |
4967 | get_dev_size(fd, NULL, &size); |
4968 | size /= 512; | |
1f24f035 | 4969 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
4970 | set_total_blocks(&dd->disk, size); |
4971 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
4972 | struct imsm_super *mpb = super->anchor; | |
4973 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
4974 | } | |
a8619d23 | 4975 | mark_spare(dd); |
c2c087e6 | 4976 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4977 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4978 | else |
b9f594fe | 4979 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4980 | |
4981 | if (st->update_tail) { | |
1a64be56 LM |
4982 | dd->next = super->disk_mgmt_list; |
4983 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4984 | } else { |
4985 | dd->next = super->disks; | |
4986 | super->disks = dd; | |
ceaf0ee1 | 4987 | super->updates_pending++; |
43dad3d6 | 4988 | } |
f20c3968 DW |
4989 | |
4990 | return 0; | |
cdddbdbc DW |
4991 | } |
4992 | ||
1a64be56 LM |
4993 | |
4994 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4995 | { | |
4996 | struct intel_super *super = st->sb; | |
4997 | struct dl *dd; | |
4998 | ||
4999 | /* remove from super works only in mdmon - for communication | |
5000 | * manager - monitor. Check if communication memory buffer | |
5001 | * is prepared. | |
5002 | */ | |
5003 | if (!st->update_tail) { | |
5004 | fprintf(stderr, | |
5005 | Name ": %s shall be used in mdmon context only" | |
5006 | "(line %d).\n", __func__, __LINE__); | |
5007 | return 1; | |
5008 | } | |
5009 | dd = malloc(sizeof(*dd)); | |
5010 | if (!dd) { | |
5011 | fprintf(stderr, | |
5012 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
5013 | return 1; | |
5014 | } | |
5015 | memset(dd, 0, sizeof(*dd)); | |
5016 | dd->major = dk->major; | |
5017 | dd->minor = dk->minor; | |
1a64be56 | 5018 | dd->fd = -1; |
a8619d23 | 5019 | mark_spare(dd); |
1a64be56 LM |
5020 | dd->action = DISK_REMOVE; |
5021 | ||
5022 | dd->next = super->disk_mgmt_list; | |
5023 | super->disk_mgmt_list = dd; | |
5024 | ||
5025 | ||
5026 | return 0; | |
5027 | } | |
5028 | ||
f796af5d DW |
5029 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5030 | ||
5031 | static union { | |
5032 | char buf[512]; | |
5033 | struct imsm_super anchor; | |
5034 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 5035 | |
d23fe947 DW |
5036 | /* spare records have their own family number and do not have any defined raid |
5037 | * devices | |
5038 | */ | |
5039 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5040 | { | |
d23fe947 | 5041 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5042 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5043 | __u32 sum; |
5044 | struct dl *d; | |
5045 | ||
f796af5d DW |
5046 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
5047 | spare->generation_num = __cpu_to_le32(1UL), | |
5048 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
5049 | spare->num_disks = 1, | |
5050 | spare->num_raid_devs = 0, | |
5051 | spare->cache_size = mpb->cache_size, | |
5052 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
5053 | ||
5054 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5055 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5056 | |
5057 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5058 | if (d->index != -1) |
d23fe947 DW |
5059 | continue; |
5060 | ||
f796af5d | 5061 | spare->disk[0] = d->disk; |
027c374f CA |
5062 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5063 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5064 | ||
f796af5d DW |
5065 | sum = __gen_imsm_checksum(spare); |
5066 | spare->family_num = __cpu_to_le32(sum); | |
5067 | spare->orig_family_num = 0; | |
5068 | sum = __gen_imsm_checksum(spare); | |
5069 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5070 | |
f796af5d | 5071 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
5072 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
5073 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 5074 | return 1; |
d23fe947 DW |
5075 | } |
5076 | if (doclose) { | |
5077 | close(d->fd); | |
5078 | d->fd = -1; | |
5079 | } | |
5080 | } | |
5081 | ||
e74255d9 | 5082 | return 0; |
d23fe947 DW |
5083 | } |
5084 | ||
36988a3d | 5085 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5086 | { |
36988a3d | 5087 | struct intel_super *super = st->sb; |
949c47a0 | 5088 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5089 | struct dl *d; |
5090 | __u32 generation; | |
5091 | __u32 sum; | |
d23fe947 | 5092 | int spares = 0; |
949c47a0 | 5093 | int i; |
a48ac0a8 | 5094 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5095 | int num_disks = 0; |
146c6260 | 5096 | int clear_migration_record = 1; |
cdddbdbc | 5097 | |
c2c087e6 DW |
5098 | /* 'generation' is incremented everytime the metadata is written */ |
5099 | generation = __le32_to_cpu(mpb->generation_num); | |
5100 | generation++; | |
5101 | mpb->generation_num = __cpu_to_le32(generation); | |
5102 | ||
148acb7b DW |
5103 | /* fix up cases where previous mdadm releases failed to set |
5104 | * orig_family_num | |
5105 | */ | |
5106 | if (mpb->orig_family_num == 0) | |
5107 | mpb->orig_family_num = mpb->family_num; | |
5108 | ||
d23fe947 | 5109 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5110 | if (d->index == -1) |
d23fe947 | 5111 | spares++; |
36988a3d | 5112 | else { |
d23fe947 | 5113 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5114 | num_disks++; |
5115 | } | |
d23fe947 | 5116 | } |
36988a3d | 5117 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5118 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5119 | num_disks++; |
5120 | } | |
5121 | mpb->num_disks = num_disks; | |
5122 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5123 | |
949c47a0 DW |
5124 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5125 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5126 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5127 | if (dev && dev2) { | |
5128 | imsm_copy_dev(dev, dev2); | |
5129 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5130 | } | |
146c6260 AK |
5131 | if (is_gen_migration(dev2)) |
5132 | clear_migration_record = 0; | |
949c47a0 | 5133 | } |
a48ac0a8 DW |
5134 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
5135 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 5136 | |
c2c087e6 | 5137 | /* recalculate checksum */ |
949c47a0 | 5138 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5139 | mpb->check_sum = __cpu_to_le32(sum); |
5140 | ||
51d83f5d AK |
5141 | if (super->clean_migration_record_by_mdmon) { |
5142 | clear_migration_record = 1; | |
5143 | super->clean_migration_record_by_mdmon = 0; | |
5144 | } | |
146c6260 | 5145 | if (clear_migration_record) |
17a4eaf9 | 5146 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); |
146c6260 | 5147 | |
d23fe947 | 5148 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5149 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5150 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5151 | continue; |
30602f53 | 5152 | |
146c6260 AK |
5153 | if (clear_migration_record) { |
5154 | unsigned long long dsize; | |
5155 | ||
5156 | get_dev_size(d->fd, NULL, &dsize); | |
5157 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
17a4eaf9 AK |
5158 | if (write(d->fd, super->migr_rec_buf, |
5159 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
9e2d750d | 5160 | perror("Write migr_rec failed"); |
146c6260 AK |
5161 | } |
5162 | } | |
51d83f5d AK |
5163 | |
5164 | if (store_imsm_mpb(d->fd, mpb)) | |
5165 | fprintf(stderr, | |
5166 | "%s: failed for device %d:%d (fd: %d)%s\n", | |
5167 | __func__, d->major, d->minor, | |
5168 | d->fd, strerror(errno)); | |
5169 | ||
c2c087e6 DW |
5170 | if (doclose) { |
5171 | close(d->fd); | |
5172 | d->fd = -1; | |
5173 | } | |
5174 | } | |
5175 | ||
d23fe947 DW |
5176 | if (spares) |
5177 | return write_super_imsm_spares(super, doclose); | |
5178 | ||
e74255d9 | 5179 | return 0; |
c2c087e6 DW |
5180 | } |
5181 | ||
0e600426 | 5182 | |
9b1fb677 | 5183 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5184 | { |
5185 | size_t len; | |
5186 | struct imsm_update_create_array *u; | |
5187 | struct intel_super *super = st->sb; | |
9b1fb677 | 5188 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5189 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5190 | struct disk_info *inf; |
5191 | struct imsm_disk *disk; | |
5192 | int i; | |
43dad3d6 | 5193 | |
54c2c1ea DW |
5194 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5195 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
5196 | u = malloc(len); |
5197 | if (!u) { | |
5198 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
5199 | __func__); | |
5200 | return 1; | |
5201 | } | |
5202 | ||
5203 | u->type = update_create_array; | |
9b1fb677 | 5204 | u->dev_idx = dev_idx; |
43dad3d6 | 5205 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5206 | inf = get_disk_info(u); |
5207 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5208 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5209 | |
54c2c1ea DW |
5210 | disk = get_imsm_disk(super, idx); |
5211 | serialcpy(inf[i].serial, disk->serial); | |
5212 | } | |
43dad3d6 DW |
5213 | append_metadata_update(st, u, len); |
5214 | ||
5215 | return 0; | |
5216 | } | |
5217 | ||
1a64be56 | 5218 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5219 | { |
5220 | struct intel_super *super = st->sb; | |
5221 | size_t len; | |
1a64be56 | 5222 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5223 | |
1a64be56 | 5224 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5225 | return 0; |
5226 | ||
5227 | len = sizeof(*u); | |
5228 | u = malloc(len); | |
5229 | if (!u) { | |
5230 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
5231 | __func__); | |
5232 | return 1; | |
5233 | } | |
5234 | ||
1a64be56 | 5235 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5236 | append_metadata_update(st, u, len); |
5237 | ||
5238 | return 0; | |
5239 | } | |
5240 | ||
c2c087e6 DW |
5241 | static int write_init_super_imsm(struct supertype *st) |
5242 | { | |
9b1fb677 DW |
5243 | struct intel_super *super = st->sb; |
5244 | int current_vol = super->current_vol; | |
5245 | ||
5246 | /* we are done with current_vol reset it to point st at the container */ | |
5247 | super->current_vol = -1; | |
5248 | ||
8273f55e | 5249 | if (st->update_tail) { |
43dad3d6 DW |
5250 | /* queue the recently created array / added disk |
5251 | * as a metadata update */ | |
43dad3d6 | 5252 | int rv; |
8273f55e | 5253 | |
43dad3d6 | 5254 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5255 | if (current_vol < 0) { |
1a64be56 LM |
5256 | /* in the mgmt (add/remove) disk case we are running |
5257 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5258 | */ |
1a64be56 | 5259 | return mgmt_disk(st); |
43dad3d6 | 5260 | } else |
9b1fb677 | 5261 | rv = create_array(st, current_vol); |
8273f55e | 5262 | |
43dad3d6 | 5263 | return rv; |
d682f344 N |
5264 | } else { |
5265 | struct dl *d; | |
5266 | for (d = super->disks; d; d = d->next) | |
5267 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 5268 | return write_super_imsm(st, 1); |
d682f344 | 5269 | } |
cdddbdbc | 5270 | } |
0e600426 | 5271 | #endif |
cdddbdbc | 5272 | |
e683ca88 | 5273 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5274 | { |
e683ca88 DW |
5275 | struct intel_super *super = st->sb; |
5276 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5277 | |
e683ca88 | 5278 | if (!mpb) |
ad97895e DW |
5279 | return 1; |
5280 | ||
1799c9e8 | 5281 | #ifndef MDASSEMBLE |
e683ca88 | 5282 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5283 | #else |
5284 | return 1; | |
5285 | #endif | |
cdddbdbc DW |
5286 | } |
5287 | ||
0e600426 N |
5288 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
5289 | { | |
5290 | return __le32_to_cpu(mpb->bbm_log_size); | |
5291 | } | |
5292 | ||
5293 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
5294 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5295 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 5296 | unsigned long long size, char *dev, |
2c514b71 NB |
5297 | unsigned long long *freesize, |
5298 | int verbose) | |
cdddbdbc | 5299 | { |
c2c087e6 DW |
5300 | int fd; |
5301 | unsigned long long ldsize; | |
f2f5c343 LM |
5302 | struct intel_super *super=NULL; |
5303 | int rv = 0; | |
cdddbdbc | 5304 | |
c2c087e6 DW |
5305 | if (level != LEVEL_CONTAINER) |
5306 | return 0; | |
5307 | if (!dev) | |
5308 | return 1; | |
5309 | ||
5310 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5311 | if (fd < 0) { | |
2c514b71 NB |
5312 | if (verbose) |
5313 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
5314 | dev, strerror(errno)); | |
c2c087e6 DW |
5315 | return 0; |
5316 | } | |
5317 | if (!get_dev_size(fd, dev, &ldsize)) { | |
5318 | close(fd); | |
5319 | return 0; | |
5320 | } | |
f2f5c343 LM |
5321 | |
5322 | /* capabilities retrieve could be possible | |
5323 | * note that there is no fd for the disks in array. | |
5324 | */ | |
5325 | super = alloc_super(); | |
5326 | if (!super) { | |
5327 | fprintf(stderr, | |
5328 | Name ": malloc of %zu failed.\n", | |
5329 | sizeof(*super)); | |
5330 | close(fd); | |
5331 | return 0; | |
5332 | } | |
5333 | ||
d424212e | 5334 | rv = find_intel_hba_capability(fd, super, verbose ? dev : NULL); |
f2f5c343 LM |
5335 | if (rv != 0) { |
5336 | #if DEBUG | |
5337 | char str[256]; | |
5338 | fd2devname(fd, str); | |
5339 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
5340 | fd, str, super->orom, rv, raiddisks); | |
5341 | #endif | |
5342 | /* no orom/efi or non-intel hba of the disk */ | |
5343 | close(fd); | |
5344 | free_imsm(super); | |
5345 | return 0; | |
5346 | } | |
c2c087e6 | 5347 | close(fd); |
9126b9a8 CA |
5348 | if (super->orom) { |
5349 | if (raiddisks > super->orom->tds) { | |
5350 | if (verbose) | |
5351 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
5352 | " platform supported disks: %d\n", | |
5353 | raiddisks, super->orom->tds); | |
5354 | free_imsm(super); | |
5355 | return 0; | |
5356 | } | |
5357 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
5358 | (ldsize >> 9) >> 32 > 0) { | |
5359 | if (verbose) | |
5360 | fprintf(stderr, Name ": %s exceeds maximum platform supported size\n", dev); | |
5361 | free_imsm(super); | |
5362 | return 0; | |
5363 | } | |
f2f5c343 | 5364 | } |
c2c087e6 DW |
5365 | |
5366 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 5367 | free_imsm(super); |
c2c087e6 DW |
5368 | |
5369 | return 1; | |
cdddbdbc DW |
5370 | } |
5371 | ||
0dcecb2e DW |
5372 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5373 | { | |
5374 | const unsigned long long base_start = e[*idx].start; | |
5375 | unsigned long long end = base_start + e[*idx].size; | |
5376 | int i; | |
5377 | ||
5378 | if (base_start == end) | |
5379 | return 0; | |
5380 | ||
5381 | *idx = *idx + 1; | |
5382 | for (i = *idx; i < num_extents; i++) { | |
5383 | /* extend overlapping extents */ | |
5384 | if (e[i].start >= base_start && | |
5385 | e[i].start <= end) { | |
5386 | if (e[i].size == 0) | |
5387 | return 0; | |
5388 | if (e[i].start + e[i].size > end) | |
5389 | end = e[i].start + e[i].size; | |
5390 | } else if (e[i].start > end) { | |
5391 | *idx = i; | |
5392 | break; | |
5393 | } | |
5394 | } | |
5395 | ||
5396 | return end - base_start; | |
5397 | } | |
5398 | ||
5399 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5400 | { | |
5401 | /* build a composite disk with all known extents and generate a new | |
5402 | * 'maxsize' given the "all disks in an array must share a common start | |
5403 | * offset" constraint | |
5404 | */ | |
5405 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
5406 | struct dl *dl; | |
5407 | int i, j; | |
5408 | int start_extent; | |
5409 | unsigned long long pos; | |
b9d77223 | 5410 | unsigned long long start = 0; |
0dcecb2e DW |
5411 | unsigned long long maxsize; |
5412 | unsigned long reserve; | |
5413 | ||
5414 | if (!e) | |
a7dd165b | 5415 | return 0; |
0dcecb2e DW |
5416 | |
5417 | /* coalesce and sort all extents. also, check to see if we need to | |
5418 | * reserve space between member arrays | |
5419 | */ | |
5420 | j = 0; | |
5421 | for (dl = super->disks; dl; dl = dl->next) { | |
5422 | if (!dl->e) | |
5423 | continue; | |
5424 | for (i = 0; i < dl->extent_cnt; i++) | |
5425 | e[j++] = dl->e[i]; | |
5426 | } | |
5427 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5428 | ||
5429 | /* merge extents */ | |
5430 | i = 0; | |
5431 | j = 0; | |
5432 | while (i < sum_extents) { | |
5433 | e[j].start = e[i].start; | |
5434 | e[j].size = find_size(e, &i, sum_extents); | |
5435 | j++; | |
5436 | if (e[j-1].size == 0) | |
5437 | break; | |
5438 | } | |
5439 | ||
5440 | pos = 0; | |
5441 | maxsize = 0; | |
5442 | start_extent = 0; | |
5443 | i = 0; | |
5444 | do { | |
5445 | unsigned long long esize; | |
5446 | ||
5447 | esize = e[i].start - pos; | |
5448 | if (esize >= maxsize) { | |
5449 | maxsize = esize; | |
5450 | start = pos; | |
5451 | start_extent = i; | |
5452 | } | |
5453 | pos = e[i].start + e[i].size; | |
5454 | i++; | |
5455 | } while (e[i-1].size); | |
5456 | free(e); | |
5457 | ||
a7dd165b DW |
5458 | if (maxsize == 0) |
5459 | return 0; | |
5460 | ||
5461 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5462 | * container | |
5463 | */ | |
0dcecb2e DW |
5464 | if (start_extent > 0) |
5465 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5466 | else | |
5467 | reserve = 0; | |
5468 | ||
5469 | if (maxsize < reserve) | |
a7dd165b | 5470 | return 0; |
0dcecb2e | 5471 | |
5551b113 | 5472 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 5473 | if (start + reserve > super->create_offset) |
a7dd165b | 5474 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5475 | super->create_offset = start + reserve; |
5476 | ||
5477 | return maxsize - reserve; | |
5478 | } | |
5479 | ||
88c32bb1 DW |
5480 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5481 | { | |
5482 | if (level < 0 || level == 6 || level == 4) | |
5483 | return 0; | |
5484 | ||
5485 | /* if we have an orom prevent invalid raid levels */ | |
5486 | if (orom) | |
5487 | switch (level) { | |
5488 | case 0: return imsm_orom_has_raid0(orom); | |
5489 | case 1: | |
5490 | if (raiddisks > 2) | |
5491 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5492 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5493 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5494 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5495 | } |
5496 | else | |
5497 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5498 | ||
5499 | return 0; | |
5500 | } | |
5501 | ||
ca9de185 LM |
5502 | |
5503 | static int | |
5504 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
5505 | int dpa, int verbose) | |
5506 | { | |
5507 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
5508 | struct mdstat_ent *memb = NULL; | |
5509 | int count = 0; | |
5510 | int num = 0; | |
5511 | struct md_list *dv = NULL; | |
5512 | int found; | |
5513 | ||
5514 | for (memb = mdstat ; memb ; memb = memb->next) { | |
5515 | if (memb->metadata_version && | |
5516 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
5517 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
5518 | !is_subarray(memb->metadata_version+9) && | |
5519 | memb->members) { | |
5520 | struct dev_member *dev = memb->members; | |
5521 | int fd = -1; | |
5522 | while(dev && (fd < 0)) { | |
5523 | char *path = malloc(strlen(dev->name) + strlen("/dev/") + 1); | |
5524 | if (path) { | |
5525 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
5526 | if (num > 0) | |
5527 | fd = open(path, O_RDONLY, 0); | |
5528 | if ((num <= 0) || (fd < 0)) { | |
5529 | pr_vrb(": Cannot open %s: %s\n", | |
5530 | dev->name, strerror(errno)); | |
5531 | } | |
5532 | free(path); | |
5533 | } | |
5534 | dev = dev->next; | |
5535 | } | |
5536 | found = 0; | |
5537 | if ((fd >= 0) && disk_attached_to_hba(fd, hba)) { | |
5538 | struct mdstat_ent *vol; | |
5539 | for (vol = mdstat ; vol ; vol = vol->next) { | |
5540 | if ((vol->active > 0) && | |
5541 | vol->metadata_version && | |
5542 | is_container_member(vol, memb->dev)) { | |
5543 | found++; | |
5544 | count++; | |
5545 | } | |
5546 | } | |
5547 | if (*devlist && (found < dpa)) { | |
5548 | dv = calloc(1, sizeof(*dv)); | |
5549 | if (dv == NULL) | |
5550 | fprintf(stderr, Name ": calloc failed\n"); | |
5551 | else { | |
5552 | dv->devname = malloc(strlen(memb->dev) + strlen("/dev/") + 1); | |
5553 | if (dv->devname != NULL) { | |
5554 | sprintf(dv->devname, "%s%s", "/dev/", memb->dev); | |
5555 | dv->found = found; | |
5556 | dv->used = 0; | |
5557 | dv->next = *devlist; | |
5558 | *devlist = dv; | |
5559 | } else | |
5560 | free(dv); | |
5561 | } | |
5562 | } | |
5563 | } | |
5564 | if (fd >= 0) | |
5565 | close(fd); | |
5566 | } | |
5567 | } | |
5568 | free_mdstat(mdstat); | |
5569 | return count; | |
5570 | } | |
5571 | ||
5572 | #ifdef DEBUG_LOOP | |
5573 | static struct md_list* | |
5574 | get_loop_devices(void) | |
5575 | { | |
5576 | int i; | |
5577 | struct md_list *devlist = NULL; | |
5578 | struct md_list *dv = NULL; | |
5579 | ||
5580 | for(i = 0; i < 12; i++) { | |
5581 | dv = calloc(1, sizeof(*dv)); | |
5582 | if (dv == NULL) { | |
5583 | fprintf(stderr, Name ": calloc failed\n"); | |
5584 | break; | |
5585 | } | |
5586 | dv->devname = malloc(40); | |
5587 | if (dv->devname == NULL) { | |
5588 | fprintf(stderr, Name ": malloc failed\n"); | |
5589 | free(dv); | |
5590 | break; | |
5591 | } | |
5592 | sprintf(dv->devname, "/dev/loop%d", i); | |
5593 | dv->next = devlist; | |
5594 | devlist = dv; | |
5595 | } | |
5596 | return devlist; | |
5597 | } | |
5598 | #endif | |
5599 | ||
5600 | static struct md_list* | |
5601 | get_devices(const char *hba_path) | |
5602 | { | |
5603 | struct md_list *devlist = NULL; | |
5604 | struct md_list *dv = NULL; | |
5605 | struct dirent *ent; | |
5606 | DIR *dir; | |
5607 | int err = 0; | |
5608 | ||
5609 | #if DEBUG_LOOP | |
5610 | devlist = get_loop_devices(); | |
5611 | return devlist; | |
5612 | #endif | |
5613 | /* scroll through /sys/dev/block looking for devices attached to | |
5614 | * this hba | |
5615 | */ | |
5616 | dir = opendir("/sys/dev/block"); | |
5617 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
5618 | int fd; | |
5619 | char buf[1024]; | |
5620 | int major, minor; | |
5621 | char *path = NULL; | |
5622 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
5623 | continue; | |
5624 | path = devt_to_devpath(makedev(major, minor)); | |
5625 | if (!path) | |
5626 | continue; | |
5627 | if (!path_attached_to_hba(path, hba_path)) { | |
5628 | free(path); | |
5629 | path = NULL; | |
5630 | continue; | |
5631 | } | |
5632 | free(path); | |
5633 | path = NULL; | |
5634 | fd = dev_open(ent->d_name, O_RDONLY); | |
5635 | if (fd >= 0) { | |
5636 | fd2devname(fd, buf); | |
5637 | close(fd); | |
5638 | } else { | |
5639 | fprintf(stderr, Name ": cannot open device: %s\n", | |
5640 | ent->d_name); | |
5641 | continue; | |
5642 | } | |
5643 | ||
5644 | ||
5645 | dv = calloc(1, sizeof(*dv)); | |
5646 | if (dv == NULL) { | |
5647 | fprintf(stderr, Name ": malloc failed\n"); | |
5648 | err = 1; | |
5649 | break; | |
5650 | } | |
5651 | dv->devname = strdup(buf); | |
5652 | if (dv->devname == NULL) { | |
5653 | fprintf(stderr, Name ": malloc failed\n"); | |
5654 | err = 1; | |
5655 | free(dv); | |
5656 | break; | |
5657 | } | |
5658 | dv->next = devlist; | |
5659 | devlist = dv; | |
5660 | } | |
5661 | if (err) { | |
5662 | while(devlist) { | |
5663 | dv = devlist; | |
5664 | devlist = devlist->next; | |
5665 | free(dv->devname); | |
5666 | free(dv); | |
5667 | } | |
5668 | } | |
5669 | return devlist; | |
5670 | } | |
5671 | ||
5672 | static int | |
5673 | count_volumes_list(struct md_list *devlist, char *homehost, | |
5674 | int verbose, int *found) | |
5675 | { | |
5676 | struct md_list *tmpdev; | |
5677 | int count = 0; | |
5678 | struct supertype *st = NULL; | |
5679 | ||
5680 | /* first walk the list of devices to find a consistent set | |
5681 | * that match the criterea, if that is possible. | |
5682 | * We flag the ones we like with 'used'. | |
5683 | */ | |
5684 | *found = 0; | |
5685 | st = match_metadata_desc_imsm("imsm"); | |
5686 | if (st == NULL) { | |
5687 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5688 | return 0; | |
5689 | } | |
5690 | ||
5691 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5692 | char *devname = tmpdev->devname; | |
5693 | struct stat stb; | |
5694 | struct supertype *tst; | |
5695 | int dfd; | |
5696 | if (tmpdev->used > 1) | |
5697 | continue; | |
5698 | tst = dup_super(st); | |
5699 | if (tst == NULL) { | |
5700 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5701 | goto err_1; | |
5702 | } | |
5703 | tmpdev->container = 0; | |
5704 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
5705 | if (dfd < 0) { | |
5706 | dprintf(": cannot open device %s: %s\n", | |
5707 | devname, strerror(errno)); | |
5708 | tmpdev->used = 2; | |
5709 | } else if (fstat(dfd, &stb)< 0) { | |
5710 | /* Impossible! */ | |
5711 | dprintf(": fstat failed for %s: %s\n", | |
5712 | devname, strerror(errno)); | |
5713 | tmpdev->used = 2; | |
5714 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
5715 | dprintf(": %s is not a block device.\n", | |
5716 | devname); | |
5717 | tmpdev->used = 2; | |
5718 | } else if (must_be_container(dfd)) { | |
5719 | struct supertype *cst; | |
5720 | cst = super_by_fd(dfd, NULL); | |
5721 | if (cst == NULL) { | |
5722 | dprintf(": cannot recognize container type %s\n", | |
