]>
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 | ||
e7b84f9d | 298 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 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; | |
d04f65f4 | 436 | unsigned long long size; |
78b10e66 N |
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 | { | |
503975b9 N |
528 | struct intel_hba *result = xmalloc(sizeof(*result)); |
529 | ||
530 | result->type = device->type; | |
531 | result->path = xstrdup(device->path); | |
532 | result->next = NULL; | |
533 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
534 | result->pci_id++; | |
535 | ||
88654014 LM |
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 | ||
503975b9 | 628 | st = xcalloc(1, sizeof(*st)); |
d1d599ea | 629 | st->container_dev = NoMdDev; |
cdddbdbc DW |
630 | st->ss = &super_imsm; |
631 | st->max_devs = IMSM_MAX_DEVICES; | |
632 | st->minor_version = 0; | |
633 | st->sb = NULL; | |
634 | return st; | |
635 | } | |
636 | ||
0e600426 | 637 | #ifndef MDASSEMBLE |
cdddbdbc DW |
638 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
639 | { | |
640 | return &mpb->sig[MPB_SIG_LEN]; | |
641 | } | |
9e2d750d | 642 | #endif |
cdddbdbc | 643 | |
949c47a0 DW |
644 | /* retrieve a disk directly from the anchor when the anchor is known to be |
645 | * up-to-date, currently only at load time | |
646 | */ | |
647 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 648 | { |
949c47a0 | 649 | if (index >= mpb->num_disks) |
cdddbdbc DW |
650 | return NULL; |
651 | return &mpb->disk[index]; | |
652 | } | |
653 | ||
95d07a2c LM |
654 | /* retrieve the disk description based on a index of the disk |
655 | * in the sub-array | |
656 | */ | |
657 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 658 | { |
b9f594fe DW |
659 | struct dl *d; |
660 | ||
661 | for (d = super->disks; d; d = d->next) | |
662 | if (d->index == index) | |
95d07a2c LM |
663 | return d; |
664 | ||
665 | return NULL; | |
666 | } | |
667 | /* retrieve a disk from the parsed metadata */ | |
668 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
669 | { | |
670 | struct dl *dl; | |
671 | ||
672 | dl = get_imsm_dl_disk(super, index); | |
673 | if (dl) | |
674 | return &dl->disk; | |
675 | ||
b9f594fe | 676 | return NULL; |
949c47a0 DW |
677 | } |
678 | ||
679 | /* generate a checksum directly from the anchor when the anchor is known to be | |
680 | * up-to-date, currently only at load or write_super after coalescing | |
681 | */ | |
682 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
683 | { |
684 | __u32 end = mpb->mpb_size / sizeof(end); | |
685 | __u32 *p = (__u32 *) mpb; | |
686 | __u32 sum = 0; | |
687 | ||
97f734fd N |
688 | while (end--) { |
689 | sum += __le32_to_cpu(*p); | |
690 | p++; | |
691 | } | |
cdddbdbc DW |
692 | |
693 | return sum - __le32_to_cpu(mpb->check_sum); | |
694 | } | |
695 | ||
a965f303 DW |
696 | static size_t sizeof_imsm_map(struct imsm_map *map) |
697 | { | |
698 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
699 | } | |
700 | ||
701 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 702 | { |
5e7b0330 AK |
703 | /* A device can have 2 maps if it is in the middle of a migration. |
704 | * If second_map is: | |
238c0a71 AK |
705 | * MAP_0 - we return the first map |
706 | * MAP_1 - we return the second map if it exists, else NULL | |
707 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 708 | */ |
a965f303 | 709 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 710 | struct imsm_map *map2 = NULL; |
a965f303 | 711 | |
9535fc47 AK |
712 | if (dev->vol.migr_state) |
713 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 714 | |
9535fc47 | 715 | switch (second_map) { |
3b451610 | 716 | case MAP_0: |
9535fc47 | 717 | break; |
3b451610 | 718 | case MAP_1: |
9535fc47 AK |
719 | map = map2; |
720 | break; | |
238c0a71 | 721 | case MAP_X: |
9535fc47 AK |
722 | if (map2) |
723 | map = map2; | |
724 | break; | |
9535fc47 AK |
725 | default: |
726 | map = NULL; | |
727 | } | |
728 | return map; | |
5e7b0330 | 729 | |
a965f303 | 730 | } |
cdddbdbc | 731 | |
3393c6af DW |
732 | /* return the size of the device. |
733 | * migr_state increases the returned size if map[0] were to be duplicated | |
734 | */ | |
735 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
736 | { |
737 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 738 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
739 | |
740 | /* migrating means an additional map */ | |
a965f303 | 741 | if (dev->vol.migr_state) |
238c0a71 | 742 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 743 | else if (migr_state) |
238c0a71 | 744 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
745 | |
746 | return size; | |
747 | } | |
748 | ||
54c2c1ea DW |
749 | #ifndef MDASSEMBLE |
750 | /* retrieve disk serial number list from a metadata update */ | |
751 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
752 | { | |
753 | void *u = update; | |
754 | struct disk_info *inf; | |
755 | ||
756 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
757 | sizeof_imsm_dev(&update->dev, 0); | |
758 | ||
759 | return inf; | |
760 | } | |
761 | #endif | |
762 | ||
949c47a0 | 763 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
764 | { |
765 | int offset; | |
766 | int i; | |
767 | void *_mpb = mpb; | |
768 | ||
949c47a0 | 769 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
770 | return NULL; |
771 | ||
772 | /* devices start after all disks */ | |
773 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
774 | ||
775 | for (i = 0; i <= index; i++) | |
776 | if (i == index) | |
777 | return _mpb + offset; | |
778 | else | |
3393c6af | 779 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
780 | |
781 | return NULL; | |
782 | } | |
783 | ||
949c47a0 DW |
784 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
785 | { | |
ba2de7ba DW |
786 | struct intel_dev *dv; |
787 | ||
949c47a0 DW |
788 | if (index >= super->anchor->num_raid_devs) |
789 | return NULL; | |
ba2de7ba DW |
790 | for (dv = super->devlist; dv; dv = dv->next) |
791 | if (dv->index == index) | |
792 | return dv->dev; | |
793 | return NULL; | |
949c47a0 DW |
794 | } |
795 | ||
98130f40 AK |
796 | /* |
797 | * for second_map: | |
238c0a71 AK |
798 | * == MAP_0 get first map |
799 | * == MAP_1 get second map | |
800 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
801 | */ |
802 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
803 | int slot, | |
804 | int second_map) | |
7eef0453 DW |
805 | { |
806 | struct imsm_map *map; | |
807 | ||
5e7b0330 | 808 | map = get_imsm_map(dev, second_map); |
7eef0453 | 809 | |
ff077194 DW |
810 | /* top byte identifies disk under rebuild */ |
811 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
812 | } | |
813 | ||
814 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 815 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 816 | { |
98130f40 | 817 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
818 | |
819 | return ord_to_idx(ord); | |
7eef0453 DW |
820 | } |
821 | ||
be73972f DW |
822 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
823 | { | |
824 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
825 | } | |
826 | ||
f21e18ca | 827 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
828 | { |
829 | int slot; | |
830 | __u32 ord; | |
831 | ||
832 | for (slot = 0; slot < map->num_members; slot++) { | |
833 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
834 | if (ord_to_idx(ord) == idx) | |
835 | return slot; | |
836 | } | |
837 | ||
838 | return -1; | |
839 | } | |
840 | ||
cdddbdbc DW |
841 | static int get_imsm_raid_level(struct imsm_map *map) |
842 | { | |
843 | if (map->raid_level == 1) { | |
844 | if (map->num_members == 2) | |
845 | return 1; | |
846 | else | |
847 | return 10; | |
848 | } | |
849 | ||
850 | return map->raid_level; | |
851 | } | |
852 | ||
c2c087e6 DW |
853 | static int cmp_extent(const void *av, const void *bv) |
854 | { | |
855 | const struct extent *a = av; | |
856 | const struct extent *b = bv; | |
857 | if (a->start < b->start) | |
858 | return -1; | |
859 | if (a->start > b->start) | |
860 | return 1; | |
861 | return 0; | |
862 | } | |
863 | ||
0dcecb2e | 864 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 865 | { |
c2c087e6 | 866 | int memberships = 0; |
620b1713 | 867 | int i; |
c2c087e6 | 868 | |
949c47a0 DW |
869 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
870 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 871 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 872 | |
620b1713 DW |
873 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
874 | memberships++; | |
c2c087e6 | 875 | } |
0dcecb2e DW |
876 | |
877 | return memberships; | |
878 | } | |
879 | ||
b81221b7 CA |
880 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
881 | ||
5551b113 CA |
882 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
883 | { | |
884 | if (lo == 0 || hi == 0) | |
885 | return 1; | |
886 | *lo = __le32_to_cpu((unsigned)n); | |
887 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
888 | return 0; | |
889 | } | |
890 | ||
891 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
892 | { | |
893 | return (unsigned long long)__le32_to_cpu(lo) | | |
894 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
895 | } | |
896 | ||
897 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
898 | { | |
899 | if (disk == NULL) | |
900 | return 0; | |
901 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
902 | } | |
903 | ||
904 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
905 | { | |
906 | if (map == NULL) | |
907 | return 0; | |
908 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
909 | } | |
910 | ||
911 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
912 | { | |
913 | if (map == NULL) | |
914 | return 0; | |
915 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
916 | } | |
917 | ||
ae2416e7 | 918 | #ifndef MDASSEMBLE |
5551b113 CA |
919 | static unsigned long long num_data_stripes(struct imsm_map *map) |
920 | { | |
921 | if (map == NULL) | |
922 | return 0; | |
923 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
924 | } | |
925 | ||
926 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
927 | { | |
928 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
929 | } | |
ae2416e7 | 930 | #endif |
5551b113 CA |
931 | |
932 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
933 | { | |
934 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
935 | } | |
936 | ||
937 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
938 | { | |
939 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
940 | } | |
941 | ||
942 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
943 | { | |
944 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
945 | } | |
946 | ||
0dcecb2e DW |
947 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
948 | { | |
949 | /* find a list of used extents on the given physical device */ | |
950 | struct extent *rv, *e; | |
620b1713 | 951 | int i; |
0dcecb2e | 952 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
953 | __u32 reservation; |
954 | ||
955 | /* trim the reserved area for spares, so they can join any array | |
956 | * regardless of whether the OROM has assigned sectors from the | |
957 | * IMSM_RESERVED_SECTORS region | |
958 | */ | |
959 | if (dl->index == -1) | |
b81221b7 | 960 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
961 | else |
962 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 963 | |
503975b9 | 964 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
965 | e = rv; |
966 | ||
949c47a0 DW |
967 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
968 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 969 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 970 | |
620b1713 | 971 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
972 | e->start = pba_of_lba0(map); |
973 | e->size = blocks_per_member(map); | |
620b1713 | 974 | e++; |
c2c087e6 DW |
975 | } |
976 | } | |
977 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
978 | ||
14e8215b DW |
979 | /* determine the start of the metadata |
980 | * when no raid devices are defined use the default | |
981 | * ...otherwise allow the metadata to truncate the value | |
982 | * as is the case with older versions of imsm | |
983 | */ | |
984 | if (memberships) { | |
985 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 986 | unsigned long long remainder; |
14e8215b | 987 | |
5551b113 | 988 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
989 | /* round down to 1k block to satisfy precision of the kernel |
990 | * 'size' interface | |
991 | */ | |
992 | remainder &= ~1UL; | |
993 | /* make sure remainder is still sane */ | |
f21e18ca | 994 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 995 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
996 | if (reservation > remainder) |
997 | reservation = remainder; | |
998 | } | |
5551b113 | 999 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1000 | e->size = 0; |
1001 | return rv; | |
1002 | } | |
1003 | ||
14e8215b DW |
1004 | /* try to determine how much space is reserved for metadata from |
1005 | * the last get_extents() entry, otherwise fallback to the | |
1006 | * default | |
1007 | */ | |
1008 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1009 | { | |
1010 | struct extent *e; | |
1011 | int i; | |
1012 | __u32 rv; | |
1013 | ||
1014 | /* for spares just return a minimal reservation which will grow | |
1015 | * once the spare is picked up by an array | |
1016 | */ | |
1017 | if (dl->index == -1) | |
1018 | return MPB_SECTOR_CNT; | |
1019 | ||
1020 | e = get_extents(super, dl); | |
1021 | if (!e) | |
1022 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1023 | ||
1024 | /* scroll to last entry */ | |
1025 | for (i = 0; e[i].size; i++) | |
1026 | continue; | |
1027 | ||
5551b113 | 1028 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1029 | |
1030 | free(e); | |
1031 | ||
1032 | return rv; | |
1033 | } | |
1034 | ||
25ed7e59 DW |
1035 | static int is_spare(struct imsm_disk *disk) |
1036 | { | |
1037 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1038 | } | |
1039 | ||
1040 | static int is_configured(struct imsm_disk *disk) | |
1041 | { | |
1042 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1043 | } | |
1044 | ||
1045 | static int is_failed(struct imsm_disk *disk) | |
1046 | { | |
1047 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1048 | } | |
1049 | ||
b81221b7 CA |
1050 | /* try to determine how much space is reserved for metadata from |
1051 | * the last get_extents() entry on the smallest active disk, | |
1052 | * otherwise fallback to the default | |
1053 | */ | |
1054 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1055 | { | |
1056 | struct extent *e; | |
1057 | int i; | |
5551b113 CA |
1058 | unsigned long long min_active; |
1059 | __u32 remainder; | |
b81221b7 CA |
1060 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1061 | struct dl *dl, *dl_min = NULL; | |
1062 | ||
1063 | if (!super) | |
1064 | return rv; | |
1065 | ||
1066 | min_active = 0; | |
1067 | for (dl = super->disks; dl; dl = dl->next) { | |
1068 | if (dl->index < 0) | |
1069 | continue; | |
5551b113 CA |
1070 | unsigned long long blocks = total_blocks(&dl->disk); |
1071 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1072 | dl_min = dl; |
5551b113 | 1073 | min_active = blocks; |
b81221b7 CA |
1074 | } |
1075 | } | |
1076 | if (!dl_min) | |
1077 | return rv; | |
1078 | ||
1079 | /* find last lba used by subarrays on the smallest active disk */ | |
1080 | e = get_extents(super, dl_min); | |
1081 | if (!e) | |
1082 | return rv; | |
1083 | for (i = 0; e[i].size; i++) | |
1084 | continue; | |
1085 | ||
1086 | remainder = min_active - e[i].start; | |
1087 | free(e); | |
1088 | ||
1089 | /* to give priority to recovery we should not require full | |
1090 | IMSM_RESERVED_SECTORS from the spare */ | |
1091 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1092 | ||
1093 | /* if real reservation is smaller use that value */ | |
1094 | return (remainder < rv) ? remainder : rv; | |
1095 | } | |
1096 | ||
80e7f8c3 AC |
1097 | /* Return minimum size of a spare that can be used in this array*/ |
1098 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1099 | { | |
1100 | struct intel_super *super = st->sb; | |
1101 | struct dl *dl; | |
1102 | struct extent *e; | |
1103 | int i; | |
1104 | unsigned long long rv = 0; | |
1105 | ||
1106 | if (!super) | |
1107 | return rv; | |
1108 | /* find first active disk in array */ | |
1109 | dl = super->disks; | |
1110 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1111 | dl = dl->next; | |
1112 | if (!dl) | |
1113 | return rv; | |
1114 | /* find last lba used by subarrays */ | |
1115 | e = get_extents(super, dl); | |
1116 | if (!e) | |
1117 | return rv; | |
1118 | for (i = 0; e[i].size; i++) | |
1119 | continue; | |
1120 | if (i > 0) | |
1121 | rv = e[i-1].start + e[i-1].size; | |
1122 | free(e); | |
b81221b7 | 1123 | |
80e7f8c3 | 1124 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1125 | rv = rv + imsm_min_reserved_sectors(super); |
1126 | ||
80e7f8c3 AC |
1127 | return rv * 512; |
1128 | } | |
1129 | ||
d1e02575 AK |
1130 | static int is_gen_migration(struct imsm_dev *dev); |
1131 | ||
1799c9e8 | 1132 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1133 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1134 | struct imsm_dev *dev); | |
1e5c6983 | 1135 | |
c47b0ff6 AK |
1136 | static void print_imsm_dev(struct intel_super *super, |
1137 | struct imsm_dev *dev, | |
1138 | char *uuid, | |
1139 | int disk_idx) | |
cdddbdbc DW |
1140 | { |
1141 | __u64 sz; | |
0d80bb2f | 1142 | int slot, i; |
238c0a71 AK |
1143 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1144 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1145 | __u32 ord; |
cdddbdbc DW |
1146 | |
1147 | printf("\n"); | |
1e7bc0ed | 1148 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1149 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1150 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1151 | if (map2) | |
1152 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1153 | printf("\n"); | |
1154 | printf(" Members : %d", map->num_members); | |
1155 | if (map2) | |
1156 | printf(" <-- %d", map2->num_members); | |
1157 | printf("\n"); | |
0d80bb2f DW |
1158 | printf(" Slots : ["); |
1159 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1160 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1161 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1162 | } | |
dd8bcb3b AK |
1163 | printf("]"); |
1164 | if (map2) { | |
1165 | printf(" <-- ["); | |
1166 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1167 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1168 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1169 | } | |
1170 | printf("]"); | |
1171 | } | |
1172 | printf("\n"); | |
7095bccb AK |
1173 | printf(" Failed disk : "); |
1174 | if (map->failed_disk_num == 0xff) | |
1175 | printf("none"); | |
1176 | else | |
1177 | printf("%i", map->failed_disk_num); | |
1178 | printf("\n"); | |
620b1713 DW |
1179 | slot = get_imsm_disk_slot(map, disk_idx); |
1180 | if (slot >= 0) { | |
238c0a71 | 1181 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1182 | printf(" This Slot : %d%s\n", slot, |
1183 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1184 | } else | |
cdddbdbc DW |
1185 | printf(" This Slot : ?\n"); |
1186 | sz = __le32_to_cpu(dev->size_high); | |
1187 | sz <<= 32; | |
1188 | sz += __le32_to_cpu(dev->size_low); | |
1189 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1190 | human_size(sz * 512)); | |
5551b113 | 1191 | sz = blocks_per_member(map); |
cdddbdbc DW |
1192 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, |
1193 | human_size(sz * 512)); | |
5551b113 CA |
1194 | printf(" Sector Offset : %llu\n", |
1195 | pba_of_lba0(map)); | |
1196 | printf(" Num Stripes : %llu\n", | |
1197 | num_data_stripes(map)); | |
dd8bcb3b | 1198 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1199 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1200 | if (map2) |
1201 | printf(" <-- %u KiB", | |
1202 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1203 | printf("\n"); | |
cdddbdbc | 1204 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1205 | printf(" Migrate State : "); |
1484e727 DW |
1206 | if (dev->vol.migr_state) { |
1207 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1208 | printf("initialize\n"); |
1484e727 | 1209 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1210 | printf("rebuild\n"); |
1484e727 | 1211 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1212 | printf("check\n"); |
1484e727 | 1213 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1214 | printf("general migration\n"); |
1484e727 | 1215 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1216 | printf("state change\n"); |
1484e727 | 1217 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1218 | printf("repair\n"); |
1484e727 | 1219 | else |
8655a7b1 DW |
1220 | printf("<unknown:%d>\n", migr_type(dev)); |
1221 | } else | |
1222 | printf("idle\n"); | |
3393c6af DW |
1223 | printf(" Map State : %s", map_state_str[map->map_state]); |
1224 | if (dev->vol.migr_state) { | |
238c0a71 | 1225 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1226 | |
b10b37b8 | 1227 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1228 | printf("\n Checkpoint : %u ", |