5723 | devname); | |
5724 | tmpdev->used = 2; | |
5725 | } else if (tst->ss != st->ss) { | |
5726 | dprintf(": non-imsm container - ignore it: %s\n", | |
5727 | devname); | |
5728 | tmpdev->used = 2; | |
5729 | } else if (!tst->ss->load_container || | |
5730 | tst->ss->load_container(tst, dfd, NULL)) | |
5731 | tmpdev->used = 2; | |
5732 | else { | |
5733 | tmpdev->container = 1; | |
5734 | } | |
5735 | if (cst) | |
5736 | cst->ss->free_super(cst); | |
5737 | } else { | |
5738 | tmpdev->st_rdev = stb.st_rdev; | |
5739 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
5740 | dprintf(": no RAID superblock on %s\n", | |
5741 | devname); | |
5742 | tmpdev->used = 2; | |
5743 | } else if (tst->ss->compare_super == NULL) { | |
5744 | dprintf(": Cannot assemble %s metadata on %s\n", | |
5745 | tst->ss->name, devname); | |
5746 | tmpdev->used = 2; | |
5747 | } | |
5748 | } | |
5749 | if (dfd >= 0) | |
5750 | close(dfd); | |
5751 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
5752 | /* Ignore unrecognised devices during auto-assembly */ | |
5753 | goto loop; | |
5754 | } | |
5755 | else { | |
5756 | struct mdinfo info; | |
5757 | tst->ss->getinfo_super(tst, &info, NULL); | |
5758 | ||
5759 | if (st->minor_version == -1) | |
5760 | st->minor_version = tst->minor_version; | |
5761 | ||
5762 | if (memcmp(info.uuid, uuid_zero, | |
5763 | sizeof(int[4])) == 0) { | |
5764 | /* this is a floating spare. It cannot define | |
5765 | * an array unless there are no more arrays of | |
5766 | * this type to be found. It can be included | |
5767 | * in an array of this type though. | |
5768 | */ | |
5769 | tmpdev->used = 3; | |
5770 | goto loop; | |
5771 | } | |
5772 | ||
5773 | if (st->ss != tst->ss || | |
5774 | st->minor_version != tst->minor_version || | |
5775 | st->ss->compare_super(st, tst) != 0) { | |
5776 | /* Some mismatch. If exactly one array matches this host, | |
5777 | * we can resolve on that one. | |
5778 | * Or, if we are auto assembling, we just ignore the second | |
5779 | * for now. | |
5780 | */ | |
5781 | dprintf(": superblock on %s doesn't match others - assembly aborted\n", | |
5782 | devname); | |
5783 | goto loop; | |
5784 | } | |
5785 | tmpdev->used = 1; | |
5786 | *found = 1; | |
5787 | dprintf("found: devname: %s\n", devname); | |
5788 | } | |
5789 | loop: | |
5790 | if (tst) | |
5791 | tst->ss->free_super(tst); | |
5792 | } | |
5793 | if (*found != 0) { | |
5794 | int err; | |
5795 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
5796 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
5797 | for (iter = head; iter; iter = iter->next) { | |
5798 | dprintf("content->text_version: %s vol\n", | |
5799 | iter->text_version); | |
5800 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
5801 | /* do not assemble arrays with unsupported | |
5802 | configurations */ | |
5803 | dprintf(": Cannot activate member %s.\n", | |
5804 | iter->text_version); | |
5805 | } else | |
5806 | count++; | |
5807 | } | |
5808 | sysfs_free(head); | |
5809 | ||
5810 | } else { | |
5811 | dprintf(" no valid super block on device list: err: %d %p\n", | |
5812 | err, st->sb); | |
5813 | } | |
5814 | } else { | |
5815 | dprintf(" no more devices to examin\n"); | |
5816 | } | |
5817 | ||
5818 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5819 | if ((tmpdev->used == 1) && (tmpdev->found)) { | |
5820 | if (count) { | |
5821 | if (count < tmpdev->found) | |
5822 | count = 0; | |
5823 | else | |
5824 | count -= tmpdev->found; | |
5825 | } | |
5826 | } | |
5827 | if (tmpdev->used == 1) | |
5828 | tmpdev->used = 4; | |
5829 | } | |
5830 | err_1: | |
5831 | if (st) | |
5832 | st->ss->free_super(st); | |
5833 | return count; | |
5834 | } | |
5835 | ||
5836 | ||
5837 | static int | |
5838 | count_volumes(char *hba, int dpa, int verbose) | |
5839 | { | |
5840 | struct md_list *devlist = NULL; | |
5841 | int count = 0; | |
5842 | int found = 0;; | |
5843 | ||
5844 | devlist = get_devices(hba); | |
5845 | /* if no intel devices return zero volumes */ | |
5846 | if (devlist == NULL) | |
5847 | return 0; | |
5848 | ||
5849 | count = active_arrays_by_format("imsm", hba, &devlist, dpa, verbose); | |
5850 | dprintf(" path: %s active arrays: %d\n", hba, count); | |
5851 | if (devlist == NULL) | |
5852 | return 0; | |
5853 | do { | |
5854 | found = 0; | |
5855 | count += count_volumes_list(devlist, | |
5856 | NULL, | |
5857 | verbose, | |
5858 | &found); | |
5859 | dprintf("found %d count: %d\n", found, count); | |
5860 | } while (found); | |
5861 | ||
5862 | dprintf("path: %s total number of volumes: %d\n", hba, count); | |
5863 | ||
5864 | while(devlist) { | |
5865 | struct md_list *dv = devlist; | |
5866 | devlist = devlist->next; | |
5867 | free(dv->devname); | |
5868 | free(dv); | |
5869 | } | |
5870 | return count; | |
5871 | } | |
5872 | ||
cd9d1ac7 DW |
5873 | static int imsm_default_chunk(const struct imsm_orom *orom) |
5874 | { | |
5875 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
5876 | * 128 if no platform detected | |
5877 | */ | |
5878 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
5879 | ||
5880 | return min(512, (1 << fs)); | |
5881 | } | |
73408129 | 5882 | |
6592ce37 DW |
5883 | static int |
5884 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 5885 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 5886 | { |
660260d0 DW |
5887 | /* check/set platform and metadata limits/defaults */ |
5888 | if (super->orom && raiddisks > super->orom->dpa) { | |
5889 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
5890 | super->orom->dpa); | |
73408129 LM |
5891 | return 0; |
5892 | } | |
5893 | ||
5894 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
660260d0 | 5895 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
5896 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
5897 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
5898 | return 0; | |
5899 | } | |
cd9d1ac7 DW |
5900 | |
5901 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
5902 | *chunk = imsm_default_chunk(super->orom); | |
5903 | ||
5904 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
5905 | pr_vrb(": platform does not support a chunk size of: " | |
5906 | "%d\n", *chunk); | |
5907 | return 0; | |
6592ce37 | 5908 | } |
cd9d1ac7 | 5909 | |
6592ce37 DW |
5910 | if (layout != imsm_level_to_layout(level)) { |
5911 | if (level == 5) | |
5912 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
5913 | else if (level == 10) | |
5914 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
5915 | else | |
5916 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
5917 | layout, level); | |
5918 | return 0; | |
5919 | } | |
2cc699af CA |
5920 | |
5921 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && chunk && | |
5922 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { | |
5923 | pr_vrb(": platform does not support a volume size over 2TB\n"); | |
5924 | return 0; | |
5925 | } | |
6592ce37 DW |
5926 | return 1; |
5927 | } | |
5928 | ||
c2c087e6 DW |
5929 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
5930 | * FIX ME add ahci details | |
5931 | */ | |
8b353278 | 5932 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 5933 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 5934 | unsigned long long size, char *dev, |
2c514b71 NB |
5935 | unsigned long long *freesize, |
5936 | int verbose) | |
cdddbdbc | 5937 | { |
c2c087e6 DW |
5938 | struct stat stb; |
5939 | struct intel_super *super = st->sb; | |
b2916f25 | 5940 | struct imsm_super *mpb; |
c2c087e6 DW |
5941 | struct dl *dl; |
5942 | unsigned long long pos = 0; | |
5943 | unsigned long long maxsize; | |
5944 | struct extent *e; | |
5945 | int i; | |
cdddbdbc | 5946 | |
88c32bb1 DW |
5947 | /* We must have the container info already read in. */ |
5948 | if (!super) | |
c2c087e6 DW |
5949 | return 0; |
5950 | ||
b2916f25 JS |
5951 | mpb = super->anchor; |
5952 | ||
2cc699af | 5953 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
d54559f0 LM |
5954 | fprintf(stderr, Name ": RAID gemetry validation failed. " |
5955 | "Cannot proceed with the action(s).\n"); | |
c2c087e6 | 5956 | return 0; |
d54559f0 | 5957 | } |
c2c087e6 DW |
5958 | if (!dev) { |
5959 | /* General test: make sure there is space for | |
2da8544a DW |
5960 | * 'raiddisks' device extents of size 'size' at a given |
5961 | * offset | |
c2c087e6 | 5962 | */ |
e46273eb | 5963 | unsigned long long minsize = size; |
b7528a20 | 5964 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5965 | int dcnt = 0; |
5966 | if (minsize == 0) | |
5967 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5968 | for (dl = super->disks; dl ; dl = dl->next) { | |
5969 | int found = 0; | |
5970 | ||
bf5a934a | 5971 | pos = 0; |
c2c087e6 DW |
5972 | i = 0; |
5973 | e = get_extents(super, dl); | |
5974 | if (!e) continue; | |
5975 | do { | |
5976 | unsigned long long esize; | |
5977 | esize = e[i].start - pos; | |
5978 | if (esize >= minsize) | |
5979 | found = 1; | |
b7528a20 | 5980 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5981 | start_offset = pos; |
5982 | break; | |
5983 | } else if (found && pos != start_offset) { | |
5984 | found = 0; | |
5985 | break; | |
5986 | } | |
c2c087e6 DW |
5987 | pos = e[i].start + e[i].size; |
5988 | i++; | |
5989 | } while (e[i-1].size); | |
5990 | if (found) | |
5991 | dcnt++; | |
5992 | free(e); | |
5993 | } | |
5994 | if (dcnt < raiddisks) { | |
2c514b71 NB |
5995 | if (verbose) |
5996 | fprintf(stderr, Name ": imsm: Not enough " | |
5997 | "devices with space for this array " | |
5998 | "(%d < %d)\n", | |
5999 | dcnt, raiddisks); | |
c2c087e6 DW |
6000 | return 0; |
6001 | } | |
6002 | return 1; | |
6003 | } | |
0dcecb2e | 6004 | |
c2c087e6 DW |
6005 | /* This device must be a member of the set */ |
6006 | if (stat(dev, &stb) < 0) | |
6007 | return 0; | |
6008 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
6009 | return 0; | |
6010 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
6011 | if (dl->major == (int)major(stb.st_rdev) && |
6012 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
6013 | break; |
6014 | } | |
6015 | if (!dl) { | |
2c514b71 NB |
6016 | if (verbose) |
6017 | fprintf(stderr, Name ": %s is not in the " | |
6018 | "same imsm set\n", dev); | |
c2c087e6 | 6019 | return 0; |
a20d2ba5 DW |
6020 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
6021 | /* If a volume is present then the current creation attempt | |
6022 | * cannot incorporate new spares because the orom may not | |
6023 | * understand this configuration (all member disks must be | |
6024 | * members of each array in the container). | |
6025 | */ | |
6026 | fprintf(stderr, Name ": %s is a spare and a volume" | |
6027 | " is already defined for this container\n", dev); | |
6028 | fprintf(stderr, Name ": The option-rom requires all member" | |
6029 | " disks to be a member of all volumes\n"); | |
6030 | return 0; | |
5fe62b94 WD |
6031 | } else if (super->orom && mpb->num_raid_devs > 0 && |
6032 | mpb->num_disks != raiddisks) { | |
6033 | fprintf(stderr, Name ": The option-rom requires all member" | |
6034 | " disks to be a member of all volumes\n"); | |
6035 | return 0; | |
c2c087e6 | 6036 | } |
0dcecb2e DW |
6037 | |
6038 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
6039 | e = get_extents(super, dl); |
6040 | maxsize = 0; | |
6041 | i = 0; | |
0dcecb2e DW |
6042 | if (e) { |
6043 | do { | |
6044 | unsigned long long esize; | |
6045 | ||
6046 | esize = e[i].start - pos; | |
6047 | if (esize >= maxsize) | |
6048 | maxsize = esize; | |
6049 | pos = e[i].start + e[i].size; | |
6050 | i++; | |
6051 | } while (e[i-1].size); | |
6052 | dl->e = e; | |
6053 | dl->extent_cnt = i; | |
6054 | } else { | |
6055 | if (verbose) | |
6056 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
6057 | dev); | |
6058 | return 0; | |
6059 | } | |
6060 | if (maxsize < size) { | |
6061 | if (verbose) | |
6062 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
6063 | dev, maxsize, size); | |
6064 | return 0; | |
6065 | } | |
6066 | ||
6067 | /* count total number of extents for merge */ | |
6068 | i = 0; | |
6069 | for (dl = super->disks; dl; dl = dl->next) | |
6070 | if (dl->e) | |
6071 | i += dl->extent_cnt; | |
6072 | ||
6073 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
6074 | |
6075 | if (!check_env("IMSM_NO_PLATFORM") && | |
6076 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
6077 | fprintf(stderr, Name ": attempting to create a second " | |
6078 | "volume with size less then remaining space. " | |
6079 | "Aborting...\n"); | |
6080 | return 0; | |
6081 | } | |
6082 | ||
a7dd165b | 6083 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
6084 | if (verbose) { |
6085 | if (maxsize == 0) | |
6086 | fprintf(stderr, Name ": no free space" | |
6087 | " left on device. Aborting...\n"); | |
6088 | else | |
6089 | fprintf(stderr, Name ": not enough space" | |
6090 | " to create volume of given size" | |
6091 | " (%llu < %llu). Aborting...\n", | |
6092 | maxsize, size); | |
6093 | } | |
0dcecb2e | 6094 | return 0; |
0dcecb2e DW |
6095 | } |
6096 | ||
c2c087e6 DW |
6097 | *freesize = maxsize; |
6098 | ||
ca9de185 LM |
6099 | if (super->orom) { |
6100 | int count = count_volumes(super->hba->path, | |
6101 | super->orom->dpa, verbose); | |
6102 | if (super->orom->vphba <= count) { | |
40110b90 | 6103 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6104 | super->orom->vphba); |
6105 | return 0; | |
6106 | } | |
6107 | } | |
c2c087e6 | 6108 | return 1; |
cdddbdbc DW |
6109 | } |
6110 | ||
efb30e7f DW |
6111 | static int reserve_space(struct supertype *st, int raiddisks, |
6112 | unsigned long long size, int chunk, | |
6113 | unsigned long long *freesize) | |
6114 | { | |
6115 | struct intel_super *super = st->sb; | |
6116 | struct imsm_super *mpb = super->anchor; | |
6117 | struct dl *dl; | |
6118 | int i; | |
6119 | int extent_cnt; | |
6120 | struct extent *e; | |
6121 | unsigned long long maxsize; | |
6122 | unsigned long long minsize; | |
6123 | int cnt; | |
6124 | int used; | |
6125 | ||
6126 | /* find the largest common start free region of the possible disks */ | |
6127 | used = 0; | |
6128 | extent_cnt = 0; | |
6129 | cnt = 0; | |
6130 | for (dl = super->disks; dl; dl = dl->next) { | |
6131 | dl->raiddisk = -1; | |
6132 | ||
6133 | if (dl->index >= 0) | |
6134 | used++; | |
6135 | ||
6136 | /* don't activate new spares if we are orom constrained | |
6137 | * and there is already a volume active in the container | |
6138 | */ | |
6139 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
6140 | continue; | |
6141 | ||
6142 | e = get_extents(super, dl); | |
6143 | if (!e) | |
6144 | continue; | |
6145 | for (i = 1; e[i-1].size; i++) | |
6146 | ; | |
6147 | dl->e = e; | |
6148 | dl->extent_cnt = i; | |
6149 | extent_cnt += i; | |
6150 | cnt++; | |
6151 | } | |
6152 | ||
6153 | maxsize = merge_extents(super, extent_cnt); | |
6154 | minsize = size; | |
6155 | if (size == 0) | |
612e59d8 CA |
6156 | /* chunk is in K */ |
6157 | minsize = chunk * 2; | |
efb30e7f DW |
6158 | |
6159 | if (cnt < raiddisks || | |
6160 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6161 | maxsize < minsize || |
6162 | maxsize == 0) { | |
efb30e7f DW |
6163 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
6164 | return 0; /* No enough free spaces large enough */ | |
6165 | } | |
6166 | ||
6167 | if (size == 0) { | |
6168 | size = maxsize; | |
6169 | if (chunk) { | |
612e59d8 CA |
6170 | size /= 2 * chunk; |
6171 | size *= 2 * chunk; | |
efb30e7f | 6172 | } |
f878b242 LM |
6173 | maxsize = size; |
6174 | } | |
6175 | if (!check_env("IMSM_NO_PLATFORM") && | |
6176 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
6177 | fprintf(stderr, Name ": attempting to create a second " | |
6178 | "volume with size less then remaining space. " | |
6179 | "Aborting...\n"); | |
6180 | return 0; | |
efb30e7f | 6181 | } |
efb30e7f DW |
6182 | cnt = 0; |
6183 | for (dl = super->disks; dl; dl = dl->next) | |
6184 | if (dl->e) | |
6185 | dl->raiddisk = cnt++; | |
6186 | ||
6187 | *freesize = size; | |
6188 | ||
6189 | return 1; | |
6190 | } | |
6191 | ||
bf5a934a | 6192 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6193 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
6194 | char *dev, unsigned long long *freesize, |
6195 | int verbose) | |
6196 | { | |
6197 | int fd, cfd; | |
6198 | struct mdinfo *sra; | |
20cbe8d2 | 6199 | int is_member = 0; |
bf5a934a | 6200 | |
d54559f0 LM |
6201 | /* load capability |
6202 | * if given unused devices create a container | |
bf5a934a DW |
6203 | * if given given devices in a container create a member volume |
6204 | */ | |
6205 | if (level == LEVEL_CONTAINER) { | |
6206 | /* Must be a fresh device to add to a container */ | |
6207 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
6208 | raiddisks, |
6209 | chunk?*chunk:0, size, | |
bf5a934a DW |
6210 | dev, freesize, |
6211 | verbose); | |
6212 | } | |
9587c373 | 6213 | |
8592f29d | 6214 | if (!dev) { |
e91a3bad | 6215 | if (st->sb) { |
ca9de185 | 6216 | struct intel_super *super = st->sb; |
e91a3bad | 6217 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 6218 | raiddisks, chunk, size, |
e91a3bad LM |
6219 | verbose)) |
6220 | return 0; | |
efb30e7f DW |
6221 | /* we are being asked to automatically layout a |
6222 | * new volume based on the current contents of | |
6223 | * the container. If the the parameters can be | |
6224 | * satisfied reserve_space will record the disks, | |
6225 | * start offset, and size of the volume to be | |
6226 | * created. add_to_super and getinfo_super | |
6227 | * detect when autolayout is in progress. | |
6228 | */ | |
ca9de185 LM |
6229 | /* assuming that freesize is always given when array is |
6230 | created */ | |
6231 | if (super->orom && freesize) { | |
6232 | int count; | |
6233 | count = count_volumes(super->hba->path, | |
6234 | super->orom->dpa, verbose); | |
6235 | if (super->orom->vphba <= count) { | |
6236 | pr_vrb(": platform does not support more" | |
40110b90 | 6237 | " than %d raid volumes.\n", |
ca9de185 LM |
6238 | super->orom->vphba); |
6239 | return 0; | |
6240 | } | |
6241 | } | |
e91a3bad LM |
6242 | if (freesize) |
6243 | return reserve_space(st, raiddisks, size, | |
6244 | chunk?*chunk:0, freesize); | |
8592f29d N |
6245 | } |
6246 | return 1; | |
6247 | } | |
bf5a934a DW |
6248 | if (st->sb) { |
6249 | /* creating in a given container */ | |
6250 | return validate_geometry_imsm_volume(st, level, layout, | |
6251 | raiddisks, chunk, size, | |
6252 | dev, freesize, verbose); | |
6253 | } | |
6254 | ||
bf5a934a DW |
6255 | /* This device needs to be a device in an 'imsm' container */ |
6256 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6257 | if (fd >= 0) { | |
6258 | if (verbose) | |
6259 | fprintf(stderr, | |
6260 | Name ": Cannot create this array on device %s\n", | |
6261 | dev); | |
6262 | close(fd); | |
6263 | return 0; | |
6264 | } | |
6265 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6266 | if (verbose) | |
6267 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
6268 | dev, strerror(errno)); | |
6269 | return 0; | |
6270 | } | |
6271 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6272 | cfd = open_container(fd); | |
20cbe8d2 | 6273 | close(fd); |
bf5a934a | 6274 | if (cfd < 0) { |
bf5a934a DW |
6275 | if (verbose) |
6276 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
6277 | dev); | |
6278 | return 0; | |
6279 | } | |
6280 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 6281 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
6282 | strcmp(sra->text_version, "imsm") == 0) |
6283 | is_member = 1; | |
6284 | sysfs_free(sra); | |
6285 | if (is_member) { | |
bf5a934a DW |
6286 | /* This is a member of a imsm container. Load the container |
6287 | * and try to create a volume | |
6288 | */ | |
6289 | struct intel_super *super; | |
6290 | ||
ec50f7b6 | 6291 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a DW |
6292 | st->sb = super; |
6293 | st->container_dev = fd2devnum(cfd); | |
6294 | close(cfd); | |
6295 | return validate_geometry_imsm_volume(st, level, layout, | |
6296 | raiddisks, chunk, | |
6297 | size, dev, | |
ecbd9e81 N |
6298 | freesize, 1) |
6299 | ? 1 : -1; | |
bf5a934a | 6300 | } |
20cbe8d2 | 6301 | } |
bf5a934a | 6302 | |
20cbe8d2 AW |
6303 | if (verbose) |
6304 | fprintf(stderr, Name ": failed container membership check\n"); | |
6305 | ||
6306 | close(cfd); | |
6307 | return 0; | |
bf5a934a | 6308 | } |
0bd16cf2 | 6309 | |
30f58b22 | 6310 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
6311 | { |
6312 | struct intel_super *super = st->sb; | |
6313 | ||
30f58b22 DW |
6314 | if (level && *level == UnSet) |
6315 | *level = LEVEL_CONTAINER; | |
6316 | ||
6317 | if (level && layout && *layout == UnSet) | |
6318 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 6319 | |
cd9d1ac7 DW |
6320 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
6321 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
6322 | } |
6323 | ||
33414a01 DW |
6324 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
6325 | ||
6326 | static int kill_subarray_imsm(struct supertype *st) | |
6327 | { | |
6328 | /* remove the subarray currently referenced by ->current_vol */ | |
6329 | __u8 i; | |
6330 | struct intel_dev **dp; | |
6331 | struct intel_super *super = st->sb; | |
6332 | __u8 current_vol = super->current_vol; | |
6333 | struct imsm_super *mpb = super->anchor; | |
6334 | ||
6335 | if (super->current_vol < 0) | |
6336 | return 2; | |
6337 | super->current_vol = -1; /* invalidate subarray cursor */ | |
6338 | ||
6339 | /* block deletions that would change the uuid of active subarrays | |
6340 | * | |
6341 | * FIXME when immutable ids are available, but note that we'll | |
6342 | * also need to fixup the invalidated/active subarray indexes in | |
6343 | * mdstat | |
6344 | */ | |
6345 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6346 | char subarray[4]; | |
6347 | ||
6348 | if (i < current_vol) | |
6349 | continue; | |
6350 | sprintf(subarray, "%u", i); | |
6351 | if (is_subarray_active(subarray, st->devname)) { | |
6352 | fprintf(stderr, | |
6353 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
6354 | current_vol, i); | |
6355 | ||
6356 | return 2; | |
6357 | } | |
6358 | } | |
6359 | ||
6360 | if (st->update_tail) { | |
6361 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
6362 | ||
6363 | if (!u) | |
6364 | return 2; | |
6365 | u->type = update_kill_array; | |
6366 | u->dev_idx = current_vol; | |
6367 | append_metadata_update(st, u, sizeof(*u)); | |
6368 | ||
6369 | return 0; | |
6370 | } | |
6371 | ||
6372 | for (dp = &super->devlist; *dp;) | |
6373 | if ((*dp)->index == current_vol) { | |
6374 | *dp = (*dp)->next; | |
6375 | } else { | |
6376 | handle_missing(super, (*dp)->dev); | |
6377 | if ((*dp)->index > current_vol) | |
6378 | (*dp)->index--; | |
6379 | dp = &(*dp)->next; | |
6380 | } | |
6381 | ||
6382 | /* no more raid devices, all active components are now spares, | |
6383 | * but of course failed are still failed | |
6384 | */ | |
6385 | if (--mpb->num_raid_devs == 0) { | |
6386 | struct dl *d; | |
6387 | ||
6388 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
6389 | if (d->index > -2) |
6390 | mark_spare(d); | |
33414a01 DW |
6391 | } |
6392 | ||
6393 | super->updates_pending++; | |
6394 | ||
6395 | return 0; | |
6396 | } | |
aa534678 | 6397 | |
a951a4f7 | 6398 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 6399 | char *update, struct mddev_ident *ident) |
aa534678 DW |
6400 | { |
6401 | /* update the subarray currently referenced by ->current_vol */ | |
6402 | struct intel_super *super = st->sb; | |
6403 | struct imsm_super *mpb = super->anchor; | |
6404 | ||
aa534678 DW |
6405 | if (strcmp(update, "name") == 0) { |
6406 | char *name = ident->name; | |
a951a4f7 N |
6407 | char *ep; |
6408 | int vol; | |
aa534678 | 6409 | |
a951a4f7 | 6410 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
6411 | fprintf(stderr, |
6412 | Name ": Unable to update name of active subarray\n"); | |
6413 | return 2; | |
6414 | } | |
6415 | ||
6416 | if (!check_name(super, name, 0)) | |
6417 | return 2; | |
6418 | ||
a951a4f7 N |
6419 | vol = strtoul(subarray, &ep, 10); |
6420 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
6421 | return 2; | |
6422 | ||
aa534678 DW |
6423 | if (st->update_tail) { |
6424 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
6425 | ||
6426 | if (!u) | |
6427 | return 2; | |
6428 | u->type = update_rename_array; | |
a951a4f7 | 6429 | u->dev_idx = vol; |
aa534678 DW |
6430 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
6431 | append_metadata_update(st, u, sizeof(*u)); | |
6432 | } else { | |
6433 | struct imsm_dev *dev; | |
6434 | int i; | |
6435 | ||
a951a4f7 | 6436 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
6437 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
6438 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6439 | dev = get_imsm_dev(super, i); | |
6440 | handle_missing(super, dev); | |
6441 | } | |
6442 | super->updates_pending++; | |
6443 | } | |
6444 | } else | |
6445 | return 2; | |
6446 | ||
6447 | return 0; | |
6448 | } | |
d1e02575 | 6449 | #endif /* MDASSEMBLE */ |
bf5a934a | 6450 | |
28bce06f AK |
6451 | static int is_gen_migration(struct imsm_dev *dev) |
6452 | { | |
7534230b AK |
6453 | if (dev == NULL) |
6454 | return 0; | |
6455 | ||
28bce06f AK |
6456 | if (!dev->vol.migr_state) |
6457 | return 0; | |
6458 | ||
6459 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
6460 | return 1; | |
6461 | ||
6462 | return 0; | |
6463 | } | |
6464 | ||