1229 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
3136abe5 | 1230 | if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0))) |
464d40e8 LD |
1231 | printf("(N/A)"); |
1232 | else | |
1233 | printf("(%llu)", (unsigned long long) | |
1234 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1235 | } |
1236 | printf("\n"); | |
cdddbdbc | 1237 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1238 | } |
1239 | ||
0ec1f4e8 | 1240 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1241 | { |
1f24f035 | 1242 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1243 | __u64 sz; |
1244 | ||
0ec1f4e8 | 1245 | if (index < -1 || !disk) |
e9d82038 DW |
1246 | return; |
1247 | ||
cdddbdbc | 1248 | printf("\n"); |
1f24f035 | 1249 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1250 | if (index >= 0) |
1251 | printf(" Disk%02d Serial : %s\n", index, str); | |
1252 | else | |
1253 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1254 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1255 | is_configured(disk) ? " active" : "", | |
1256 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1257 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1258 | sz = total_blocks(disk) - reserved; |
cdddbdbc DW |
1259 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1260 | human_size(sz * 512)); | |
1261 | } | |
1262 | ||
520e69e2 AK |
1263 | void examine_migr_rec_imsm(struct intel_super *super) |
1264 | { | |
1265 | struct migr_record *migr_rec = super->migr_rec; | |
1266 | struct imsm_super *mpb = super->anchor; | |
1267 | int i; | |
1268 | ||
1269 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1270 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1271 | struct imsm_map *map; |
b4ab44d8 | 1272 | int slot = -1; |
3136abe5 | 1273 | |
520e69e2 AK |
1274 | if (is_gen_migration(dev) == 0) |
1275 | continue; | |
1276 | ||
1277 | printf("\nMigration Record Information:"); | |
3136abe5 | 1278 | |
44bfe6df AK |
1279 | /* first map under migration */ |
1280 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1281 | if (map) |
1282 | slot = get_imsm_disk_slot(map, super->disks->index); | |
1283 | if ((map == NULL) || (slot > 1) || (slot < 0)) { | |
520e69e2 AK |
1284 | printf(" Empty\n "); |
1285 | printf("Examine one of first two disks in array\n"); | |
1286 | break; | |
1287 | } | |
1288 | printf("\n Status : "); | |
1289 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1290 | printf("Normal\n"); | |
1291 | else | |
1292 | printf("Contains Data\n"); | |
1293 | printf(" Current Unit : %u\n", | |
1294 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1295 | printf(" Family : %u\n", | |
1296 | __le32_to_cpu(migr_rec->family_num)); | |
1297 | printf(" Ascending : %u\n", | |
1298 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1299 | printf(" Blocks Per Unit : %u\n", | |
1300 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1301 | printf(" Dest. Depth Per Unit : %u\n", | |
1302 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1303 | printf(" Checkpoint Area pba : %u\n", | |
1304 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1305 | printf(" First member lba : %u\n", | |
1306 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1307 | printf(" Total Number of Units : %u\n", | |
1308 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1309 | printf(" Size of volume : %u\n", | |
1310 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1311 | printf(" Expansion space for LBA64 : %u\n", | |
1312 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1313 | printf(" Record was read from : %u\n", | |
1314 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1315 | ||
1316 | break; | |
1317 | } | |
1318 | } | |
9e2d750d | 1319 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1320 | /******************************************************************************* |
1321 | * function: imsm_check_attributes | |
1322 | * Description: Function checks if features represented by attributes flags | |
1323 | * are supported by mdadm. | |
1324 | * Parameters: | |
1325 | * attributes - Attributes read from metadata | |
1326 | * Returns: | |
1327 | * 0 - passed attributes contains unsupported features flags | |
1328 | * 1 - all features are supported | |
1329 | ******************************************************************************/ | |
1330 | static int imsm_check_attributes(__u32 attributes) | |
1331 | { | |
1332 | int ret_val = 1; | |
418f9b36 N |
1333 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1334 | ||
1335 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1336 | |
1337 | not_supported &= attributes; | |
1338 | if (not_supported) { | |
e7b84f9d | 1339 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1340 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1341 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1342 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1343 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1344 | } | |
1345 | if (not_supported & MPB_ATTRIB_2TB) { | |
1346 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1347 | not_supported ^= MPB_ATTRIB_2TB; | |
1348 | } | |
1349 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1350 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1351 | not_supported ^= MPB_ATTRIB_RAID0; | |
1352 | } | |
1353 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1354 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1355 | not_supported ^= MPB_ATTRIB_RAID1; | |
1356 | } | |
1357 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1358 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1359 | not_supported ^= MPB_ATTRIB_RAID10; | |
1360 | } | |
1361 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1362 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1363 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1364 | } | |
1365 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1366 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1367 | not_supported ^= MPB_ATTRIB_RAID5; | |
1368 | } | |
1369 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1370 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1371 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1372 | } | |
1373 | if (not_supported & MPB_ATTRIB_BBM) { | |
1374 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1375 | not_supported ^= MPB_ATTRIB_BBM; | |
1376 | } | |
1377 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1378 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1379 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1380 | } | |
1381 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1382 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1383 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1384 | } | |
1385 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1386 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1387 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1388 | } | |
1389 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1390 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1391 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1392 | } | |
1393 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1394 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1395 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1396 | } | |
1397 | ||
1398 | if (not_supported) | |
1399 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1400 | ||
1401 | ret_val = 0; | |
1402 | } | |
1403 | ||
1404 | return ret_val; | |
1405 | } | |
1406 | ||
9e2d750d | 1407 | #ifndef MDASSEMBLE |
a5d85af7 | 1408 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1409 | |
cdddbdbc DW |
1410 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1411 | { | |
1412 | struct intel_super *super = st->sb; | |
949c47a0 | 1413 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1414 | char str[MAX_SIGNATURE_LENGTH]; |
1415 | int i; | |
27fd6274 DW |
1416 | struct mdinfo info; |
1417 | char nbuf[64]; | |
cdddbdbc | 1418 | __u32 sum; |
14e8215b | 1419 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1420 | struct dl *dl; |
27fd6274 | 1421 | |
cdddbdbc DW |
1422 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1423 | printf(" Magic : %s\n", str); | |
1424 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1425 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1426 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1427 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1428 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1429 | printf(" Attributes : "); |
1430 | if (imsm_check_attributes(mpb->attributes)) | |
1431 | printf("All supported\n"); | |
1432 | else | |
1433 | printf("not supported\n"); | |
a5d85af7 | 1434 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1435 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1436 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1437 | sum = __le32_to_cpu(mpb->check_sum); |
1438 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1439 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1440 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1441 | printf(" Disks : %d\n", mpb->num_disks); |
1442 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1443 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1444 | if (super->bbm_log) { |
1445 | struct bbm_log *log = super->bbm_log; | |
1446 | ||
1447 | printf("\n"); | |
1448 | printf("Bad Block Management Log:\n"); | |
1449 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1450 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1451 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1452 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1453 | printf(" First Spare : %llx\n", |
1454 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1455 | } |
44470971 DW |
1456 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1457 | struct mdinfo info; | |
1458 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1459 | ||
1460 | super->current_vol = i; | |
a5d85af7 | 1461 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1462 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1463 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1464 | } |
cdddbdbc DW |
1465 | for (i = 0; i < mpb->num_disks; i++) { |
1466 | if (i == super->disks->index) | |
1467 | continue; | |
0ec1f4e8 | 1468 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1469 | } |
94827db3 | 1470 | |
0ec1f4e8 DW |
1471 | for (dl = super->disks; dl; dl = dl->next) |
1472 | if (dl->index == -1) | |
1473 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1474 | |
1475 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1476 | } |
1477 | ||
061f2c6a | 1478 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1479 | { |
27fd6274 | 1480 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1481 | struct mdinfo info; |
1482 | char nbuf[64]; | |
1e7bc0ed | 1483 | struct intel_super *super = st->sb; |
1e7bc0ed | 1484 | |
0d5a423f DW |
1485 | if (!super->anchor->num_raid_devs) { |
1486 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1487 | return; |
0d5a423f | 1488 | } |
ff54de6e | 1489 | |
a5d85af7 | 1490 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1491 | fname_from_uuid(st, &info, nbuf, ':'); |
1492 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1493 | } | |
1494 | ||
1495 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1496 | { | |
1497 | /* We just write a generic IMSM ARRAY entry */ | |
1498 | struct mdinfo info; | |
1499 | char nbuf[64]; | |
1500 | char nbuf1[64]; | |
1501 | struct intel_super *super = st->sb; | |
1502 | int i; | |
1503 | ||
1504 | if (!super->anchor->num_raid_devs) | |
1505 | return; | |
1506 | ||
a5d85af7 | 1507 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1508 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1509 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1510 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1511 | ||
1512 | super->current_vol = i; | |
a5d85af7 | 1513 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1514 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1515 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1516 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1517 | } |
cdddbdbc DW |
1518 | } |
1519 | ||
9d84c8ea DW |
1520 | static void export_examine_super_imsm(struct supertype *st) |
1521 | { | |
1522 | struct intel_super *super = st->sb; | |
1523 | struct imsm_super *mpb = super->anchor; | |
1524 | struct mdinfo info; | |
1525 | char nbuf[64]; | |
1526 | ||
a5d85af7 | 1527 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1528 | fname_from_uuid(st, &info, nbuf, ':'); |
1529 | printf("MD_METADATA=imsm\n"); | |
1530 | printf("MD_LEVEL=container\n"); | |
1531 | printf("MD_UUID=%s\n", nbuf+5); | |
1532 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1533 | } | |
1534 | ||
cdddbdbc DW |
1535 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1536 | { | |
3ebe00a1 DW |
1537 | struct mdinfo info; |
1538 | char nbuf[64]; | |
1539 | ||
a5d85af7 | 1540 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1541 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1542 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1543 | } |
1544 | ||
1545 | static void brief_detail_super_imsm(struct supertype *st) | |
1546 | { | |
ff54de6e N |
1547 | struct mdinfo info; |
1548 | char nbuf[64]; | |
a5d85af7 | 1549 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1550 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1551 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1552 | } |
d665cc31 DW |
1553 | |
1554 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1555 | static void fd2devname(int fd, char *name); | |
1556 | ||
120dc887 | 1557 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1558 | { |
120dc887 LM |
1559 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1560 | * controller, as well as non-connected ports | |
d665cc31 DW |
1561 | */ |
1562 | int hba_len = strlen(hba_path) + 1; | |
1563 | struct dirent *ent; | |
1564 | DIR *dir; | |
1565 | char *path = NULL; | |
1566 | int err = 0; | |
1567 | unsigned long port_mask = (1 << port_count) - 1; | |
1568 | ||
f21e18ca | 1569 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 1570 | if (verbose > 0) |
e7b84f9d | 1571 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
1572 | return 2; |
1573 | } | |
1574 | ||
1575 | /* scroll through /sys/dev/block looking for devices attached to | |
1576 | * this hba | |
1577 | */ | |
1578 | dir = opendir("/sys/dev/block"); | |
1579 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1580 | int fd; | |
1581 | char model[64]; | |
1582 | char vendor[64]; | |
1583 | char buf[1024]; | |
1584 | int major, minor; | |
1585 | char *device; | |
1586 | char *c; | |
1587 | int port; | |
1588 | int type; | |
1589 | ||
1590 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1591 | continue; | |
1592 | path = devt_to_devpath(makedev(major, minor)); | |
1593 | if (!path) | |
1594 | continue; | |
1595 | if (!path_attached_to_hba(path, hba_path)) { | |
1596 | free(path); | |
1597 | path = NULL; | |
1598 | continue; | |
1599 | } | |
1600 | ||
1601 | /* retrieve the scsi device type */ | |
1602 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 1603 | if (verbose > 0) |
e7b84f9d | 1604 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
1605 | err = 2; |
1606 | break; | |
1607 | } | |
1608 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1609 | if (load_sys(device, buf) != 0) { | |
ba728be7 | 1610 | if (verbose > 0) |
e7b84f9d | 1611 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
1612 | path); |
1613 | err = 2; | |
1614 | free(device); | |
1615 | break; | |
1616 | } | |
1617 | type = strtoul(buf, NULL, 10); | |
1618 | ||
1619 | /* if it's not a disk print the vendor and model */ | |
1620 | if (!(type == 0 || type == 7 || type == 14)) { | |
1621 | vendor[0] = '\0'; | |
1622 | model[0] = '\0'; | |
1623 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1624 | if (load_sys(device, buf) == 0) { | |
1625 | strncpy(vendor, buf, sizeof(vendor)); | |
1626 | vendor[sizeof(vendor) - 1] = '\0'; | |
1627 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1628 | while (isspace(*c) || *c == '\0') | |
1629 | *c-- = '\0'; | |
1630 | ||
1631 | } | |
1632 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1633 | if (load_sys(device, buf) == 0) { | |
1634 | strncpy(model, buf, sizeof(model)); | |
1635 | model[sizeof(model) - 1] = '\0'; | |
1636 | c = (char *) &model[sizeof(model) - 1]; | |
1637 | while (isspace(*c) || *c == '\0') | |
1638 | *c-- = '\0'; | |
1639 | } | |
1640 | ||
1641 | if (vendor[0] && model[0]) | |
1642 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1643 | else | |
1644 | switch (type) { /* numbers from hald/linux/device.c */ | |
1645 | case 1: sprintf(buf, "tape"); break; | |
1646 | case 2: sprintf(buf, "printer"); break; | |
1647 | case 3: sprintf(buf, "processor"); break; | |
1648 | case 4: | |
1649 | case 5: sprintf(buf, "cdrom"); break; | |
1650 | case 6: sprintf(buf, "scanner"); break; | |
1651 | case 8: sprintf(buf, "media_changer"); break; | |
1652 | case 9: sprintf(buf, "comm"); break; | |
1653 | case 12: sprintf(buf, "raid"); break; | |
1654 | default: sprintf(buf, "unknown"); | |
1655 | } | |
1656 | } else | |
1657 | buf[0] = '\0'; | |
1658 | free(device); | |
1659 | ||
1660 | /* chop device path to 'host%d' and calculate the port number */ | |
1661 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 1662 | if (!c) { |
ba728be7 | 1663 | if (verbose > 0) |
e7b84f9d | 1664 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
1665 | err = 2; |
1666 | break; | |
1667 | } | |
d665cc31 DW |
1668 | *c = '\0'; |
1669 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1670 | port -= host_base; | |
1671 | else { | |
ba728be7 | 1672 | if (verbose > 0) { |
d665cc31 | 1673 | *c = '/'; /* repair the full string */ |
e7b84f9d | 1674 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
1675 | path); |
1676 | } | |
1677 | err = 2; | |
1678 | break; | |
1679 | } | |
1680 | ||
1681 | /* mark this port as used */ | |
1682 | port_mask &= ~(1 << port); | |
1683 | ||
1684 | /* print out the device information */ | |
1685 | if (buf[0]) { | |
1686 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1687 | continue; | |
1688 | } | |
1689 | ||
1690 | fd = dev_open(ent->d_name, O_RDONLY); | |
1691 | if (fd < 0) | |
1692 | printf(" Port%d : - disk info unavailable -\n", port); | |
1693 | else { | |
1694 | fd2devname(fd, buf); | |
1695 | printf(" Port%d : %s", port, buf); | |
1696 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1697 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1698 | else |
664d5325 | 1699 | printf(" ()\n"); |
4dab422a | 1700 | close(fd); |
d665cc31 | 1701 | } |
d665cc31 DW |
1702 | free(path); |
1703 | path = NULL; | |
1704 | } | |
1705 | if (path) | |
1706 | free(path); | |
1707 | if (dir) | |
1708 | closedir(dir); | |
1709 | if (err == 0) { | |
1710 | int i; | |
1711 | ||
1712 | for (i = 0; i < port_count; i++) | |
1713 | if (port_mask & (1 << i)) | |
1714 | printf(" Port%d : - no device attached -\n", i); | |
1715 | } | |
1716 | ||
1717 | return err; | |
1718 | } | |
1719 | ||
120dc887 LM |
1720 | static void print_found_intel_controllers(struct sys_dev *elem) |
1721 | { | |
1722 | for (; elem; elem = elem->next) { | |
e7b84f9d | 1723 | pr_err("found Intel(R) "); |
120dc887 LM |
1724 | if (elem->type == SYS_DEV_SATA) |
1725 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1726 | else if (elem->type == SYS_DEV_SAS) |
1727 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1728 | fprintf(stderr, "RAID controller"); |
1729 | if (elem->pci_id) | |
1730 | fprintf(stderr, " at %s", elem->pci_id); | |
1731 | fprintf(stderr, ".\n"); | |
1732 | } | |
1733 | fflush(stderr); | |
1734 | } | |
1735 | ||
120dc887 LM |
1736 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1737 | { | |
1738 | struct dirent *ent; | |
1739 | DIR *dir; | |
1740 | int host_base = -1; | |
1741 | ||
1742 | *port_count = 0; | |
1743 | if ((dir = opendir(hba_path)) == NULL) | |
1744 | return -1; | |
1745 | ||
1746 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1747 | int host; | |
1748 | ||
1749 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1750 | continue; | |
1751 | if (*port_count == 0) | |
1752 | host_base = host; | |
1753 | else if (host < host_base) | |
1754 | host_base = host; | |
1755 | ||
1756 | if (host + 1 > *port_count + host_base) | |
1757 | *port_count = host + 1 - host_base; | |
1758 | } | |
1759 | closedir(dir); | |
1760 | return host_base; | |
1761 | } | |
1762 | ||
a891a3c2 LM |
1763 | static void print_imsm_capability(const struct imsm_orom *orom) |
1764 | { | |
1765 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1766 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1767 | orom->hotfix_ver, orom->build); | |