1e5c6983 DW |
6465 | static int is_rebuilding(struct imsm_dev *dev) |
6466 | { | |
6467 | struct imsm_map *migr_map; | |
6468 | ||
6469 | if (!dev->vol.migr_state) | |
6470 | return 0; | |
6471 | ||
6472 | if (migr_type(dev) != MIGR_REBUILD) | |
6473 | return 0; | |
6474 | ||
238c0a71 | 6475 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
6476 | |
6477 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
6478 | return 1; | |
6479 | else | |
6480 | return 0; | |
6481 | } | |
6482 | ||
b4ab44d8 | 6483 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
6484 | static int is_initializing(struct imsm_dev *dev) |
6485 | { | |
6486 | struct imsm_map *migr_map; | |
6487 | ||
6488 | if (!dev->vol.migr_state) | |
6489 | return 0; | |
6490 | ||
6491 | if (migr_type(dev) != MIGR_INIT) | |
6492 | return 0; | |
6493 | ||
238c0a71 | 6494 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
6495 | |
6496 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6497 | return 1; | |
6498 | ||
6499 | return 0; | |
6ce1fbf1 | 6500 | } |
b4ab44d8 | 6501 | #endif |
6ce1fbf1 | 6502 | |
c47b0ff6 AK |
6503 | static void update_recovery_start(struct intel_super *super, |
6504 | struct imsm_dev *dev, | |
6505 | struct mdinfo *array) | |
1e5c6983 DW |
6506 | { |
6507 | struct mdinfo *rebuild = NULL; | |
6508 | struct mdinfo *d; | |
6509 | __u32 units; | |
6510 | ||
6511 | if (!is_rebuilding(dev)) | |
6512 | return; | |
6513 | ||
6514 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
6515 | for (d = array->devs; d; d = d->next) | |
6516 | if (d->recovery_start == 0) { | |
6517 | if (rebuild) | |
6518 | return; | |
6519 | rebuild = d; | |
6520 | } | |
6521 | ||
4363fd80 DW |
6522 | if (!rebuild) { |
6523 | /* (?) none of the disks are marked with | |
6524 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
6525 | * disk_ord_tbl was not correctly updated | |
6526 | */ | |
6527 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
6528 | return; | |
6529 | } | |
6530 | ||
1e5c6983 | 6531 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 6532 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
6533 | } |
6534 | ||
9e2d750d | 6535 | #ifndef MDASSEMBLE |
276d77db | 6536 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 6537 | #endif |
1e5c6983 | 6538 | |
00bbdbda | 6539 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 6540 | { |
4f5bc454 DW |
6541 | /* Given a container loaded by load_super_imsm_all, |
6542 | * extract information about all the arrays into | |
6543 | * an mdinfo tree. | |
00bbdbda | 6544 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
6545 | * |
6546 | * For each imsm_dev create an mdinfo, fill it in, | |
6547 | * then look for matching devices in super->disks | |
6548 | * and create appropriate device mdinfo. | |
6549 | */ | |
6550 | struct intel_super *super = st->sb; | |
949c47a0 | 6551 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 6552 | struct mdinfo *rest = NULL; |
00bbdbda | 6553 | unsigned int i; |
81219e70 | 6554 | int sb_errors = 0; |
abef11a3 AK |
6555 | struct dl *d; |
6556 | int spare_disks = 0; | |
cdddbdbc | 6557 | |
19482bcc AK |
6558 | /* do not assemble arrays when not all attributes are supported */ |
6559 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 LM |
6560 | sb_errors = 1; |
6561 | fprintf(stderr, Name ": Unsupported attributes in IMSM metadata." | |
6562 | "Arrays activation is blocked.\n"); | |
19482bcc AK |
6563 | } |
6564 | ||
a06d022d | 6565 | /* check for bad blocks */ |
81219e70 LM |
6566 | if (imsm_bbm_log_size(super->anchor)) { |
6567 | fprintf(stderr, Name ": BBM log found in IMSM metadata." | |
6568 | "Arrays activation is blocked.\n"); | |
6569 | sb_errors = 1; | |
6570 | } | |
6571 | ||
604b746f | 6572 | |
abef11a3 AK |
6573 | /* count spare devices, not used in maps |
6574 | */ | |
6575 | for (d = super->disks; d; d = d->next) | |
6576 | if (d->index == -1) | |
6577 | spare_disks++; | |
6578 | ||
4f5bc454 | 6579 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
6580 | struct imsm_dev *dev; |
6581 | struct imsm_map *map; | |
86e3692b | 6582 | struct imsm_map *map2; |
4f5bc454 | 6583 | struct mdinfo *this; |
a6482415 N |
6584 | int slot; |
6585 | #ifndef MDASSEMBLE | |
6586 | int chunk; | |
6587 | #endif | |
00bbdbda N |
6588 | char *ep; |
6589 | ||
6590 | if (subarray && | |
6591 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
6592 | continue; | |
6593 | ||
6594 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
6595 | map = get_imsm_map(dev, MAP_0); |
6596 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 6597 | |
1ce0101c DW |
6598 | /* do not publish arrays that are in the middle of an |
6599 | * unsupported migration | |
6600 | */ | |
6601 | if (dev->vol.migr_state && | |
28bce06f | 6602 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
6603 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
6604 | " unsupported migration in progress\n", | |
6605 | dev->volume); | |
6606 | continue; | |
6607 | } | |
2db86302 LM |
6608 | /* do not publish arrays that are not support by controller's |
6609 | * OROM/EFI | |
6610 | */ | |
1ce0101c | 6611 | |
4f5bc454 | 6612 | this = malloc(sizeof(*this)); |
0fbd635c | 6613 | if (!this) { |
cf1be220 | 6614 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
6615 | sizeof(*this)); |
6616 | break; | |
6617 | } | |
4f5bc454 | 6618 | |
301406c9 | 6619 | super->current_vol = i; |
a5d85af7 | 6620 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 6621 | this->next = rest; |
81219e70 | 6622 | #ifndef MDASSEMBLE |
a6482415 | 6623 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
6624 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
6625 | if (!validate_geometry_imsm_orom(super, | |
6626 | get_imsm_raid_level(map), /* RAID level */ | |
6627 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
6628 | map->num_members, /* raid disks */ | |
2cc699af | 6629 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 6630 | 1 /* verbose */)) { |
446894ea N |
6631 | fprintf(stderr, Name ": IMSM RAID geometry validation" |
6632 | " failed. Array %s activation is blocked.\n", | |
81219e70 LM |
6633 | dev->volume); |
6634 | this->array.state |= | |
6635 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6636 | (1<<MD_SB_BLOCK_VOLUME); | |
6637 | } | |
6638 | #endif | |
6639 | ||
6640 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
6641 | if (sb_errors) | |
6642 | this->array.state |= | |
6643 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6644 | (1<<MD_SB_BLOCK_VOLUME); | |
6645 | ||
4f5bc454 | 6646 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 6647 | unsigned long long recovery_start; |
4f5bc454 DW |
6648 | struct mdinfo *info_d; |
6649 | struct dl *d; | |
6650 | int idx; | |
9a1608e5 | 6651 | int skip; |
7eef0453 | 6652 | __u32 ord; |
4f5bc454 | 6653 | |
9a1608e5 | 6654 | skip = 0; |
238c0a71 AK |
6655 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
6656 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
6657 | for (d = super->disks; d ; d = d->next) |
6658 | if (d->index == idx) | |
0fbd635c | 6659 | break; |
4f5bc454 | 6660 | |
1e5c6983 | 6661 | recovery_start = MaxSector; |
4f5bc454 | 6662 | if (d == NULL) |
9a1608e5 | 6663 | skip = 1; |
25ed7e59 | 6664 | if (d && is_failed(&d->disk)) |
9a1608e5 | 6665 | skip = 1; |
7eef0453 | 6666 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 6667 | recovery_start = 0; |
9a1608e5 DW |
6668 | |
6669 | /* | |
6670 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
6671 | * reset resync start to avoid a dirty-degraded |
6672 | * situation when performing the intial sync | |
9a1608e5 DW |
6673 | * |
6674 | * FIXME handle dirty degraded | |
6675 | */ | |
1e5c6983 | 6676 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 6677 | this->resync_start = MaxSector; |
9a1608e5 DW |
6678 | if (skip) |
6679 | continue; | |
4f5bc454 | 6680 | |
1e5c6983 | 6681 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
6682 | if (!info_d) { |
6683 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 6684 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
6685 | info_d = this->devs; |
6686 | while (info_d) { | |
6687 | struct mdinfo *d = info_d->next; | |
6688 | ||
6689 | free(info_d); | |
6690 | info_d = d; | |
6691 | } | |
9a1608e5 DW |
6692 | free(this); |
6693 | this = rest; | |
6694 | break; | |
6695 | } | |
4f5bc454 DW |
6696 | info_d->next = this->devs; |
6697 | this->devs = info_d; | |
6698 | ||
4f5bc454 DW |
6699 | info_d->disk.number = d->index; |
6700 | info_d->disk.major = d->major; | |
6701 | info_d->disk.minor = d->minor; | |
6702 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 6703 | info_d->recovery_start = recovery_start; |
86e3692b AK |
6704 | if (map2) { |
6705 | if (slot < map2->num_members) | |
6706 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6707 | else |
6708 | this->array.spare_disks++; | |
86e3692b AK |
6709 | } else { |
6710 | if (slot < map->num_members) | |
6711 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6712 | else |
6713 | this->array.spare_disks++; | |
86e3692b | 6714 | } |
1e5c6983 DW |
6715 | if (info_d->recovery_start == MaxSector) |
6716 | this->array.working_disks++; | |
4f5bc454 DW |
6717 | |
6718 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 CA |
6719 | info_d->data_offset = pba_of_lba0(map); |
6720 | info_d->component_size = blocks_per_member(map); | |
4f5bc454 | 6721 | } |
1e5c6983 | 6722 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 6723 | update_recovery_start(super, dev, this); |
abef11a3 | 6724 | this->array.spare_disks += spare_disks; |
276d77db | 6725 | |
9e2d750d | 6726 | #ifndef MDASSEMBLE |
276d77db AK |
6727 | /* check for reshape */ |
6728 | if (this->reshape_active == 1) | |
6729 | recover_backup_imsm(st, this); | |
9e2d750d | 6730 | #endif |
9a1608e5 | 6731 | rest = this; |
4f5bc454 DW |
6732 | } |
6733 | ||
6734 | return rest; | |
cdddbdbc DW |
6735 | } |
6736 | ||
845dea95 | 6737 | |
3b451610 AK |
6738 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
6739 | int failed, int look_in_map) | |
c2a1e7da | 6740 | { |
3b451610 AK |
6741 | struct imsm_map *map; |
6742 | ||
6743 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
6744 | |
6745 | if (!failed) | |
3393c6af DW |
6746 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
6747 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
6748 | |
6749 | switch (get_imsm_raid_level(map)) { | |
6750 | case 0: | |
6751 | return IMSM_T_STATE_FAILED; | |
6752 | break; | |
6753 | case 1: | |
6754 | if (failed < map->num_members) | |
6755 | return IMSM_T_STATE_DEGRADED; | |
6756 | else | |
6757 | return IMSM_T_STATE_FAILED; | |
6758 | break; | |
6759 | case 10: | |
6760 | { | |
6761 | /** | |
c92a2527 DW |
6762 | * check to see if any mirrors have failed, otherwise we |
6763 | * are degraded. Even numbered slots are mirrored on | |
6764 | * slot+1 | |
c2a1e7da | 6765 | */ |
c2a1e7da | 6766 | int i; |
d9b420a5 N |
6767 | /* gcc -Os complains that this is unused */ |
6768 | int insync = insync; | |
c2a1e7da DW |
6769 | |
6770 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6771 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6772 | int idx = ord_to_idx(ord); |
6773 | struct imsm_disk *disk; | |
c2a1e7da | 6774 | |
c92a2527 DW |
6775 | /* reset the potential in-sync count on even-numbered |
6776 | * slots. num_copies is always 2 for imsm raid10 | |
6777 | */ | |
6778 | if ((i & 1) == 0) | |
6779 | insync = 2; | |
c2a1e7da | 6780 | |
c92a2527 | 6781 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6782 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6783 | insync--; |
c2a1e7da | 6784 | |
c92a2527 DW |
6785 | /* no in-sync disks left in this mirror the |
6786 | * array has failed | |
6787 | */ | |
6788 | if (insync == 0) | |
6789 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6790 | } |
6791 | ||
6792 | return IMSM_T_STATE_DEGRADED; | |
6793 | } | |
6794 | case 5: | |
6795 | if (failed < 2) | |
6796 | return IMSM_T_STATE_DEGRADED; | |
6797 | else | |
6798 | return IMSM_T_STATE_FAILED; | |
6799 | break; | |
6800 | default: | |
6801 | break; | |
6802 | } | |
6803 | ||
6804 | return map->map_state; | |
6805 | } | |
6806 | ||
3b451610 AK |
6807 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6808 | int look_in_map) | |
c2a1e7da DW |
6809 | { |
6810 | int i; | |
6811 | int failed = 0; | |
6812 | struct imsm_disk *disk; | |
d5985138 AK |
6813 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6814 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6815 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6816 | __u32 ord; |
6817 | int idx; | |
d5985138 | 6818 | int idx_1; |
c2a1e7da | 6819 | |
0556e1a2 DW |
6820 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6821 | * disks that are being rebuilt. New failures are recorded to | |
6822 | * map[0]. So we look through all the disks we started with and | |
6823 | * see if any failures are still present, or if any new ones | |
6824 | * have arrived | |
0556e1a2 | 6825 | */ |
d5985138 AK |
6826 | map_for_loop = map; |
6827 | if (prev && (map->num_members < prev->num_members)) | |
6828 | map_for_loop = prev; | |
68fe4598 LD |
6829 | |
6830 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6831 | idx_1 = -255; |
238c0a71 AK |
6832 | /* when MAP_X is passed both maps failures are counted |
6833 | */ | |
d5985138 | 6834 | if (prev && |
238c0a71 AK |
6835 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6836 | (i < prev->num_members)) { | |
d5985138 AK |
6837 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6838 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6839 | |
d5985138 AK |
6840 | disk = get_imsm_disk(super, idx_1); |
6841 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6842 | failed++; | |
6843 | } | |
238c0a71 AK |
6844 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6845 | (i < map->num_members)) { | |
d5985138 AK |
6846 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6847 | idx = ord_to_idx(ord); | |
6848 | ||
6849 | if (idx != idx_1) { | |
6850 | disk = get_imsm_disk(super, idx); | |
6851 | if (!disk || is_failed(disk) || | |
6852 | ord & IMSM_ORD_REBUILD) | |
6853 | failed++; | |
6854 | } | |
6855 | } | |
c2a1e7da DW |
6856 | } |
6857 | ||
6858 | return failed; | |
845dea95 NB |
6859 | } |
6860 | ||
97b4d0e9 DW |
6861 | #ifndef MDASSEMBLE |
6862 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
6863 | char *inst) | |
6864 | { | |
6865 | struct intel_super *super = c->sb; | |
6866 | struct imsm_super *mpb = super->anchor; | |
9587c373 | 6867 | |
97b4d0e9 DW |
6868 | if (atoi(inst) >= mpb->num_raid_devs) { |
6869 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
6870 | __func__, atoi(inst)); | |
6871 | return -ENODEV; | |
6872 | } | |
6873 | ||
6874 | dprintf("imsm: open_new %s\n", inst); | |
6875 | a->info.container_member = atoi(inst); | |
6876 | return 0; | |
6877 | } | |
6878 | ||
0c046afd DW |
6879 | static int is_resyncing(struct imsm_dev *dev) |
6880 | { | |
6881 | struct imsm_map *migr_map; | |
6882 | ||
6883 | if (!dev->vol.migr_state) | |
6884 | return 0; | |
6885 | ||
1484e727 DW |
6886 | if (migr_type(dev) == MIGR_INIT || |
6887 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
6888 | return 1; |
6889 | ||
4c9bc37b AK |
6890 | if (migr_type(dev) == MIGR_GEN_MIGR) |
6891 | return 0; | |
6892 | ||
238c0a71 | 6893 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 6894 | |
4c9bc37b AK |
6895 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
6896 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
6897 | return 1; |
6898 | else | |
6899 | return 0; | |
6900 | } | |
6901 | ||
0556e1a2 DW |
6902 | /* return true if we recorded new information */ |
6903 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 6904 | { |
0556e1a2 DW |
6905 | __u32 ord; |
6906 | int slot; | |
6907 | struct imsm_map *map; | |
86c54047 DW |
6908 | char buf[MAX_RAID_SERIAL_LEN+3]; |
6909 | unsigned int len, shift = 0; | |
0556e1a2 DW |
6910 | |
6911 | /* new failures are always set in map[0] */ | |
238c0a71 | 6912 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
6913 | |
6914 | slot = get_imsm_disk_slot(map, idx); | |
6915 | if (slot < 0) | |
6916 | return 0; | |
6917 | ||
6918 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 6919 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
6920 | return 0; |
6921 | ||
7d0c5e24 LD |
6922 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
6923 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
6924 | strcat(buf, ":0"); | |
86c54047 DW |
6925 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
6926 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
6927 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
6928 | ||
f2f27e63 | 6929 | disk->status |= FAILED_DISK; |
0556e1a2 | 6930 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
6931 | /* mark failures in second map if second map exists and this disk |
6932 | * in this slot. | |
6933 | * This is valid for migration, initialization and rebuild | |
6934 | */ | |
6935 | if (dev->vol.migr_state) { | |
238c0a71 | 6936 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
6937 | int slot2 = get_imsm_disk_slot(map2, idx); |
6938 | ||
6939 | if ((slot2 < map2->num_members) && | |
6940 | (slot2 >= 0)) | |
6941 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
6942 | idx | IMSM_ORD_REBUILD); |
6943 | } | |
f21e18ca | 6944 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
6945 | map->failed_disk_num = slot; |
6946 | return 1; | |
6947 | } | |
6948 | ||
6949 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
6950 | { | |
6951 | mark_failure(dev, disk, idx); | |
6952 | ||
6953 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
6954 | return; | |
6955 | ||
47ee5a45 DW |
6956 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
6957 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
6958 | } | |
6959 | ||
33414a01 DW |
6960 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
6961 | { | |
33414a01 | 6962 | struct dl *dl; |
33414a01 DW |
6963 | |
6964 | if (!super->missing) | |
6965 | return; | |
33414a01 DW |
6966 | |
6967 | dprintf("imsm: mark missing\n"); | |
3d59f0c0 AK |
6968 | /* end process for initialization and rebuild only |
6969 | */ | |
6970 | if (is_gen_migration(dev) == 0) { | |
6971 | __u8 map_state; | |
6972 | int failed; | |
6973 | ||
6974 | failed = imsm_count_failed(super, dev, MAP_0); | |
6975 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
6976 | ||
6977 | end_migration(dev, super, map_state); | |
6978 | } | |
33414a01 DW |
6979 | for (dl = super->missing; dl; dl = dl->next) |
6980 | mark_missing(dev, &dl->disk, dl->index); | |
6981 | super->updates_pending++; | |
6982 | } | |
6983 | ||
f3871fdc AK |
6984 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
6985 | long long new_size) | |
70bdf0dc | 6986 | { |
238c0a71 | 6987 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
6988 | unsigned long long array_blocks; |
6989 | struct imsm_map *map; | |
6990 | ||
6991 | if (used_disks == 0) { | |
6992 | /* when problems occures | |
6993 | * return current array_blocks value | |
6994 | */ | |
6995 | array_blocks = __le32_to_cpu(dev->size_high); | |
6996 | array_blocks = array_blocks << 32; | |
6997 | array_blocks += __le32_to_cpu(dev->size_low); | |
6998 | ||
6999 | return array_blocks; | |
7000 | } | |
7001 | ||
7002 | /* set array size in metadata | |
7003 | */ | |
f3871fdc AK |
7004 | if (new_size <= 0) { |
7005 | /* OLCE size change is caused by added disks | |
7006 | */ | |
7007 | map = get_imsm_map(dev, MAP_0); | |
7008 | array_blocks = blocks_per_member(map) * used_disks; | |
7009 | } else { | |
7010 | /* Online Volume Size Change | |
7011 | * Using available free space | |
7012 | */ | |
7013 | array_blocks = new_size; | |
7014 | } | |
70bdf0dc AK |
7015 | |
7016 | /* round array size down to closest MB | |
7017 | */ | |
7018 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
7019 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
7020 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
7021 | ||
7022 | return array_blocks; | |
7023 | } | |
7024 | ||
28bce06f AK |
7025 | static void imsm_set_disk(struct active_array *a, int n, int state); |
7026 | ||
0e2d1a4e AK |
7027 | static void imsm_progress_container_reshape(struct intel_super *super) |
7028 | { | |
7029 | /* if no device has a migr_state, but some device has a | |
7030 | * different number of members than the previous device, start | |
7031 | * changing the number of devices in this device to match | |
7032 | * previous. | |
7033 | */ | |
7034 | struct imsm_super *mpb = super->anchor; | |
7035 | int prev_disks = -1; | |
7036 | int i; | |
1dfaa380 | 7037 | int copy_map_size; |
0e2d1a4e AK |
7038 | |
7039 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7040 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 7041 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
7042 | struct imsm_map *map2; |
7043 | int prev_num_members; | |
0e2d1a4e AK |
7044 | |
7045 | if (dev->vol.migr_state) | |
7046 | return; | |
7047 | ||
7048 | if (prev_disks == -1) | |
7049 | prev_disks = map->num_members; | |
7050 | if (prev_disks == map->num_members) | |
7051 | continue; | |
7052 | ||
7053 | /* OK, this array needs to enter reshape mode. | |
7054 | * i.e it needs a migr_state | |
7055 | */ | |
7056 | ||
1dfaa380 | 7057 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
7058 | prev_num_members = map->num_members; |
7059 | map->num_members = prev_disks; | |
7060 | dev->vol.migr_state = 1; | |
7061 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7062 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
7063 | for (i = prev_num_members; |
7064 | i < map->num_members; i++) | |
7065 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 7066 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 7067 | /* Copy the current map */ |
1dfaa380 | 7068 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
7069 | map2->num_members = prev_num_members; |
7070 | ||
f3871fdc | 7071 | imsm_set_array_size(dev, -1); |
51d83f5d | 7072 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
7073 | super->updates_pending++; |
7074 | } | |
7075 | } | |
7076 | ||
aad6f216 | 7077 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
7078 | * states are handled in imsm_set_disk() with one exception, when a |
7079 | * resync is stopped due to a new failure this routine will set the | |
7080 | * 'degraded' state for the array. | |
7081 | */ | |
01f157d7 | 7082 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
7083 | { |
7084 | int inst = a->info.container_member; | |
7085 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7086 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7087 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
7088 | int failed = imsm_count_failed(super, dev, MAP_0); |
7089 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 7090 | __u32 blocks_per_unit; |
a862209d | 7091 | |
1af97990 AK |
7092 | if (dev->vol.migr_state && |
7093 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
7094 | /* array state change is blocked due to reshape action | |
aad6f216 N |
7095 | * We might need to |
7096 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
7097 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
7098 | * - update curr_migr_unit | |
1af97990 | 7099 | */ |
aad6f216 N |
7100 | if (a->curr_action == reshape) { |
7101 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 7102 | goto mark_checkpoint; |
aad6f216 N |
7103 | } else { |
7104 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
7105 | /* for some reason we aborted the reshape. | |
b66e591b AK |
7106 | * |
7107 | * disable automatic metadata rollback | |
7108 | * user action is required to recover process | |
aad6f216 | 7109 | */ |
b66e591b | 7110 | if (0) { |
238c0a71 AK |
7111 | struct imsm_map *map2 = |
7112 | get_imsm_map(dev, MAP_1); | |
7113 | dev->vol.migr_state = 0; | |
7114 | set_migr_type(dev, 0); | |
7115 | dev->vol.curr_migr_unit = 0; | |
7116 | memcpy(map, map2, | |
7117 | sizeof_imsm_map(map2)); | |
7118 | super->updates_pending++; | |
b66e591b | 7119 | } |
aad6f216 N |
7120 | } |
7121 | if (a->last_checkpoint >= a->info.component_size) { | |
7122 | unsigned long long array_blocks; | |
7123 | int used_disks; | |
e154ced3 | 7124 | struct mdinfo *mdi; |
aad6f216 | 7125 | |
238c0a71 | 7126 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
7127 | if (used_disks > 0) { |
7128 | array_blocks = | |
5551b113 | 7129 | blocks_per_member(map) * |
d55adef9 AK |
7130 | used_disks; |
7131 | /* round array size down to closest MB | |
7132 | */ | |
7133 | array_blocks = (array_blocks | |
7134 | >> SECT_PER_MB_SHIFT) | |
7135 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
7136 | a->info.custom_array_size = array_blocks; |
7137 | /* encourage manager to update array | |
7138 | * size | |
7139 | */ | |
e154ced3 | 7140 | |
d55adef9 | 7141 | a->check_reshape = 1; |
633b5610 | 7142 | } |
e154ced3 AK |
7143 | /* finalize online capacity expansion/reshape */ |
7144 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
7145 | imsm_set_disk(a, | |
7146 | mdi->disk.raid_disk, | |
7147 | mdi->curr_state); | |
7148 | ||
0e2d1a4e | 7149 | imsm_progress_container_reshape(super); |
e154ced3 | 7150 | } |
aad6f216 | 7151 | } |
1af97990 AK |
7152 | } |
7153 | ||
47ee5a45 | 7154 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
7155 | if (consistent == 2) |
7156 | handle_missing(super, dev); | |
1e5c6983 | 7157 | |
0c046afd | 7158 | if (consistent == 2 && |