1768 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1769 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1770 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1771 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1772 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1773 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1774 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1775 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1776 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1777 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1778 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1779 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1780 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1781 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1782 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1783 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1784 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1785 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1786 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1787 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1788 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1789 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1790 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
1791 | printf(" 2TB volumes :%s supported\n", |
1792 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
1793 | printf(" 2TB disks :%s supported\n", | |
1794 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 LM |
1795 | printf(" Max Disks : %d\n", orom->tds); |
1796 | printf(" Max Volumes : %d per array, %d per controller\n", | |
1797 | orom->vpa, orom->vphba); | |
a891a3c2 LM |
1798 | return; |
1799 | } | |
1800 | ||
e50cf220 MN |
1801 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
1802 | { | |
1803 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
1804 | printf("IMSM_VERSION=%d.%d.%d.%d\n",orom->major_ver, orom->minor_ver, | |
1805 | orom->hotfix_ver, orom->build); | |
1806 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", | |
1807 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
1808 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
1809 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
1810 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
1811 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
1812 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1813 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
1814 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
1815 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
1816 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
1817 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
1818 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
1819 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
1820 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
1821 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
1822 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
1823 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
1824 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
1825 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
1826 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
1827 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
1828 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
1829 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
1830 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
1831 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
1832 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
1833 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
1834 | } | |
1835 | ||
5615172f | 1836 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1837 | { |
1838 | /* There are two components to imsm platform support, the ahci SATA | |
1839 | * controller and the option-rom. To find the SATA controller we | |
1840 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1841 | * controller with the Intel vendor id is present. This approach | |
1842 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1843 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1844 | * detect platform raid capabilities. The option-rom resides in a | |
1845 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1846 | * platform capabilities. If raid support is disabled in the BIOS the | |
1847 | * option-rom capability structure will not be available. | |
1848 | */ | |
1849 | const struct imsm_orom *orom; | |
1850 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1851 | int host_base = 0; |
1852 | int port_count = 0; | |
120dc887 | 1853 | int result=0; |
d665cc31 | 1854 | |
5615172f | 1855 | if (enumerate_only) { |
a891a3c2 | 1856 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1857 | return 0; |
a891a3c2 LM |
1858 | list = find_intel_devices(); |
1859 | if (!list) | |
1860 | return 2; | |
1861 | for (hba = list; hba; hba = hba->next) { | |
1862 | orom = find_imsm_capability(hba->type); | |
1863 | if (!orom) { | |
1864 | result = 2; | |
1865 | break; | |
1866 | } | |
1867 | } | |
1868 | free_sys_dev(&list); | |
1869 | return result; | |
5615172f DW |
1870 | } |
1871 | ||
155cbb4c LM |
1872 | list = find_intel_devices(); |
1873 | if (!list) { | |
ba728be7 | 1874 | if (verbose > 0) |
e7b84f9d | 1875 | pr_err("no active Intel(R) RAID " |
155cbb4c | 1876 | "controller found.\n"); |
d665cc31 DW |
1877 | free_sys_dev(&list); |
1878 | return 2; | |
ba728be7 | 1879 | } else if (verbose > 0) |
155cbb4c | 1880 | print_found_intel_controllers(list); |
d665cc31 | 1881 | |
a891a3c2 LM |
1882 | for (hba = list; hba; hba = hba->next) { |
1883 | orom = find_imsm_capability(hba->type); | |
1884 | if (!orom) | |
e7b84f9d | 1885 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
a891a3c2 LM |
1886 | hba->path, get_sys_dev_type(hba->type)); |
1887 | else | |
1888 | print_imsm_capability(orom); | |
d665cc31 DW |
1889 | } |
1890 | ||
120dc887 LM |
1891 | for (hba = list; hba; hba = hba->next) { |
1892 | printf(" I/O Controller : %s (%s)\n", | |
1893 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1894 | |
120dc887 LM |
1895 | if (hba->type == SYS_DEV_SATA) { |
1896 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1897 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
ba728be7 | 1898 | if (verbose > 0) |
e7b84f9d | 1899 | pr_err("failed to enumerate " |
120dc887 LM |
1900 | "ports on SATA controller at %s.", hba->pci_id); |
1901 | result |= 2; | |
1902 | } | |
1903 | } | |
d665cc31 | 1904 | } |
155cbb4c | 1905 | |
120dc887 LM |
1906 | free_sys_dev(&list); |
1907 | return result; | |
d665cc31 | 1908 | } |
e50cf220 MN |
1909 | |
1910 | static int export_detail_platform_imsm(int verbose) | |
1911 | { | |
1912 | const struct imsm_orom *orom; | |
1913 | struct sys_dev *list, *hba; | |
1914 | int result=1; | |
1915 | ||
1916 | list = find_intel_devices(); | |
1917 | if (!list) { | |
1918 | if (verbose > 0) | |
1919 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
1920 | result = 2; | |
1921 | free_sys_dev(&list); | |
1922 | return result; | |
1923 | } | |
1924 | ||
1925 | for (hba = list; hba; hba = hba->next) { | |
1926 | orom = find_imsm_capability(hba->type); | |
1927 | if (!orom) { | |
1928 | if (verbose > 0) | |
1929 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",hba->path); | |
1930 | } | |
1931 | else { | |
1932 | print_imsm_capability_export(orom); | |
1933 | result = 0; | |
1934 | } | |
1935 | } | |
1936 | ||
1937 | if (result == 1 && verbose > 0) | |
1938 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICES\n"); | |
1939 | ||
1940 | return result; | |
1941 | } | |
1942 | ||
cdddbdbc DW |
1943 | #endif |
1944 | ||
1945 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1946 | { | |
5115ca67 DW |
1947 | /* the imsm metadata format does not specify any host |
1948 | * identification information. We return -1 since we can never | |
1949 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1950 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1951 | * exclude member disks that do not belong, and we rely on |
1952 | * mdadm.conf to specify the arrays that should be assembled. | |
1953 | * Auto-assembly may still pick up "foreign" arrays. | |
1954 | */ | |
cdddbdbc | 1955 | |
9362c1c8 | 1956 | return -1; |
cdddbdbc DW |
1957 | } |
1958 | ||
1959 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1960 | { | |
51006d85 N |
1961 | /* The uuid returned here is used for: |
1962 | * uuid to put into bitmap file (Create, Grow) | |
1963 | * uuid for backup header when saving critical section (Grow) | |
1964 | * comparing uuids when re-adding a device into an array | |
1965 | * In these cases the uuid required is that of the data-array, | |
1966 | * not the device-set. | |
1967 | * uuid to recognise same set when adding a missing device back | |
1968 | * to an array. This is a uuid for the device-set. | |
1969 | * | |
1970 | * For each of these we can make do with a truncated | |
1971 | * or hashed uuid rather than the original, as long as | |
1972 | * everyone agrees. | |
1973 | * In each case the uuid required is that of the data-array, | |
1974 | * not the device-set. | |
43dad3d6 | 1975 | */ |
51006d85 N |
1976 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1977 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1978 | * - the orig_family_num of the container |
51006d85 N |
1979 | * - the index number of the volume |
1980 | * - the 'serial' number of the volume. | |
1981 | * Hopefully these are all constant. | |
1982 | */ | |
1983 | struct intel_super *super = st->sb; | |
43dad3d6 | 1984 | |
51006d85 N |
1985 | char buf[20]; |
1986 | struct sha1_ctx ctx; | |
1987 | struct imsm_dev *dev = NULL; | |
148acb7b | 1988 | __u32 family_num; |
51006d85 | 1989 | |
148acb7b DW |
1990 | /* some mdadm versions failed to set ->orig_family_num, in which |
1991 | * case fall back to ->family_num. orig_family_num will be | |
1992 | * fixed up with the first metadata update. | |
1993 | */ | |
1994 | family_num = super->anchor->orig_family_num; | |
1995 | if (family_num == 0) | |
1996 | family_num = super->anchor->family_num; | |
51006d85 | 1997 | sha1_init_ctx(&ctx); |
92bd8f8d | 1998 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1999 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2000 | if (super->current_vol >= 0) |
2001 | dev = get_imsm_dev(super, super->current_vol); | |
2002 | if (dev) { | |
2003 | __u32 vol = super->current_vol; | |
2004 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2005 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2006 | } | |
2007 | sha1_finish_ctx(&ctx, buf); | |
2008 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2009 | } |
2010 | ||
0d481d37 | 2011 | #if 0 |
4f5bc454 DW |
2012 | static void |
2013 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2014 | { |
cdddbdbc DW |
2015 | __u8 *v = get_imsm_version(mpb); |
2016 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2017 | char major[] = { 0, 0, 0 }; | |
2018 | char minor[] = { 0 ,0, 0 }; | |
2019 | char patch[] = { 0, 0, 0 }; | |
2020 | char *ver_parse[] = { major, minor, patch }; | |
2021 | int i, j; | |
2022 | ||
2023 | i = j = 0; | |
2024 | while (*v != '\0' && v < end) { | |
2025 | if (*v != '.' && j < 2) | |
2026 | ver_parse[i][j++] = *v; | |
2027 | else { | |
2028 | i++; | |
2029 | j = 0; | |
2030 | } | |
2031 | v++; | |
2032 | } | |
2033 | ||
4f5bc454 DW |
2034 | *m = strtol(minor, NULL, 0); |
2035 | *p = strtol(patch, NULL, 0); | |
2036 | } | |
0d481d37 | 2037 | #endif |
4f5bc454 | 2038 | |
1e5c6983 DW |
2039 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2040 | { | |
2041 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2042 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2043 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2044 | ||
2045 | switch (get_imsm_raid_level(map)) { | |
2046 | case 5: | |
2047 | case 10: | |
2048 | return chunk; | |
2049 | default: | |
2050 | return 128*1024 >> 9; | |
2051 | } | |
2052 | } | |
2053 | ||
2054 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2055 | { | |
2056 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2057 | * this is different than migr_strip_size_resync(), but it's good | |
2058 | * to be compatible | |
2059 | */ | |
238c0a71 | 2060 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2061 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2062 | ||
2063 | switch (get_imsm_raid_level(map)) { | |
2064 | case 1: | |
2065 | case 10: | |
2066 | if (map->num_members % map->num_domains == 0) | |
2067 | return 128*1024 >> 9; | |
2068 | else | |
2069 | return chunk; | |
2070 | case 5: | |
2071 | return max((__u32) 64*1024 >> 9, chunk); | |
2072 | default: | |
2073 | return 128*1024 >> 9; | |
2074 | } | |
2075 | } | |
2076 | ||
2077 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2078 | { | |
238c0a71 AK |
2079 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2080 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2081 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2082 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2083 | ||
2084 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2085 | } | |
2086 | ||
2087 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2088 | { | |
238c0a71 | 2089 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2090 | int level = get_imsm_raid_level(lo); |
2091 | ||
2092 | if (level == 1 || level == 10) { | |
238c0a71 | 2093 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2094 | |
2095 | return hi->num_domains; | |
2096 | } else | |
2097 | return num_stripes_per_unit_resync(dev); | |
2098 | } | |
2099 | ||
98130f40 | 2100 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2101 | { |
2102 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2103 | * counter-intuitively have the same number of data disks | |
2104 | */ | |
98130f40 | 2105 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2106 | |
2107 | switch (get_imsm_raid_level(map)) { | |
2108 | case 0: | |
36fd8ccc AK |
2109 | return map->num_members; |
2110 | break; | |
1e5c6983 DW |
2111 | case 1: |
2112 | case 10: | |
36fd8ccc | 2113 | return map->num_members/2; |
1e5c6983 DW |
2114 | case 5: |
2115 | return map->num_members - 1; | |
2116 | default: | |
2117 | dprintf("%s: unsupported raid level\n", __func__); | |
2118 | return 0; | |
2119 | } | |
2120 | } | |
2121 | ||
2122 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2123 | { | |
238c0a71 | 2124 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2125 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2126 | ||
2127 | switch(get_imsm_raid_level(map)) { | |
2128 | case 1: | |
2129 | case 10: | |
2130 | return chunk * map->num_domains; | |
2131 | case 5: | |
2132 | return chunk * map->num_members; | |
2133 | default: | |
2134 | return chunk; | |
2135 | } | |
2136 | } | |
2137 | ||
2138 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2139 | { | |
238c0a71 | 2140 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2141 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2142 | __u32 strip = block / chunk; | |
2143 | ||
2144 | switch (get_imsm_raid_level(map)) { | |
2145 | case 1: | |
2146 | case 10: { | |
2147 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2148 | __u32 vol_stripe = vol_strip / map->num_members; | |
2149 | ||
2150 | return vol_stripe * chunk + block % chunk; | |
2151 | } case 5: { | |
2152 | __u32 stripe = strip / (map->num_members - 1); | |
2153 | ||
2154 | return stripe * chunk + block % chunk; | |
2155 | } | |
2156 | default: | |
2157 | return 0; | |
2158 | } | |
2159 | } | |
2160 | ||
c47b0ff6 AK |
2161 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2162 | struct imsm_dev *dev) | |
1e5c6983 DW |
2163 | { |
2164 | /* calculate the conversion factor between per member 'blocks' | |
2165 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2166 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2167 | */ | |
2168 | if (!dev->vol.migr_state) | |
2169 | return 0; | |
2170 | ||
2171 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2172 | case MIGR_GEN_MIGR: { |
2173 | struct migr_record *migr_rec = super->migr_rec; | |
2174 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2175 | } | |
1e5c6983 DW |
2176 | case MIGR_VERIFY: |
2177 | case MIGR_REPAIR: | |
2178 | case MIGR_INIT: { | |
238c0a71 | 2179 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2180 | __u32 stripes_per_unit; |
2181 | __u32 blocks_per_unit; | |
2182 | __u32 parity_depth; | |
2183 | __u32 migr_chunk; | |
2184 | __u32 block_map; | |
2185 | __u32 block_rel; | |
2186 | __u32 segment; | |
2187 | __u32 stripe; | |
2188 | __u8 disks; | |
2189 | ||
2190 | /* yes, this is really the translation of migr_units to | |
2191 | * per-member blocks in the 'resync' case | |
2192 | */ | |
2193 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2194 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2195 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2196 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2197 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2198 | segment = blocks_per_unit / stripe; |
2199 | block_rel = blocks_per_unit - segment * stripe; | |
2200 | parity_depth = parity_segment_depth(dev); | |
2201 | block_map = map_migr_block(dev, block_rel); | |
2202 | return block_map + parity_depth * segment; | |
2203 | } | |
2204 | case MIGR_REBUILD: { | |
2205 | __u32 stripes_per_unit; | |
2206 | __u32 migr_chunk; | |
2207 | ||
2208 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2209 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2210 | return migr_chunk * stripes_per_unit; | |
2211 | } | |
1e5c6983 DW |
2212 | case MIGR_STATE_CHANGE: |
2213 | default: | |
2214 | return 0; | |
2215 | } | |
2216 | } | |
2217 | ||
c2c087e6 DW |
2218 | static int imsm_level_to_layout(int level) |
2219 | { | |
2220 | switch (level) { | |
2221 | case 0: | |
2222 | case 1: | |
2223 | return 0; | |
2224 | case 5: | |
2225 | case 6: | |
a380c027 | 2226 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2227 | case 10: |
c92a2527 | 2228 | return 0x102; |
c2c087e6 | 2229 | } |
a18a888e | 2230 | return UnSet; |
c2c087e6 DW |
2231 | } |
2232 | ||
8e59f3d8 AK |
2233 | /******************************************************************************* |
2234 | * Function: read_imsm_migr_rec | |
2235 | * Description: Function reads imsm migration record from last sector of disk | |
2236 | * Parameters: | |
2237 | * fd : disk descriptor | |
2238 | * super : metadata info | |
2239 | * Returns: | |
2240 | * 0 : success, | |
2241 | * -1 : fail | |
2242 | ******************************************************************************/ | |
2243 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2244 | { | |
2245 | int ret_val = -1; | |
2246 | unsigned long long dsize; | |
2247 | ||
2248 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2249 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2250 | pr_err("Cannot seek to anchor block: %s\n", |
2251 | strerror(errno)); | |
8e59f3d8 AK |
2252 | goto out; |
2253 | } | |
17a4eaf9 AK |
2254 | if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2255 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2256 | pr_err("Cannot read migr record block: %s\n", |
2257 | strerror(errno)); | |
8e59f3d8 AK |
2258 | goto out; |
2259 | } | |
2260 | ret_val = 0; | |
2261 | ||
2262 | out: | |
2263 | return ret_val; | |
2264 | } | |
2265 | ||
3136abe5 AK |
2266 | static struct imsm_dev *imsm_get_device_during_migration( |
2267 | struct intel_super *super) | |
2268 | { | |
2269 | ||
2270 | struct intel_dev *dv; | |
2271 | ||
2272 | for (dv = super->devlist; dv; dv = dv->next) { | |
2273 | if (is_gen_migration(dv->dev)) | |
2274 | return dv->dev; | |
2275 | } | |
2276 | return NULL; | |
2277 | } | |
2278 | ||
8e59f3d8 AK |
2279 | /******************************************************************************* |
2280 | * Function: load_imsm_migr_rec | |
2281 | * Description: Function reads imsm migration record (it is stored at the last | |
2282 | * sector of disk) | |
2283 | * Parameters: | |
2284 | * super : imsm internal array info | |
2285 | * info : general array info | |
2286 | * Returns: | |
2287 | * 0 : success | |
2288 | * -1 : fail | |
4c965cc9 | 2289 | * -2 : no migration in progress |
8e59f3d8 AK |
2290 | ******************************************************************************/ |
2291 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2292 | { | |
2293 | struct mdinfo *sd; | |
2294 | struct dl *dl = NULL; | |
2295 | char nm[30]; | |
2296 | int retval = -1; | |
2297 | int fd = -1; | |
3136abe5 AK |
2298 | struct imsm_dev *dev; |
2299 | struct imsm_map *map = NULL; | |
b4ab44d8 | 2300 | int slot = -1; |
3136abe5 AK |
2301 | |
2302 | /* find map under migration */ | |
2303 | dev = imsm_get_device_during_migration(super); | |
2304 | /* nothing to load,no migration in progress? | |
2305 | */ | |
2306 | if (dev == NULL) | |
4c965cc9 | 2307 | return -2; |
44bfe6df | 2308 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 AK |
2309 | |
2310 | if (info) { | |
2311 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3136abe5 AK |
2312 | /* skip spare and failed disks |
2313 | */ | |
2314 | if (sd->disk.raid_disk < 0) | |
2315 | continue; | |
8e59f3d8 | 2316 | /* read only from one of the first two slots */ |
3136abe5 AK |
2317 | if (map) |
2318 | slot = get_imsm_disk_slot(map, | |
2319 | sd->disk.raid_disk); | |
2320 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 | 2321 | continue; |
3136abe5 | 2322 | |
8e59f3d8 AK |
2323 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2324 | fd = dev_open(nm, O_RDONLY); | |
2325 | if (fd >= 0) | |
2326 | break; | |
2327 | } | |
2328 | } | |
2329 | if (fd < 0) { | |
2330 | for (dl = super->disks; dl; dl = dl->next) { | |
3136abe5 AK |
2331 | /* skip spare and failed disks |
2332 | */ | |
2333 | if (dl->index < 0) | |
2334 | continue; | |
8e59f3d8 | 2335 | /* read only from one of the first two slots */ |
3136abe5 AK |
2336 | if (map) |
2337 | slot = get_imsm_disk_slot(map, dl->index); | |
2338 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 AK |
2339 | continue; |
2340 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2341 | fd = dev_open(nm, O_RDONLY); | |
2342 | if (fd >= 0) | |
2343 | break; | |
2344 | } | |
2345 | } | |
2346 | if (fd < 0) | |
2347 | goto out; | |
2348 | retval = read_imsm_migr_rec(fd, super); | |
2349 | ||
2350 | out: | |
2351 | if (fd >= 0) | |
2352 | close(fd); | |
2353 | return retval; | |
2354 | } | |
2355 | ||
9e2d750d | 2356 | #ifndef MDASSEMBLE |
c17608ea AK |
2357 | /******************************************************************************* |
2358 | * function: imsm_create_metadata_checkpoint_update | |
2359 | * Description: It creates update for checkpoint change. | |
2360 | * Parameters: | |
2361 | * super : imsm internal array info | |
2362 | * u : pointer to prepared update | |
2363 | * Returns: | |
2364 | * Uptate length. | |
2365 | * If length is equal to 0, input pointer u contains no update | |
2366 | ******************************************************************************/ | |
2367 | static int imsm_create_metadata_checkpoint_update( | |
2368 | struct intel_super *super, | |
2369 | struct imsm_update_general_migration_checkpoint **u) | |
2370 | { | |
2371 | ||
2372 | int update_memory_size = 0; | |
2373 | ||
2374 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2375 | ||
2376 | if (u == NULL) | |
2377 | return 0; | |
2378 | *u = NULL; | |
2379 | ||
2380 | /* size of all update data without anchor */ | |
2381 | update_memory_size = | |
2382 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2383 | ||
503975b9 | 2384 | *u = xcalloc(1, update_memory_size); |
c17608ea AK |
2385 | if (*u == NULL) { |
2386 | dprintf("error: cannot get memory for " | |
2387 | "imsm_create_metadata_checkpoint_update update\n"); | |
2388 | return 0; | |
2389 | } | |
2390 | (*u)->type = update_general_migration_checkpoint; | |
2391 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2392 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2393 | (*u)->curr_migr_unit); | |
2394 | ||
2395 | return update_memory_size; | |
2396 | } | |
2397 | ||
2398 | ||
2399 | static void imsm_update_metadata_locally(struct supertype *st, | |
2400 | void *buf, int len); | |
2401 | ||
687629c2 AK |
2402 | /******************************************************************************* |
2403 | * Function: write_imsm_migr_rec | |
2404 | * Description: Function writes imsm migration record | |
2405 | * (at the last sector of disk) | |
2406 | * Parameters: | |
2407 | * super : imsm internal array info | |
2408 | * Returns: | |
2409 | * 0 : success | |
2410 | * -1 : if fail | |
2411 | ******************************************************************************/ | |
2412 | static int write_imsm_migr_rec(struct supertype *st) | |
2413 | { | |
2414 | struct intel_super *super = st->sb; | |
2415 | unsigned long long dsize; | |
2416 | char nm[30]; | |
2417 | int fd = -1; | |
2418 | int retval = -1; | |
2419 | struct dl *sd; | |
c17608ea AK |
2420 | int len; |
2421 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 AK |
2422 | struct imsm_dev *dev; |
2423 | struct imsm_map *map = NULL; | |
2424 | ||
2425 | /* find map under migration */ | |
2426 | dev = imsm_get_device_during_migration(super); | |
2427 | /* if no migration, write buffer anyway to clear migr_record | |
2428 | * on disk based on first available device | |
2429 | */ | |
2430 | if (dev == NULL) | |
2431 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2432 | super->current_vol); | |
2433 | ||
44bfe6df | 2434 | map = get_imsm_map(dev, MAP_0); |
687629c2 AK |
2435 | |
2436 | for (sd = super->disks ; sd ; sd = sd->next) { | |
b4ab44d8 | 2437 | int slot = -1; |
3136abe5 AK |
2438 | |
2439 | /* skip failed and spare devices */ | |
2440 | if (sd->index < 0) | |
2441 | continue; | |
687629c2 | 2442 | /* write to 2 first slots only */ |
3136abe5 AK |
2443 | if (map) |
2444 | slot = get_imsm_disk_slot(map, sd->index); | |
2445 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
687629c2 | 2446 | continue; |
3136abe5 | 2447 | |
687629c2 AK |
2448 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
2449 | fd = dev_open(nm, O_RDWR); | |
2450 | if (fd < 0) | |
2451 | continue; | |
2452 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2453 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2454 | pr_err("Cannot seek to anchor block: %s\n", |
2455 | strerror(errno)); | |
687629c2 AK |
2456 | goto out; |
2457 | } | |
17a4eaf9 AK |
2458 | if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2459 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2460 | pr_err("Cannot write migr record block: %s\n", |
2461 | strerror(errno)); | |
687629c2 AK |
2462 | goto out; |
2463 | } | |
2464 | close(fd); | |
2465 | fd = -1; | |
2466 | } | |
c17608ea AK |
2467 | /* update checkpoint information in metadata */ |
2468 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2469 | ||
2470 | if (len <= 0) { | |
2471 | dprintf("imsm: Cannot prepare update\n"); | |
2472 | goto out; | |
2473 | } | |
2474 | /* update metadata locally */ | |
2475 | imsm_update_metadata_locally(st, u, len); | |
2476 | /* and possibly remotely */ | |
2477 | if (st->update_tail) { | |
2478 | append_metadata_update(st, u, len); | |
2479 | /* during reshape we do all work inside metadata handler | |
2480 | * manage_reshape(), so metadata update has to be triggered | |
2481 | * insida it | |
2482 | */ | |
2483 | flush_metadata_updates(st); | |
2484 | st->update_tail = &st->updates; | |
2485 | } else | |
2486 | free(u); | |
687629c2 AK |
2487 | |
2488 | retval = 0; | |
2489 | out: | |
2490 | if (fd >= 0) | |
2491 | close(fd); | |
2492 | return retval; | |
2493 | } | |
9e2d750d | 2494 | #endif /* MDASSEMBLE */ |
687629c2 | 2495 | |
e2962bfc AK |
2496 | /* spare/missing disks activations are not allowe when |
2497 | * array/container performs reshape operation, because | |
2498 | * all arrays in container works on the same disks set | |
2499 | */ | |
2500 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2501 | { | |
2502 | int rv = 0; | |
2503 | struct intel_dev *i_dev; | |
2504 | struct imsm_dev *dev; | |
2505 | ||
2506 | /* check whole container | |
2507 | */ | |
2508 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2509 | dev = i_dev->dev; | |
3ad25638 | 2510 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2511 | /* No repair during any migration in container |
2512 | */ | |
2513 | rv = 1; | |
2514 | break; | |
2515 | } | |
2516 | } | |
2517 | return rv; | |
2518 | } | |
c41e00b2 AK |
2519 | static unsigned long long imsm_component_size_aligment_check(int level, |
2520 | int chunk_size, | |
2521 | unsigned long long component_size) | |
2522 | { | |
2523 | unsigned int component_size_alligment; | |
2524 | ||
2525 | /* check component size aligment | |
2526 | */ | |
2527 | component_size_alligment = component_size % (chunk_size/512); | |
2528 | ||
2529 | dprintf("imsm_component_size_aligment_check(Level: %i, " | |
2530 | "chunk_size = %i, component_size = %llu), " | |
2531 | "component_size_alligment = %u\n", | |
2532 | level, chunk_size, component_size, | |
2533 | component_size_alligment); | |
2534 | ||
2535 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
2536 | dprintf("imsm: reported component size alligned from %llu ", | |
2537 | component_size); | |
2538 | component_size -= component_size_alligment; | |
2539 | dprintf("to %llu (%i).\n", | |
2540 | component_size, component_size_alligment); | |
2541 | } | |
2542 | ||
2543 | return component_size; | |
2544 | } | |
e2962bfc | 2545 | |
a5d85af7 | 2546 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2547 | { |
2548 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2549 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2550 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2551 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2552 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2553 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2554 | struct dl *dl; |
e207da2f | 2555 | char *devname; |
a5d85af7 | 2556 | int map_disks = info->array.raid_disks; |
bf5a934a | 2557 | |
95eeceeb | 2558 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2559 | if (prev_map) |
2560 | map_to_analyse = prev_map; | |
2561 | ||
ca0748fa | 2562 | dl = super->current_disk; |
9894ec0d | 2563 | |
bf5a934a | 2564 | info->container_member = super->current_vol; |
cd0430a1 | 2565 | info->array.raid_disks = map->num_members; |
b335e593 | 2566 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2567 | info->array.layout = imsm_level_to_layout(info->array.level); |
2568 | info->array.md_minor = -1; | |
2569 | info->array.ctime = 0; | |
2570 | info->array.utime = 0; | |
b335e593 AK |
2571 | info->array.chunk_size = |
2572 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2573 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2574 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2575 | info->custom_array_size <<= 32; | |
2576 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2577 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2578 | ||
3f510843 | 2579 | if (is_gen_migration(dev)) { |
3f83228a | 2580 | info->reshape_active = 1; |
b335e593 AK |
2581 | info->new_level = get_imsm_raid_level(map); |
2582 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2583 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2584 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2585 | if (info->delta_disks) { |
2586 | /* this needs to be applied to every array | |
2587 | * in the container. | |
2588 | */ | |
81219e70 | 2589 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2590 | } |
3f83228a N |
2591 | /* We shape information that we give to md might have to be |
2592 | * modify to cope with md's requirement for reshaping arrays. | |
2593 | * For example, when reshaping a RAID0, md requires it to be | |
2594 | * presented as a degraded RAID4. | |
2595 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2596 | * the array as already being RAID5, but the 'before' layout | |
2597 | * is a RAID4-like layout. | |
2598 | */ | |
2599 | switch (info->array.level) { | |
2600 | case 0: | |
2601 | switch(info->new_level) { | |
2602 | case 0: | |
2603 | /* conversion is happening as RAID4 */ | |
2604 | info->array.level = 4; | |
2605 | info->array.raid_disks += 1; | |
2606 | break; | |
2607 | case 5: | |
2608 | /* conversion is happening as RAID5 */ | |
2609 | info->array.level = 5; | |
2610 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2611 | info->delta_disks -= 1; |
2612 | break; | |
2613 | default: | |
2614 | /* FIXME error message */ | |
2615 | info->array.level = UnSet; | |
2616 | break; | |
2617 | } | |
2618 | break; | |
2619 | } | |
b335e593 AK |
2620 | } else { |
2621 | info->new_level = UnSet; | |
2622 | info->new_layout = UnSet; | |
2623 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2624 | info->delta_disks = 0; |
b335e593 | 2625 | } |
ca0748fa | 2626 | |
efb30e7f DW |
2627 | if (dl) { |
2628 | info->disk.major = dl->major; | |
2629 | info->disk.minor = dl->minor; | |
ca0748fa | 2630 | info->disk.number = dl->index; |
656b6b5a N |
2631 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2632 | dl->index); | |
efb30e7f | 2633 | } |
bf5a934a | 2634 | |
5551b113 CA |
2635 | info->data_offset = pba_of_lba0(map_to_analyse); |
2636 | info->component_size = blocks_per_member(map_to_analyse); | |
139dae11 | 2637 | |
c41e00b2 AK |
2638 | info->component_size = imsm_component_size_aligment_check( |
2639 | info->array.level, | |
2640 | info->array.chunk_size, | |
2641 | info->component_size); | |
139dae11 | 2642 | |
301406c9 | 2643 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2644 | info->recovery_start = MaxSector; |
bf5a934a | 2645 | |
d2e6d5d6 | 2646 | info->reshape_progress = 0; |
b6796ce1 | 2647 | info->resync_start = MaxSector; |
b9172665 AK |
2648 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2649 | dev->vol.dirty) && | |
2650 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2651 | info->resync_start = 0; |
b6796ce1 AK |
2652 | } |
2653 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2654 | switch (migr_type(dev)) { |
2655 | case MIGR_REPAIR: | |
2656 | case MIGR_INIT: { | |
c47b0ff6 AK |
2657 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2658 | dev); | |
1e5c6983 DW |
2659 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2660 | ||
2661 | info->resync_start = blocks_per_unit * units; | |
2662 | break; | |
2663 | } | |
d2e6d5d6 | 2664 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2665 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2666 | dev); | |
2667 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2668 | unsigned long long array_blocks; |
2669 | int used_disks; | |
d2e6d5d6 | 2670 | |
befb629b AK |
2671 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2672 | (units < | |
2673 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2674 | (super->migr_rec->rec_status == | |
2675 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2676 | units++; | |
2677 | ||
d2e6d5d6 | 2678 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2679 | |
d2e6d5d6 AK |
2680 | dprintf("IMSM: General Migration checkpoint : %llu " |
2681 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2682 | (unsigned long long)units, |
2683 | (unsigned long long)blocks_per_unit, | |
2684 | info->reshape_progress); | |
75156c46 | 2685 | |
238c0a71 | 2686 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 2687 | if (used_disks > 0) { |
5551b113 | 2688 | array_blocks = blocks_per_member(map) * |
75156c46 AK |
2689 | used_disks; |
2690 | /* round array size down to closest MB | |
2691 | */ | |
2692 | info->custom_array_size = (array_blocks | |
2693 | >> SECT_PER_MB_SHIFT) | |
2694 | << SECT_PER_MB_SHIFT; | |
2695 | } | |
d2e6d5d6 | 2696 | } |
1e5c6983 DW |
2697 | case MIGR_VERIFY: |
2698 | /* we could emulate the checkpointing of | |
2699 | * 'sync_action=check' migrations, but for now | |
2700 | * we just immediately complete them | |
2701 | */ | |
2702 | case MIGR_REBUILD: | |
2703 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2704 | case MIGR_STATE_CHANGE: |
2705 | /* FIXME handle other migrations */ | |
2706 | default: | |
2707 | /* we are not dirty, so... */ | |
2708 | info->resync_start = MaxSector; | |
2709 | } | |
b6796ce1 | 2710 | } |
301406c9 DW |
2711 | |
2712 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2713 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2714 | |
f35f2525 N |
2715 | info->array.major_version = -1; |
2716 | info->array.minor_version = -2; | |
e207da2f AW |
2717 | devname = devnum2devname(st->container_dev); |
2718 | *info->text_version = '\0'; | |
2719 | if (devname) | |
2720 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2721 | free(devname); | |
a67dd8cc | 2722 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2723 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2724 | |
2725 | if (dmap) { | |
2726 | int i, j; | |
2727 | for (i=0; i<map_disks; i++) { | |
2728 | dmap[i] = 0; | |
2729 | if (i < info->array.raid_disks) { | |
2730 | struct imsm_disk *dsk; | |
238c0a71 | 2731 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2732 | dsk = get_imsm_disk(super, j); |
2733 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2734 | dmap[i] = 1; | |
2735 | } | |
2736 | } | |
2737 | } | |
81ac8b4d | 2738 | } |
bf5a934a | 2739 | |
3b451610 AK |
2740 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2741 | int failed, int look_in_map); | |
2742 | ||
2743 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2744 | int look_in_map); | |
2745 | ||
b4ab44d8 AK |
2746 | |
2747 | #ifndef MDASSEMBLE | |
3b451610 AK |
2748 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
2749 | { | |
2750 | if (is_gen_migration(dev)) { | |
2751 | int failed; | |
2752 | __u8 map_state; | |
2753 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2754 | ||
2755 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2756 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2757 | if (map2->map_state != map_state) { |
2758 | map2->map_state = map_state; | |
2759 | super->updates_pending++; | |
2760 | } | |
2761 | } | |
2762 | } | |
b4ab44d8 | 2763 | #endif |
97b4d0e9 DW |
2764 | |
2765 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2766 | { | |
2767 | struct dl *d; | |
2768 | ||
2769 | for (d = super->missing; d; d = d->next) | |
2770 | if (d->index == index) | |
2771 | return &d->disk; | |
2772 | return NULL; | |
2773 | } | |
2774 | ||
a5d85af7 | 2775 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2776 | { |
2777 | struct intel_super *super = st->sb; | |
4f5bc454 | 2778 | struct imsm_disk *disk; |
a5d85af7 | 2779 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2780 | int max_enough = -1; |
2781 | int i; | |
2782 | struct imsm_super *mpb; | |
4f5bc454 | 2783 | |
bf5a934a | 2784 | if (super->current_vol >= 0) { |
a5d85af7 | 2785 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2786 | return; |
2787 | } | |
95eeceeb | 2788 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2789 | |
2790 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2791 | * spares | |
2792 | */ | |
2793 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2794 | info->array.level = LEVEL_CONTAINER; |
2795 | info->array.layout = 0; | |
2796 | info->array.md_minor = -1; | |
c2c087e6 | 2797 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2798 | info->array.utime = 0; |
2799 | info->array.chunk_size = 0; | |
2800 | ||
2801 | info->disk.major = 0; | |
2802 | info->disk.minor = 0; | |
cdddbdbc | 2803 | info->disk.raid_disk = -1; |
c2c087e6 | 2804 | info->reshape_active = 0; |
f35f2525 N |
2805 | info->array.major_version = -1; |
2806 | info->array.minor_version = -2; | |
c2c087e6 | 2807 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2808 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2809 | info->disk.number = -1; |
2810 | info->disk.state = 0; | |
c5afc314 | 2811 | info->name[0] = 0; |
921d9e16 | 2812 | info->recovery_start = MaxSector; |
3ad25638 | 2813 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2814 | |
97b4d0e9 | 2815 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2816 | mpb = super->anchor; |
97b4d0e9 | 2817 | |
ab3cb6b3 N |
2818 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2819 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2820 | int failed, enough, j, missing = 0; | |
2821 | struct imsm_map *map; | |
2822 | __u8 state; | |
97b4d0e9 | 2823 | |
3b451610 AK |
2824 | failed = imsm_count_failed(super, dev, MAP_0); |
2825 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2826 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2827 | |
2828 | /* any newly missing disks? | |
2829 | * (catches single-degraded vs double-degraded) | |
2830 | */ | |
2831 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2832 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2833 | __u32 idx = ord_to_idx(ord); |
2834 | ||
2835 | if (!(ord & IMSM_ORD_REBUILD) && | |
2836 | get_imsm_missing(super, idx)) { | |
2837 | missing = 1; | |
2838 | break; | |
2839 | } | |
97b4d0e9 | 2840 | } |
ab3cb6b3 N |
2841 | |
2842 | if (state == IMSM_T_STATE_FAILED) | |
2843 | enough = -1; | |
2844 | else if (state == IMSM_T_STATE_DEGRADED && | |
2845 | (state != map->map_state || missing)) | |
2846 | enough = 0; | |
2847 | else /* we're normal, or already degraded */ | |
2848 | enough = 1; | |
d2bde6d3 AK |
2849 | if (is_gen_migration(dev) && missing) { |
2850 | /* during general migration we need all disks | |
2851 | * that process is running on. | |
2852 | * No new missing disk is allowed. | |
2853 | */ | |
2854 | max_enough = -1; | |
2855 | enough = -1; | |
2856 | /* no more checks necessary | |
2857 | */ | |
2858 | break; | |
2859 | } | |
ab3cb6b3 N |
2860 | /* in the missing/failed disk case check to see |
2861 | * if at least one array is runnable | |
2862 | */ | |
2863 | max_enough = max(max_enough, enough); | |
2864 | } | |
2865 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2866 | info->container_enough = max_enough; | |
97b4d0e9 | 2867 | |
4a04ec6c | 2868 | if (super->disks) { |
14e8215b DW |
2869 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2870 | ||
b9f594fe | 2871 | disk = &super->disks->disk; |
5551b113 | 2872 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 2873 | info->component_size = reserved; |
25ed7e59 | 2874 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2875 | /* we don't change info->disk.raid_disk here because |
2876 | * this state will be finalized in mdmon after we have | |
2877 | * found the 'most fresh' version of the metadata | |
2878 | */ | |
25ed7e59 DW |
2879 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2880 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2881 | } |
a575e2a7 DW |
2882 | |
2883 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2884 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2885 | */ | |
2886 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2887 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2888 | else |
2889 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2890 | |
2891 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2892 | if (map) { | |