b7941fd6 | 7159 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7160 | map_state != IMSM_T_STATE_NORMAL || |
7161 | dev->vol.migr_state)) | |
01f157d7 | 7162 | consistent = 0; |
272906ef | 7163 | |
b7941fd6 | 7164 | if (is_resync_complete(&a->info)) { |
0c046afd | 7165 | /* complete intialization / resync, |
0556e1a2 DW |
7166 | * recovery and interrupted recovery is completed in |
7167 | * ->set_disk | |
0c046afd DW |
7168 | */ |
7169 | if (is_resyncing(dev)) { | |
7170 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7171 | end_migration(dev, super, map_state); |
115c3803 | 7172 | super->updates_pending++; |
484240d8 | 7173 | a->last_checkpoint = 0; |
115c3803 | 7174 | } |
b9172665 AK |
7175 | } else if ((!is_resyncing(dev) && !failed) && |
7176 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7177 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7178 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7179 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7180 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7181 | else |
8e59f3d8 | 7182 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7183 | super->updates_pending++; |
115c3803 | 7184 | } |
a862209d | 7185 | |
633b5610 | 7186 | mark_checkpoint: |
5b83bacf AK |
7187 | /* skip checkpointing for general migration, |
7188 | * it is controlled in mdadm | |
7189 | */ | |
7190 | if (is_gen_migration(dev)) | |
7191 | goto skip_mark_checkpoint; | |
7192 | ||
1e5c6983 | 7193 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7194 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7195 | if (blocks_per_unit) { |
1e5c6983 DW |
7196 | __u32 units32; |
7197 | __u64 units; | |
7198 | ||
4f0a7acc | 7199 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7200 | units32 = units; |
7201 | ||
7202 | /* check that we did not overflow 32-bits, and that | |
7203 | * curr_migr_unit needs updating | |
7204 | */ | |
7205 | if (units32 == units && | |
bfd80a56 | 7206 | units32 != 0 && |
1e5c6983 DW |
7207 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7208 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7209 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7210 | super->updates_pending++; | |
7211 | } | |
7212 | } | |
f8f603f1 | 7213 | |
5b83bacf | 7214 | skip_mark_checkpoint: |
3393c6af | 7215 | /* mark dirty / clean */ |
0c046afd | 7216 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7217 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7218 | if (consistent) |
7219 | dev->vol.dirty = 0; | |
7220 | else | |
7221 | dev->vol.dirty = 1; | |
a862209d DW |
7222 | super->updates_pending++; |
7223 | } | |
28bce06f | 7224 | |
01f157d7 | 7225 | return consistent; |
a862209d DW |
7226 | } |
7227 | ||
8d45d196 | 7228 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7229 | { |
8d45d196 DW |
7230 | int inst = a->info.container_member; |
7231 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7232 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7233 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7234 | struct imsm_disk *disk; |
0c046afd | 7235 | int failed; |
b10b37b8 | 7236 | __u32 ord; |
0c046afd | 7237 | __u8 map_state; |
8d45d196 DW |
7238 | |
7239 | if (n > map->num_members) | |
7240 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
7241 | n, map->num_members - 1); | |
7242 | ||
7243 | if (n < 0) | |
7244 | return; | |
7245 | ||
4e6e574a | 7246 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 7247 | |
238c0a71 | 7248 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 7249 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7250 | |
5802a811 | 7251 | /* check for new failures */ |
0556e1a2 DW |
7252 | if (state & DS_FAULTY) { |
7253 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
7254 | super->updates_pending++; | |
8d45d196 | 7255 | } |
47ee5a45 | 7256 | |
19859edc | 7257 | /* check if in_sync */ |
0556e1a2 | 7258 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7259 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7260 | |
7261 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7262 | super->updates_pending++; |
7263 | } | |
8d45d196 | 7264 | |
3b451610 AK |
7265 | failed = imsm_count_failed(super, dev, MAP_0); |
7266 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7267 | |
0c046afd | 7268 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7269 | dprintf("imsm: Detected transition to state "); |
7270 | switch (map_state) { | |
7271 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
7272 | dprintf("normal: "); | |
7273 | if (is_rebuilding(dev)) { | |
7274 | dprintf("while rebuilding"); | |
7275 | end_migration(dev, super, map_state); | |
238c0a71 | 7276 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
7277 | map->failed_disk_num = ~0; |
7278 | super->updates_pending++; | |
7279 | a->last_checkpoint = 0; | |
7280 | break; | |
7281 | } | |
7282 | if (is_gen_migration(dev)) { | |
7283 | dprintf("while general migration"); | |
bf2f0071 | 7284 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7285 | end_migration(dev, super, map_state); |
94002678 AK |
7286 | else |
7287 | map->map_state = map_state; | |
238c0a71 | 7288 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 7289 | map->failed_disk_num = ~0; |
94002678 | 7290 | super->updates_pending++; |
bf2f0071 | 7291 | break; |
94002678 AK |
7292 | } |
7293 | break; | |
7294 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
7295 | dprintf("degraded: "); | |
7296 | if ((map->map_state != map_state) && | |
7297 | !dev->vol.migr_state) { | |
7298 | dprintf("mark degraded"); | |
7299 | map->map_state = map_state; | |
7300 | super->updates_pending++; | |
7301 | a->last_checkpoint = 0; | |
7302 | break; | |
7303 | } | |
7304 | if (is_rebuilding(dev)) { | |
7305 | dprintf("while rebuilding."); | |
7306 | if (map->map_state != map_state) { | |
7307 | dprintf(" Map state change"); | |
7308 | end_migration(dev, super, map_state); | |
7309 | super->updates_pending++; | |
7310 | } | |
7311 | break; | |
7312 | } | |
7313 | if (is_gen_migration(dev)) { | |
7314 | dprintf("while general migration"); | |
bf2f0071 | 7315 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7316 | end_migration(dev, super, map_state); |
94002678 AK |
7317 | else { |
7318 | map->map_state = map_state; | |
3b451610 | 7319 | manage_second_map(super, dev); |
94002678 AK |
7320 | } |
7321 | super->updates_pending++; | |
bf2f0071 | 7322 | break; |
28bce06f | 7323 | } |
6ce1fbf1 AK |
7324 | if (is_initializing(dev)) { |
7325 | dprintf("while initialization."); | |
7326 | map->map_state = map_state; | |
7327 | super->updates_pending++; | |
7328 | break; | |
7329 | } | |
94002678 AK |
7330 | break; |
7331 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
7332 | dprintf("failed: "); | |
7333 | if (is_gen_migration(dev)) { | |
7334 | dprintf("while general migration"); | |
7335 | map->map_state = map_state; | |
7336 | super->updates_pending++; | |
7337 | break; | |
7338 | } | |
7339 | if (map->map_state != map_state) { | |
7340 | dprintf("mark failed"); | |
7341 | end_migration(dev, super, map_state); | |
7342 | super->updates_pending++; | |
7343 | a->last_checkpoint = 0; | |
7344 | break; | |
7345 | } | |
7346 | break; | |
7347 | default: | |
7348 | dprintf("state %i\n", map_state); | |
5802a811 | 7349 | } |
94002678 AK |
7350 | dprintf("\n"); |
7351 | ||
845dea95 NB |
7352 | } |
7353 | ||
f796af5d | 7354 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 7355 | { |
f796af5d | 7356 | void *buf = mpb; |
c2a1e7da DW |
7357 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
7358 | unsigned long long dsize; | |
7359 | unsigned long long sectors; | |
7360 | ||
7361 | get_dev_size(fd, NULL, &dsize); | |
7362 | ||
272f648f DW |
7363 | if (mpb_size > 512) { |
7364 | /* -1 to account for anchor */ | |
7365 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 7366 | |
272f648f DW |
7367 | /* write the extended mpb to the sectors preceeding the anchor */ |
7368 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
7369 | return 1; | |
c2a1e7da | 7370 | |
f21e18ca N |
7371 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
7372 | != 512 * sectors) | |
272f648f DW |
7373 | return 1; |
7374 | } | |
c2a1e7da | 7375 | |
272f648f DW |
7376 | /* first block is stored on second to last sector of the disk */ |
7377 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
7378 | return 1; |
7379 | ||
f796af5d | 7380 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
7381 | return 1; |
7382 | ||
c2a1e7da DW |
7383 | return 0; |
7384 | } | |
7385 | ||
2e735d19 | 7386 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 7387 | { |
2e735d19 | 7388 | struct intel_super *super = container->sb; |
c2a1e7da | 7389 | |
1a64be56 | 7390 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
7391 | if (!super->updates_pending) |
7392 | return; | |
7393 | ||
36988a3d | 7394 | write_super_imsm(container, 0); |
c2a1e7da DW |
7395 | |
7396 | super->updates_pending = 0; | |
845dea95 NB |
7397 | } |
7398 | ||
272906ef DW |
7399 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
7400 | { | |
7401 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7402 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
7403 | struct dl *dl; |
7404 | ||
7405 | for (dl = super->disks; dl; dl = dl->next) | |
7406 | if (dl->index == i) | |
7407 | break; | |
7408 | ||
25ed7e59 | 7409 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
7410 | dl = NULL; |
7411 | ||
7412 | if (dl) | |
7413 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
7414 | ||
7415 | return dl; | |
7416 | } | |
7417 | ||
a20d2ba5 | 7418 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
7419 | struct active_array *a, int activate_new, |
7420 | struct mdinfo *additional_test_list) | |
272906ef DW |
7421 | { |
7422 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7423 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
7424 | struct imsm_super *mpb = super->anchor; |
7425 | struct imsm_map *map; | |
272906ef DW |
7426 | unsigned long long pos; |
7427 | struct mdinfo *d; | |
7428 | struct extent *ex; | |
a20d2ba5 | 7429 | int i, j; |
272906ef | 7430 | int found; |
569cc43f DW |
7431 | __u32 array_start = 0; |
7432 | __u32 array_end = 0; | |
272906ef | 7433 | struct dl *dl; |
6c932028 | 7434 | struct mdinfo *test_list; |
272906ef DW |
7435 | |
7436 | for (dl = super->disks; dl; dl = dl->next) { | |
7437 | /* If in this array, skip */ | |
7438 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
7439 | if (d->state_fd >= 0 && |
7440 | d->disk.major == dl->major && | |
272906ef | 7441 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
7442 | dprintf("%x:%x already in array\n", |
7443 | dl->major, dl->minor); | |
272906ef DW |
7444 | break; |
7445 | } | |
7446 | if (d) | |
7447 | continue; | |
6c932028 AK |
7448 | test_list = additional_test_list; |
7449 | while (test_list) { | |
7450 | if (test_list->disk.major == dl->major && | |
7451 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
7452 | dprintf("%x:%x already in additional test list\n", |
7453 | dl->major, dl->minor); | |
7454 | break; | |
7455 | } | |
6c932028 | 7456 | test_list = test_list->next; |
8ba77d32 | 7457 | } |
6c932028 | 7458 | if (test_list) |
8ba77d32 | 7459 | continue; |
272906ef | 7460 | |
e553d2a4 | 7461 | /* skip in use or failed drives */ |
25ed7e59 | 7462 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
7463 | dl->index == -2) { |
7464 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 7465 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
7466 | continue; |
7467 | } | |
7468 | ||
a20d2ba5 DW |
7469 | /* skip pure spares when we are looking for partially |
7470 | * assimilated drives | |
7471 | */ | |
7472 | if (dl->index == -1 && !activate_new) | |
7473 | continue; | |
7474 | ||
272906ef | 7475 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 7476 | * It needs to be able to cover all member volumes |
272906ef DW |
7477 | */ |
7478 | ex = get_extents(super, dl); | |
7479 | if (!ex) { | |
7480 | dprintf("cannot get extents\n"); | |
7481 | continue; | |
7482 | } | |
a20d2ba5 DW |
7483 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7484 | dev = get_imsm_dev(super, i); | |
238c0a71 | 7485 | map = get_imsm_map(dev, MAP_0); |
272906ef | 7486 | |
a20d2ba5 DW |
7487 | /* check if this disk is already a member of |
7488 | * this array | |
272906ef | 7489 | */ |
620b1713 | 7490 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
7491 | continue; |
7492 | ||
7493 | found = 0; | |
7494 | j = 0; | |
7495 | pos = 0; | |
5551b113 | 7496 | array_start = pba_of_lba0(map); |
329c8278 | 7497 | array_end = array_start + |
5551b113 | 7498 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
7499 | |
7500 | do { | |
7501 | /* check that we can start at pba_of_lba0 with | |
7502 | * blocks_per_member of space | |
7503 | */ | |
329c8278 | 7504 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
7505 | found = 1; |
7506 | break; | |
7507 | } | |
7508 | pos = ex[j].start + ex[j].size; | |
7509 | j++; | |
7510 | } while (ex[j-1].size); | |
7511 | ||
7512 | if (!found) | |
272906ef | 7513 | break; |
a20d2ba5 | 7514 | } |
272906ef DW |
7515 | |
7516 | free(ex); | |
a20d2ba5 | 7517 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
7518 | dprintf("%x:%x does not have %u to %u available\n", |
7519 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
7520 | /* No room */ |
7521 | continue; | |
a20d2ba5 DW |
7522 | } |
7523 | return dl; | |
272906ef DW |
7524 | } |
7525 | ||
7526 | return dl; | |
7527 | } | |
7528 | ||
95d07a2c LM |
7529 | |
7530 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
7531 | { | |
7532 | struct imsm_dev *dev2; | |
7533 | struct imsm_map *map; | |
7534 | struct dl *idisk; | |
7535 | int slot; | |
7536 | int idx; | |
7537 | __u8 state; | |
7538 | ||
7539 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
7540 | if (dev2) { | |
238c0a71 | 7541 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 7542 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 7543 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
7544 | if (!map) |
7545 | return 1; | |
7546 | for (slot = 0; slot < map->num_members; slot++) { | |
7547 | /* | |
7548 | * Check if failed disks are deleted from intel | |
7549 | * disk list or are marked to be deleted | |
7550 | */ | |
238c0a71 | 7551 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
7552 | idisk = get_imsm_dl_disk(cont->sb, idx); |
7553 | /* | |
7554 | * Do not rebuild the array if failed disks | |
7555 | * from failed sub-array are not removed from | |
7556 | * container. | |
7557 | */ | |
7558 | if (idisk && | |
7559 | is_failed(&idisk->disk) && | |
7560 | (idisk->action != DISK_REMOVE)) | |
7561 | return 0; | |
7562 | } | |
7563 | } | |
7564 | } | |
7565 | return 1; | |
7566 | } | |
7567 | ||
88758e9d DW |
7568 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
7569 | struct metadata_update **updates) | |
7570 | { | |
7571 | /** | |
d23fe947 DW |
7572 | * Find a device with unused free space and use it to replace a |
7573 | * failed/vacant region in an array. We replace failed regions one a | |
7574 | * array at a time. The result is that a new spare disk will be added | |
7575 | * to the first failed array and after the monitor has finished | |
7576 | * propagating failures the remainder will be consumed. | |
88758e9d | 7577 | * |
d23fe947 DW |
7578 | * FIXME add a capability for mdmon to request spares from another |
7579 | * container. | |
88758e9d DW |
7580 | */ |
7581 | ||
7582 | struct intel_super *super = a->container->sb; | |
88758e9d | 7583 | int inst = a->info.container_member; |
949c47a0 | 7584 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7585 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
7586 | int failed = a->info.array.raid_disks; |
7587 | struct mdinfo *rv = NULL; | |
7588 | struct mdinfo *d; | |
7589 | struct mdinfo *di; | |
7590 | struct metadata_update *mu; | |
7591 | struct dl *dl; | |
7592 | struct imsm_update_activate_spare *u; | |
7593 | int num_spares = 0; | |
7594 | int i; | |
95d07a2c | 7595 | int allowed; |
88758e9d DW |
7596 | |
7597 | for (d = a->info.devs ; d ; d = d->next) { | |
7598 | if ((d->curr_state & DS_FAULTY) && | |
7599 | d->state_fd >= 0) | |
7600 | /* wait for Removal to happen */ | |
7601 | return NULL; | |
7602 | if (d->state_fd >= 0) | |
7603 | failed--; | |
7604 | } | |
7605 | ||
7606 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
7607 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 7608 | |
e2962bfc AK |
7609 | if (imsm_reshape_blocks_arrays_changes(super)) |
7610 | return NULL; | |
1af97990 | 7611 | |
fc8ca064 AK |
7612 | /* Cannot activate another spare if rebuild is in progress already |
7613 | */ | |
7614 | if (is_rebuilding(dev)) { | |
7615 | dprintf("imsm: No spare activation allowed. " | |
7616 | "Rebuild in progress already.\n"); | |
7617 | return NULL; | |
7618 | } | |
7619 | ||
89c67882 AK |
7620 | if (a->info.array.level == 4) |
7621 | /* No repair for takeovered array | |
7622 | * imsm doesn't support raid4 | |
7623 | */ | |
7624 | return NULL; | |
7625 | ||
3b451610 AK |
7626 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
7627 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
7628 | return NULL; |
7629 | ||
95d07a2c LM |
7630 | /* |
7631 | * If there are any failed disks check state of the other volume. | |
7632 | * Block rebuild if the another one is failed until failed disks | |
7633 | * are removed from container. | |
7634 | */ | |
7635 | if (failed) { | |
c4acd1e5 | 7636 | dprintf("found failed disks in %.*s, check if there another" |
95d07a2c | 7637 | "failed sub-array.\n", |
c4acd1e5 | 7638 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
7639 | /* check if states of the other volumes allow for rebuild */ |
7640 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
7641 | if (i != inst) { | |
7642 | allowed = imsm_rebuild_allowed(a->container, | |
7643 | i, failed); | |
7644 | if (!allowed) | |
7645 | return NULL; | |
7646 | } | |
7647 | } | |
7648 | } | |
7649 | ||
88758e9d | 7650 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
7651 | for (i = 0; i < a->info.array.raid_disks; i++) { |
7652 | for (d = a->info.devs ; d ; d = d->next) | |
7653 | if (d->disk.raid_disk == i) | |
7654 | break; | |
7655 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
7656 | if (d && (d->state_fd >= 0)) | |
7657 | continue; | |
7658 | ||
272906ef | 7659 | /* |
a20d2ba5 DW |
7660 | * OK, this device needs recovery. Try to re-add the |
7661 | * previous occupant of this slot, if this fails see if | |
7662 | * we can continue the assimilation of a spare that was | |
7663 | * partially assimilated, finally try to activate a new | |
7664 | * spare. | |
272906ef DW |
7665 | */ |
7666 | dl = imsm_readd(super, i, a); | |
7667 | if (!dl) | |
b303fe21 | 7668 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 7669 | if (!dl) |
b303fe21 | 7670 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
7671 | if (!dl) |
7672 | continue; | |
7673 | ||
7674 | /* found a usable disk with enough space */ | |
7675 | di = malloc(sizeof(*di)); | |
79244939 DW |
7676 | if (!di) |
7677 | continue; | |
272906ef DW |
7678 | memset(di, 0, sizeof(*di)); |
7679 | ||
7680 | /* dl->index will be -1 in the case we are activating a | |
7681 | * pristine spare. imsm_process_update() will create a | |
7682 | * new index in this case. Once a disk is found to be | |
7683 | * failed in all member arrays it is kicked from the | |
7684 | * metadata | |
7685 | */ | |
7686 | di->disk.number = dl->index; | |
d23fe947 | 7687 | |
272906ef DW |
7688 | /* (ab)use di->devs to store a pointer to the device |
7689 | * we chose | |
7690 | */ | |
7691 | di->devs = (struct mdinfo *) dl; | |
7692 | ||
7693 | di->disk.raid_disk = i; | |
7694 | di->disk.major = dl->major; | |
7695 | di->disk.minor = dl->minor; | |
7696 | di->disk.state = 0; | |
d23534e4 | 7697 | di->recovery_start = 0; |
5551b113 | 7698 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
7699 | di->component_size = a->info.component_size; |
7700 | di->container_member = inst; | |
148acb7b | 7701 | super->random = random32(); |
272906ef DW |
7702 | di->next = rv; |
7703 | rv = di; | |
7704 | num_spares++; | |
7705 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
7706 | i, di->data_offset); | |
88758e9d DW |
7707 | } |
7708 | ||
7709 | if (!rv) | |
7710 | /* No spares found */ | |
7711 | return rv; | |
7712 | /* Now 'rv' has a list of devices to return. | |
7713 | * Create a metadata_update record to update the | |
7714 | * disk_ord_tbl for the array | |
7715 | */ | |
7716 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
7717 | if (mu) { |
7718 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
7719 | if (mu->buf == NULL) { | |
7720 | free(mu); | |
7721 | mu = NULL; | |
7722 | } | |
7723 | } | |
7724 | if (!mu) { | |
7725 | while (rv) { | |
7726 | struct mdinfo *n = rv->next; | |
7727 | ||
7728 | free(rv); | |
7729 | rv = n; | |
7730 | } | |
7731 | return NULL; | |
7732 | } | |
ca9de185 | 7733 | |
88758e9d | 7734 | mu->space = NULL; |
cb23f1f4 | 7735 | mu->space_list = NULL; |
88758e9d DW |
7736 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
7737 | mu->next = *updates; | |
7738 | u = (struct imsm_update_activate_spare *) mu->buf; | |
7739 | ||
7740 | for (di = rv ; di ; di = di->next) { | |
7741 | u->type = update_activate_spare; | |
d23fe947 DW |
7742 | u->dl = (struct dl *) di->devs; |
7743 | di->devs = NULL; | |
88758e9d DW |
7744 | u->slot = di->disk.raid_disk; |
7745 | u->array = inst; | |
7746 | u->next = u + 1; | |
7747 | u++; | |
7748 | } | |
7749 | (u-1)->next = NULL; | |
7750 | *updates = mu; | |
7751 | ||
7752 | return rv; | |
7753 | } | |
7754 | ||
54c2c1ea | 7755 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 7756 | { |
54c2c1ea | 7757 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
7758 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7759 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
7760 | struct disk_info *inf = get_disk_info(u); |
7761 | struct imsm_disk *disk; | |
8273f55e DW |
7762 | int i; |
7763 | int j; | |
8273f55e | 7764 | |
54c2c1ea | 7765 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 7766 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
7767 | for (j = 0; j < new_map->num_members; j++) |
7768 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
7769 | return 1; |
7770 | } | |
7771 | ||
7772 | return 0; | |
7773 | } | |
7774 | ||
1a64be56 LM |
7775 | |
7776 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
7777 | { | |
7778 | struct dl *dl = NULL; | |
7779 | for (dl = super->disks; dl; dl = dl->next) | |
7780 | if ((dl->major == major) && (dl->minor == minor)) | |
7781 | return dl; | |
7782 | return NULL; | |
7783 | } | |
7784 | ||
7785 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7786 | { | |
7787 | struct dl *prev = NULL; | |
7788 | struct dl *dl; | |
7789 | ||
7790 | prev = NULL; | |
7791 | for (dl = super->disks; dl; dl = dl->next) { | |
7792 | if ((dl->major == major) && (dl->minor == minor)) { | |
7793 | /* remove */ | |
7794 | if (prev) | |
7795 | prev->next = dl->next; | |
7796 | else | |
7797 | super->disks = dl->next; | |
7798 | dl->next = NULL; | |
7799 | __free_imsm_disk(dl); | |
7800 | dprintf("%s: removed %x:%x\n", | |
7801 | __func__, major, minor); | |
7802 | break; | |
7803 | } | |
7804 | prev = dl; | |
7805 | } | |
7806 | return 0; | |
7807 | } | |
7808 | ||
f21e18ca | 7809 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7810 | |
1a64be56 LM |
7811 | static int add_remove_disk_update(struct intel_super *super) |
7812 | { | |
7813 | int check_degraded = 0; | |
7814 | struct dl *disk = NULL; | |
7815 | /* add/remove some spares to/from the metadata/contrainer */ | |
7816 | while (super->disk_mgmt_list) { | |
7817 | struct dl *disk_cfg; | |
7818 | ||
7819 | disk_cfg = super->disk_mgmt_list; | |
7820 | super->disk_mgmt_list = disk_cfg->next; | |
7821 | disk_cfg->next = NULL; | |
7822 | ||
7823 | if (disk_cfg->action == DISK_ADD) { | |
7824 | disk_cfg->next = super->disks; | |
7825 | super->disks = disk_cfg; | |
7826 | check_degraded = 1; | |
7827 | dprintf("%s: added %x:%x\n", | |
7828 | __func__, disk_cfg->major, | |
7829 | disk_cfg->minor); | |
7830 | } else if (disk_cfg->action == DISK_REMOVE) { | |
7831 | dprintf("Disk remove action processed: %x.%x\n", | |
7832 | disk_cfg->major, disk_cfg->minor); | |
7833 | disk = get_disk_super(super, | |
7834 | disk_cfg->major, | |
7835 | disk_cfg->minor); | |
7836 | if (disk) { | |
7837 | /* store action status */ | |
7838 | disk->action = DISK_REMOVE; | |
7839 | /* remove spare disks only */ | |
7840 | if (disk->index == -1) { | |
7841 | remove_disk_super(super, | |
7842 | disk_cfg->major, | |
7843 | disk_cfg->minor); | |
7844 | } | |
7845 | } | |
7846 | /* release allocate disk structure */ | |
7847 | __free_imsm_disk(disk_cfg); | |
7848 | } | |
7849 | } | |
7850 | return check_degraded; | |
7851 | } | |
7852 | ||
a29911da PC |
7853 | |
7854 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
7855 | struct intel_super *super, | |
7856 | void ***space_list) | |
7857 | { | |
7858 | struct intel_dev *id; | |
7859 | void **tofree = NULL; | |
7860 | int ret_val = 0; | |
7861 | ||
7862 | dprintf("apply_reshape_migration_update()\n"); | |
7863 | if ((u->subdev < 0) || | |
7864 | (u->subdev > 1)) { | |
7865 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7866 | return ret_val; | |
7867 | } | |
7868 | if ((space_list == NULL) || (*space_list == NULL)) { | |
7869 | dprintf("imsm: Error: Memory is not allocated\n"); | |