2893 | int i; | |
2894 | for (i = 0; i < map_disks; i++) | |
2895 | map[i] = 1; | |
2896 | } | |
2897 | ||
cdddbdbc DW |
2898 | } |
2899 | ||
5c4cd5da AC |
2900 | /* allocates memory and fills disk in mdinfo structure |
2901 | * for each disk in array */ | |
2902 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2903 | { | |
2904 | struct mdinfo *mddev = NULL; | |
2905 | struct intel_super *super = st->sb; | |
2906 | struct imsm_disk *disk; | |
2907 | int count = 0; | |
2908 | struct dl *dl; | |
2909 | if (!super || !super->disks) | |
2910 | return NULL; | |
2911 | dl = super->disks; | |
503975b9 | 2912 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
2913 | while (dl) { |
2914 | struct mdinfo *tmp; | |
2915 | disk = &dl->disk; | |
503975b9 | 2916 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
2917 | if (mddev->devs) |
2918 | tmp->next = mddev->devs; | |
2919 | mddev->devs = tmp; | |
2920 | tmp->disk.number = count++; | |
2921 | tmp->disk.major = dl->major; | |
2922 | tmp->disk.minor = dl->minor; | |
2923 | tmp->disk.state = is_configured(disk) ? | |
2924 | (1 << MD_DISK_ACTIVE) : 0; | |
2925 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2926 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2927 | tmp->disk.raid_disk = -1; | |
2928 | dl = dl->next; | |
2929 | } | |
2930 | return mddev; | |
2931 | } | |
2932 | ||
cdddbdbc DW |
2933 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2934 | char *update, char *devname, int verbose, | |
2935 | int uuid_set, char *homehost) | |
2936 | { | |
f352c545 DW |
2937 | /* For 'assemble' and 'force' we need to return non-zero if any |
2938 | * change was made. For others, the return value is ignored. | |
2939 | * Update options are: | |
2940 | * force-one : This device looks a bit old but needs to be included, | |
2941 | * update age info appropriately. | |
2942 | * assemble: clear any 'faulty' flag to allow this device to | |
2943 | * be assembled. | |
2944 | * force-array: Array is degraded but being forced, mark it clean | |
2945 | * if that will be needed to assemble it. | |
2946 | * | |
2947 | * newdev: not used ???? | |
2948 | * grow: Array has gained a new device - this is currently for | |
2949 | * linear only | |
2950 | * resync: mark as dirty so a resync will happen. | |
2951 | * name: update the name - preserving the homehost | |
6e46bf34 | 2952 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2953 | * |
2954 | * Following are not relevant for this imsm: | |
2955 | * sparc2.2 : update from old dodgey metadata | |
2956 | * super-minor: change the preferred_minor number | |
2957 | * summaries: update redundant counters. | |
f352c545 DW |
2958 | * homehost: update the recorded homehost |
2959 | * _reshape_progress: record new reshape_progress position. | |
2960 | */ | |
6e46bf34 DW |
2961 | int rv = 1; |
2962 | struct intel_super *super = st->sb; | |
2963 | struct imsm_super *mpb; | |
f352c545 | 2964 | |
6e46bf34 DW |
2965 | /* we can only update container info */ |
2966 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2967 | return 1; | |
2968 | ||
2969 | mpb = super->anchor; | |
2970 | ||
81a5b4f5 N |
2971 | if (strcmp(update, "uuid") == 0) { |
2972 | /* We take this to mean that the family_num should be updated. | |
2973 | * However that is much smaller than the uuid so we cannot really | |
2974 | * allow an explicit uuid to be given. And it is hard to reliably | |
2975 | * know if one was. | |
2976 | * So if !uuid_set we know the current uuid is random and just used | |
2977 | * the first 'int' and copy it to the other 3 positions. | |
2978 | * Otherwise we require the 4 'int's to be the same as would be the | |
2979 | * case if we are using a random uuid. So an explicit uuid will be | |
2980 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 2981 | */ |
81a5b4f5 N |
2982 | if (!uuid_set) { |
2983 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 2984 | rv = 0; |
81a5b4f5 N |
2985 | } else { |
2986 | if (info->uuid[0] != info->uuid[1] || | |
2987 | info->uuid[1] != info->uuid[2] || | |
2988 | info->uuid[2] != info->uuid[3]) | |
2989 | rv = -1; | |
2990 | else | |
2991 | rv = 0; | |
6e46bf34 | 2992 | } |
81a5b4f5 N |
2993 | if (rv == 0) |
2994 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
2995 | } else if (strcmp(update, "assemble") == 0) |
2996 | rv = 0; | |
2997 | else | |
1e2b2765 | 2998 | rv = -1; |
f352c545 | 2999 | |
6e46bf34 DW |
3000 | /* successful update? recompute checksum */ |
3001 | if (rv == 0) | |
3002 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3003 | |
3004 | return rv; | |
cdddbdbc DW |
3005 | } |
3006 | ||
c2c087e6 | 3007 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3008 | { |
c2c087e6 | 3009 | size_t size; |
cdddbdbc | 3010 | |
c2c087e6 DW |
3011 | size = sizeof(struct imsm_super); |
3012 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3013 | size += 2 * sizeof(struct imsm_dev); | |
3014 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3015 | size += (4 - 2) * sizeof(struct imsm_map); | |
3016 | /* 4 possible disk_ord_tbl's */ | |
3017 | size += 4 * (disks - 1) * sizeof(__u32); | |
3018 | ||
3019 | return size; | |
3020 | } | |
3021 | ||
3022 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
3023 | { | |
3024 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3025 | return 0; | |
3026 | ||
3027 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3028 | } |
3029 | ||
ba2de7ba DW |
3030 | static void free_devlist(struct intel_super *super) |
3031 | { | |
3032 | struct intel_dev *dv; | |
3033 | ||
3034 | while (super->devlist) { | |
3035 | dv = super->devlist->next; | |
3036 | free(super->devlist->dev); | |
3037 | free(super->devlist); | |
3038 | super->devlist = dv; | |
3039 | } | |
3040 | } | |
3041 | ||
3042 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3043 | { | |
3044 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3045 | } | |
3046 | ||
cdddbdbc DW |
3047 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3048 | { | |
3049 | /* | |
3050 | * return: | |
3051 | * 0 same, or first was empty, and second was copied | |
3052 | * 1 second had wrong number | |
3053 | * 2 wrong uuid | |
3054 | * 3 wrong other info | |
3055 | */ | |
3056 | struct intel_super *first = st->sb; | |
3057 | struct intel_super *sec = tst->sb; | |
3058 | ||
3059 | if (!first) { | |
3060 | st->sb = tst->sb; | |
3061 | tst->sb = NULL; | |
3062 | return 0; | |
3063 | } | |
8603ea6f LM |
3064 | /* in platform dependent environment test if the disks |
3065 | * use the same Intel hba | |
3066 | */ | |
3067 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
3068 | if (!first->hba || !sec->hba || |
3069 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
3070 | fprintf(stderr, |
3071 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
3072 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
3073 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
3074 | return 3; |
3075 | } | |
3076 | } | |
cdddbdbc | 3077 | |
d23fe947 DW |
3078 | /* if an anchor does not have num_raid_devs set then it is a free |
3079 | * floating spare | |
3080 | */ | |
3081 | if (first->anchor->num_raid_devs > 0 && | |
3082 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3083 | /* Determine if these disks might ever have been |
3084 | * related. Further disambiguation can only take place | |
3085 | * in load_super_imsm_all | |
3086 | */ | |
3087 | __u32 first_family = first->anchor->orig_family_num; | |
3088 | __u32 sec_family = sec->anchor->orig_family_num; | |
3089 | ||
f796af5d DW |
3090 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3091 | MAX_SIGNATURE_LENGTH) != 0) | |
3092 | return 3; | |
3093 | ||
a2b97981 DW |
3094 | if (first_family == 0) |
3095 | first_family = first->anchor->family_num; | |
3096 | if (sec_family == 0) | |
3097 | sec_family = sec->anchor->family_num; | |
3098 | ||
3099 | if (first_family != sec_family) | |
d23fe947 | 3100 | return 3; |
f796af5d | 3101 | |
d23fe947 | 3102 | } |
cdddbdbc | 3103 | |
f796af5d | 3104 | |
3e372e5a DW |
3105 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3106 | * family number | |
3107 | */ | |
3108 | if (first->anchor->num_raid_devs == 0 && | |
3109 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3110 | int i; |
ba2de7ba DW |
3111 | struct intel_dev *dv; |
3112 | struct imsm_dev *dev; | |
78d30f94 DW |
3113 | |
3114 | /* we need to copy raid device info from sec if an allocation | |
3115 | * fails here we don't associate the spare | |
3116 | */ | |
3117 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3118 | dv = xmalloc(sizeof(*dv)); |
3119 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3120 | dv->dev = dev; |
3121 | dv->index = i; | |
3122 | dv->next = first->devlist; | |
3123 | first->devlist = dv; | |
78d30f94 | 3124 | } |
709743c5 | 3125 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3126 | /* allocation failure */ |
3127 | free_devlist(first); | |
3128 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
3129 | return 3; | |
78d30f94 | 3130 | } |
3e372e5a | 3131 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3132 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3133 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3134 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3135 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3136 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3137 | } |
3138 | ||
cdddbdbc DW |
3139 | return 0; |
3140 | } | |
3141 | ||
0030e8d6 DW |
3142 | static void fd2devname(int fd, char *name) |
3143 | { | |
3144 | struct stat st; | |
3145 | char path[256]; | |
33a6535d | 3146 | char dname[PATH_MAX]; |
0030e8d6 DW |
3147 | char *nm; |
3148 | int rv; | |
3149 | ||
3150 | name[0] = '\0'; | |
3151 | if (fstat(fd, &st) != 0) | |
3152 | return; | |
3153 | sprintf(path, "/sys/dev/block/%d:%d", | |
3154 | major(st.st_rdev), minor(st.st_rdev)); | |
3155 | ||
9cf014ec | 3156 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3157 | if (rv <= 0) |
3158 | return; | |
9587c373 | 3159 | |
0030e8d6 DW |
3160 | dname[rv] = '\0'; |
3161 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3162 | if (nm) { |
3163 | nm++; | |
3164 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3165 | } | |
0030e8d6 DW |
3166 | } |
3167 | ||
cdddbdbc DW |
3168 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3169 | ||
3170 | static int imsm_read_serial(int fd, char *devname, | |
3171 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3172 | { | |
3173 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
3174 | int rv; |
3175 | int rsp_len; | |
1f24f035 | 3176 | int len; |
316e2bf4 DW |
3177 | char *dest; |
3178 | char *src; | |
3179 | char *rsp_buf; | |
3180 | int i; | |
cdddbdbc DW |
3181 | |
3182 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 3183 | |
f9ba0ff1 DW |
3184 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
3185 | ||
40ebbb9c | 3186 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3187 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3188 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3189 | return 0; |
3190 | } | |
3191 | ||
cdddbdbc DW |
3192 | if (rv != 0) { |
3193 | if (devname) | |
e7b84f9d N |
3194 | pr_err("Failed to retrieve serial for %s\n", |
3195 | devname); | |
cdddbdbc DW |
3196 | return rv; |
3197 | } | |
3198 | ||
3199 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
3200 | if (!rsp_len) { |
3201 | if (devname) | |
e7b84f9d N |
3202 | pr_err("Failed to retrieve serial for %s\n", |
3203 | devname); | |
03cd4cc8 DW |
3204 | return 2; |
3205 | } | |
1f24f035 | 3206 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 3207 | |
316e2bf4 DW |
3208 | /* trim all whitespace and non-printable characters and convert |
3209 | * ':' to ';' | |
3210 | */ | |
3211 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
3212 | src = &rsp_buf[i]; | |
3213 | if (*src > 0x20) { | |
3214 | /* ':' is reserved for use in placeholder serial | |
3215 | * numbers for missing disks | |
3216 | */ | |
3217 | if (*src == ':') | |
3218 | *dest++ = ';'; | |
3219 | else | |
3220 | *dest++ = *src; | |
3221 | } | |
3222 | } | |
3223 | len = dest - rsp_buf; | |
3224 | dest = rsp_buf; | |
3225 | ||
3226 | /* truncate leading characters */ | |
3227 | if (len > MAX_RAID_SERIAL_LEN) { | |
3228 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3229 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3230 | } |
5c3db629 | 3231 | |
5c3db629 | 3232 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3233 | memcpy(serial, dest, len); |
cdddbdbc DW |
3234 | |
3235 | return 0; | |
3236 | } | |
3237 | ||
1f24f035 DW |
3238 | static int serialcmp(__u8 *s1, __u8 *s2) |
3239 | { | |
3240 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3241 | } | |
3242 | ||
3243 | static void serialcpy(__u8 *dest, __u8 *src) | |
3244 | { | |
3245 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3246 | } | |
3247 | ||
54c2c1ea DW |
3248 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3249 | { | |
3250 | struct dl *dl; | |
3251 | ||
3252 | for (dl = super->disks; dl; dl = dl->next) | |
3253 | if (serialcmp(dl->serial, serial) == 0) | |
3254 | break; | |
3255 | ||
3256 | return dl; | |
3257 | } | |
3258 | ||
a2b97981 DW |
3259 | static struct imsm_disk * |
3260 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3261 | { | |
3262 | int i; | |
3263 | ||
3264 | for (i = 0; i < mpb->num_disks; i++) { | |
3265 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3266 | ||
3267 | if (serialcmp(disk->serial, serial) == 0) { | |
3268 | if (idx) | |
3269 | *idx = i; | |
3270 | return disk; | |
3271 | } | |
3272 | } | |
3273 | ||
3274 | return NULL; | |
3275 | } | |
3276 | ||
cdddbdbc DW |
3277 | static int |
3278 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3279 | { | |
a2b97981 | 3280 | struct imsm_disk *disk; |
cdddbdbc DW |
3281 | struct dl *dl; |
3282 | struct stat stb; | |
cdddbdbc | 3283 | int rv; |
a2b97981 | 3284 | char name[40]; |
d23fe947 DW |
3285 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3286 | ||
3287 | rv = imsm_read_serial(fd, devname, serial); | |
3288 | ||
3289 | if (rv != 0) | |
3290 | return 2; | |
3291 | ||
503975b9 | 3292 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3293 | |
a2b97981 DW |
3294 | fstat(fd, &stb); |
3295 | dl->major = major(stb.st_rdev); | |
3296 | dl->minor = minor(stb.st_rdev); | |
3297 | dl->next = super->disks; | |
3298 | dl->fd = keep_fd ? fd : -1; | |
3299 | assert(super->disks == NULL); | |
3300 | super->disks = dl; | |
3301 | serialcpy(dl->serial, serial); | |
3302 | dl->index = -2; | |
3303 | dl->e = NULL; | |
3304 | fd2devname(fd, name); | |
3305 | if (devname) | |
503975b9 | 3306 | dl->devname = xstrdup(devname); |
a2b97981 | 3307 | else |
503975b9 | 3308 | dl->devname = xstrdup(name); |
cdddbdbc | 3309 | |
d23fe947 | 3310 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3311 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3312 | if (disk) { | |
3313 | dl->disk = *disk; | |
3314 | /* only set index on disks that are a member of a | |
3315 | * populated contianer, i.e. one with raid_devs | |
3316 | */ | |
3317 | if (is_failed(&dl->disk)) | |
3f6efecc | 3318 | dl->index = -2; |
a2b97981 DW |
3319 | else if (is_spare(&dl->disk)) |
3320 | dl->index = -1; | |
3f6efecc DW |
3321 | } |
3322 | ||
949c47a0 DW |
3323 | return 0; |
3324 | } | |
3325 | ||
0e600426 | 3326 | #ifndef MDASSEMBLE |
0c046afd DW |
3327 | /* When migrating map0 contains the 'destination' state while map1 |
3328 | * contains the current state. When not migrating map0 contains the | |
3329 | * current state. This routine assumes that map[0].map_state is set to | |
3330 | * the current array state before being called. | |
3331 | * | |
3332 | * Migration is indicated by one of the following states | |
3333 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3334 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3335 | * map1state=unitialized) |
1484e727 | 3336 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3337 | * map1state=normal) |
e3bba0e0 | 3338 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3339 | * map1state=degraded) |
8e59f3d8 AK |
3340 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3341 | * map1state=normal) | |
0c046afd | 3342 | */ |
8e59f3d8 AK |
3343 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3344 | __u8 to_state, int migr_type) | |
3393c6af | 3345 | { |
0c046afd | 3346 | struct imsm_map *dest; |
238c0a71 | 3347 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3348 | |
0c046afd | 3349 | dev->vol.migr_state = 1; |
1484e727 | 3350 | set_migr_type(dev, migr_type); |
f8f603f1 | 3351 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3352 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3353 | |
0556e1a2 | 3354 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3355 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3356 | if ((migr_type == MIGR_REBUILD) || |
3357 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3358 | __u32 ord; |
3359 | int i; | |
3360 | ||
3361 | for (i = 0; i < src->num_members; i++) { | |
3362 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3363 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3364 | } | |
3365 | } | |
3366 | ||
8e59f3d8 AK |
3367 | if (migr_type == MIGR_GEN_MIGR) |
3368 | /* Clear migration record */ | |
3369 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3370 | ||
0c046afd | 3371 | src->map_state = to_state; |
949c47a0 | 3372 | } |
f8f603f1 | 3373 | |
809da78e AK |
3374 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3375 | __u8 map_state) | |
f8f603f1 | 3376 | { |
238c0a71 AK |
3377 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3378 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3379 | MAP_0 : MAP_1); | |
28bce06f | 3380 | int i, j; |
0556e1a2 DW |
3381 | |
3382 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3383 | * completed in the last migration. | |
3384 | * | |
28bce06f | 3385 | * FIXME add support for raid-level-migration |
0556e1a2 | 3386 | */ |
809da78e AK |
3387 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3388 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3389 | /* when final map state is other than expected | |
3390 | * merge maps (not for migration) | |
3391 | */ | |
3392 | int failed; | |
3393 | ||
3394 | for (i = 0; i < prev->num_members; i++) | |
3395 | for (j = 0; j < map->num_members; j++) | |
3396 | /* during online capacity expansion | |
3397 | * disks position can be changed | |
3398 | * if takeover is used | |
3399 | */ | |
3400 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3401 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3402 | map->disk_ord_tbl[j] |= | |
3403 | prev->disk_ord_tbl[i]; | |
3404 | break; | |
3405 | } | |
3406 | failed = imsm_count_failed(super, dev, MAP_0); | |
3407 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3408 | } | |
f8f603f1 DW |
3409 | |
3410 | dev->vol.migr_state = 0; | |
ea672ee1 | 3411 | set_migr_type(dev, 0); |
f8f603f1 DW |
3412 | dev->vol.curr_migr_unit = 0; |
3413 | map->map_state = map_state; | |
3414 | } | |
0e600426 | 3415 | #endif |
949c47a0 DW |
3416 | |
3417 | static int parse_raid_devices(struct intel_super *super) | |
3418 | { | |
3419 | int i; | |
3420 | struct imsm_dev *dev_new; | |
4d7b1503 | 3421 | size_t len, len_migr; |
401d313b | 3422 | size_t max_len = 0; |
4d7b1503 DW |
3423 | size_t space_needed = 0; |
3424 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3425 | |
3426 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3427 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3428 | struct intel_dev *dv; |
949c47a0 | 3429 | |
4d7b1503 DW |
3430 | len = sizeof_imsm_dev(dev_iter, 0); |
3431 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3432 | if (len_migr > len) | |
3433 | space_needed += len_migr - len; | |
ca9de185 | 3434 | |
503975b9 | 3435 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
3436 | if (max_len < len_migr) |
3437 | max_len = len_migr; | |
3438 | if (max_len > len_migr) | |
3439 | space_needed += max_len - len_migr; | |
503975b9 | 3440 | dev_new = xmalloc(max_len); |
949c47a0 | 3441 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3442 | dv->dev = dev_new; |
3443 | dv->index = i; | |
3444 | dv->next = super->devlist; | |
3445 | super->devlist = dv; | |
949c47a0 | 3446 | } |
cdddbdbc | 3447 | |
4d7b1503 DW |
3448 | /* ensure that super->buf is large enough when all raid devices |
3449 | * are migrating | |
3450 | */ | |
3451 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3452 | void *buf; | |
3453 | ||
3454 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3455 | if (posix_memalign(&buf, 512, len) != 0) | |
3456 | return 1; | |
3457 | ||
1f45a8ad DW |
3458 | memcpy(buf, super->buf, super->len); |
3459 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3460 | free(super->buf); |
3461 | super->buf = buf; | |
3462 | super->len = len; | |
3463 | } | |
ca9de185 | 3464 | |
cdddbdbc DW |
3465 | return 0; |
3466 | } | |
3467 | ||
604b746f JD |
3468 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3469 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3470 | { | |
3471 | void *ptr = NULL; | |
3472 | ||
3473 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3474 | ptr = mpb; | |
3475 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3476 | } | |
3477 | ||
3478 | return ptr; | |
3479 | } | |
3480 | ||