7870 | return ret_val; | |
7871 | } | |
7872 | ||
7873 | for (id = super->devlist ; id; id = id->next) { | |
7874 | if (id->index == (unsigned)u->subdev) { | |
7875 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7876 | struct imsm_map *map; | |
7877 | struct imsm_dev *new_dev = | |
7878 | (struct imsm_dev *)*space_list; | |
238c0a71 | 7879 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
7880 | int to_state; |
7881 | struct dl *new_disk; | |
7882 | ||
7883 | if (new_dev == NULL) | |
7884 | return ret_val; | |
7885 | *space_list = **space_list; | |
7886 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 7887 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
7888 | if (migr_map) { |
7889 | dprintf("imsm: Error: migration in progress"); | |
7890 | return ret_val; | |
7891 | } | |
7892 | ||
7893 | to_state = map->map_state; | |
7894 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
7895 | map->num_members++; | |
7896 | /* this should not happen */ | |
7897 | if (u->new_disks[0] < 0) { | |
7898 | map->failed_disk_num = | |
7899 | map->num_members - 1; | |
7900 | to_state = IMSM_T_STATE_DEGRADED; | |
7901 | } else | |
7902 | to_state = IMSM_T_STATE_NORMAL; | |
7903 | } | |
8e59f3d8 | 7904 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
7905 | if (u->new_level > -1) |
7906 | map->raid_level = u->new_level; | |
238c0a71 | 7907 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
7908 | if ((u->new_level == 5) && |
7909 | (migr_map->raid_level == 0)) { | |
7910 | int ord = map->num_members - 1; | |
7911 | migr_map->num_members--; | |
7912 | if (u->new_disks[0] < 0) | |
7913 | ord |= IMSM_ORD_REBUILD; | |
7914 | set_imsm_ord_tbl_ent(map, | |
7915 | map->num_members - 1, | |
7916 | ord); | |
7917 | } | |
7918 | id->dev = new_dev; | |
7919 | tofree = (void **)dev; | |
7920 | ||
4bba0439 PC |
7921 | /* update chunk size |
7922 | */ | |
7923 | if (u->new_chunksize > 0) | |
7924 | map->blocks_per_strip = | |
7925 | __cpu_to_le16(u->new_chunksize * 2); | |
7926 | ||
a29911da PC |
7927 | /* add disk |
7928 | */ | |
7929 | if ((u->new_level != 5) || | |
7930 | (migr_map->raid_level != 0) || | |
7931 | (migr_map->raid_level == map->raid_level)) | |
7932 | goto skip_disk_add; | |
7933 | ||
7934 | if (u->new_disks[0] >= 0) { | |
7935 | /* use passes spare | |
7936 | */ | |
7937 | new_disk = get_disk_super(super, | |
7938 | major(u->new_disks[0]), | |
7939 | minor(u->new_disks[0])); | |
7940 | dprintf("imsm: new disk for reshape is: %i:%i " | |
7941 | "(%p, index = %i)\n", | |
7942 | major(u->new_disks[0]), | |
7943 | minor(u->new_disks[0]), | |
7944 | new_disk, new_disk->index); | |
7945 | if (new_disk == NULL) | |
7946 | goto error_disk_add; | |
7947 | ||
7948 | new_disk->index = map->num_members - 1; | |
7949 | /* slot to fill in autolayout | |
7950 | */ | |
7951 | new_disk->raiddisk = new_disk->index; | |
7952 | new_disk->disk.status |= CONFIGURED_DISK; | |
7953 | new_disk->disk.status &= ~SPARE_DISK; | |
7954 | } else | |
7955 | goto error_disk_add; | |
7956 | ||
7957 | skip_disk_add: | |
7958 | *tofree = *space_list; | |
7959 | /* calculate new size | |
7960 | */ | |
f3871fdc | 7961 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
7962 | |
7963 | ret_val = 1; | |
7964 | } | |
7965 | } | |
7966 | ||
7967 | if (tofree) | |
7968 | *space_list = tofree; | |
7969 | return ret_val; | |
7970 | ||
7971 | error_disk_add: | |
7972 | dprintf("Error: imsm: Cannot find disk.\n"); | |
7973 | return ret_val; | |
7974 | } | |
7975 | ||
f3871fdc AK |
7976 | static int apply_size_change_update(struct imsm_update_size_change *u, |
7977 | struct intel_super *super) | |
7978 | { | |
7979 | struct intel_dev *id; | |
7980 | int ret_val = 0; | |
7981 | ||
7982 | dprintf("apply_size_change_update()\n"); | |
7983 | if ((u->subdev < 0) || | |
7984 | (u->subdev > 1)) { | |
7985 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7986 | return ret_val; | |
7987 | } | |
7988 | ||
7989 | for (id = super->devlist ; id; id = id->next) { | |
7990 | if (id->index == (unsigned)u->subdev) { | |
7991 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7992 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
7993 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
7994 | unsigned long long blocks_per_member; | |
7995 | ||
7996 | /* calculate new size | |
7997 | */ | |
7998 | blocks_per_member = u->new_size / used_disks; | |
7999 | dprintf("imsm: apply_size_change_update(size: %llu, " | |
8000 | "blocks per member: %llu)\n", | |
8001 | u->new_size, blocks_per_member); | |
8002 | set_blocks_per_member(map, blocks_per_member); | |
8003 | imsm_set_array_size(dev, u->new_size); | |
8004 | ||
8005 | ret_val = 1; | |
8006 | break; | |
8007 | } | |
8008 | } | |
8009 | ||
8010 | return ret_val; | |
8011 | } | |
8012 | ||
8013 | ||
061d7da3 | 8014 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 8015 | struct intel_super *super, |
061d7da3 LO |
8016 | struct active_array *active_array) |
8017 | { | |
8018 | struct imsm_super *mpb = super->anchor; | |
8019 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 8020 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
8021 | struct imsm_map *migr_map; |
8022 | struct active_array *a; | |
8023 | struct imsm_disk *disk; | |
8024 | __u8 to_state; | |
8025 | struct dl *dl; | |
8026 | unsigned int found; | |
8027 | int failed; | |
5961eeec | 8028 | int victim; |
061d7da3 | 8029 | int i; |
5961eeec | 8030 | int second_map_created = 0; |
061d7da3 | 8031 | |
5961eeec | 8032 | for (; u; u = u->next) { |
238c0a71 | 8033 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 8034 | |
5961eeec | 8035 | if (victim < 0) |
8036 | return 0; | |
061d7da3 | 8037 | |
5961eeec | 8038 | for (dl = super->disks; dl; dl = dl->next) |
8039 | if (dl == u->dl) | |
8040 | break; | |
061d7da3 | 8041 | |
5961eeec | 8042 | if (!dl) { |
8043 | fprintf(stderr, "error: imsm_activate_spare passed " | |
8044 | "an unknown disk (index: %d)\n", | |
8045 | u->dl->index); | |
8046 | return 0; | |
8047 | } | |
061d7da3 | 8048 | |
5961eeec | 8049 | /* count failures (excluding rebuilds and the victim) |
8050 | * to determine map[0] state | |
8051 | */ | |
8052 | failed = 0; | |
8053 | for (i = 0; i < map->num_members; i++) { | |
8054 | if (i == u->slot) | |
8055 | continue; | |
8056 | disk = get_imsm_disk(super, | |
238c0a71 | 8057 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 8058 | if (!disk || is_failed(disk)) |
8059 | failed++; | |
8060 | } | |
061d7da3 | 8061 | |
5961eeec | 8062 | /* adding a pristine spare, assign a new index */ |
8063 | if (dl->index < 0) { | |
8064 | dl->index = super->anchor->num_disks; | |
8065 | super->anchor->num_disks++; | |
8066 | } | |
8067 | disk = &dl->disk; | |
8068 | disk->status |= CONFIGURED_DISK; | |
8069 | disk->status &= ~SPARE_DISK; | |
8070 | ||
8071 | /* mark rebuild */ | |
238c0a71 | 8072 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 8073 | if (!second_map_created) { |
8074 | second_map_created = 1; | |
8075 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8076 | migrate(dev, super, to_state, MIGR_REBUILD); | |
8077 | } else | |
8078 | map->map_state = to_state; | |
238c0a71 | 8079 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 8080 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
8081 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
8082 | dl->index | IMSM_ORD_REBUILD); | |
8083 | ||
8084 | /* update the family_num to mark a new container | |
8085 | * generation, being careful to record the existing | |
8086 | * family_num in orig_family_num to clean up after | |
8087 | * earlier mdadm versions that neglected to set it. | |
8088 | */ | |
8089 | if (mpb->orig_family_num == 0) | |
8090 | mpb->orig_family_num = mpb->family_num; | |
8091 | mpb->family_num += super->random; | |
8092 | ||
8093 | /* count arrays using the victim in the metadata */ | |
8094 | found = 0; | |
8095 | for (a = active_array; a ; a = a->next) { | |
8096 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8097 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 8098 | |
5961eeec | 8099 | if (get_imsm_disk_slot(map, victim) >= 0) |
8100 | found++; | |
8101 | } | |
061d7da3 | 8102 | |
5961eeec | 8103 | /* delete the victim if it is no longer being |
8104 | * utilized anywhere | |
061d7da3 | 8105 | */ |
5961eeec | 8106 | if (!found) { |
8107 | struct dl **dlp; | |
061d7da3 | 8108 | |
5961eeec | 8109 | /* We know that 'manager' isn't touching anything, |
8110 | * so it is safe to delete | |
8111 | */ | |
8112 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
8113 | if ((*dlp)->index == victim) |
8114 | break; | |
5961eeec | 8115 | |
8116 | /* victim may be on the missing list */ | |
8117 | if (!*dlp) | |
8118 | for (dlp = &super->missing; *dlp; | |
8119 | dlp = &(*dlp)->next) | |
8120 | if ((*dlp)->index == victim) | |
8121 | break; | |
8122 | imsm_delete(super, dlp, victim); | |
8123 | } | |
061d7da3 LO |
8124 | } |
8125 | ||
8126 | return 1; | |
8127 | } | |
a29911da | 8128 | |
2e5dc010 N |
8129 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
8130 | struct intel_super *super, | |
8131 | void ***space_list) | |
8132 | { | |
8133 | struct dl *new_disk; | |
8134 | struct intel_dev *id; | |
8135 | int i; | |
8136 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 8137 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
8138 | void **tofree = NULL; |
8139 | int devices_to_reshape = 1; | |
8140 | struct imsm_super *mpb = super->anchor; | |
8141 | int ret_val = 0; | |
d098291a | 8142 | unsigned int dev_id; |
2e5dc010 | 8143 | |
ed7333bd | 8144 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
8145 | |
8146 | /* enable spares to use in array */ | |
8147 | for (i = 0; i < delta_disks; i++) { | |
8148 | new_disk = get_disk_super(super, | |
8149 | major(u->new_disks[i]), | |
8150 | minor(u->new_disks[i])); | |
ed7333bd AK |
8151 | dprintf("imsm: new disk for reshape is: %i:%i " |
8152 | "(%p, index = %i)\n", | |
2e5dc010 N |
8153 | major(u->new_disks[i]), minor(u->new_disks[i]), |
8154 | new_disk, new_disk->index); | |
8155 | if ((new_disk == NULL) || | |
8156 | ((new_disk->index >= 0) && | |
8157 | (new_disk->index < u->old_raid_disks))) | |
8158 | goto update_reshape_exit; | |
ee4beede | 8159 | new_disk->index = disk_count++; |
2e5dc010 N |
8160 | /* slot to fill in autolayout |
8161 | */ | |
8162 | new_disk->raiddisk = new_disk->index; | |
8163 | new_disk->disk.status |= | |
8164 | CONFIGURED_DISK; | |
8165 | new_disk->disk.status &= ~SPARE_DISK; | |
8166 | } | |
8167 | ||
ed7333bd AK |
8168 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
8169 | mpb->num_raid_devs); | |
2e5dc010 N |
8170 | /* manage changes in volume |
8171 | */ | |
d098291a | 8172 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
8173 | void **sp = *space_list; |
8174 | struct imsm_dev *newdev; | |
8175 | struct imsm_map *newmap, *oldmap; | |
8176 | ||
d098291a AK |
8177 | for (id = super->devlist ; id; id = id->next) { |
8178 | if (id->index == dev_id) | |
8179 | break; | |
8180 | } | |
8181 | if (id == NULL) | |
8182 | break; | |
2e5dc010 N |
8183 | if (!sp) |
8184 | continue; | |
8185 | *space_list = *sp; | |
8186 | newdev = (void*)sp; | |
8187 | /* Copy the dev, but not (all of) the map */ | |
8188 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
8189 | oldmap = get_imsm_map(id->dev, MAP_0); |
8190 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
8191 | /* Copy the current map */ |
8192 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
8193 | /* update one device only | |
8194 | */ | |
8195 | if (devices_to_reshape) { | |
ed7333bd AK |
8196 | dprintf("imsm: modifying subdev: %i\n", |
8197 | id->index); | |
2e5dc010 N |
8198 | devices_to_reshape--; |
8199 | newdev->vol.migr_state = 1; | |
8200 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8201 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8202 | newmap->num_members = u->new_raid_disks; |
8203 | for (i = 0; i < delta_disks; i++) { | |
8204 | set_imsm_ord_tbl_ent(newmap, | |
8205 | u->old_raid_disks + i, | |
8206 | u->old_raid_disks + i); | |
8207 | } | |
8208 | /* New map is correct, now need to save old map | |
8209 | */ | |
238c0a71 | 8210 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8211 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8212 | ||
f3871fdc | 8213 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
8214 | } |
8215 | ||
8216 | sp = (void **)id->dev; | |
8217 | id->dev = newdev; | |
8218 | *sp = tofree; | |
8219 | tofree = sp; | |
8e59f3d8 AK |
8220 | |
8221 | /* Clear migration record */ | |
8222 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8223 | } |
819bc634 AK |
8224 | if (tofree) |
8225 | *space_list = tofree; | |
2e5dc010 N |
8226 | ret_val = 1; |
8227 | ||
8228 | update_reshape_exit: | |
8229 | ||
8230 | return ret_val; | |
8231 | } | |
8232 | ||
bb025c2f | 8233 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8234 | struct intel_super *super, |
8235 | void ***space_list) | |
bb025c2f KW |
8236 | { |
8237 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8238 | struct intel_dev *dv; |
8239 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8240 | struct imsm_map *map; |
8241 | struct dl *dm, *du; | |
8ca6df95 | 8242 | int i; |
bb025c2f KW |
8243 | |
8244 | for (dv = super->devlist; dv; dv = dv->next) | |
8245 | if (dv->index == (unsigned int)u->subarray) { | |
8246 | dev = dv->dev; | |
8247 | break; | |
8248 | } | |
8249 | ||
8250 | if (dev == NULL) | |
8251 | return 0; | |
8252 | ||
238c0a71 | 8253 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8254 | |
8255 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 8256 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
8257 | if (imsm_count_failed(super, dev, MAP_0) != |
8258 | (map->num_members / 2)) | |
43d5ec18 KW |
8259 | return 0; |
8260 | ||
bb025c2f KW |
8261 | /* iterate through devices to mark removed disks as spare */ |
8262 | for (dm = super->disks; dm; dm = dm->next) { | |
8263 | if (dm->disk.status & FAILED_DISK) { | |
8264 | int idx = dm->index; | |
8265 | /* update indexes on the disk list */ | |
8266 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
8267 | the index values will end up being correct.... NB */ | |
8268 | for (du = super->disks; du; du = du->next) | |
8269 | if (du->index > idx) | |
8270 | du->index--; | |
8271 | /* mark as spare disk */ | |
a8619d23 | 8272 | mark_spare(dm); |
bb025c2f KW |
8273 | } |
8274 | } | |
bb025c2f KW |
8275 | /* update map */ |
8276 | map->num_members = map->num_members / 2; | |
8277 | map->map_state = IMSM_T_STATE_NORMAL; | |
8278 | map->num_domains = 1; | |
8279 | map->raid_level = 0; | |
8280 | map->failed_disk_num = -1; | |
8281 | } | |
8282 | ||
8ca6df95 KW |
8283 | if (u->direction == R0_TO_R10) { |
8284 | void **space; | |
8285 | /* update slots in current disk list */ | |
8286 | for (dm = super->disks; dm; dm = dm->next) { | |
8287 | if (dm->index >= 0) | |
8288 | dm->index *= 2; | |
8289 | } | |
8290 | /* create new *missing* disks */ | |
8291 | for (i = 0; i < map->num_members; i++) { | |
8292 | space = *space_list; | |
8293 | if (!space) | |
8294 | continue; | |
8295 | *space_list = *space; | |
8296 | du = (void *)space; | |
8297 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
8298 | du->fd = -1; |
8299 | du->minor = 0; | |
8300 | du->major = 0; | |
8301 | du->index = (i * 2) + 1; | |
8302 | sprintf((char *)du->disk.serial, | |
8303 | " MISSING_%d", du->index); | |
8304 | sprintf((char *)du->serial, | |
8305 | "MISSING_%d", du->index); | |
8306 | du->next = super->missing; | |
8307 | super->missing = du; | |
8308 | } | |
8309 | /* create new dev and map */ | |
8310 | space = *space_list; | |
8311 | if (!space) | |
8312 | return 0; | |
8313 | *space_list = *space; | |
8314 | dev_new = (void *)space; | |
8315 | memcpy(dev_new, dev, sizeof(*dev)); | |
8316 | /* update new map */ | |
238c0a71 | 8317 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 8318 | map->num_members = map->num_members * 2; |
1a2487c2 | 8319 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
8320 | map->num_domains = 2; |
8321 | map->raid_level = 1; | |
8322 | /* replace dev<->dev_new */ | |
8323 | dv->dev = dev_new; | |
8324 | } | |
bb025c2f KW |
8325 | /* update disk order table */ |
8326 | for (du = super->disks; du; du = du->next) | |
8327 | if (du->index >= 0) | |
8328 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 8329 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
8330 | if (du->index >= 0) { |
8331 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 8332 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 8333 | } |
bb025c2f KW |
8334 | |
8335 | return 1; | |
8336 | } | |
8337 | ||
e8319a19 DW |
8338 | static void imsm_process_update(struct supertype *st, |
8339 | struct metadata_update *update) | |
8340 | { | |
8341 | /** | |
8342 | * crack open the metadata_update envelope to find the update record | |
8343 | * update can be one of: | |
d195167d AK |
8344 | * update_reshape_container_disks - all the arrays in the container |
8345 | * are being reshaped to have more devices. We need to mark | |
8346 | * the arrays for general migration and convert selected spares | |
8347 | * into active devices. | |
8348 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
8349 | * device in an array, update the disk_ord_tbl. If this disk is |
8350 | * present in all member arrays then also clear the SPARE_DISK | |
8351 | * flag | |
d195167d AK |
8352 | * update_create_array |
8353 | * update_kill_array | |
8354 | * update_rename_array | |
8355 | * update_add_remove_disk | |
e8319a19 DW |
8356 | */ |
8357 | struct intel_super *super = st->sb; | |
4d7b1503 | 8358 | struct imsm_super *mpb; |
e8319a19 DW |
8359 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
8360 | ||
4d7b1503 DW |
8361 | /* update requires a larger buf but the allocation failed */ |
8362 | if (super->next_len && !super->next_buf) { | |
8363 | super->next_len = 0; | |
8364 | return; | |
8365 | } | |
8366 | ||
8367 | if (super->next_buf) { | |
8368 | memcpy(super->next_buf, super->buf, super->len); | |
8369 | free(super->buf); | |
8370 | super->len = super->next_len; | |
8371 | super->buf = super->next_buf; | |
8372 | ||
8373 | super->next_len = 0; | |
8374 | super->next_buf = NULL; | |
8375 | } | |
8376 | ||
8377 | mpb = super->anchor; | |
8378 | ||
e8319a19 | 8379 | switch (type) { |
0ec5d470 AK |
8380 | case update_general_migration_checkpoint: { |
8381 | struct intel_dev *id; | |
8382 | struct imsm_update_general_migration_checkpoint *u = | |
8383 | (void *)update->buf; | |
8384 | ||
8385 | dprintf("imsm: process_update() " | |
8386 | "for update_general_migration_checkpoint called\n"); | |
8387 | ||
8388 | /* find device under general migration */ | |
8389 | for (id = super->devlist ; id; id = id->next) { | |
8390 | if (is_gen_migration(id->dev)) { | |
8391 | id->dev->vol.curr_migr_unit = | |
8392 | __cpu_to_le32(u->curr_migr_unit); | |
8393 | super->updates_pending++; | |
8394 | } | |
8395 | } | |
8396 | break; | |
8397 | } | |
bb025c2f KW |
8398 | case update_takeover: { |
8399 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
8400 | if (apply_takeover_update(u, super, &update->space_list)) { |
8401 | imsm_update_version_info(super); | |
bb025c2f | 8402 | super->updates_pending++; |
1a2487c2 | 8403 | } |
bb025c2f KW |
8404 | break; |
8405 | } | |
8406 | ||
78b10e66 | 8407 | case update_reshape_container_disks: { |
d195167d | 8408 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
8409 | if (apply_reshape_container_disks_update( |
8410 | u, super, &update->space_list)) | |
8411 | super->updates_pending++; | |
78b10e66 N |
8412 | break; |
8413 | } | |
48c5303a | 8414 | case update_reshape_migration: { |
a29911da PC |
8415 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
8416 | if (apply_reshape_migration_update( | |
8417 | u, super, &update->space_list)) | |
8418 | super->updates_pending++; | |
48c5303a PC |
8419 | break; |
8420 | } | |
f3871fdc AK |
8421 | case update_size_change: { |
8422 | struct imsm_update_size_change *u = (void *)update->buf; | |
8423 | if (apply_size_change_update(u, super)) | |
8424 | super->updates_pending++; | |
8425 | break; | |
8426 | } | |
e8319a19 DW |
8427 | case update_activate_spare: { |
8428 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
061d7da3 LO |
8429 | if (apply_update_activate_spare(u, super, st->arrays)) |
8430 | super->updates_pending++; | |
8273f55e DW |
8431 | break; |
8432 | } | |
8433 | case update_create_array: { | |
8434 | /* someone wants to create a new array, we need to be aware of | |
8435 | * a few races/collisions: | |
8436 | * 1/ 'Create' called by two separate instances of mdadm | |
8437 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
8438 | * devices that have since been assimilated via | |
8439 | * activate_spare. | |
8440 | * In the event this update can not be carried out mdadm will | |
8441 | * (FIX ME) notice that its update did not take hold. | |
8442 | */ | |
8443 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8444 | struct intel_dev *dv; |
8273f55e DW |
8445 | struct imsm_dev *dev; |
8446 | struct imsm_map *map, *new_map; | |
8447 | unsigned long long start, end; | |
8448 | unsigned long long new_start, new_end; | |
8449 | int i; | |
54c2c1ea DW |
8450 | struct disk_info *inf; |
8451 | struct dl *dl; | |
8273f55e DW |
8452 | |
8453 | /* handle racing creates: first come first serve */ | |
8454 | if (u->dev_idx < mpb->num_raid_devs) { | |
8455 | dprintf("%s: subarray %d already defined\n", | |
8456 | __func__, u->dev_idx); | |
ba2de7ba | 8457 | goto create_error; |
8273f55e DW |
8458 | } |
8459 | ||
8460 | /* check update is next in sequence */ | |
8461 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
8462 | dprintf("%s: can not create array %d expected index %d\n", |
8463 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 8464 | goto create_error; |
8273f55e DW |
8465 | } |
8466 | ||
238c0a71 | 8467 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
8468 | new_start = pba_of_lba0(new_map); |
8469 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 8470 | inf = get_disk_info(u); |
8273f55e DW |
8471 | |
8472 | /* handle activate_spare versus create race: | |
8473 | * check to make sure that overlapping arrays do not include | |
8474 | * overalpping disks | |
8475 | */ | |
8476 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 8477 | dev = get_imsm_dev(super, i); |
238c0a71 | 8478 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
8479 | start = pba_of_lba0(map); |
8480 | end = start + blocks_per_member(map); | |
8273f55e DW |
8481 | if ((new_start >= start && new_start <= end) || |
8482 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
8483 | /* overlap */; |
8484 | else | |
8485 | continue; | |
8486 | ||
8487 | if (disks_overlap(super, i, u)) { | |
8273f55e | 8488 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 8489 | goto create_error; |
8273f55e DW |
8490 | } |
8491 | } | |
8273f55e | 8492 | |
949c47a0 DW |
8493 | /* check that prepare update was successful */ |
8494 | if (!update->space) { | |
8495 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 8496 | goto create_error; |
949c47a0 DW |
8497 | } |
8498 | ||
54c2c1ea DW |
8499 | /* check that all disks are still active before committing |
8500 | * changes. FIXME: could we instead handle this by creating a | |
8501 | * degraded array? That's probably not what the user expects, | |
8502 | * so better to drop this update on the floor. | |
8503 | */ | |
8504 | for (i = 0; i < new_map->num_members; i++) { | |
8505 | dl = serial_to_dl(inf[i].serial, super); | |
8506 | if (!dl) { | |
8507 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 8508 | goto create_error; |
54c2c1ea | 8509 | } |
949c47a0 DW |
8510 | } |
8511 | ||
8273f55e | 8512 | super->updates_pending++; |
54c2c1ea DW |
8513 | |
8514 | /* convert spares to members and fixup ord_tbl */ | |
8515 | for (i = 0; i < new_map->num_members; i++) { | |
8516 | dl = serial_to_dl(inf[i].serial, super); | |
8517 | if (dl->index == -1) { | |
8518 | dl->index = mpb->num_disks; | |
8519 | mpb->num_disks++; | |
8520 | dl->disk.status |= CONFIGURED_DISK; | |
8521 | dl->disk.status &= ~SPARE_DISK; | |
8522 | } | |
8523 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
8524 | } | |
8525 | ||
ba2de7ba DW |
8526 | dv = update->space; |
8527 | dev = dv->dev; | |
949c47a0 DW |
8528 | update->space = NULL; |
8529 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
8530 | dv->index = u->dev_idx; |
8531 | dv->next = super->devlist; | |
8532 | super->devlist = dv; | |
8273f55e | 8533 | mpb->num_raid_devs++; |
8273f55e | 8534 | |
4d1313e9 | 8535 | imsm_update_version_info(super); |
8273f55e | 8536 | break; |
ba2de7ba DW |
8537 | create_error: |
8538 | /* mdmon knows how to release update->space, but not | |
8539 | * ((struct intel_dev *) update->space)->dev | |
8540 | */ | |
8541 | if (update->space) { | |
8542 | dv = update->space; | |
8543 | free(dv->dev); | |
8544 | } | |
8273f55e | 8545 | break; |
e8319a19 | 8546 | } |
33414a01 DW |
8547 | case update_kill_array: { |
8548 | struct imsm_update_kill_array *u = (void *) update->buf; | |
8549 | int victim = u->dev_idx; | |
8550 | struct active_array *a; | |
8551 | struct intel_dev **dp; | |
8552 | struct imsm_dev *dev; | |
8553 | ||
8554 | /* sanity check that we are not affecting the uuid of | |
8555 | * active arrays, or deleting an active array | |
8556 | * | |
8557 | * FIXME when immutable ids are available, but note that | |