e2f41b2c AK |
3481 | /******************************************************************************* |
3482 | * Function: check_mpb_migr_compatibility | |
3483 | * Description: Function checks for unsupported migration features: | |
3484 | * - migration optimization area (pba_of_lba0) | |
3485 | * - descending reshape (ascending_migr) | |
3486 | * Parameters: | |
3487 | * super : imsm metadata information | |
3488 | * Returns: | |
3489 | * 0 : migration is compatible | |
3490 | * -1 : migration is not compatible | |
3491 | ******************************************************************************/ | |
3492 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3493 | { | |
3494 | struct imsm_map *map0, *map1; | |
3495 | struct migr_record *migr_rec = super->migr_rec; | |
3496 | int i; | |
3497 | ||
3498 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3499 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3500 | ||
3501 | if (dev_iter && | |
3502 | dev_iter->vol.migr_state == 1 && | |
3503 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3504 | /* This device is migrating */ | |
238c0a71 AK |
3505 | map0 = get_imsm_map(dev_iter, MAP_0); |
3506 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 3507 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
3508 | /* migration optimization area was used */ |
3509 | return -1; | |
3510 | if (migr_rec->ascending_migr == 0 | |
3511 | && migr_rec->dest_depth_per_unit > 0) | |
3512 | /* descending reshape not supported yet */ | |
3513 | return -1; | |
3514 | } | |
3515 | } | |
3516 | return 0; | |
3517 | } | |
3518 | ||
d23fe947 | 3519 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3520 | |
cdddbdbc | 3521 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3522 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3523 | */ |
3524 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3525 | { | |
3526 | unsigned long long dsize; | |
cdddbdbc DW |
3527 | unsigned long long sectors; |
3528 | struct stat; | |
6416d527 | 3529 | struct imsm_super *anchor; |
cdddbdbc DW |
3530 | __u32 check_sum; |
3531 | ||
cdddbdbc | 3532 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3533 | if (dsize < 1024) { |
3534 | if (devname) | |
e7b84f9d N |
3535 | pr_err("%s: device to small for imsm\n", |
3536 | devname); | |
64436f06 N |
3537 | return 1; |
3538 | } | |
cdddbdbc DW |
3539 | |
3540 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3541 | if (devname) | |
e7b84f9d N |
3542 | pr_err("Cannot seek to anchor block on %s: %s\n", |
3543 | devname, strerror(errno)); | |
cdddbdbc DW |
3544 | return 1; |
3545 | } | |
3546 | ||
949c47a0 | 3547 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e | 3548 | if (devname) |
e7b84f9d N |
3549 | pr_err("Failed to allocate imsm anchor buffer" |
3550 | " on %s\n", devname); | |
ad97895e DW |
3551 | return 1; |
3552 | } | |
949c47a0 | 3553 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc | 3554 | if (devname) |
e7b84f9d N |
3555 | pr_err("Cannot read anchor block on %s: %s\n", |
3556 | devname, strerror(errno)); | |
6416d527 | 3557 | free(anchor); |
cdddbdbc DW |
3558 | return 1; |
3559 | } | |
3560 | ||
6416d527 | 3561 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 3562 | if (devname) |
e7b84f9d | 3563 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 3564 | free(anchor); |
cdddbdbc DW |
3565 | return 2; |
3566 | } | |
3567 | ||
d23fe947 | 3568 | __free_imsm(super, 0); |
f2f5c343 LM |
3569 | /* reload capability and hba */ |
3570 | ||
3571 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3572 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3573 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3574 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc | 3575 | if (devname) |
e7b84f9d N |
3576 | pr_err("unable to allocate %zu byte mpb buffer\n", |
3577 | super->len); | |
6416d527 | 3578 | free(anchor); |
cdddbdbc DW |
3579 | return 2; |
3580 | } | |
949c47a0 | 3581 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3582 | |
6416d527 NB |
3583 | sectors = mpb_sectors(anchor) - 1; |
3584 | free(anchor); | |
8e59f3d8 | 3585 | |
17a4eaf9 | 3586 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
e7b84f9d | 3587 | pr_err("%s could not allocate migr_rec buffer\n", __func__); |
8e59f3d8 AK |
3588 | free(super->buf); |
3589 | return 2; | |
3590 | } | |
51d83f5d | 3591 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 3592 | |
949c47a0 | 3593 | if (!sectors) { |
ecf45690 DW |
3594 | check_sum = __gen_imsm_checksum(super->anchor); |
3595 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3596 | if (devname) | |
e7b84f9d N |
3597 | pr_err("IMSM checksum %x != %x on %s\n", |
3598 | check_sum, | |
3599 | __le32_to_cpu(super->anchor->check_sum), | |
3600 | devname); | |
ecf45690 DW |
3601 | return 2; |
3602 | } | |
3603 | ||
a2b97981 | 3604 | return 0; |
949c47a0 | 3605 | } |
cdddbdbc DW |
3606 | |
3607 | /* read the extended mpb */ | |
3608 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3609 | if (devname) | |
e7b84f9d N |
3610 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
3611 | devname, strerror(errno)); | |
cdddbdbc DW |
3612 | return 1; |
3613 | } | |
3614 | ||
f21e18ca | 3615 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc | 3616 | if (devname) |
e7b84f9d N |
3617 | pr_err("Cannot read extended mpb on %s: %s\n", |
3618 | devname, strerror(errno)); | |
cdddbdbc DW |
3619 | return 2; |
3620 | } | |
3621 | ||
949c47a0 DW |
3622 | check_sum = __gen_imsm_checksum(super->anchor); |
3623 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 3624 | if (devname) |
e7b84f9d N |
3625 | pr_err("IMSM checksum %x != %x on %s\n", |
3626 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
3627 | devname); | |
db575f3b | 3628 | return 3; |
cdddbdbc DW |
3629 | } |
3630 | ||
604b746f JD |
3631 | /* FIXME the BBM log is disk specific so we cannot use this global |
3632 | * buffer for all disks. Ok for now since we only look at the global | |
3633 | * bbm_log_size parameter to gate assembly | |
3634 | */ | |
3635 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3636 | ||
a2b97981 DW |
3637 | return 0; |
3638 | } | |
3639 | ||
8e59f3d8 AK |
3640 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3641 | ||
97f81ee2 CA |
3642 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
3643 | static void clear_hi(struct intel_super *super) | |
3644 | { | |
3645 | struct imsm_super *mpb = super->anchor; | |
3646 | int i, n; | |
3647 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
3648 | return; | |
3649 | for (i = 0; i < mpb->num_disks; ++i) { | |
3650 | struct imsm_disk *disk = &mpb->disk[i]; | |
3651 | disk->total_blocks_hi = 0; | |
3652 | } | |
3653 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
3654 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3655 | if (!dev) | |
3656 | return; | |
3657 | for (n = 0; n < 2; ++n) { | |
3658 | struct imsm_map *map = get_imsm_map(dev, n); | |
3659 | if (!map) | |
3660 | continue; | |
3661 | map->pba_of_lba0_hi = 0; | |
3662 | map->blocks_per_member_hi = 0; | |
3663 | map->num_data_stripes_hi = 0; | |
3664 | } | |
3665 | } | |
3666 | } | |
3667 | ||
a2b97981 DW |
3668 | static int |
3669 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3670 | { | |
3671 | int err; | |
3672 | ||
3673 | err = load_imsm_mpb(fd, super, devname); | |
3674 | if (err) | |
3675 | return err; | |
3676 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3677 | if (err) | |
3678 | return err; | |
3679 | err = parse_raid_devices(super); | |
97f81ee2 | 3680 | clear_hi(super); |
a2b97981 | 3681 | return err; |
cdddbdbc DW |
3682 | } |
3683 | ||
ae6aad82 DW |
3684 | static void __free_imsm_disk(struct dl *d) |
3685 | { | |
3686 | if (d->fd >= 0) | |
3687 | close(d->fd); | |
3688 | if (d->devname) | |
3689 | free(d->devname); | |
0dcecb2e DW |
3690 | if (d->e) |
3691 | free(d->e); | |
ae6aad82 DW |
3692 | free(d); |
3693 | ||
3694 | } | |
1a64be56 | 3695 | |
cdddbdbc DW |
3696 | static void free_imsm_disks(struct intel_super *super) |
3697 | { | |
47ee5a45 | 3698 | struct dl *d; |
cdddbdbc | 3699 | |
47ee5a45 DW |
3700 | while (super->disks) { |
3701 | d = super->disks; | |
cdddbdbc | 3702 | super->disks = d->next; |
ae6aad82 | 3703 | __free_imsm_disk(d); |
cdddbdbc | 3704 | } |
cb82edca AK |
3705 | while (super->disk_mgmt_list) { |
3706 | d = super->disk_mgmt_list; | |
3707 | super->disk_mgmt_list = d->next; | |
3708 | __free_imsm_disk(d); | |
3709 | } | |
47ee5a45 DW |
3710 | while (super->missing) { |
3711 | d = super->missing; | |
3712 | super->missing = d->next; | |
3713 | __free_imsm_disk(d); | |
3714 | } | |
3715 | ||
cdddbdbc DW |
3716 | } |
3717 | ||
9ca2c81c | 3718 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3719 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3720 | { |
88654014 LM |
3721 | struct intel_hba *elem, *next; |
3722 | ||
9ca2c81c | 3723 | if (super->buf) { |
949c47a0 | 3724 | free(super->buf); |
9ca2c81c DW |
3725 | super->buf = NULL; |
3726 | } | |
f2f5c343 LM |
3727 | /* unlink capability description */ |
3728 | super->orom = NULL; | |
8e59f3d8 AK |
3729 | if (super->migr_rec_buf) { |
3730 | free(super->migr_rec_buf); | |
3731 | super->migr_rec_buf = NULL; | |
3732 | } | |
d23fe947 DW |
3733 | if (free_disks) |
3734 | free_imsm_disks(super); | |
ba2de7ba | 3735 | free_devlist(super); |
88654014 LM |
3736 | elem = super->hba; |
3737 | while (elem) { | |
3738 | if (elem->path) | |
3739 | free((void *)elem->path); | |
3740 | next = elem->next; | |
3741 | free(elem); | |
3742 | elem = next; | |
88c32bb1 | 3743 | } |
88654014 | 3744 | super->hba = NULL; |
cdddbdbc DW |
3745 | } |
3746 | ||
9ca2c81c DW |
3747 | static void free_imsm(struct intel_super *super) |
3748 | { | |
d23fe947 | 3749 | __free_imsm(super, 1); |
9ca2c81c DW |
3750 | free(super); |
3751 | } | |
cdddbdbc DW |
3752 | |
3753 | static void free_super_imsm(struct supertype *st) | |
3754 | { | |
3755 | struct intel_super *super = st->sb; | |
3756 | ||
3757 | if (!super) | |
3758 | return; | |
3759 | ||
3760 | free_imsm(super); | |
3761 | st->sb = NULL; | |
3762 | } | |
3763 | ||
49133e57 | 3764 | static struct intel_super *alloc_super(void) |
c2c087e6 | 3765 | { |
503975b9 | 3766 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 3767 | |
503975b9 N |
3768 | super->current_vol = -1; |
3769 | super->create_offset = ~((unsigned long long) 0); | |
c2c087e6 DW |
3770 | return super; |
3771 | } | |
3772 | ||
f0f5a016 LM |
3773 | /* |
3774 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3775 | */ | |
d424212e | 3776 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3777 | { |
3778 | struct sys_dev *hba_name; | |
3779 | int rv = 0; | |
3780 | ||
3781 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3782 | super->orom = NULL; |
f0f5a016 LM |
3783 | super->hba = NULL; |
3784 | return 0; | |
3785 | } | |
3786 | hba_name = find_disk_attached_hba(fd, NULL); | |
3787 | if (!hba_name) { | |
d424212e | 3788 | if (devname) |
e7b84f9d N |
3789 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
3790 | devname); | |
f0f5a016 LM |
3791 | return 1; |
3792 | } | |
3793 | rv = attach_hba_to_super(super, hba_name); | |
3794 | if (rv == 2) { | |
d424212e N |
3795 | if (devname) { |
3796 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3797 | |
e7b84f9d | 3798 | pr_err("%s is attached to Intel(R) %s RAID " |
f0f5a016 LM |
3799 | "controller (%s),\n" |
3800 | " but the container is assigned to Intel(R) " | |
3801 | "%s RAID controller (", | |
d424212e | 3802 | devname, |
f0f5a016 LM |
3803 | hba_name->path, |
3804 | hba_name->pci_id ? : "Err!", | |
3805 | get_sys_dev_type(hba_name->type)); | |
3806 | ||
f0f5a016 LM |
3807 | while (hba) { |
3808 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3809 | if (hba->next) | |
3810 | fprintf(stderr, ", "); | |
3811 | hba = hba->next; | |
3812 | } | |
3813 | ||
3814 | fprintf(stderr, ").\n" | |
3815 | " Mixing devices attached to different controllers " | |
3816 | "is not allowed.\n"); | |
3817 | } | |
3818 | free_sys_dev(&hba_name); | |
3819 | return 2; | |
3820 | } | |
f2f5c343 | 3821 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3822 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3823 | if (!super->orom) |
3824 | return 3; | |
f0f5a016 LM |
3825 | return 0; |
3826 | } | |
3827 | ||
47ee5a45 DW |
3828 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3829 | * disks that have disappeared from the system. This routine relies on | |
3830 | * the mpb being uptodate, which it is at load time. | |
3831 | */ | |
3832 | static int find_missing(struct intel_super *super) | |
3833 | { | |
3834 | int i; | |
3835 | struct imsm_super *mpb = super->anchor; | |
3836 | struct dl *dl; | |
3837 | struct imsm_disk *disk; | |
47ee5a45 DW |
3838 | |
3839 | for (i = 0; i < mpb->num_disks; i++) { | |
3840 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3841 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3842 | if (dl) |
3843 | continue; | |
47ee5a45 | 3844 | |
503975b9 | 3845 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
3846 | dl->major = 0; |
3847 | dl->minor = 0; | |
3848 | dl->fd = -1; | |
503975b9 | 3849 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
3850 | dl->index = i; |
3851 | serialcpy(dl->serial, disk->serial); | |
3852 | dl->disk = *disk; | |
689c9bf3 | 3853 | dl->e = NULL; |
47ee5a45 DW |
3854 | dl->next = super->missing; |
3855 | super->missing = dl; | |
3856 | } | |
3857 | ||
3858 | return 0; | |
3859 | } | |
3860 | ||
3960e579 | 3861 | #ifndef MDASSEMBLE |
a2b97981 DW |
3862 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3863 | { | |
3864 | struct intel_disk *idisk = disk_list; | |
3865 | ||
3866 | while (idisk) { | |
3867 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3868 | break; | |
3869 | idisk = idisk->next; | |
3870 | } | |
3871 | ||
3872 | return idisk; | |
3873 | } | |
3874 | ||
3875 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3876 | struct intel_super *super, | |
3877 | struct intel_disk **disk_list) | |
3878 | { | |
3879 | struct imsm_disk *d = &super->disks->disk; | |
3880 | struct imsm_super *mpb = super->anchor; | |
3881 | int i, j; | |
3882 | ||
3883 | for (i = 0; i < tbl_size; i++) { | |
3884 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3885 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3886 | ||
3887 | if (tbl_mpb->family_num == mpb->family_num) { | |
3888 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3889 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3890 | __func__, super->disks->major, | |
3891 | super->disks->minor, | |
3892 | table[i]->disks->major, | |
3893 | table[i]->disks->minor); | |
3894 | break; | |
3895 | } | |
3896 | ||
3897 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3898 | is_configured(d) == is_configured(tbl_d)) && | |
3899 | tbl_mpb->generation_num < mpb->generation_num) { | |
3900 | /* current version of the mpb is a | |
3901 | * better candidate than the one in | |
3902 | * super_table, but copy over "cross | |
3903 | * generational" status | |
3904 | */ | |
3905 | struct intel_disk *idisk; | |
3906 | ||
3907 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3908 | __func__, super->disks->major, | |
3909 | super->disks->minor, | |
3910 | table[i]->disks->major, | |
3911 | table[i]->disks->minor); | |
3912 | ||
3913 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3914 | if (idisk && is_failed(&idisk->disk)) | |
3915 | tbl_d->status |= FAILED_DISK; | |
3916 | break; | |
3917 | } else { | |
3918 | struct intel_disk *idisk; | |
3919 | struct imsm_disk *disk; | |
3920 | ||
3921 | /* tbl_mpb is more up to date, but copy | |
3922 | * over cross generational status before | |
3923 | * returning | |
3924 | */ | |
3925 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3926 | if (disk && is_failed(disk)) | |
3927 | d->status |= FAILED_DISK; | |
3928 | ||
3929 | idisk = disk_list_get(d->serial, *disk_list); | |
3930 | if (idisk) { | |
3931 | idisk->owner = i; | |
3932 | if (disk && is_configured(disk)) | |
3933 | idisk->disk.status |= CONFIGURED_DISK; | |
3934 | } | |
3935 | ||
3936 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3937 | __func__, super->disks->major, | |
3938 | super->disks->minor, | |
3939 | table[i]->disks->major, | |
3940 | table[i]->disks->minor); | |
3941 | ||
3942 | return tbl_size; | |
3943 | } | |
3944 | } | |
3945 | } | |
3946 | ||
3947 | if (i >= tbl_size) | |
3948 | table[tbl_size++] = super; | |
3949 | else | |
3950 | table[i] = super; | |
3951 | ||
3952 | /* update/extend the merged list of imsm_disk records */ | |
3953 | for (j = 0; j < mpb->num_disks; j++) { | |
3954 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3955 | struct intel_disk *idisk; | |
3956 | ||
3957 | idisk = disk_list_get(disk->serial, *disk_list); | |
3958 | if (idisk) { | |
3959 | idisk->disk.status |= disk->status; | |
3960 | if (is_configured(&idisk->disk) || | |
3961 | is_failed(&idisk->disk)) | |
3962 | idisk->disk.status &= ~(SPARE_DISK); | |
3963 | } else { | |
503975b9 | 3964 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
3965 | idisk->owner = IMSM_UNKNOWN_OWNER; |
3966 | idisk->disk = *disk; | |
3967 | idisk->next = *disk_list; | |
3968 | *disk_list = idisk; | |
3969 | } | |
3970 | ||
3971 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3972 | idisk->owner = i; | |
3973 | } | |
3974 | ||
3975 | return tbl_size; | |
3976 | } | |
3977 | ||
3978 | static struct intel_super * | |
3979 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3980 | const int owner) | |
3981 | { | |
3982 | struct imsm_super *mpb = super->anchor; | |
3983 | int ok_count = 0; | |
3984 | int i; | |
3985 | ||
3986 | for (i = 0; i < mpb->num_disks; i++) { | |
3987 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3988 | struct intel_disk *idisk; | |
3989 | ||
3990 | idisk = disk_list_get(disk->serial, disk_list); | |
3991 | if (idisk) { | |
3992 | if (idisk->owner == owner || | |
3993 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3994 | ok_count++; | |
3995 | else | |
3996 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3997 | __func__, disk->serial, idisk->owner, | |
3998 | owner); | |
3999 | } else { | |
4000 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
4001 | __func__, __le32_to_cpu(mpb->family_num), i, | |
4002 | disk->serial); | |
4003 | break; | |
4004 | } | |
4005 | } | |
4006 | ||
4007 | if (ok_count == mpb->num_disks) | |
4008 | return super; | |
4009 | return NULL; | |
4010 | } | |
4011 | ||
4012 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4013 | { | |
4014 | struct intel_super *s; | |
4015 | ||
4016 | for (s = super_list; s; s = s->next) { | |
4017 | if (family_num != s->anchor->family_num) | |
4018 | continue; | |
4019 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
4020 | __le32_to_cpu(family_num), s->disks->devname); | |
4021 | } | |
4022 | } | |
4023 | ||
4024 | static struct intel_super * | |
4025 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4026 | { | |
4027 | struct intel_super *super_table[len]; | |
4028 | struct intel_disk *disk_list = NULL; | |
4029 | struct intel_super *champion, *spare; | |
4030 | struct intel_super *s, **del; | |
4031 | int tbl_size = 0; | |
4032 | int conflict; | |
4033 | int i; | |
4034 | ||
4035 | memset(super_table, 0, sizeof(super_table)); | |
4036 | for (s = *super_list; s; s = s->next) | |
4037 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4038 | ||
4039 | for (i = 0; i < tbl_size; i++) { | |
4040 | struct imsm_disk *d; | |
4041 | struct intel_disk *idisk; | |
4042 | struct imsm_super *mpb = super_table[i]->anchor; | |
4043 | ||
4044 | s = super_table[i]; | |
4045 | d = &s->disks->disk; | |
4046 | ||
4047 | /* 'd' must appear in merged disk list for its | |
4048 | * configuration to be valid | |
4049 | */ | |
4050 | idisk = disk_list_get(d->serial, disk_list); | |
4051 | if (idisk && idisk->owner == i) | |
4052 | s = validate_members(s, disk_list, i); | |
4053 | else | |
4054 | s = NULL; | |
4055 | ||
4056 | if (!s) | |
4057 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
4058 | __func__, mpb->family_num, | |
4059 | super_table[i]->disks->major, | |
4060 | super_table[i]->disks->minor); | |
4061 | super_table[i] = s; | |
4062 | } | |
4063 | ||
4064 | /* This is where the mdadm implementation differs from the Windows | |
4065 | * driver which has no strict concept of a container. We can only | |
4066 | * assemble one family from a container, so when returning a prodigal | |
4067 | * array member to this system the code will not be able to disambiguate | |
4068 | * the container contents that should be assembled ("foreign" versus | |
4069 | * "local"). It requires user intervention to set the orig_family_num | |
4070 | * to a new value to establish a new container. The Windows driver in | |
4071 | * this situation fixes up the volume name in place and manages the | |
4072 | * foreign array as an independent entity. | |
4073 | */ | |
4074 | s = NULL; | |
4075 | spare = NULL; | |
4076 | conflict = 0; | |
4077 | for (i = 0; i < tbl_size; i++) { | |
4078 | struct intel_super *tbl_ent = super_table[i]; | |
4079 | int is_spare = 0; | |
4080 | ||
4081 | if (!tbl_ent) | |
4082 | continue; | |
4083 | ||
4084 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4085 | spare = tbl_ent; | |
4086 | is_spare = 1; | |
4087 | } | |
4088 | ||
4089 | if (s && !is_spare) { | |