8558 | * we'll also need to fixup the invalidated/active | |
8559 | * subarray indexes in mdstat | |
8560 | */ | |
8561 | for (a = st->arrays; a; a = a->next) | |
8562 | if (a->info.container_member >= victim) | |
8563 | break; | |
8564 | /* by definition if mdmon is running at least one array | |
8565 | * is active in the container, so checking | |
8566 | * mpb->num_raid_devs is just extra paranoia | |
8567 | */ | |
8568 | dev = get_imsm_dev(super, victim); | |
8569 | if (a || !dev || mpb->num_raid_devs == 1) { | |
8570 | dprintf("failed to delete subarray-%d\n", victim); | |
8571 | break; | |
8572 | } | |
8573 | ||
8574 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 8575 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
8576 | *dp = (*dp)->next; |
8577 | } else { | |
f21e18ca | 8578 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
8579 | (*dp)->index--; |
8580 | dp = &(*dp)->next; | |
8581 | } | |
8582 | mpb->num_raid_devs--; | |
8583 | super->updates_pending++; | |
8584 | break; | |
8585 | } | |
aa534678 DW |
8586 | case update_rename_array: { |
8587 | struct imsm_update_rename_array *u = (void *) update->buf; | |
8588 | char name[MAX_RAID_SERIAL_LEN+1]; | |
8589 | int target = u->dev_idx; | |
8590 | struct active_array *a; | |
8591 | struct imsm_dev *dev; | |
8592 | ||
8593 | /* sanity check that we are not affecting the uuid of | |
8594 | * an active array | |
8595 | */ | |
8596 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
8597 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
8598 | for (a = st->arrays; a; a = a->next) | |
8599 | if (a->info.container_member == target) | |
8600 | break; | |
8601 | dev = get_imsm_dev(super, u->dev_idx); | |
8602 | if (a || !dev || !check_name(super, name, 1)) { | |
8603 | dprintf("failed to rename subarray-%d\n", target); | |
8604 | break; | |
8605 | } | |
8606 | ||
cdbe98cd | 8607 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
8608 | super->updates_pending++; |
8609 | break; | |
8610 | } | |
1a64be56 | 8611 | case update_add_remove_disk: { |
43dad3d6 | 8612 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
8613 | * being added, check teh status of add_remove_disk |
8614 | * if discs has been added. | |
8615 | */ | |
8616 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 8617 | struct active_array *a; |
072b727f DW |
8618 | |
8619 | super->updates_pending++; | |
1a64be56 | 8620 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
8621 | a->check_degraded = 1; |
8622 | } | |
43dad3d6 | 8623 | break; |
e8319a19 | 8624 | } |
1a64be56 LM |
8625 | default: |
8626 | fprintf(stderr, "error: unsuported process update type:" | |
8627 | "(type: %d)\n", type); | |
8628 | } | |
e8319a19 | 8629 | } |
88758e9d | 8630 | |
bc0b9d34 PC |
8631 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
8632 | ||
8273f55e DW |
8633 | static void imsm_prepare_update(struct supertype *st, |
8634 | struct metadata_update *update) | |
8635 | { | |
949c47a0 | 8636 | /** |
4d7b1503 DW |
8637 | * Allocate space to hold new disk entries, raid-device entries or a new |
8638 | * mpb if necessary. The manager synchronously waits for updates to | |
8639 | * complete in the monitor, so new mpb buffers allocated here can be | |
8640 | * integrated by the monitor thread without worrying about live pointers | |
8641 | * in the manager thread. | |
8273f55e | 8642 | */ |
949c47a0 | 8643 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
8644 | struct intel_super *super = st->sb; |
8645 | struct imsm_super *mpb = super->anchor; | |
8646 | size_t buf_len; | |
8647 | size_t len = 0; | |
949c47a0 DW |
8648 | |
8649 | switch (type) { | |
0ec5d470 AK |
8650 | case update_general_migration_checkpoint: |
8651 | dprintf("imsm: prepare_update() " | |
8652 | "for update_general_migration_checkpoint called\n"); | |
8653 | break; | |
abedf5fc KW |
8654 | case update_takeover: { |
8655 | struct imsm_update_takeover *u = (void *)update->buf; | |
8656 | if (u->direction == R0_TO_R10) { | |
8657 | void **tail = (void **)&update->space_list; | |
8658 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 8659 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
8660 | int num_members = map->num_members; |
8661 | void *space; | |
8662 | int size, i; | |
8663 | int err = 0; | |
8664 | /* allocate memory for added disks */ | |
8665 | for (i = 0; i < num_members; i++) { | |
8666 | size = sizeof(struct dl); | |
8667 | space = malloc(size); | |
8668 | if (!space) { | |
8669 | err++; | |
8670 | break; | |
8671 | } | |
8672 | *tail = space; | |
8673 | tail = space; | |
8674 | *tail = NULL; | |
8675 | } | |
8676 | /* allocate memory for new device */ | |
8677 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
8678 | (num_members * sizeof(__u32)); | |
8679 | space = malloc(size); | |
8680 | if (!space) | |
8681 | err++; | |
8682 | else { | |
8683 | *tail = space; | |
8684 | tail = space; | |
8685 | *tail = NULL; | |
8686 | } | |
8687 | if (!err) { | |
8688 | len = disks_to_mpb_size(num_members * 2); | |
8689 | } else { | |
8690 | /* if allocation didn't success, free buffer */ | |
8691 | while (update->space_list) { | |
8692 | void **sp = update->space_list; | |
8693 | update->space_list = *sp; | |
8694 | free(sp); | |
8695 | } | |
8696 | } | |
8697 | } | |
8698 | ||
8699 | break; | |
8700 | } | |
78b10e66 | 8701 | case update_reshape_container_disks: { |
d195167d AK |
8702 | /* Every raid device in the container is about to |
8703 | * gain some more devices, and we will enter a | |
8704 | * reconfiguration. | |
8705 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
8706 | * will now hold 2 of them. | |
8707 | * Thus we need new 'struct imsm_dev' allocations sized | |
8708 | * as sizeof_imsm_dev but with more devices in both maps. | |
8709 | */ | |
8710 | struct imsm_update_reshape *u = (void *)update->buf; | |
8711 | struct intel_dev *dl; | |
8712 | void **space_tail = (void**)&update->space_list; | |
8713 | ||
8714 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8715 | ||
8716 | for (dl = super->devlist; dl; dl = dl->next) { | |
8717 | int size = sizeof_imsm_dev(dl->dev, 1); | |
8718 | void *s; | |
d677e0b8 AK |
8719 | if (u->new_raid_disks > u->old_raid_disks) |
8720 | size += sizeof(__u32)*2* | |
8721 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
8722 | s = malloc(size); |
8723 | if (!s) | |
8724 | break; | |
8725 | *space_tail = s; | |
8726 | space_tail = s; | |
8727 | *space_tail = NULL; | |
8728 | } | |
8729 | ||
8730 | len = disks_to_mpb_size(u->new_raid_disks); | |
8731 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
8732 | break; |
8733 | } | |
48c5303a | 8734 | case update_reshape_migration: { |
bc0b9d34 PC |
8735 | /* for migration level 0->5 we need to add disks |
8736 | * so the same as for container operation we will copy | |
8737 | * device to the bigger location. | |
8738 | * in memory prepared device and new disk area are prepared | |
8739 | * for usage in process update | |
8740 | */ | |
8741 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
8742 | struct intel_dev *id; | |
8743 | void **space_tail = (void **)&update->space_list; | |
8744 | int size; | |
8745 | void *s; | |
8746 | int current_level = -1; | |
8747 | ||
8748 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8749 | ||
8750 | /* add space for bigger array in update | |
8751 | */ | |
8752 | for (id = super->devlist; id; id = id->next) { | |
8753 | if (id->index == (unsigned)u->subdev) { | |
8754 | size = sizeof_imsm_dev(id->dev, 1); | |
8755 | if (u->new_raid_disks > u->old_raid_disks) | |
8756 | size += sizeof(__u32)*2* | |
8757 | (u->new_raid_disks - u->old_raid_disks); | |
8758 | s = malloc(size); | |
8759 | if (!s) | |
8760 | break; | |
8761 | *space_tail = s; | |
8762 | space_tail = s; | |
8763 | *space_tail = NULL; | |
8764 | break; | |
8765 | } | |
8766 | } | |
8767 | if (update->space_list == NULL) | |
8768 | break; | |
8769 | ||
8770 | /* add space for disk in update | |
8771 | */ | |
8772 | size = sizeof(struct dl); | |
8773 | s = malloc(size); | |
8774 | if (!s) { | |
8775 | free(update->space_list); | |
8776 | update->space_list = NULL; | |
8777 | break; | |
8778 | } | |
8779 | *space_tail = s; | |
8780 | space_tail = s; | |
8781 | *space_tail = NULL; | |
8782 | ||
8783 | /* add spare device to update | |
8784 | */ | |
8785 | for (id = super->devlist ; id; id = id->next) | |
8786 | if (id->index == (unsigned)u->subdev) { | |
8787 | struct imsm_dev *dev; | |
8788 | struct imsm_map *map; | |
8789 | ||
8790 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 8791 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
8792 | current_level = map->raid_level; |
8793 | break; | |
8794 | } | |
8795 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
8796 | struct mdinfo *spares; | |
8797 | ||
8798 | spares = get_spares_for_grow(st); | |
8799 | if (spares) { | |
8800 | struct dl *dl; | |
8801 | struct mdinfo *dev; | |
8802 | ||
8803 | dev = spares->devs; | |
8804 | if (dev) { | |
8805 | u->new_disks[0] = | |
8806 | makedev(dev->disk.major, | |
8807 | dev->disk.minor); | |
8808 | dl = get_disk_super(super, | |
8809 | dev->disk.major, | |
8810 | dev->disk.minor); | |
8811 | dl->index = u->old_raid_disks; | |
8812 | dev = dev->next; | |
8813 | } | |
8814 | sysfs_free(spares); | |
8815 | } | |
8816 | } | |
8817 | len = disks_to_mpb_size(u->new_raid_disks); | |
8818 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8819 | break; |
8820 | } | |
f3871fdc AK |
8821 | case update_size_change: { |
8822 | break; | |
8823 | } | |
949c47a0 DW |
8824 | case update_create_array: { |
8825 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8826 | struct intel_dev *dv; |
54c2c1ea | 8827 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8828 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8829 | struct dl *dl; |
8830 | struct disk_info *inf; | |
8831 | int i; | |
8832 | int activate = 0; | |
949c47a0 | 8833 | |
54c2c1ea DW |
8834 | inf = get_disk_info(u); |
8835 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
8836 | /* allocate a new super->devlist entry */ |
8837 | dv = malloc(sizeof(*dv)); | |
8838 | if (dv) { | |
8839 | dv->dev = malloc(len); | |
8840 | if (dv->dev) | |
8841 | update->space = dv; | |
8842 | else { | |
8843 | free(dv); | |
8844 | update->space = NULL; | |
8845 | } | |
8846 | } | |
949c47a0 | 8847 | |
54c2c1ea DW |
8848 | /* count how many spares will be converted to members */ |
8849 | for (i = 0; i < map->num_members; i++) { | |
8850 | dl = serial_to_dl(inf[i].serial, super); | |
8851 | if (!dl) { | |
8852 | /* hmm maybe it failed?, nothing we can do about | |
8853 | * it here | |
8854 | */ | |
8855 | continue; | |
8856 | } | |
8857 | if (count_memberships(dl, super) == 0) | |
8858 | activate++; | |
8859 | } | |
8860 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
8861 | break; |
8862 | default: | |
8863 | break; | |
8864 | } | |
8865 | } | |
8273f55e | 8866 | |
4d7b1503 DW |
8867 | /* check if we need a larger metadata buffer */ |
8868 | if (super->next_buf) | |
8869 | buf_len = super->next_len; | |
8870 | else | |
8871 | buf_len = super->len; | |
8872 | ||
8873 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
8874 | /* ok we need a larger buf than what is currently allocated | |
8875 | * if this allocation fails process_update will notice that | |
8876 | * ->next_len is set and ->next_buf is NULL | |
8877 | */ | |
8878 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
8879 | if (super->next_buf) | |
8880 | free(super->next_buf); | |
8881 | ||
8882 | super->next_len = buf_len; | |
1f45a8ad DW |
8883 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
8884 | memset(super->next_buf, 0, buf_len); | |
8885 | else | |
4d7b1503 DW |
8886 | super->next_buf = NULL; |
8887 | } | |
8273f55e DW |
8888 | } |
8889 | ||
ae6aad82 | 8890 | /* must be called while manager is quiesced */ |
f21e18ca | 8891 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
8892 | { |
8893 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
8894 | struct dl *iter; |
8895 | struct imsm_dev *dev; | |
8896 | struct imsm_map *map; | |
24565c9a DW |
8897 | int i, j, num_members; |
8898 | __u32 ord; | |
ae6aad82 | 8899 | |
24565c9a DW |
8900 | dprintf("%s: deleting device[%d] from imsm_super\n", |
8901 | __func__, index); | |
ae6aad82 DW |
8902 | |
8903 | /* shift all indexes down one */ | |
8904 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 8905 | if (iter->index > (int)index) |
ae6aad82 | 8906 | iter->index--; |
47ee5a45 | 8907 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 8908 | if (iter->index > (int)index) |
47ee5a45 | 8909 | iter->index--; |
ae6aad82 DW |
8910 | |
8911 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8912 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8913 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
8914 | num_members = map->num_members; |
8915 | for (j = 0; j < num_members; j++) { | |
8916 | /* update ord entries being careful not to propagate | |
8917 | * ord-flags to the first map | |
8918 | */ | |
238c0a71 | 8919 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 8920 | |
24565c9a DW |
8921 | if (ord_to_idx(ord) <= index) |
8922 | continue; | |
ae6aad82 | 8923 | |
238c0a71 | 8924 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 8925 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 8926 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
8927 | if (map) |
8928 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
8929 | } |
8930 | } | |
8931 | ||
8932 | mpb->num_disks--; | |
8933 | super->updates_pending++; | |
24565c9a DW |
8934 | if (*dlp) { |
8935 | struct dl *dl = *dlp; | |
8936 | ||
8937 | *dlp = (*dlp)->next; | |
8938 | __free_imsm_disk(dl); | |
8939 | } | |
ae6aad82 | 8940 | } |
9e2d750d | 8941 | #endif /* MDASSEMBLE */ |
9a717282 AK |
8942 | |
8943 | static void close_targets(int *targets, int new_disks) | |
8944 | { | |
8945 | int i; | |
8946 | ||
8947 | if (!targets) | |
8948 | return; | |
8949 | ||
8950 | for (i = 0; i < new_disks; i++) { | |
8951 | if (targets[i] >= 0) { | |
8952 | close(targets[i]); | |
8953 | targets[i] = -1; | |
8954 | } | |
8955 | } | |
8956 | } | |
8957 | ||
8958 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
8959 | struct intel_super *super, | |
8960 | struct imsm_dev *dev) | |
8961 | { | |
8962 | switch (level) { | |
bf5cf7c7 | 8963 | case 1: |
9a717282 AK |
8964 | case 10:{ |
8965 | int ret_val = 0; | |
8966 | struct imsm_map *map; | |
8967 | int i; | |
8968 | ||
8969 | ret_val = raid_disks/2; | |
8970 | /* check map if all disks pairs not failed | |
8971 | * in both maps | |
8972 | */ | |
238c0a71 | 8973 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
8974 | for (i = 0; i < ret_val; i++) { |
8975 | int degradation = 0; | |
8976 | if (get_imsm_disk(super, i) == NULL) | |
8977 | degradation++; | |
8978 | if (get_imsm_disk(super, i + 1) == NULL) | |
8979 | degradation++; | |
8980 | if (degradation == 2) | |
8981 | return 0; | |
8982 | } | |
238c0a71 | 8983 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
8984 | /* if there is no second map |
8985 | * result can be returned | |
8986 | */ | |
8987 | if (map == NULL) | |
8988 | return ret_val; | |
8989 | /* check degradation in second map | |
8990 | */ | |
8991 | for (i = 0; i < ret_val; i++) { | |
8992 | int degradation = 0; | |
8993 | if (get_imsm_disk(super, i) == NULL) | |
8994 | degradation++; | |
8995 | if (get_imsm_disk(super, i + 1) == NULL) | |
8996 | degradation++; | |
8997 | if (degradation == 2) | |
8998 | return 0; | |
8999 | } | |
9000 | return ret_val; | |
9001 | } | |
9002 | case 5: | |
9003 | return 1; | |
9004 | case 6: | |
9005 | return 2; | |
9006 | default: | |
9007 | return 0; | |
9008 | } | |
9009 | } | |
9010 | ||
9011 | ||
687629c2 AK |
9012 | /******************************************************************************* |
9013 | * Function: open_backup_targets | |
9014 | * Description: Function opens file descriptors for all devices given in | |
9015 | * info->devs | |
9016 | * Parameters: | |
9017 | * info : general array info | |
9018 | * raid_disks : number of disks | |
9019 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
9020 | * super : intel super for raid10 degradation check |
9021 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
9022 | * Returns: |
9023 | * 0 : success | |
9024 | * -1 : fail | |
9025 | ******************************************************************************/ | |
9a717282 AK |
9026 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
9027 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
9028 | { |
9029 | struct mdinfo *sd; | |
f627f5ad | 9030 | int i; |
9a717282 | 9031 | int opened = 0; |
f627f5ad AK |
9032 | |
9033 | for (i = 0; i < raid_disks; i++) | |
9034 | raid_fds[i] = -1; | |
687629c2 AK |
9035 | |
9036 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9037 | char *dn; | |
9038 | ||
9039 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
9040 | dprintf("disk is faulty!!\n"); | |
9041 | continue; | |
9042 | } | |
9043 | ||
9044 | if ((sd->disk.raid_disk >= raid_disks) || | |
9045 | (sd->disk.raid_disk < 0)) | |
9046 | continue; | |
9047 | ||
9048 | dn = map_dev(sd->disk.major, | |
9049 | sd->disk.minor, 1); | |
9050 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
9051 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
9052 | fprintf(stderr, "cannot open component\n"); | |
9a717282 | 9053 | continue; |
687629c2 | 9054 | } |
9a717282 AK |
9055 | opened++; |
9056 | } | |
9057 | /* check if maximum array degradation level is not exceeded | |
9058 | */ | |
9059 | if ((raid_disks - opened) > | |
9060 | imsm_get_allowed_degradation(info->new_level, | |
9061 | raid_disks, | |
9062 | super, dev)) { | |
9063 | fprintf(stderr, "Not enough disks can be opened.\n"); | |
9064 | close_targets(raid_fds, raid_disks); | |
9065 | return -2; | |
687629c2 AK |
9066 | } |
9067 | return 0; | |
9068 | } | |
9069 | ||
9e2d750d | 9070 | #ifndef MDASSEMBLE |
687629c2 AK |
9071 | /******************************************************************************* |
9072 | * Function: init_migr_record_imsm | |
9073 | * Description: Function inits imsm migration record | |
9074 | * Parameters: | |
9075 | * super : imsm internal array info | |
9076 | * dev : device under migration | |
9077 | * info : general array info to find the smallest device | |
9078 | * Returns: | |
9079 | * none | |
9080 | ******************************************************************************/ | |
9081 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
9082 | struct mdinfo *info) | |
9083 | { | |
9084 | struct intel_super *super = st->sb; | |
9085 | struct migr_record *migr_rec = super->migr_rec; | |
9086 | int new_data_disks; | |
9087 | unsigned long long dsize, dev_sectors; | |
9088 | long long unsigned min_dev_sectors = -1LLU; | |
9089 | struct mdinfo *sd; | |
9090 | char nm[30]; | |
9091 | int fd; | |
238c0a71 AK |
9092 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
9093 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 9094 | unsigned long long num_migr_units; |
3ef4403c | 9095 | unsigned long long array_blocks; |
687629c2 AK |
9096 | |
9097 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
9098 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
9099 | ||
9100 | /* only ascending reshape supported now */ | |
9101 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
9102 | ||
9103 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
9104 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
9105 | migr_rec->dest_depth_per_unit *= |
9106 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 9107 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
9108 | migr_rec->blocks_per_unit = |
9109 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
9110 | migr_rec->dest_depth_per_unit = | |
9111 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 9112 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
9113 | num_migr_units = |
9114 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
9115 | ||
9116 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
9117 | num_migr_units++; | |
9118 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
9119 | ||
9120 | migr_rec->post_migr_vol_cap = dev->size_low; | |
9121 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
9122 | ||
9123 | ||
9124 | /* Find the smallest dev */ | |
9125 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9126 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
9127 | fd = dev_open(nm, O_RDONLY); | |
9128 | if (fd < 0) | |
9129 | continue; | |
9130 | get_dev_size(fd, NULL, &dsize); | |
9131 | dev_sectors = dsize / 512; | |
9132 | if (dev_sectors < min_dev_sectors) | |
9133 | min_dev_sectors = dev_sectors; | |
9134 | close(fd); | |
9135 | } | |
9136 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
9137 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
9138 | ||
9139 | write_imsm_migr_rec(st); | |
9140 | ||
9141 | return; | |
9142 | } | |
9143 | ||
9144 | /******************************************************************************* | |
9145 | * Function: save_backup_imsm | |
9146 | * Description: Function saves critical data stripes to Migration Copy Area | |
9147 | * and updates the current migration unit status. | |
9148 | * Use restore_stripes() to form a destination stripe, | |
9149 | * and to write it to the Copy Area. | |
9150 | * Parameters: | |
9151 | * st : supertype information | |
aea93171 | 9152 | * dev : imsm device that backup is saved for |
687629c2 AK |
9153 | * info : general array info |
9154 | * buf : input buffer | |
687629c2 AK |
9155 | * length : length of data to backup (blocks_per_unit) |
9156 | * Returns: | |
9157 | * 0 : success | |
9158 | *, -1 : fail | |
9159 | ******************************************************************************/ | |
9160 | int save_backup_imsm(struct supertype *st, | |
9161 | struct imsm_dev *dev, | |
9162 | struct mdinfo *info, | |
9163 | void *buf, | |
687629c2 AK |
9164 | int length) |
9165 | { | |
9166 | int rv = -1; | |
9167 | struct intel_super *super = st->sb; | |
9168 | unsigned long long *target_offsets = NULL; | |
9169 | int *targets = NULL; | |
9170 | int i; | |
238c0a71 | 9171 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 9172 | int new_disks = map_dest->num_members; |
ab724b98 AK |
9173 | int dest_layout = 0; |
9174 | int dest_chunk; | |
d1877f69 | 9175 | unsigned long long start; |
238c0a71 | 9176 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
9177 | |
9178 | targets = malloc(new_disks * sizeof(int)); | |
9179 | if (!targets) | |
9180 | goto abort; | |
9181 | ||
7e45b550 AK |
9182 | for (i = 0; i < new_disks; i++) |
9183 | targets[i] = -1; | |
9184 | ||
687629c2 AK |
9185 | target_offsets = malloc(new_disks * sizeof(unsigned long long)); |
9186 | if (!target_offsets) | |
9187 | goto abort; | |
9188 | ||
d1877f69 | 9189 | start = info->reshape_progress * 512; |
687629c2 | 9190 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
9191 | target_offsets[i] = (unsigned long long) |
9192 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
9193 | /* move back copy area adderss, it will be moved forward |
9194 | * in restore_stripes() using start input variable | |
9195 | */ | |
9196 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
9197 | } |
9198 | ||
9a717282 AK |
9199 | if (open_backup_targets(info, new_disks, targets, |
9200 | super, dev)) | |
687629c2 AK |
9201 | goto abort; |
9202 | ||
68eb8bc6 | 9203 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
9204 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
9205 | ||
687629c2 AK |
9206 | if (restore_stripes(targets, /* list of dest devices */ |
9207 | target_offsets, /* migration record offsets */ | |
9208 | new_disks, | |
ab724b98 AK |
9209 | dest_chunk, |
9210 | map_dest->raid_level, | |
9211 | dest_layout, | |
9212 | -1, /* source backup file descriptor */ | |
9213 | 0, /* input buf offset | |
9214 | * always 0 buf is already offseted */ | |
d1877f69 | 9215 | start, |
687629c2 AK |
9216 | length, |
9217 | buf) != 0) { | |
9218 | fprintf(stderr, Name ": Error restoring stripes\n"); | |
9219 | goto abort; | |
9220 | } | |
9221 | ||
9222 | rv = 0; | |
9223 | ||
9224 | abort: | |
9225 | if (targets) { | |
9a717282 | 9226 | close_targets(targets, new_disks); |
687629c2 AK |
9227 | free(targets); |
9228 | } | |
9229 | free(target_offsets); | |
9230 | ||
9231 | return rv; | |
9232 | } | |
9233 | ||
9234 | /******************************************************************************* | |
9235 | * Function: save_checkpoint_imsm | |
9236 | * Description: Function called for current unit status update | |
9237 | * in the migration record. It writes it to disk. | |
9238 | * Parameters: | |
9239 | * super : imsm internal array info | |
9240 | * info : general array info | |
9241 | * Returns: | |
9242 | * 0: success | |
9243 | * 1: failure | |
0228d92c AK |
9244 | * 2: failure, means no valid migration record |
9245 | * / no general migration in progress / | |
687629c2 AK |
9246 | ******************************************************************************/ |
9247 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
9248 | { | |
9249 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
9250 | unsigned long long blocks_per_unit; |
9251 | unsigned long long curr_migr_unit; | |
9252 | ||
2e062e82 AK |
9253 | if (load_imsm_migr_rec(super, info) != 0) { |
9254 | dprintf("imsm: ERROR: Cannot read migration record " | |
9255 | "for checkpoint save.\n"); | |
9256 | return 1; | |
9257 | } | |
9258 | ||
f8b72ef5 AK |
9259 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
9260 | if (blocks_per_unit == 0) { | |
0228d92c AK |
9261 | dprintf("imsm: no migration in progress.\n"); |
9262 | return 2; | |
687629c2 | 9263 | } |
f8b72ef5 AK |
9264 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
9265 | /* check if array is alligned to copy area | |
9266 | * if it is not alligned, add one to current migration unit value | |
9267 | * this can happend on array reshape finish only | |
9268 | */ | |
9269 | if (info->reshape_progress % blocks_per_unit) | |
9270 | curr_migr_unit++; | |
687629c2 AK |
9271 | |
9272 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 9273 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
9274 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9275 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
9276 | __cpu_to_le32(curr_migr_unit * |
9277 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
9278 | if (write_imsm_migr_rec(st) < 0) { |
9279 | dprintf("imsm: Cannot write migration record " | |
9280 | "outside backup area\n"); | |
9281 | return 1; | |
9282 | } | |
9283 | ||
9284 | return 0; | |
9285 | } | |
9286 | ||
276d77db AK |
9287 | /******************************************************************************* |
9288 | * Function: recover_backup_imsm | |
9289 | * Description: Function recovers critical data from the Migration Copy Area | |
9290 | * while assembling an array. | |
9291 | * Parameters: | |
9292 | * super : imsm internal array info | |
9293 | * info : general array info | |
9294 | * Returns: | |
9295 | * 0 : success (or there is no data to recover) | |
9296 | * 1 : fail | |
9297 | ******************************************************************************/ | |
9298 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
9299 | { | |
9300 | struct intel_super *super = st->sb; | |
9301 | struct migr_record *migr_rec = super->migr_rec; | |
9302 | struct imsm_map *map_dest = NULL; | |
9303 | struct intel_dev *id = NULL; | |
9304 | unsigned long long read_offset; | |
9305 | unsigned long long write_offset; | |
9306 | unsigned unit_len; | |
9307 | int *targets = NULL; | |
9308 | int new_disks, i, err; | |
9309 | char *buf = NULL; | |
9310 | int retval = 1; | |
9311 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
9312 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 9313 | char buffer[20]; |
6c3560c0 | 9314 | int skipped_disks = 0; |
276d77db AK |
9315 | |
9316 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
9317 | if (err < 1) | |
9318 | return 1; | |
9319 | ||
9320 | /* recover data only during assemblation */ | |
9321 | if (strncmp(buffer, "inactive", 8) != 0) | |
9322 | return 0; | |
9323 | /* no data to recover */ | |
9324 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
9325 | return 0; | |
9326 | if (curr_migr_unit >= num_migr_units) | |
9327 | return 1; | |
9328 | ||
9329 | /* find device during reshape */ | |
9330 | for (id = super->devlist; id; id = id->next) | |
9331 | if (is_gen_migration(id->dev)) | |
9332 | break; | |
9333 | if (id == NULL) | |
9334 | return 1; | |
9335 | ||
238c0a71 | 9336 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
9337 | new_disks = map_dest->num_members; |
9338 | ||
9339 | read_offset = (unsigned long long) | |
9340 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
9341 | ||
9342 | write_offset = ((unsigned long long) | |
9343 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 9344 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
9345 | |
9346 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9347 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
9348 | goto abort; | |
9349 | targets = malloc(new_disks * sizeof(int)); | |
9350 | if (!targets) | |
9351 | goto abort; | |
9352 | ||
9a717282 | 9353 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
f627f5ad AK |
9354 | fprintf(stderr, |
9355 | Name ": Cannot open some devices belonging to array.\n"); | |
9356 | goto abort; | |
9357 | } | |
276d77db AK |
9358 | |
9359 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
9360 | if (targets[i] < 0) { |
9361 | skipped_disks++; | |
9362 | continue; | |
9363 | } | |
276d77db AK |
9364 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
9365 | fprintf(stderr, | |
9366 | Name ": Cannot seek to block: %s\n", | |
9367 | strerror(errno)); | |
137debce AK |
9368 | skipped_disks++; |
9369 | continue; | |
276d77db | 9370 | } |
9ec11d1a | 9371 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
9372 | fprintf(stderr, |
9373 | Name ": Cannot read copy area block: %s\n", | |
9374 | strerror(errno)); | |
137debce AK |
9375 | skipped_disks++; |
9376 | continue; | |
276d77db AK |
9377 | } |
9378 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
9379 | fprintf(stderr, | |
9380 | Name ": Cannot seek to block: %s\n", | |
9381 | strerror(errno)); | |
137debce AK |
9382 | skipped_disks++; |
9383 | continue; | |
276d77db | 9384 | } |
9ec11d1a | 9385 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
9386 | fprintf(stderr, |
9387 | Name ": Cannot restore block: %s\n", | |
9388 | strerror(errno)); | |
137debce AK |
9389 | skipped_disks++; |
9390 | continue; | |
276d77db AK |
9391 | } |
9392 | } | |
9393 | ||
137debce AK |
9394 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
9395 | new_disks, | |
9396 | super, | |
9397 | id->dev)) { | |
6c3560c0 AK |
9398 | fprintf(stderr, |
9399 | Name ": Cannot restore data from backup." | |
9400 | " Too many failed disks\n"); | |
9401 | goto abort; | |
9402 | } | |
9403 | ||
befb629b AK |
9404 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
9405 | /* ignore error == 2, this can mean end of reshape here | |
9406 | */ | |
9407 | dprintf("imsm: Cannot write checkpoint to " | |
9408 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
9409 | } else | |
276d77db | 9410 | retval = 0; |
276d77db AK |
9411 | |
9412 | abort: | |
9413 | if (targets) { | |
9414 | for (i = 0; i < new_disks; i++) | |
9415 | if (targets[i]) | |
9416 | close(targets[i]); | |
9417 | free(targets); | |
9418 | } | |
9419 | free(buf); | |
9420 | return retval; | |
9421 | } | |
9422 | ||
2cda7640 ML |
9423 | static char disk_by_path[] = "/dev/disk/by-path/"; |
9424 | ||
9425 | static const char *imsm_get_disk_controller_domain(const char *path) | |
9426 | { | |
2cda7640 | 9427 | char disk_path[PATH_MAX]; |
96234762 LM |
9428 | char *drv=NULL; |
9429 | struct stat st; | |
2cda7640 | 9430 | |
96234762 LM |
9431 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
9432 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
9433 | if (stat(disk_path, &st) == 0) { | |
9434 | struct sys_dev* hba; | |
9435 | char *path=NULL; | |
9436 | ||
9437 | path = devt_to_devpath(st.st_rdev); | |
9438 | if (path == NULL) | |
9439 | return "unknown"; | |
9440 | hba = find_disk_attached_hba(-1, path); | |
9441 | if (hba && hba->type == SYS_DEV_SAS) | |
9442 | drv = "isci"; | |
9443 | else if (hba && hba->type == SYS_DEV_SATA) | |
9444 | drv = "ahci"; | |
9445 | else | |
9446 | drv = "unknown"; | |
9447 | dprintf("path: %s hba: %s attached: %s\n", | |
9448 | path, (hba) ? hba->path : "NULL", drv); | |
9449 | free(path); | |
9450 | if (hba) | |
9451 | free_sys_dev(&hba); | |
2cda7640 | 9452 | } |
96234762 | 9453 | return drv; |
2cda7640 ML |
9454 | } |
9455 | ||
78b10e66 N |
9456 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
9457 | { | |
9458 | char subdev_name[20]; | |
9459 | struct mdstat_ent *mdstat; | |
9460 | ||
9461 | sprintf(subdev_name, "%d", subdev); | |
9462 | mdstat = mdstat_by_subdev(subdev_name, container); | |
9463 | if (!mdstat) | |
9464 | return -1; | |
9465 | ||
9466 | *minor = mdstat->devnum; | |
9467 | free_mdstat(mdstat); | |
9468 | return 0; | |
9469 | } | |
9470 | ||
9471 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
9472 | struct geo_params *geo, | |
fbf3d202 AK |
9473 | int *old_raid_disks, |
9474 | int direction) | |
78b10e66 | 9475 | { |
694575e7 KW |
9476 | /* currently we only support increasing the number of devices |
9477 | * for a container. This increases the number of device for each | |
9478 | * member array. They must all be RAID0 or RAID5. | |
9479 | */ | |
78b10e66 N |
9480 | int ret_val = 0; |
9481 | struct mdinfo *info, *member; | |
9482 | int devices_that_can_grow = 0; | |
9483 | ||
9484 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
9485 | "st->devnum = (%i)\n", | |
9486 | st->devnum); | |
9487 | ||
9488 | if (geo->size != -1 || | |
9489 | geo->level != UnSet || | |
9490 | geo->layout != UnSet || | |
9491 | geo->chunksize != 0 || | |
9492 | geo->raid_disks == UnSet) { | |
9493 | dprintf("imsm: Container operation is allowed for " | |
9494 | "raid disks number change only.\n"); | |
9495 | return ret_val; | |
9496 | } | |
9497 | ||
fbf3d202 AK |
9498 | if (direction == ROLLBACK_METADATA_CHANGES) { |
9499 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9500 | "container operation.\n"); | |
9501 | return ret_val; | |
9502 | } | |
9503 | ||
78b10e66 N |
9504 | info = container_content_imsm(st, NULL); |
9505 | for (member = info; member; member = member->next) { | |
9506 | int result; | |
9507 | int minor; | |
9508 | ||
9509 | dprintf("imsm: checking device_num: %i\n", | |
9510 | member->container_member); | |
9511 | ||
d7d205bd | 9512 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
9513 | /* we work on container for Online Capacity Expansion |
9514 | * only so raid_disks has to grow | |
9515 | */ | |
9516 | dprintf("imsm: for container operation raid disks " | |
9517 | "increase is required\n"); | |
9518 | break; | |
9519 | } | |
9520 | ||
9521 | if ((info->array.level != 0) && | |
9522 | (info->array.level != 5)) { | |
9523 | /* we cannot use this container with other raid level | |
9524 | */ | |
690aae1a | 9525 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
9526 | " raid level (%i) detected\n", |
9527 | info->array.level); | |
9528 | break; | |
9529 | } else { | |
9530 | /* check for platform support | |
9531 | * for this raid level configuration | |
9532 | */ | |
9533 | struct intel_super *super = st->sb; | |
9534 | if (!is_raid_level_supported(super->orom, | |
9535 | member->array.level, | |
9536 | geo->raid_disks)) { | |
690aae1a | 9537 | dprintf("platform does not support raid%d with" |
78b10e66 N |
9538 | " %d disk%s\n", |
9539 | info->array.level, | |
9540 | geo->raid_disks, | |
9541 | geo->raid_disks > 1 ? "s" : ""); | |
9542 | break; | |
9543 | } | |
2a4a08e7 AK |
9544 | /* check if component size is aligned to chunk size |
9545 | */ | |
9546 | if (info->component_size % | |
9547 | (info->array.chunk_size/512)) { | |
9548 | dprintf("Component size is not aligned to " | |
9549 | "chunk size\n"); | |
9550 | break; | |
9551 | } | |
78b10e66 N |
9552 | } |
9553 | ||
9554 | if (*old_raid_disks && | |
9555 | info->array.raid_disks != *old_raid_disks) | |
9556 | break; | |
9557 | *old_raid_disks = info->array.raid_disks; | |
9558 | ||
9559 | /* All raid5 and raid0 volumes in container | |
9560 | * have to be ready for Online Capacity Expansion | |
9561 | * so they need to be assembled. We have already | |
9562 | * checked that no recovery etc is happening. | |
9563 | */ | |
9564 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
9565 | st->container_dev, | |
9566 | &minor); | |
9567 | if (result < 0) { | |
9568 | dprintf("imsm: cannot find array\n"); | |
9569 | break; | |
9570 | } | |
9571 | devices_that_can_grow++; | |
9572 | } | |
9573 | sysfs_free(info); | |
9574 | if (!member && devices_that_can_grow) | |
9575 | ret_val = 1; | |
9576 | ||
9577 | if (ret_val) | |
9578 | dprintf("\tContainer operation allowed\n"); | |
9579 | else | |
9580 | dprintf("\tError: %i\n", ret_val); | |
9581 | ||
9582 | return ret_val; | |
9583 | } | |
9584 | ||
9585 | /* Function: get_spares_for_grow | |
9586 | * Description: Allocates memory and creates list of spare devices | |
9587 | * avaliable in container. Checks if spare drive size is acceptable. | |
9588 | * Parameters: Pointer to the supertype structure | |
9589 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
9590 | * NULL if fail | |
9591 | */ | |
9592 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
9593 | { | |
78b10e66 | 9594 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 9595 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
9596 | } |
9597 | ||
9598 | /****************************************************************************** | |
9599 | * function: imsm_create_metadata_update_for_reshape | |
9600 | * Function creates update for whole IMSM container. | |
9601 | * | |
9602 | ******************************************************************************/ | |
9603 | static int imsm_create_metadata_update_for_reshape( | |
9604 | struct supertype *st, | |
9605 | struct geo_params *geo, | |
9606 | int old_raid_disks, | |
9607 | struct imsm_update_reshape **updatep) | |
9608 | { | |
9609 | struct intel_super *super = st->sb; | |
9610 | struct imsm_super *mpb = super->anchor; | |
9611 | int update_memory_size = 0; | |
9612 | struct imsm_update_reshape *u = NULL; | |
9613 | struct mdinfo *spares = NULL; | |
9614 | int i; | |
9615 | int delta_disks = 0; | |
bbd24d86 | 9616 | struct mdinfo *dev; |
78b10e66 N |
9617 | |
9618 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
9619 | geo->raid_disks); | |
9620 | ||
9621 | delta_disks = geo->raid_disks - old_raid_disks; | |
9622 | ||
9623 | /* size of all update data without anchor */ | |
9624 | update_memory_size = sizeof(struct imsm_update_reshape); | |
9625 | ||
9626 | /* now add space for spare disks that we need to add. */ | |
9627 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
9628 | ||
9629 | u = calloc(1, update_memory_size); | |
9630 | if (u == NULL) { | |
9631 | dprintf("error: " | |
9632 | "cannot get memory for imsm_update_reshape update\n"); | |
9633 | return 0; | |
9634 | } | |
9635 | u->type = update_reshape_container_disks; | |
9636 | u->old_raid_disks = old_raid_disks; | |
9637 | u->new_raid_disks = geo->raid_disks; | |
9638 | ||
9639 | /* now get spare disks list | |
9640 | */ | |
9641 | spares = get_spares_for_grow(st); | |
9642 | ||
9643 | if (spares == NULL | |
9644 | || delta_disks > spares->array.spare_disks) { | |
e14e5960 KW |
9645 | fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices " |
9646 | "for %s.\n", geo->dev_name); | |
e4c72d1d | 9647 | i = -1; |
78b10e66 N |
9648 | goto abort; |
9649 | } | |
9650 | ||
9651 | /* we have got spares | |
9652 | * update disk list in imsm_disk list table in anchor | |
9653 | */ | |
9654 | dprintf("imsm: %i spares are available.\n\n", | |
9655 | spares->array.spare_disks); | |
9656 | ||
bbd24d86 | 9657 | dev = spares->devs; |
78b10e66 | 9658 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
9659 | struct dl *dl; |
9660 | ||
bbd24d86 AK |
9661 | if (dev == NULL) |
9662 | break; | |
78b10e66 N |
9663 | u->new_disks[i] = makedev(dev->disk.major, |
9664 | dev->disk.minor); | |
9665 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
9666 | dl->index = mpb->num_disks; |
9667 | mpb->num_disks++; | |
bbd24d86 | 9668 | dev = dev->next; |
78b10e66 | 9669 | } |
78b10e66 N |
9670 | |
9671 | abort: | |
9672 | /* free spares | |
9673 | */ | |
9674 | sysfs_free(spares); | |
9675 | ||
d677e0b8 | 9676 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 9677 | if (i == delta_disks) { |
d677e0b8 | 9678 | dprintf(" OK\n"); |
78b10e66 N |
9679 | *updatep = u; |
9680 | return update_memory_size; | |
9681 | } | |
9682 | free(u); | |
d677e0b8 | 9683 | dprintf(" Error\n"); |
78b10e66 N |
9684 | |
9685 | return 0; | |
9686 | } | |
9687 | ||
f3871fdc AK |
9688 | |
9689 | /****************************************************************************** | |
9690 | * function: imsm_create_metadata_update_for_size_change() | |
9691 | * Creates update for IMSM array for array size change. | |
9692 | * | |
9693 | ******************************************************************************/ | |
9694 | static int imsm_create_metadata_update_for_size_change( | |
9695 | struct supertype *st, | |
9696 | struct geo_params *geo, | |
9697 | struct imsm_update_size_change **updatep) | |
9698 | { | |
9699 | struct intel_super *super = st->sb; | |
9700 | int update_memory_size = 0; | |
9701 | struct imsm_update_size_change *u = NULL; | |
9702 | ||
9703 | dprintf("imsm_create_metadata_update_for_size_change(enter)" | |
9704 | " New size = %llu\n", geo->size); | |
9705 | ||
9706 | /* size of all update data without anchor */ | |
9707 | update_memory_size = sizeof(struct imsm_update_size_change); | |
9708 | ||
9709 | u = calloc(1, update_memory_size); | |
9710 | if (u == NULL) { | |
9711 | dprintf("error: cannot get memory for " | |
9712 | "imsm_create_metadata_update_for_size_change\n"); | |
9713 | return 0; | |
9714 | } | |
9715 | u->type = update_size_change; | |
9716 | u->subdev = super->current_vol; | |
9717 | u->new_size = geo->size; | |
9718 | ||
9719 | dprintf("imsm: reshape update preparation : OK\n"); | |
9720 | *updatep = u; | |
9721 | ||
9722 | return update_memory_size; | |
9723 | } | |
9724 | ||
48c5303a PC |
9725 | /****************************************************************************** |
9726 | * function: imsm_create_metadata_update_for_migration() | |
9727 | * Creates update for IMSM array. | |
9728 | * | |
9729 | ******************************************************************************/ | |
9730 | static int imsm_create_metadata_update_for_migration( | |
9731 | struct supertype *st, | |
9732 | struct geo_params *geo, | |
9733 | struct imsm_update_reshape_migration **updatep) | |
9734 | { | |
9735 | struct intel_super *super = st->sb; | |
9736 | int update_memory_size = 0; | |
9737 | struct imsm_update_reshape_migration *u = NULL; | |
9738 | struct imsm_dev *dev; | |
9739 | int previous_level = -1; | |
9740 | ||
9741 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
9742 | " New Level = %i\n", geo->level); | |
9743 | ||
9744 | /* size of all update data without anchor */ | |
9745 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
9746 | ||
9747 | u = calloc(1, update_memory_size); | |
9748 | if (u == NULL) { | |
9749 | dprintf("error: cannot get memory for " | |
9750 | "imsm_create_metadata_update_for_migration\n"); | |
9751 | return 0; | |
9752 | } | |
9753 | u->type = update_reshape_migration; | |
9754 | u->subdev = super->current_vol; | |
9755 | u->new_level = geo->level; | |
9756 | u->new_layout = geo->layout; | |
9757 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
9758 | u->new_disks[0] = -1; | |
4bba0439 | 9759 | u->new_chunksize = -1; |
48c5303a PC |
9760 | |
9761 | dev = get_imsm_dev(super, u->subdev); | |
9762 | if (dev) { | |
9763 | struct imsm_map *map; | |
9764 | ||
238c0a71 | 9765 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
9766 | if (map) { |
9767 | int current_chunk_size = | |
9768 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
9769 | ||
9770 | if (geo->chunksize != current_chunk_size) { | |
9771 | u->new_chunksize = geo->chunksize / 1024; | |
9772 | dprintf("imsm: " | |
9773 | "chunk size change from %i to %i\n", | |
9774 | current_chunk_size, u->new_chunksize); | |
9775 | } | |
48c5303a | 9776 | previous_level = map->raid_level; |
4bba0439 | 9777 | } |
48c5303a PC |
9778 | } |
9779 | if ((geo->level == 5) && (previous_level == 0)) { | |
9780 | struct mdinfo *spares = NULL; | |
9781 | ||
9782 | u->new_raid_disks++; | |
9783 | spares = get_spares_for_grow(st); | |
9784 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
9785 | free(u); | |
9786 | sysfs_free(spares); | |
9787 | update_memory_size = 0; | |
9788 | dprintf("error: cannot get spare device " | |
9789 | "for requested migration"); | |
9790 | return 0; | |
9791 | } | |
9792 | sysfs_free(spares); | |
9793 | } | |
9794 | dprintf("imsm: reshape update preparation : OK\n"); | |
9795 | *updatep = u; | |
9796 | ||
9797 | return update_memory_size; | |
9798 | } | |
9799 | ||
8dd70bce AK |
9800 | static void imsm_update_metadata_locally(struct supertype *st, |
9801 | void *buf, int len) | |
9802 | { | |
9803 | struct metadata_update mu; | |
9804 | ||
9805 | mu.buf = buf; | |
9806 | mu.len = len; | |
9807 | mu.space = NULL; | |
9808 | mu.space_list = NULL; | |
9809 | mu.next = NULL; | |
9810 | imsm_prepare_update(st, &mu); | |
9811 | imsm_process_update(st, &mu); | |
9812 | ||
9813 | while (mu.space_list) { | |
9814 | void **space = mu.space_list; | |
9815 | mu.space_list = *space; | |
9816 | free(space); | |
9817 | } | |
9818 | } | |
78b10e66 | 9819 | |
471bceb6 | 9820 | /*************************************************************************** |
694575e7 | 9821 | * Function: imsm_analyze_change |
471bceb6 KW |
9822 | * Description: Function analyze change for single volume |
9823 | * and validate if transition is supported | |
fbf3d202 AK |
9824 | * Parameters: Geometry parameters, supertype structure, |
9825 | * metadata change direction (apply/rollback) | |
694575e7 | 9826 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
9827 | ****************************************************************************/ |
9828 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
9829 | struct geo_params *geo, |
9830 | int direction) | |
694575e7 | 9831 | { |
471bceb6 KW |
9832 | struct mdinfo info; |
9833 | int change = -1; | |
9834 | int check_devs = 0; | |
c21e737b | 9835 | int chunk; |
67a2db32 AK |
9836 | /* number of added/removed disks in operation result */ |
9837 | int devNumChange = 0; | |
9838 | /* imsm compatible layout value for array geometry verification */ | |
9839 | int imsm_layout = -1; | |
7abc9871 AK |
9840 | int data_disks; |
9841 | struct imsm_dev *dev; | |
9842 | struct intel_super *super; | |
9843 | long long current_size; | |
471bceb6 KW |
9844 | |
9845 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
9846 | if ((geo->level != info.array.level) && |
9847 | (geo->level >= 0) && | |
9848 | (geo->level != UnSet)) { | |
9849 | switch (info.array.level) { | |
9850 | case 0: | |
9851 | if (geo->level == 5) { | |
b5347799 | 9852 | change = CH_MIGRATION; |
e13ce846 AK |
9853 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
9854 | fprintf(stderr, | |
9855 | Name " Error. Requested Layout " | |
9856 | "not supported (left-asymmetric layout " | |
9857 | "is supported only)!\n"); | |
9858 | change = -1; | |
9859 | goto analyse_change_exit; | |
9860 | } | |
67a2db32 | 9861 | imsm_layout = geo->layout; |
471bceb6 | 9862 | check_devs = 1; |
e91a3bad LM |
9863 | devNumChange = 1; /* parity disk added */ |
9864 | } else if (geo->level == 10) { | |
471bceb6 KW |
9865 | change = CH_TAKEOVER; |
9866 | check_devs = 1; | |
e91a3bad | 9867 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 9868 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 9869 | } |
dfe77a9e KW |
9870 | break; |
9871 | case 1: | |
471bceb6 KW |
9872 | case 10: |
9873 | if (geo->level == 0) { | |
9874 | change = CH_TAKEOVER; | |
9875 | check_devs = 1; | |
e91a3bad | 9876 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 9877 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
9878 | } |
9879 | break; | |
9880 | } | |
9881 | if (change == -1) { | |
9882 | fprintf(stderr, | |
9883 | Name " Error. Level Migration from %d to %d " | |
9884 | "not supported!\n", | |
9885 | info.array.level, geo->level); | |
9886 | goto analyse_change_exit; | |
9887 | } | |
9888 | } else | |
9889 | geo->level = info.array.level; | |
9890 | ||
9891 | if ((geo->layout != info.array.layout) | |
9892 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 9893 | change = CH_MIGRATION; |
471bceb6 KW |
9894 | if ((info.array.layout == 0) |
9895 | && (info.array.level == 5) | |
9896 | && (geo->layout == 5)) { | |
9897 | /* reshape 5 -> 4 */ | |
9898 | } else if ((info.array.layout == 5) | |
9899 | && (info.array.level == 5) | |
9900 | && (geo->layout == 0)) { | |
9901 | /* reshape 4 -> 5 */ | |
9902 | geo->layout = 0; | |
9903 | geo->level = 5; | |
9904 | } else { | |
9905 | fprintf(stderr, | |
9906 | Name " Error. Layout Migration from %d to %d " | |
9907 | "not supported!\n", | |
9908 | info.array.layout, geo->layout); | |
9909 | change = -1; | |
9910 | goto analyse_change_exit; | |
9911 | } | |
67a2db32 | 9912 | } else { |
471bceb6 | 9913 | geo->layout = info.array.layout; |
67a2db32 AK |
9914 | if (imsm_layout == -1) |
9915 | imsm_layout = info.array.layout; | |
9916 | } | |
471bceb6 KW |
9917 | |
9918 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
9919 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 9920 | change = CH_MIGRATION; |
471bceb6 KW |
9921 | else |
9922 | geo->chunksize = info.array.chunk_size; | |
9923 | ||
c21e737b | 9924 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
9925 | |
9926 | super = st->sb; | |
9927 | dev = get_imsm_dev(super, super->current_vol); | |
9928 | data_disks = imsm_num_data_members(dev , MAP_0); | |
9929 | /* compute current size in K per disk member | |
9930 | */ | |
9931 | current_size = info.custom_array_size / 2 / data_disks; | |
9932 | ||
9933 | if ((current_size != geo->size) && (geo->size > 0)) { | |
9934 | if (change != -1) { | |
9935 | fprintf(stderr, | |
9936 | Name " Error. Size change should be the only " | |
9937 | "one at a time.\n"); | |
9938 | change = -1; | |
9939 | goto analyse_change_exit; | |
9940 | } | |
9941 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
9942 | fprintf(stderr, | |
9943 | Name " Error. The last volume in container " | |
9944 | "can be expanded only (%i/%i).\n", | |
9945 | super->current_vol, st->devnum); | |
9946 | goto analyse_change_exit; | |
9947 | } | |
9948 | geo->size *= 2; | |