4090 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4091 | conflict++; | |
4092 | } else if (!s && !is_spare) | |
4093 | s = tbl_ent; | |
4094 | } | |
4095 | ||
4096 | if (!s) | |
4097 | s = spare; | |
4098 | if (!s) { | |
4099 | champion = NULL; | |
4100 | goto out; | |
4101 | } | |
4102 | champion = s; | |
4103 | ||
4104 | if (conflict) | |
4105 | fprintf(stderr, "Chose family %#x on '%s', " | |
4106 | "assemble conflicts to new container with '--update=uuid'\n", | |
4107 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
4108 | ||
4109 | /* collect all dl's onto 'champion', and update them to | |
4110 | * champion's version of the status | |
4111 | */ | |
4112 | for (s = *super_list; s; s = s->next) { | |
4113 | struct imsm_super *mpb = champion->anchor; | |
4114 | struct dl *dl = s->disks; | |
4115 | ||
4116 | if (s == champion) | |
4117 | continue; | |
4118 | ||
5d7b407a CA |
4119 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4120 | ||
a2b97981 DW |
4121 | for (i = 0; i < mpb->num_disks; i++) { |
4122 | struct imsm_disk *disk; | |
4123 | ||
4124 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4125 | if (disk) { | |
4126 | dl->disk = *disk; | |
4127 | /* only set index on disks that are a member of | |
4128 | * a populated contianer, i.e. one with | |
4129 | * raid_devs | |
4130 | */ | |
4131 | if (is_failed(&dl->disk)) | |
4132 | dl->index = -2; | |
4133 | else if (is_spare(&dl->disk)) | |
4134 | dl->index = -1; | |
4135 | break; | |
4136 | } | |
4137 | } | |
4138 | ||
4139 | if (i >= mpb->num_disks) { | |
4140 | struct intel_disk *idisk; | |
4141 | ||
4142 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4143 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4144 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4145 | dl->index = -1; | |
4146 | else { | |
4147 | dl->index = -2; | |
4148 | continue; | |
4149 | } | |
4150 | } | |
4151 | ||
4152 | dl->next = champion->disks; | |
4153 | champion->disks = dl; | |
4154 | s->disks = NULL; | |
4155 | } | |
4156 | ||
4157 | /* delete 'champion' from super_list */ | |
4158 | for (del = super_list; *del; ) { | |
4159 | if (*del == champion) { | |
4160 | *del = (*del)->next; | |
4161 | break; | |
4162 | } else | |
4163 | del = &(*del)->next; | |
4164 | } | |
4165 | champion->next = NULL; | |
4166 | ||
4167 | out: | |
4168 | while (disk_list) { | |
4169 | struct intel_disk *idisk = disk_list; | |
4170 | ||
4171 | disk_list = disk_list->next; | |
4172 | free(idisk); | |
4173 | } | |
4174 | ||
4175 | return champion; | |
4176 | } | |
4177 | ||
9587c373 LM |
4178 | |
4179 | static int | |
4180 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
9587c373 LM |
4181 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, |
4182 | int major, int minor, int keep_fd); | |
ec50f7b6 LM |
4183 | static int |
4184 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4185 | int *max, int keep_fd); | |
4186 | ||
9587c373 | 4187 | |
cdddbdbc | 4188 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4189 | char *devname, struct md_list *devlist, |
4190 | int keep_fd) | |
cdddbdbc | 4191 | { |
a2b97981 DW |
4192 | struct intel_super *super_list = NULL; |
4193 | struct intel_super *super = NULL; | |
a2b97981 | 4194 | int err = 0; |
9587c373 | 4195 | int i = 0; |
dab4a513 | 4196 | |
9587c373 LM |
4197 | if (fd >= 0) |
4198 | /* 'fd' is an opened container */ | |
4199 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4200 | else | |
ec50f7b6 LM |
4201 | /* get super block from devlist devices */ |
4202 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4203 | if (err) |
1602d52c | 4204 | goto error; |
a2b97981 DW |
4205 | /* all mpbs enter, maybe one leaves */ |
4206 | super = imsm_thunderdome(&super_list, i); | |
4207 | if (!super) { | |
4208 | err = 1; | |
4209 | goto error; | |
cdddbdbc DW |
4210 | } |
4211 | ||
47ee5a45 DW |
4212 | if (find_missing(super) != 0) { |
4213 | free_imsm(super); | |
a2b97981 DW |
4214 | err = 2; |
4215 | goto error; | |
47ee5a45 | 4216 | } |
8e59f3d8 AK |
4217 | |
4218 | /* load migration record */ | |
4219 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4220 | if (err == -1) { |
4221 | /* migration is in progress, | |
4222 | * but migr_rec cannot be loaded, | |
4223 | */ | |
8e59f3d8 AK |
4224 | err = 4; |
4225 | goto error; | |
4226 | } | |
e2f41b2c AK |
4227 | |
4228 | /* Check migration compatibility */ | |
4c965cc9 | 4229 | if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) { |
e7b84f9d | 4230 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4231 | if (devname) |
4232 | fprintf(stderr, " on %s\n", devname); | |
4233 | else | |
4234 | fprintf(stderr, " (IMSM).\n"); | |
4235 | ||
4236 | err = 5; | |
4237 | goto error; | |
4238 | } | |
4239 | ||
a2b97981 DW |
4240 | err = 0; |
4241 | ||
4242 | error: | |
4243 | while (super_list) { | |
4244 | struct intel_super *s = super_list; | |
4245 | ||
4246 | super_list = super_list->next; | |
4247 | free_imsm(s); | |
4248 | } | |
9587c373 | 4249 | |
a2b97981 DW |
4250 | |
4251 | if (err) | |
4252 | return err; | |
f7e7067b | 4253 | |
cdddbdbc | 4254 | *sbp = super; |
9587c373 LM |
4255 | if (fd >= 0) |
4256 | st->container_dev = fd2devnum(fd); | |
4257 | else | |
4258 | st->container_dev = NoMdDev; | |
a2b97981 | 4259 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4260 | st->ss = &super_imsm; |
cdddbdbc DW |
4261 | st->minor_version = 0; |
4262 | st->max_devs = IMSM_MAX_DEVICES; | |
4263 | } | |
cdddbdbc DW |
4264 | return 0; |
4265 | } | |
2b959fbf | 4266 | |
9587c373 | 4267 | |
ec50f7b6 LM |
4268 | static int |
4269 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4270 | int *max, int keep_fd) | |
4271 | { | |
4272 | struct md_list *tmpdev; | |
4273 | int err = 0; | |
4274 | int i = 0; | |
9587c373 | 4275 | |
ec50f7b6 LM |
4276 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4277 | if (tmpdev->used != 1) | |
4278 | continue; | |
4279 | if (tmpdev->container == 1) { | |
ca9de185 | 4280 | int lmax = 0; |
ec50f7b6 LM |
4281 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4282 | if (fd < 0) { | |
e7b84f9d | 4283 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4284 | tmpdev->devname, strerror(errno)); |
4285 | err = 8; | |
4286 | goto error; | |
4287 | } | |
4288 | err = get_sra_super_block(fd, super_list, | |
4289 | tmpdev->devname, &lmax, | |
4290 | keep_fd); | |
4291 | i += lmax; | |
4292 | close(fd); | |
4293 | if (err) { | |
4294 | err = 7; | |
4295 | goto error; | |
4296 | } | |
4297 | } else { | |
4298 | int major = major(tmpdev->st_rdev); | |
4299 | int minor = minor(tmpdev->st_rdev); | |
4300 | err = get_super_block(super_list, | |
4301 | -1, | |
4302 | tmpdev->devname, | |
4303 | major, minor, | |
4304 | keep_fd); | |
4305 | i++; | |
4306 | if (err) { | |
4307 | err = 6; | |
4308 | goto error; | |
4309 | } | |
4310 | } | |
4311 | } | |
4312 | error: | |
4313 | *max = i; | |
4314 | return err; | |
4315 | } | |
9587c373 LM |
4316 | |
4317 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, | |
4318 | int major, int minor, int keep_fd) | |
4319 | { | |
4320 | struct intel_super*s = NULL; | |
4321 | char nm[32]; | |
4322 | int dfd = -1; | |
4323 | int rv; | |
4324 | int err = 0; | |
4325 | int retry; | |
4326 | ||
4327 | s = alloc_super(); | |
4328 | if (!s) { | |
4329 | err = 1; | |
4330 | goto error; | |
4331 | } | |
4332 | ||
4333 | sprintf(nm, "%d:%d", major, minor); | |
4334 | dfd = dev_open(nm, O_RDWR); | |
4335 | if (dfd < 0) { | |
4336 | err = 2; | |
4337 | goto error; | |
4338 | } | |
4339 | ||
4340 | rv = find_intel_hba_capability(dfd, s, devname); | |
4341 | /* no orom/efi or non-intel hba of the disk */ | |
4342 | if (rv != 0) { | |
4343 | err = 4; | |
4344 | goto error; | |
4345 | } | |
4346 | ||
4347 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4348 | ||
4349 | /* retry the load if we might have raced against mdmon */ | |
4350 | if (err == 3 && (devnum != -1) && mdmon_running(devnum)) | |
4351 | for (retry = 0; retry < 3; retry++) { | |
4352 | usleep(3000); | |
4353 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4354 | if (err != 3) | |
4355 | break; | |
4356 | } | |
4357 | error: | |
4358 | if (!err) { | |
4359 | s->next = *super_list; | |
4360 | *super_list = s; | |
4361 | } else { | |
4362 | if (s) | |
4363 | free(s); | |
36614e95 | 4364 | if (dfd >= 0) |
9587c373 LM |
4365 | close(dfd); |
4366 | } | |
4367 | if ((dfd >= 0) && (!keep_fd)) | |
4368 | close(dfd); | |
4369 | return err; | |
4370 | ||
4371 | } | |
4372 | ||
4373 | static int | |
4374 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4375 | { | |
4376 | struct mdinfo *sra; | |
4377 | int devnum; | |
4378 | struct mdinfo *sd; | |
4379 | int err = 0; | |
4380 | int i = 0; | |
4381 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
4382 | if (!sra) | |
4383 | return 1; | |
4384 | ||
4385 | if (sra->array.major_version != -1 || | |
4386 | sra->array.minor_version != -2 || | |
4387 | strcmp(sra->text_version, "imsm") != 0) { | |
4388 | err = 1; | |
4389 | goto error; | |
4390 | } | |
4391 | /* load all mpbs */ | |
4392 | devnum = fd2devnum(fd); | |
4393 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
4394 | if (get_super_block(super_list, devnum, devname, | |
4395 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { | |
4396 | err = 7; | |
4397 | goto error; | |
4398 | } | |
4399 | } | |
4400 | error: | |
4401 | sysfs_free(sra); | |
4402 | *max = i; | |
4403 | return err; | |
4404 | } | |
4405 | ||
2b959fbf N |
4406 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4407 | { | |
ec50f7b6 | 4408 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4409 | } |
cdddbdbc DW |
4410 | #endif |
4411 | ||
4412 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4413 | { | |
4414 | struct intel_super *super; | |
4415 | int rv; | |
4416 | ||
691c6ee1 N |
4417 | if (test_partition(fd)) |
4418 | /* IMSM not allowed on partitions */ | |
4419 | return 1; | |
4420 | ||
37424f13 DW |
4421 | free_super_imsm(st); |
4422 | ||
49133e57 | 4423 | super = alloc_super(); |
ea2bc72b LM |
4424 | /* Load hba and capabilities if they exist. |
4425 | * But do not preclude loading metadata in case capabilities or hba are | |
4426 | * non-compliant and ignore_hw_compat is set. | |
4427 | */ | |
d424212e | 4428 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4429 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4430 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 | 4431 | if (devname) |
e7b84f9d | 4432 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
4433 | free_imsm(super); |
4434 | return 2; | |
4435 | } | |
a2b97981 | 4436 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
4437 | |
4438 | if (rv) { | |
4439 | if (devname) | |
e7b84f9d N |
4440 | pr_err("Failed to load all information " |
4441 | "sections on %s\n", devname); | |
cdddbdbc DW |
4442 | free_imsm(super); |
4443 | return rv; | |
4444 | } | |
4445 | ||
4446 | st->sb = super; | |
4447 | if (st->ss == NULL) { | |
4448 | st->ss = &super_imsm; | |
4449 | st->minor_version = 0; | |
4450 | st->max_devs = IMSM_MAX_DEVICES; | |
4451 | } | |
8e59f3d8 AK |
4452 | |
4453 | /* load migration record */ | |
2e062e82 AK |
4454 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4455 | /* Check for unsupported migration features */ | |
4456 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 4457 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
4458 | if (devname) |
4459 | fprintf(stderr, " on %s\n", devname); | |
4460 | else | |
4461 | fprintf(stderr, " (IMSM).\n"); | |
4462 | return 3; | |
4463 | } | |
e2f41b2c AK |
4464 | } |
4465 | ||
cdddbdbc DW |
4466 | return 0; |
4467 | } | |
4468 | ||
ef6ffade DW |
4469 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4470 | { | |
4471 | if (info->level == 1) | |
4472 | return 128; | |
4473 | return info->chunk_size >> 9; | |
4474 | } | |
4475 | ||
5551b113 CA |
4476 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
4477 | unsigned long long size) | |
fcfd9599 | 4478 | { |
4025c288 | 4479 | if (info->level == 1) |
5551b113 | 4480 | return size * 2; |
4025c288 | 4481 | else |
5551b113 | 4482 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
4483 | } |
4484 | ||
4d1313e9 DW |
4485 | static void imsm_update_version_info(struct intel_super *super) |
4486 | { | |
4487 | /* update the version and attributes */ | |
4488 | struct imsm_super *mpb = super->anchor; | |
4489 | char *version; | |
4490 | struct imsm_dev *dev; | |
4491 | struct imsm_map *map; | |
4492 | int i; | |
4493 | ||
4494 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4495 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4496 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4497 | if (__le32_to_cpu(dev->size_high) > 0) |
4498 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4499 | ||
4500 | /* FIXME detect when an array spans a port multiplier */ | |
4501 | #if 0 | |
4502 | mpb->attributes |= MPB_ATTRIB_PM; | |
4503 | #endif | |
4504 | ||
4505 | if (mpb->num_raid_devs > 1 || | |
4506 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4507 | version = MPB_VERSION_ATTRIBS; | |
4508 | switch (get_imsm_raid_level(map)) { | |
4509 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4510 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4511 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4512 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4513 | } | |
4514 | } else { | |
4515 | if (map->num_members >= 5) | |
4516 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4517 | else if (dev->status == DEV_CLONE_N_GO) | |
4518 | version = MPB_VERSION_CNG; | |
4519 | else if (get_imsm_raid_level(map) == 5) | |
4520 | version = MPB_VERSION_RAID5; | |
4521 | else if (map->num_members >= 3) | |
4522 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4523 | else if (get_imsm_raid_level(map) == 1) | |
4524 | version = MPB_VERSION_RAID1; | |
4525 | else | |
4526 | version = MPB_VERSION_RAID0; | |
4527 | } | |
4528 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4529 | } | |
4530 | } | |
4531 | ||
aa534678 DW |
4532 | static int check_name(struct intel_super *super, char *name, int quiet) |
4533 | { | |
4534 | struct imsm_super *mpb = super->anchor; | |
4535 | char *reason = NULL; | |
4536 | int i; | |
4537 | ||
4538 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4539 | reason = "must be 16 characters or less"; | |
4540 | ||
4541 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4542 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4543 | ||
4544 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4545 | reason = "already exists"; | |
4546 | break; | |
4547 | } | |
4548 | } | |
4549 | ||
4550 | if (reason && !quiet) | |
e7b84f9d | 4551 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
4552 | |
4553 | return !reason; | |
4554 | } | |
4555 | ||
8b353278 DW |
4556 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4557 | unsigned long long size, char *name, | |
4558 | char *homehost, int *uuid) | |
cdddbdbc | 4559 | { |
c2c087e6 DW |
4560 | /* We are creating a volume inside a pre-existing container. |
4561 | * so st->sb is already set. | |
4562 | */ | |
4563 | struct intel_super *super = st->sb; | |
949c47a0 | 4564 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4565 | struct intel_dev *dv; |
c2c087e6 DW |
4566 | struct imsm_dev *dev; |
4567 | struct imsm_vol *vol; | |
4568 | struct imsm_map *map; | |
4569 | int idx = mpb->num_raid_devs; | |
4570 | int i; | |
4571 | unsigned long long array_blocks; | |
2c092cad | 4572 | size_t size_old, size_new; |
5551b113 | 4573 | unsigned long long num_data_stripes; |
cdddbdbc | 4574 | |
88c32bb1 | 4575 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
e7b84f9d | 4576 | pr_err("This imsm-container already has the " |
88c32bb1 | 4577 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4578 | return 0; |
4579 | } | |
4580 | ||
2c092cad DW |
4581 | /* ensure the mpb is large enough for the new data */ |
4582 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4583 | size_new = disks_to_mpb_size(info->nr_disks); | |
4584 | if (size_new > size_old) { | |
4585 | void *mpb_new; | |
4586 | size_t size_round = ROUND_UP(size_new, 512); | |
4587 | ||
4588 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
e7b84f9d | 4589 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
4590 | return 0; |
4591 | } | |
17a4eaf9 AK |
4592 | if (posix_memalign(&super->migr_rec_buf, 512, |
4593 | MIGR_REC_BUF_SIZE) != 0) { | |
e7b84f9d N |
4594 | pr_err("%s could not allocate migr_rec buffer\n", |
4595 | __func__); | |
8e59f3d8 AK |
4596 | free(super->buf); |
4597 | free(super); | |
ea944c8f | 4598 | free(mpb_new); |
8e59f3d8 AK |
4599 | return 0; |
4600 | } | |
2c092cad DW |
4601 | memcpy(mpb_new, mpb, size_old); |
4602 | free(mpb); | |
4603 | mpb = mpb_new; | |
949c47a0 | 4604 | super->anchor = mpb_new; |
2c092cad DW |
4605 | mpb->mpb_size = __cpu_to_le32(size_new); |
4606 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4607 | } | |
bf5a934a | 4608 | super->current_vol = idx; |
3960e579 DW |
4609 | |
4610 | /* handle 'failed_disks' by either: | |
4611 | * a) create dummy disk entries in the table if this the first | |
4612 | * volume in the array. We add them here as this is the only | |
4613 | * opportunity to add them. add_to_super_imsm_volume() | |
4614 | * handles the non-failed disks and continues incrementing | |
4615 | * mpb->num_disks. | |
4616 | * b) validate that 'failed_disks' matches the current number | |
4617 | * of missing disks if the container is populated | |
d23fe947 | 4618 | */ |
3960e579 | 4619 | if (super->current_vol == 0) { |
d23fe947 | 4620 | mpb->num_disks = 0; |
3960e579 DW |
4621 | for (i = 0; i < info->failed_disks; i++) { |
4622 | struct imsm_disk *disk; | |
4623 | ||
4624 | mpb->num_disks++; | |
4625 | disk = __get_imsm_disk(mpb, i); | |
4626 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4627 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4628 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4629 | "missing:%d", i); | |
4630 | } | |
4631 | find_missing(super); | |
4632 | } else { | |
4633 | int missing = 0; | |
4634 | struct dl *d; | |
4635 | ||
4636 | for (d = super->missing; d; d = d->next) | |
4637 | missing++; | |
4638 | if (info->failed_disks > missing) { | |
e7b84f9d | 4639 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
4640 | return 0; |
4641 | } | |
4642 | } | |
5a038140 | 4643 | |
aa534678 DW |
4644 | if (!check_name(super, name, 0)) |
4645 | return 0; | |
503975b9 N |
4646 | dv = xmalloc(sizeof(*dv)); |
4647 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 4648 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 4649 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 4650 | info->layout, info->chunk_size, |
5551b113 | 4651 | size * 2); |
979d38be DW |
4652 | /* round array size down to closest MB */ |
4653 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4654 | ||
c2c087e6 DW |
4655 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4656 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4657 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4658 | vol = &dev->vol; |
4659 | vol->migr_state = 0; | |
1484e727 | 4660 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4661 | vol->dirty = !info->state; |
f8f603f1 | 4662 | vol->curr_migr_unit = 0; |
238c0a71 | 4663 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
4664 | set_pba_of_lba0(map, super->create_offset); |
4665 | set_blocks_per_member(map, info_to_blocks_per_member(info, size)); | |
ef6ffade | 4666 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4667 | map->failed_disk_num = ~0; |
bf4442ab AK |
4668 | if (info->level > 0) |
4669 | map->map_state = IMSM_T_STATE_UNINITIALIZED; | |
4670 | else | |
4671 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4672 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4673 | map->ddf = 1; |
ef6ffade DW |
4674 | |
4675 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4676 | free(dev); |
4677 | free(dv); | |
e7b84f9d | 4678 | pr_err("imsm does not support more than 2 disks" |
ef6ffade DW |
4679 | "in a raid1 volume\n"); |
4680 | return 0; | |
4681 | } | |
81062a36 DW |
4682 | |
4683 | map->raid_level = info->level; | |
4d1313e9 | 4684 | if (info->level == 10) { |
c2c087e6 | 4685 | map->raid_level = 1; |
4d1313e9 | 4686 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4687 | } else if (info->level == 1) |
4688 | map->num_domains = info->raid_disks; | |
4689 | else | |
ff596308 | 4690 | map->num_domains = 1; |
81062a36 | 4691 | |
5551b113 CA |
4692 | /* info->size is only int so use the 'size' parameter instead */ |
4693 | num_data_stripes = (size * 2) / info_to_blocks_per_strip(info); | |
4694 | num_data_stripes /= map->num_domains; | |
4695 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 4696 | |
c2c087e6 DW |
4697 | map->num_members = info->raid_disks; |
4698 | for (i = 0; i < map->num_members; i++) { | |
4699 | /* initialized in add_to_super */ | |
4eb26970 | 4700 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4701 | } |
949c47a0 | 4702 | mpb->num_raid_devs++; |
ba2de7ba DW |
4703 | |
4704 | dv->dev = dev; | |
4705 | dv->index = super->current_vol; | |
4706 | dv->next = super->devlist; | |
4707 | super->devlist = dv; | |
c2c087e6 | 4708 | |
4d1313e9 DW |
4709 | imsm_update_version_info(super); |
4710 | ||
c2c087e6 | 4711 | return 1; |
cdddbdbc DW |
4712 | } |
4713 | ||
bf5a934a DW |