fbf3d202 AK |
9949 | if ((direction == ROLLBACK_METADATA_CHANGES)) { |
9950 | /* accept size for rollback only | |
9951 | */ | |
9952 | } else { | |
9953 | /* round size due to metadata compatibility | |
9954 | */ | |
9955 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
9956 | << SECT_PER_MB_SHIFT; | |
9957 | dprintf("Prepare update for size change to %llu\n", | |
9958 | geo->size ); | |
9959 | if (current_size >= geo->size) { | |
9960 | fprintf(stderr, | |
9961 | Name " Error. Size expanssion is " | |
9962 | "supported only (current size is %llu, " | |
9963 | "requested size /rounded/ is %llu).\n", | |
9964 | current_size, geo->size); | |
9965 | goto analyse_change_exit; | |
9966 | } | |
7abc9871 AK |
9967 | } |
9968 | geo->size *= data_disks; | |
9969 | geo->raid_disks = dev->vol.map->num_members; | |
9970 | change = CH_ARRAY_SIZE; | |
9971 | } | |
471bceb6 KW |
9972 | if (!validate_geometry_imsm(st, |
9973 | geo->level, | |
67a2db32 | 9974 | imsm_layout, |
e91a3bad | 9975 | geo->raid_disks + devNumChange, |
c21e737b | 9976 | &chunk, |
471bceb6 KW |
9977 | geo->size, |
9978 | 0, 0, 1)) | |
9979 | change = -1; | |
9980 | ||
9981 | if (check_devs) { | |
9982 | struct intel_super *super = st->sb; | |
9983 | struct imsm_super *mpb = super->anchor; | |
9984 | ||
9985 | if (mpb->num_raid_devs > 1) { | |
9986 | fprintf(stderr, | |
9987 | Name " Error. Cannot perform operation on %s" | |
9988 | "- for this operation it MUST be single " | |
9989 | "array in container\n", | |
9990 | geo->dev_name); | |
9991 | change = -1; | |
9992 | } | |
9993 | } | |
9994 | ||
9995 | analyse_change_exit: | |
fbf3d202 AK |
9996 | if ((direction == ROLLBACK_METADATA_CHANGES) && |
9997 | ((change == CH_MIGRATION) || (change == CH_TAKEOVER))) { | |
9998 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9999 | "migration and takeover operations.\n"); | |
10000 | change = -1; | |
10001 | } | |
471bceb6 | 10002 | return change; |
694575e7 KW |
10003 | } |
10004 | ||
bb025c2f KW |
10005 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
10006 | { | |
10007 | struct intel_super *super = st->sb; | |
10008 | struct imsm_update_takeover *u; | |
10009 | ||
10010 | u = malloc(sizeof(struct imsm_update_takeover)); | |
10011 | if (u == NULL) | |
10012 | return 1; | |
10013 | ||
10014 | u->type = update_takeover; | |
10015 | u->subarray = super->current_vol; | |
10016 | ||
10017 | /* 10->0 transition */ | |
10018 | if (geo->level == 0) | |
10019 | u->direction = R10_TO_R0; | |
10020 | ||
0529c688 KW |
10021 | /* 0->10 transition */ |
10022 | if (geo->level == 10) | |
10023 | u->direction = R0_TO_R10; | |
10024 | ||
bb025c2f KW |
10025 | /* update metadata locally */ |
10026 | imsm_update_metadata_locally(st, u, | |
10027 | sizeof(struct imsm_update_takeover)); | |
10028 | /* and possibly remotely */ | |
10029 | if (st->update_tail) | |
10030 | append_metadata_update(st, u, | |
10031 | sizeof(struct imsm_update_takeover)); | |
10032 | else | |
10033 | free(u); | |
10034 | ||
10035 | return 0; | |
10036 | } | |
10037 | ||
78b10e66 N |
10038 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
10039 | int layout, int chunksize, int raid_disks, | |
41784c88 | 10040 | int delta_disks, char *backup, char *dev, |
016e00f5 | 10041 | int direction, int verbose) |
78b10e66 | 10042 | { |
78b10e66 N |
10043 | int ret_val = 1; |
10044 | struct geo_params geo; | |
10045 | ||
10046 | dprintf("imsm: reshape_super called.\n"); | |
10047 | ||
71204a50 | 10048 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
10049 | |
10050 | geo.dev_name = dev; | |
694575e7 | 10051 | geo.dev_id = st->devnum; |
78b10e66 N |
10052 | geo.size = size; |
10053 | geo.level = level; | |
10054 | geo.layout = layout; | |
10055 | geo.chunksize = chunksize; | |
10056 | geo.raid_disks = raid_disks; | |
41784c88 AK |
10057 | if (delta_disks != UnSet) |
10058 | geo.raid_disks += delta_disks; | |
78b10e66 N |
10059 | |
10060 | dprintf("\tfor level : %i\n", geo.level); | |
10061 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
10062 | ||
10063 | if (experimental() == 0) | |
10064 | return ret_val; | |
10065 | ||
78b10e66 | 10066 | if (st->container_dev == st->devnum) { |
694575e7 KW |
10067 | /* On container level we can only increase number of devices. */ |
10068 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 10069 | int old_raid_disks = 0; |
6dc0be30 | 10070 | |
78b10e66 | 10071 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 10072 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
10073 | struct imsm_update_reshape *u = NULL; |
10074 | int len; | |
10075 | ||
10076 | len = imsm_create_metadata_update_for_reshape( | |
10077 | st, &geo, old_raid_disks, &u); | |
10078 | ||
ed08d51c AK |
10079 | if (len <= 0) { |
10080 | dprintf("imsm: Cannot prepare update\n"); | |
10081 | goto exit_imsm_reshape_super; | |
10082 | } | |
10083 | ||
8dd70bce AK |
10084 | ret_val = 0; |
10085 | /* update metadata locally */ | |
10086 | imsm_update_metadata_locally(st, u, len); | |
10087 | /* and possibly remotely */ | |
10088 | if (st->update_tail) | |
10089 | append_metadata_update(st, u, len); | |
10090 | else | |
ed08d51c | 10091 | free(u); |
8dd70bce | 10092 | |
694575e7 | 10093 | } else { |
e7ff7e40 AK |
10094 | fprintf(stderr, Name ": (imsm) Operation " |
10095 | "is not allowed on this container\n"); | |
694575e7 KW |
10096 | } |
10097 | } else { | |
10098 | /* On volume level we support following operations | |
471bceb6 KW |
10099 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
10100 | * - chunk size migration | |
10101 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
10102 | */ | |
10103 | struct intel_super *super = st->sb; | |
10104 | struct intel_dev *dev = super->devlist; | |
10105 | int change, devnum; | |
694575e7 | 10106 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
10107 | /* find requested device */ |
10108 | while (dev) { | |
19986c72 MB |
10109 | if (imsm_find_array_minor_by_subdev( |
10110 | dev->index, st->container_dev, &devnum) == 0 | |
10111 | && devnum == geo.dev_id) | |
471bceb6 KW |
10112 | break; |
10113 | dev = dev->next; | |
10114 | } | |
10115 | if (dev == NULL) { | |
10116 | fprintf(stderr, Name " Cannot find %s (%i) subarray\n", | |
10117 | geo.dev_name, geo.dev_id); | |
10118 | goto exit_imsm_reshape_super; | |
10119 | } | |
10120 | super->current_vol = dev->index; | |
fbf3d202 | 10121 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 10122 | switch (change) { |
471bceb6 | 10123 | case CH_TAKEOVER: |
bb025c2f | 10124 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 10125 | break; |
48c5303a PC |
10126 | case CH_MIGRATION: { |
10127 | struct imsm_update_reshape_migration *u = NULL; | |
10128 | int len = | |
10129 | imsm_create_metadata_update_for_migration( | |
10130 | st, &geo, &u); | |
10131 | if (len < 1) { | |
10132 | dprintf("imsm: " | |
10133 | "Cannot prepare update\n"); | |
10134 | break; | |
10135 | } | |
471bceb6 | 10136 | ret_val = 0; |
48c5303a PC |
10137 | /* update metadata locally */ |
10138 | imsm_update_metadata_locally(st, u, len); | |
10139 | /* and possibly remotely */ | |
10140 | if (st->update_tail) | |
10141 | append_metadata_update(st, u, len); | |
10142 | else | |
10143 | free(u); | |
10144 | } | |
10145 | break; | |
7abc9871 | 10146 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
10147 | struct imsm_update_size_change *u = NULL; |
10148 | int len = | |
10149 | imsm_create_metadata_update_for_size_change( | |
10150 | st, &geo, &u); | |
10151 | if (len < 1) { | |
10152 | dprintf("imsm: " | |
10153 | "Cannot prepare update\n"); | |
10154 | break; | |
10155 | } | |
10156 | ret_val = 0; | |
10157 | /* update metadata locally */ | |
10158 | imsm_update_metadata_locally(st, u, len); | |
10159 | /* and possibly remotely */ | |
10160 | if (st->update_tail) | |
10161 | append_metadata_update(st, u, len); | |
10162 | else | |
10163 | free(u); | |
7abc9871 AK |
10164 | } |
10165 | break; | |
471bceb6 KW |
10166 | default: |
10167 | ret_val = 1; | |
694575e7 | 10168 | } |
694575e7 | 10169 | } |
78b10e66 | 10170 | |
ed08d51c | 10171 | exit_imsm_reshape_super: |
78b10e66 N |
10172 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
10173 | return ret_val; | |
10174 | } | |
2cda7640 | 10175 | |
eee67a47 AK |
10176 | /******************************************************************************* |
10177 | * Function: wait_for_reshape_imsm | |
10178 | * Description: Function writes new sync_max value and waits until | |
10179 | * reshape process reach new position | |
10180 | * Parameters: | |
10181 | * sra : general array info | |
eee67a47 AK |
10182 | * ndata : number of disks in new array's layout |
10183 | * Returns: | |
10184 | * 0 : success, | |
10185 | * 1 : there is no reshape in progress, | |
10186 | * -1 : fail | |
10187 | ******************************************************************************/ | |
ae9f01f8 | 10188 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
10189 | { |
10190 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
10191 | unsigned long long completed; | |
ae9f01f8 AK |
10192 | /* to_complete : new sync_max position */ |
10193 | unsigned long long to_complete = sra->reshape_progress; | |
10194 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 10195 | |
ae9f01f8 AK |
10196 | if (fd < 0) { |
10197 | dprintf("imsm: wait_for_reshape_imsm() " | |
10198 | "cannot open reshape_position\n"); | |
eee67a47 | 10199 | return 1; |
ae9f01f8 | 10200 | } |
eee67a47 | 10201 | |
ae9f01f8 AK |
10202 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
10203 | dprintf("imsm: wait_for_reshape_imsm() " | |
10204 | "cannot read reshape_position (no reshape in progres)\n"); | |
10205 | close(fd); | |
10206 | return 0; | |
10207 | } | |
eee67a47 | 10208 | |
ae9f01f8 AK |
10209 | if (completed > to_complete) { |
10210 | dprintf("imsm: wait_for_reshape_imsm() " | |
10211 | "wrong next position to set %llu (%llu)\n", | |
10212 | to_complete, completed); | |
10213 | close(fd); | |
10214 | return -1; | |
10215 | } | |
10216 | dprintf("Position set: %llu\n", position_to_set); | |
10217 | if (sysfs_set_num(sra, NULL, "sync_max", | |
10218 | position_to_set) != 0) { | |
10219 | dprintf("imsm: wait_for_reshape_imsm() " | |
10220 | "cannot set reshape position to %llu\n", | |
10221 | position_to_set); | |
10222 | close(fd); | |
10223 | return -1; | |
eee67a47 AK |
10224 | } |
10225 | ||
eee67a47 AK |
10226 | do { |
10227 | char action[20]; | |
10228 | fd_set rfds; | |
10229 | FD_ZERO(&rfds); | |
10230 | FD_SET(fd, &rfds); | |
a47e44fb AK |
10231 | select(fd+1, &rfds, NULL, NULL, NULL); |
10232 | if (sysfs_get_str(sra, NULL, "sync_action", | |
10233 | action, 20) > 0 && | |
10234 | strncmp(action, "reshape", 7) != 0) | |
10235 | break; | |
eee67a47 | 10236 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
10237 | dprintf("imsm: wait_for_reshape_imsm() " |
10238 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
10239 | close(fd); |
10240 | return 1; | |
10241 | } | |
eee67a47 AK |
10242 | } while (completed < to_complete); |
10243 | close(fd); | |
10244 | return 0; | |
10245 | ||
10246 | } | |
10247 | ||
b915c95f AK |
10248 | /******************************************************************************* |
10249 | * Function: check_degradation_change | |
10250 | * Description: Check that array hasn't become failed. | |
10251 | * Parameters: | |
10252 | * info : for sysfs access | |
10253 | * sources : source disks descriptors | |
10254 | * degraded: previous degradation level | |
10255 | * Returns: | |
10256 | * degradation level | |
10257 | ******************************************************************************/ | |
10258 | int check_degradation_change(struct mdinfo *info, | |
10259 | int *sources, | |
10260 | int degraded) | |
10261 | { | |
10262 | unsigned long long new_degraded; | |
10263 | sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
10264 | if (new_degraded != (unsigned long long)degraded) { | |
10265 | /* check each device to ensure it is still working */ | |
10266 | struct mdinfo *sd; | |
10267 | new_degraded = 0; | |
10268 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10269 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
10270 | continue; | |
10271 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
10272 | char sbuf[20]; | |
10273 | if (sysfs_get_str(info, | |
10274 | sd, "state", sbuf, 20) < 0 || | |
10275 | strstr(sbuf, "faulty") || | |
10276 | strstr(sbuf, "in_sync") == NULL) { | |
10277 | /* this device is dead */ | |
10278 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
10279 | if (sd->disk.raid_disk >= 0 && | |
10280 | sources[sd->disk.raid_disk] >= 0) { | |
10281 | close(sources[ | |
10282 | sd->disk.raid_disk]); | |
10283 | sources[sd->disk.raid_disk] = | |
10284 | -1; | |
10285 | } | |
10286 | new_degraded++; | |
10287 | } | |
10288 | } | |
10289 | } | |
10290 | } | |
10291 | ||
10292 | return new_degraded; | |
10293 | } | |
10294 | ||
10f22854 AK |
10295 | /******************************************************************************* |
10296 | * Function: imsm_manage_reshape | |
10297 | * Description: Function finds array under reshape and it manages reshape | |
10298 | * process. It creates stripes backups (if required) and sets | |
10299 | * checheckpoits. | |
10300 | * Parameters: | |
10301 | * afd : Backup handle (nattive) - not used | |
10302 | * sra : general array info | |
10303 | * reshape : reshape parameters - not used | |
10304 | * st : supertype structure | |
10305 | * blocks : size of critical section [blocks] | |
10306 | * fds : table of source device descriptor | |
10307 | * offsets : start of array (offest per devices) | |
10308 | * dests : not used | |
10309 | * destfd : table of destination device descriptor | |
10310 | * destoffsets : table of destination offsets (per device) | |
10311 | * Returns: | |
10312 | * 1 : success, reshape is done | |
10313 | * 0 : fail | |
10314 | ******************************************************************************/ | |
999b4972 N |
10315 | static int imsm_manage_reshape( |
10316 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 10317 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
10318 | int *fds, unsigned long long *offsets, |
10319 | int dests, int *destfd, unsigned long long *destoffsets) | |
10320 | { | |
10f22854 AK |
10321 | int ret_val = 0; |
10322 | struct intel_super *super = st->sb; | |
10323 | struct intel_dev *dv = NULL; | |
10324 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 10325 | struct imsm_map *map_src; |
10f22854 AK |
10326 | int migr_vol_qan = 0; |
10327 | int ndata, odata; /* [bytes] */ | |
10328 | int chunk; /* [bytes] */ | |
10329 | struct migr_record *migr_rec; | |
10330 | char *buf = NULL; | |
10331 | unsigned int buf_size; /* [bytes] */ | |
10332 | unsigned long long max_position; /* array size [bytes] */ | |
10333 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
10334 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
10335 | unsigned long long start_src; /* [bytes] */ |
10336 | unsigned long long start; /* [bytes] */ | |
10337 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 10338 | int degraded = 0; |
ab724b98 | 10339 | int source_layout = 0; |
10f22854 | 10340 | |
1ab242d8 | 10341 | if (!fds || !offsets || !sra) |
10f22854 AK |
10342 | goto abort; |
10343 | ||
10344 | /* Find volume during the reshape */ | |
10345 | for (dv = super->devlist; dv; dv = dv->next) { | |
10346 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
10347 | && dv->dev->vol.migr_state == 1) { | |
10348 | dev = dv->dev; | |
10349 | migr_vol_qan++; | |
10350 | } | |
10351 | } | |
10352 | /* Only one volume can migrate at the same time */ | |
10353 | if (migr_vol_qan != 1) { | |
10354 | fprintf(stderr, Name " : %s", migr_vol_qan ? | |
10355 | "Number of migrating volumes greater than 1\n" : | |
10356 | "There is no volume during migrationg\n"); | |
10357 | goto abort; | |
10358 | } | |
10359 | ||
238c0a71 | 10360 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
10361 | if (map_src == NULL) |
10362 | goto abort; | |
10f22854 | 10363 | |
238c0a71 AK |
10364 | ndata = imsm_num_data_members(dev, MAP_0); |
10365 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 10366 | |
7b1ab482 | 10367 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
10368 | old_data_stripe_length = odata * chunk; |
10369 | ||
10370 | migr_rec = super->migr_rec; | |
10371 | ||
10f22854 AK |
10372 | /* initialize migration record for start condition */ |
10373 | if (sra->reshape_progress == 0) | |
10374 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
10375 | else { |
10376 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
10377 | dprintf("imsm: cannot restart migration when data " | |
10378 | "are present in copy area.\n"); | |
10379 | goto abort; | |
10380 | } | |
6a75c8ca AK |
10381 | /* Save checkpoint to update migration record for current |
10382 | * reshape position (in md). It can be farther than current | |
10383 | * reshape position in metadata. | |
10384 | */ | |
10385 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
10386 | /* ignore error == 2, this can mean end of reshape here | |
10387 | */ | |
10388 | dprintf("imsm: Cannot write checkpoint to " | |
10389 | "migration record (UNIT_SRC_NORMAL, " | |
10390 | "initial save)\n"); | |
10391 | goto abort; | |
10392 | } | |
b2c59438 | 10393 | } |
10f22854 AK |
10394 | |
10395 | /* size for data */ | |
10396 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
10397 | /* extend buffer size for parity disk */ | |
10398 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
10399 | /* add space for stripe aligment */ | |
10400 | buf_size += old_data_stripe_length; | |
10401 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
10402 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
10403 | goto abort; | |
10404 | } | |
10405 | ||
3ef4403c | 10406 | max_position = sra->component_size * ndata; |
68eb8bc6 | 10407 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
10408 | |
10409 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
10410 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
10411 | /* current reshape position [blocks] */ | |
10412 | unsigned long long current_position = | |
10413 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
10414 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
10415 | unsigned long long border; | |
10416 | ||
b915c95f AK |
10417 | /* Check that array hasn't become failed. |
10418 | */ | |
10419 | degraded = check_degradation_change(sra, fds, degraded); | |
10420 | if (degraded > 1) { | |
10421 | dprintf("imsm: Abort reshape due to degradation" | |
10422 | " level (%i)\n", degraded); | |
10423 | goto abort; | |
10424 | } | |
10425 | ||
10f22854 AK |
10426 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
10427 | ||
10428 | if ((current_position + next_step) > max_position) | |
10429 | next_step = max_position - current_position; | |
10430 | ||
92144abf | 10431 | start = current_position * 512; |
10f22854 AK |
10432 | |
10433 | /* allign reading start to old geometry */ | |
10434 | start_buf_shift = start % old_data_stripe_length; | |
10435 | start_src = start - start_buf_shift; | |
10436 | ||
10437 | border = (start_src / odata) - (start / ndata); | |
10438 | border /= 512; | |
10439 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
10440 | /* save critical stripes to buf | |
10441 | * start - start address of current unit | |
10442 | * to backup [bytes] | |
10443 | * start_src - start address of current unit | |
10444 | * to backup alligned to source array | |
10445 | * [bytes] | |
10446 | */ | |
10447 | unsigned long long next_step_filler = 0; | |
10448 | unsigned long long copy_length = next_step * 512; | |
10449 | ||
10450 | /* allign copy area length to stripe in old geometry */ | |
10451 | next_step_filler = ((copy_length + start_buf_shift) | |
10452 | % old_data_stripe_length); | |
10453 | if (next_step_filler) | |
10454 | next_step_filler = (old_data_stripe_length | |
10455 | - next_step_filler); | |
10456 | dprintf("save_stripes() parameters: start = %llu," | |
10457 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
10458 | "\tstart_in_buf_shift = %llu," | |
10459 | "\tnext_step_filler = %llu\n", | |
10460 | start, start_src, copy_length, | |
10461 | start_buf_shift, next_step_filler); | |
10462 | ||
10463 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
10464 | chunk, map_src->raid_level, |
10465 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
10466 | copy_length + |
10467 | next_step_filler + start_buf_shift, | |
10468 | buf)) { | |
10469 | dprintf("imsm: Cannot save stripes" | |
10470 | " to buffer\n"); | |
10471 | goto abort; | |
10472 | } | |
10473 | /* Convert data to destination format and store it | |
10474 | * in backup general migration area | |
10475 | */ | |
10476 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 10477 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
10478 | dprintf("imsm: Cannot save stripes to " |
10479 | "target devices\n"); | |
10480 | goto abort; | |
10481 | } | |
10482 | if (save_checkpoint_imsm(st, sra, | |
10483 | UNIT_SRC_IN_CP_AREA)) { | |
10484 | dprintf("imsm: Cannot write checkpoint to " | |
10485 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
10486 | goto abort; | |
10487 | } | |
8016a6d4 AK |
10488 | } else { |
10489 | /* set next step to use whole border area */ | |
10490 | border /= next_step; | |
10491 | if (border > 1) | |
10492 | next_step *= border; | |
10f22854 AK |
10493 | } |
10494 | /* When data backed up, checkpoint stored, | |
10495 | * kick the kernel to reshape unit of data | |
10496 | */ | |
10497 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
10498 | /* limit next step to array max position */ |
10499 | if (next_step > max_position) | |
10500 | next_step = max_position; | |
10f22854 AK |
10501 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
10502 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 10503 | sra->reshape_progress = next_step; |
10f22854 AK |
10504 | |
10505 | /* wait until reshape finish */ | |
ae9f01f8 | 10506 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
10507 | dprintf("wait_for_reshape_imsm returned error!\n"); |
10508 | goto abort; | |
10509 | } | |
10f22854 | 10510 | |
0228d92c AK |
10511 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
10512 | /* ignore error == 2, this can mean end of reshape here | |
10513 | */ | |
10f22854 AK |
10514 | dprintf("imsm: Cannot write checkpoint to " |
10515 | "migration record (UNIT_SRC_NORMAL)\n"); | |
10516 | goto abort; | |
10517 | } | |
10518 | ||
10519 | } | |
10520 | ||
10521 | /* return '1' if done */ | |
10522 | ret_val = 1; | |
10523 | abort: | |
10524 | free(buf); | |
10525 | abort_reshape(sra); | |
10526 | ||
10527 | return ret_val; | |
999b4972 | 10528 | } |
71204a50 | 10529 | #endif /* MDASSEMBLE */ |
999b4972 | 10530 | |
cdddbdbc DW |
10531 | struct superswitch super_imsm = { |
10532 | #ifndef MDASSEMBLE | |
10533 | .examine_super = examine_super_imsm, | |
10534 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 10535 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 10536 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
10537 | .detail_super = detail_super_imsm, |
10538 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 10539 | .write_init_super = write_init_super_imsm, |
0e600426 N |
10540 | .validate_geometry = validate_geometry_imsm, |
10541 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 10542 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 10543 | .detail_platform = detail_platform_imsm, |
33414a01 | 10544 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 10545 | .update_subarray = update_subarray_imsm, |
2b959fbf | 10546 | .load_container = load_container_imsm, |
71204a50 N |
10547 | .default_geometry = default_geometry_imsm, |
10548 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
10549 | .reshape_super = imsm_reshape_super, | |
10550 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 10551 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
10552 | #endif |
10553 | .match_home = match_home_imsm, | |
10554 | .uuid_from_super= uuid_from_super_imsm, | |
10555 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 10556 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
10557 | .update_super = update_super_imsm, |
10558 | ||
10559 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 10560 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
10561 | |
10562 | .compare_super = compare_super_imsm, | |
10563 | ||
10564 | .load_super = load_super_imsm, | |
bf5a934a | 10565 | .init_super = init_super_imsm, |
e683ca88 | 10566 | .store_super = store_super_imsm, |
cdddbdbc DW |
10567 | .free_super = free_super_imsm, |
10568 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 10569 | .container_content = container_content_imsm, |
cdddbdbc | 10570 | |
276d77db | 10571 | |
cdddbdbc | 10572 | .external = 1, |
4cce4069 | 10573 | .name = "imsm", |
845dea95 | 10574 | |
0e600426 | 10575 | #ifndef MDASSEMBLE |
845dea95 NB |
10576 | /* for mdmon */ |
10577 | .open_new = imsm_open_new, | |
ed9d66aa | 10578 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
10579 | .set_disk = imsm_set_disk, |
10580 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 10581 | .activate_spare = imsm_activate_spare, |
e8319a19 | 10582 | .process_update = imsm_process_update, |
8273f55e | 10583 | .prepare_update = imsm_prepare_update, |
0e600426 | 10584 | #endif /* MDASSEMBLE */ |
cdddbdbc | 10585 | }; |