4714 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4715 | unsigned long long size, char *name, | |
4716 | char *homehost, int *uuid) | |
4717 | { | |
4718 | /* This is primarily called by Create when creating a new array. | |
4719 | * We will then get add_to_super called for each component, and then | |
4720 | * write_init_super called to write it out to each device. | |
4721 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4722 | * array. | |
4723 | * To create on a pre-existing array a different method will be called. | |
4724 | * This one is just for fresh drives. | |
4725 | */ | |
4726 | struct intel_super *super; | |
4727 | struct imsm_super *mpb; | |
4728 | size_t mpb_size; | |
4d1313e9 | 4729 | char *version; |
bf5a934a | 4730 | |
bf5a934a | 4731 | if (st->sb) |
e683ca88 DW |
4732 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
4733 | ||
4734 | if (info) | |
4735 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4736 | else | |
4737 | mpb_size = 512; | |
bf5a934a | 4738 | |
49133e57 | 4739 | super = alloc_super(); |
e683ca88 | 4740 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4741 | free(super); |
e683ca88 DW |
4742 | super = NULL; |
4743 | } | |
4744 | if (!super) { | |
e7b84f9d | 4745 | pr_err("%s could not allocate superblock\n", __func__); |
bf5a934a DW |
4746 | return 0; |
4747 | } | |
17a4eaf9 | 4748 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
e7b84f9d | 4749 | pr_err("%s could not allocate migr_rec buffer\n", __func__); |
8e59f3d8 AK |
4750 | free(super->buf); |
4751 | free(super); | |
4752 | return 0; | |
4753 | } | |
e683ca88 | 4754 | memset(super->buf, 0, mpb_size); |
ef649044 | 4755 | mpb = super->buf; |
e683ca88 DW |
4756 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4757 | st->sb = super; | |
4758 | ||
4759 | if (info == NULL) { | |
4760 | /* zeroing superblock */ | |
4761 | return 0; | |
4762 | } | |
bf5a934a | 4763 | |
4d1313e9 DW |
4764 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4765 | ||
4766 | version = (char *) mpb->sig; | |
4767 | strcpy(version, MPB_SIGNATURE); | |
4768 | version += strlen(MPB_SIGNATURE); | |
4769 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4770 | |
bf5a934a DW |
4771 | return 1; |
4772 | } | |
4773 | ||
0e600426 | 4774 | #ifndef MDASSEMBLE |
f20c3968 | 4775 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4776 | int fd, char *devname) |
4777 | { | |
4778 | struct intel_super *super = st->sb; | |
d23fe947 | 4779 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4780 | struct imsm_disk *_disk; |
bf5a934a DW |
4781 | struct imsm_dev *dev; |
4782 | struct imsm_map *map; | |
3960e579 | 4783 | struct dl *dl, *df; |
4eb26970 | 4784 | int slot; |
bf5a934a | 4785 | |
949c47a0 | 4786 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4787 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4788 | |
208933a7 | 4789 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 4790 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
4791 | devname); |
4792 | return 1; | |
4793 | } | |
4794 | ||
efb30e7f DW |
4795 | if (fd == -1) { |
4796 | /* we're doing autolayout so grab the pre-marked (in | |
4797 | * validate_geometry) raid_disk | |
4798 | */ | |
4799 | for (dl = super->disks; dl; dl = dl->next) | |
4800 | if (dl->raiddisk == dk->raid_disk) | |
4801 | break; | |
4802 | } else { | |
4803 | for (dl = super->disks; dl ; dl = dl->next) | |
4804 | if (dl->major == dk->major && | |
4805 | dl->minor == dk->minor) | |
4806 | break; | |
4807 | } | |
d23fe947 | 4808 | |
208933a7 | 4809 | if (!dl) { |
e7b84f9d | 4810 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 4811 | return 1; |
208933a7 | 4812 | } |
bf5a934a | 4813 | |
d23fe947 DW |
4814 | /* add a pristine spare to the metadata */ |
4815 | if (dl->index < 0) { | |
4816 | dl->index = super->anchor->num_disks; | |
4817 | super->anchor->num_disks++; | |
4818 | } | |
4eb26970 DW |
4819 | /* Check the device has not already been added */ |
4820 | slot = get_imsm_disk_slot(map, dl->index); | |
4821 | if (slot >= 0 && | |
238c0a71 | 4822 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 4823 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
4824 | devname); |
4825 | return 1; | |
4826 | } | |
656b6b5a | 4827 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4828 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4829 | |
3960e579 DW |
4830 | /* update size of 'missing' disks to be at least as large as the |
4831 | * largest acitve member (we only have dummy missing disks when | |
4832 | * creating the first volume) | |
4833 | */ | |
4834 | if (super->current_vol == 0) { | |
4835 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
4836 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
4837 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
4838 | _disk = __get_imsm_disk(mpb, df->index); |
4839 | *_disk = df->disk; | |
4840 | } | |
4841 | } | |
4842 | ||
4843 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
4844 | for (df = super->missing; df; df = df->next) | |
4845 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 4846 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
4847 | |
4848 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
4849 | continue; | |
4850 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 4851 | if (is_gen_migration(dev)) { |
238c0a71 AK |
4852 | struct imsm_map *map2 = get_imsm_map(dev, |
4853 | MAP_1); | |
0a108d63 AK |
4854 | int slot2 = get_imsm_disk_slot(map2, df->index); |
4855 | if ((slot2 < map2->num_members) && | |
4856 | (slot2 >= 0)) { | |
1ace8403 | 4857 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
4858 | slot2, |
4859 | MAP_1); | |
1ace8403 AK |
4860 | if ((unsigned)df->index == |
4861 | ord_to_idx(ord2)) | |
4862 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 4863 | slot2, |
1ace8403 AK |
4864 | df->index | |
4865 | IMSM_ORD_REBUILD); | |
4866 | } | |
4867 | } | |
3960e579 DW |
4868 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
4869 | break; | |
4870 | } | |
4871 | ||
d23fe947 DW |
4872 | /* if we are creating the first raid device update the family number */ |
4873 | if (super->current_vol == 0) { | |
4874 | __u32 sum; | |
4875 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 4876 | |
3960e579 | 4877 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 4878 | if (!_dev || !_disk) { |
e7b84f9d | 4879 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
4880 | return 1; |
4881 | } | |
d23fe947 DW |
4882 | *_dev = *dev; |
4883 | *_disk = dl->disk; | |
148acb7b DW |
4884 | sum = random32(); |
4885 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4886 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4887 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4888 | } |
ca0748fa | 4889 | super->current_disk = dl; |
f20c3968 | 4890 | return 0; |
bf5a934a DW |
4891 | } |
4892 | ||
a8619d23 AK |
4893 | /* mark_spare() |
4894 | * Function marks disk as spare and restores disk serial | |
4895 | * in case it was previously marked as failed by takeover operation | |
4896 | * reruns: | |
4897 | * -1 : critical error | |
4898 | * 0 : disk is marked as spare but serial is not set | |
4899 | * 1 : success | |
4900 | */ | |
4901 | int mark_spare(struct dl *disk) | |
4902 | { | |
4903 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
4904 | int ret_val = -1; | |
4905 | ||
4906 | if (!disk) | |
4907 | return ret_val; | |
4908 | ||
4909 | ret_val = 0; | |
4910 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
4911 | /* Restore disk serial number, because takeover marks disk | |
4912 | * as failed and adds to serial ':0' before it becomes | |
4913 | * a spare disk. | |
4914 | */ | |
4915 | serialcpy(disk->serial, serial); | |
4916 | serialcpy(disk->disk.serial, serial); | |
4917 | ret_val = 1; | |
4918 | } | |
4919 | disk->disk.status = SPARE_DISK; | |
4920 | disk->index = -1; | |
4921 | ||
4922 | return ret_val; | |
4923 | } | |
88654014 | 4924 | |
f20c3968 | 4925 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4926 | int fd, char *devname) |
cdddbdbc | 4927 | { |
c2c087e6 | 4928 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4929 | struct dl *dd; |
4930 | unsigned long long size; | |
f2f27e63 | 4931 | __u32 id; |
c2c087e6 DW |
4932 | int rv; |
4933 | struct stat stb; | |
4934 | ||
88654014 LM |
4935 | /* If we are on an RAID enabled platform check that the disk is |
4936 | * attached to the raid controller. | |
4937 | * We do not need to test disks attachment for container based additions, | |
4938 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4939 | */ |
d424212e | 4940 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4941 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4942 | if (rv != 0) { |
4943 | dprintf("capability: %p fd: %d ret: %d\n", | |
4944 | super->orom, fd, rv); | |
4945 | return 1; | |
88c32bb1 DW |
4946 | } |
4947 | ||
f20c3968 DW |
4948 | if (super->current_vol >= 0) |
4949 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4950 | |
c2c087e6 | 4951 | fstat(fd, &stb); |
503975b9 | 4952 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
4953 | dd->major = major(stb.st_rdev); |
4954 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 4955 | dd->devname = devname ? xstrdup(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) { |
e7b84f9d | 4961 | pr_err("failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4962 | free(dd); |
0030e8d6 | 4963 | abort(); |
c2c087e6 DW |
4964 | } |
4965 | ||
c2c087e6 DW |
4966 | get_dev_size(fd, NULL, &size); |
4967 | size /= 512; | |
1f24f035 | 4968 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
4969 | set_total_blocks(&dd->disk, size); |
4970 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
4971 | struct imsm_super *mpb = super->anchor; | |
4972 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
4973 | } | |
a8619d23 | 4974 | mark_spare(dd); |
c2c087e6 | 4975 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4976 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4977 | else |
b9f594fe | 4978 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4979 | |
4980 | if (st->update_tail) { | |
1a64be56 LM |
4981 | dd->next = super->disk_mgmt_list; |
4982 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4983 | } else { |
4984 | dd->next = super->disks; | |
4985 | super->disks = dd; | |
ceaf0ee1 | 4986 | super->updates_pending++; |
43dad3d6 | 4987 | } |
f20c3968 DW |
4988 | |
4989 | return 0; | |
cdddbdbc DW |
4990 | } |
4991 | ||
1a64be56 LM |
4992 | |
4993 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4994 | { | |
4995 | struct intel_super *super = st->sb; | |
4996 | struct dl *dd; | |
4997 | ||
4998 | /* remove from super works only in mdmon - for communication | |
4999 | * manager - monitor. Check if communication memory buffer | |
5000 | * is prepared. | |
5001 | */ | |
5002 | if (!st->update_tail) { | |
e7b84f9d N |
5003 | pr_err("%s shall be used in mdmon context only" |
5004 | "(line %d).\n", __func__, __LINE__); | |
1a64be56 LM |
5005 | return 1; |
5006 | } | |
503975b9 | 5007 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5008 | dd->major = dk->major; |
5009 | dd->minor = dk->minor; | |
1a64be56 | 5010 | dd->fd = -1; |
a8619d23 | 5011 | mark_spare(dd); |
1a64be56 LM |
5012 | dd->action = DISK_REMOVE; |
5013 | ||
5014 | dd->next = super->disk_mgmt_list; | |
5015 | super->disk_mgmt_list = dd; | |
5016 | ||
5017 | ||
5018 | return 0; | |
5019 | } | |
5020 | ||
f796af5d DW |
5021 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5022 | ||
5023 | static union { | |
5024 | char buf[512]; | |
5025 | struct imsm_super anchor; | |
5026 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 5027 | |
d23fe947 DW |
5028 | /* spare records have their own family number and do not have any defined raid |
5029 | * devices | |
5030 | */ | |
5031 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5032 | { | |
d23fe947 | 5033 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5034 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5035 | __u32 sum; |
5036 | struct dl *d; | |
5037 | ||
f796af5d DW |
5038 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
5039 | spare->generation_num = __cpu_to_le32(1UL), | |
5040 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
5041 | spare->num_disks = 1, | |
5042 | spare->num_raid_devs = 0, | |
5043 | spare->cache_size = mpb->cache_size, | |
5044 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
5045 | ||
5046 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5047 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5048 | |
5049 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5050 | if (d->index != -1) |
d23fe947 DW |
5051 | continue; |
5052 | ||
f796af5d | 5053 | spare->disk[0] = d->disk; |
027c374f CA |
5054 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5055 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5056 | ||
f796af5d DW |
5057 | sum = __gen_imsm_checksum(spare); |
5058 | spare->family_num = __cpu_to_le32(sum); | |
5059 | spare->orig_family_num = 0; | |
5060 | sum = __gen_imsm_checksum(spare); | |
5061 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5062 | |
f796af5d | 5063 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
5064 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
5065 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 5066 | return 1; |
d23fe947 DW |
5067 | } |
5068 | if (doclose) { | |
5069 | close(d->fd); | |
5070 | d->fd = -1; | |
5071 | } | |
5072 | } | |
5073 | ||
e74255d9 | 5074 | return 0; |
d23fe947 DW |
5075 | } |
5076 | ||
36988a3d | 5077 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5078 | { |
36988a3d | 5079 | struct intel_super *super = st->sb; |
949c47a0 | 5080 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5081 | struct dl *d; |
5082 | __u32 generation; | |
5083 | __u32 sum; | |
d23fe947 | 5084 | int spares = 0; |
949c47a0 | 5085 | int i; |
a48ac0a8 | 5086 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5087 | int num_disks = 0; |
146c6260 | 5088 | int clear_migration_record = 1; |
cdddbdbc | 5089 | |
c2c087e6 DW |
5090 | /* 'generation' is incremented everytime the metadata is written */ |
5091 | generation = __le32_to_cpu(mpb->generation_num); | |
5092 | generation++; | |
5093 | mpb->generation_num = __cpu_to_le32(generation); | |
5094 | ||
148acb7b DW |
5095 | /* fix up cases where previous mdadm releases failed to set |
5096 | * orig_family_num | |
5097 | */ | |
5098 | if (mpb->orig_family_num == 0) | |
5099 | mpb->orig_family_num = mpb->family_num; | |
5100 | ||
d23fe947 | 5101 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5102 | if (d->index == -1) |
d23fe947 | 5103 | spares++; |
36988a3d | 5104 | else { |
d23fe947 | 5105 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5106 | num_disks++; |
5107 | } | |
d23fe947 | 5108 | } |
36988a3d | 5109 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5110 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5111 | num_disks++; |
5112 | } | |
5113 | mpb->num_disks = num_disks; | |
5114 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5115 | |
949c47a0 DW |
5116 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5117 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5118 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5119 | if (dev && dev2) { | |
5120 | imsm_copy_dev(dev, dev2); | |
5121 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5122 | } | |
146c6260 AK |
5123 | if (is_gen_migration(dev2)) |
5124 | clear_migration_record = 0; | |
949c47a0 | 5125 | } |
a48ac0a8 DW |
5126 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
5127 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 5128 | |
c2c087e6 | 5129 | /* recalculate checksum */ |
949c47a0 | 5130 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5131 | mpb->check_sum = __cpu_to_le32(sum); |
5132 | ||
51d83f5d AK |
5133 | if (super->clean_migration_record_by_mdmon) { |
5134 | clear_migration_record = 1; | |
5135 | super->clean_migration_record_by_mdmon = 0; | |
5136 | } | |
146c6260 | 5137 | if (clear_migration_record) |
17a4eaf9 | 5138 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); |
146c6260 | 5139 | |
d23fe947 | 5140 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5141 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5142 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5143 | continue; |
30602f53 | 5144 | |
146c6260 AK |
5145 | if (clear_migration_record) { |
5146 | unsigned long long dsize; | |
5147 | ||
5148 | get_dev_size(d->fd, NULL, &dsize); | |
5149 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
17a4eaf9 AK |
5150 | if (write(d->fd, super->migr_rec_buf, |
5151 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
9e2d750d | 5152 | perror("Write migr_rec failed"); |
146c6260 AK |
5153 | } |
5154 | } | |
51d83f5d AK |
5155 | |
5156 | if (store_imsm_mpb(d->fd, mpb)) | |
5157 | fprintf(stderr, | |
5158 | "%s: failed for device %d:%d (fd: %d)%s\n", | |
5159 | __func__, d->major, d->minor, | |
5160 | d->fd, strerror(errno)); | |
5161 | ||
c2c087e6 DW |
5162 | if (doclose) { |
5163 | close(d->fd); | |
5164 | d->fd = -1; | |
5165 | } | |
5166 | } | |
5167 | ||
d23fe947 DW |
5168 | if (spares) |
5169 | return write_super_imsm_spares(super, doclose); | |
5170 | ||
e74255d9 | 5171 | return 0; |
c2c087e6 DW |
5172 | } |
5173 | ||
0e600426 | 5174 | |
9b1fb677 | 5175 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5176 | { |
5177 | size_t len; | |
5178 | struct imsm_update_create_array *u; | |
5179 | struct intel_super *super = st->sb; | |
9b1fb677 | 5180 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5181 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5182 | struct disk_info *inf; |
5183 | struct imsm_disk *disk; | |
5184 | int i; | |
43dad3d6 | 5185 | |
54c2c1ea DW |
5186 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5187 | sizeof(*inf) * map->num_members; | |
503975b9 | 5188 | u = xmalloc(len); |
43dad3d6 | 5189 | u->type = update_create_array; |
9b1fb677 | 5190 | u->dev_idx = dev_idx; |
43dad3d6 | 5191 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5192 | inf = get_disk_info(u); |
5193 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5194 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5195 | |
54c2c1ea DW |
5196 | disk = get_imsm_disk(super, idx); |
5197 | serialcpy(inf[i].serial, disk->serial); | |
5198 | } | |
43dad3d6 DW |
5199 | append_metadata_update(st, u, len); |
5200 | ||
5201 | return 0; | |
5202 | } | |
5203 | ||
1a64be56 | 5204 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5205 | { |
5206 | struct intel_super *super = st->sb; | |
5207 | size_t len; | |
1a64be56 | 5208 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5209 | |
1a64be56 | 5210 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5211 | return 0; |
5212 | ||
5213 | len = sizeof(*u); | |
503975b9 | 5214 | u = xmalloc(len); |
1a64be56 | 5215 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5216 | append_metadata_update(st, u, len); |
5217 | ||
5218 | return 0; | |
5219 | } | |
5220 | ||
c2c087e6 DW |
5221 | static int write_init_super_imsm(struct supertype *st) |
5222 | { | |
9b1fb677 DW |
5223 | struct intel_super *super = st->sb; |
5224 | int current_vol = super->current_vol; | |
5225 | ||
5226 | /* we are done with current_vol reset it to point st at the container */ | |
5227 | super->current_vol = -1; | |
5228 | ||
8273f55e | 5229 | if (st->update_tail) { |
43dad3d6 DW |
5230 | /* queue the recently created array / added disk |
5231 | * as a metadata update */ | |
43dad3d6 | 5232 | int rv; |
8273f55e | 5233 | |
43dad3d6 | 5234 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5235 | if (current_vol < 0) { |
1a64be56 LM |
5236 | /* in the mgmt (add/remove) disk case we are running |
5237 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5238 | */ |
1a64be56 | 5239 | return mgmt_disk(st); |
43dad3d6 | 5240 | } else |
9b1fb677 | 5241 | rv = create_array(st, current_vol); |
8273f55e | 5242 | |
43dad3d6 | 5243 | return rv; |
d682f344 N |
5244 | } else { |
5245 | struct dl *d; | |
5246 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 5247 | Kill(d->devname, NULL, 0, -1, 1); |
36988a3d | 5248 | return write_super_imsm(st, 1); |
d682f344 | 5249 | } |
cdddbdbc | 5250 | } |
0e600426 | 5251 | #endif |
cdddbdbc | 5252 | |
e683ca88 | 5253 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5254 | { |
e683ca88 DW |
5255 | struct intel_super *super = st->sb; |
5256 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5257 | |
e683ca88 | 5258 | if (!mpb) |
ad97895e DW |
5259 | return 1; |
5260 | ||
1799c9e8 | 5261 | #ifndef MDASSEMBLE |
e683ca88 | 5262 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5263 | #else |
5264 | return 1; | |
5265 | #endif | |
cdddbdbc DW |
5266 | } |
5267 | ||
0e600426 N |
5268 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
5269 | { | |
5270 | return __le32_to_cpu(mpb->bbm_log_size); | |
5271 | } | |
5272 | ||
5273 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
5274 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5275 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 5276 | unsigned long long size, char *dev, |
2c514b71 NB |
5277 | unsigned long long *freesize, |
5278 | int verbose) | |
cdddbdbc | 5279 | { |
c2c087e6 DW |
5280 | int fd; |
5281 | unsigned long long ldsize; | |
f2f5c343 LM |
5282 | struct intel_super *super=NULL; |
5283 | int rv = 0; | |
cdddbdbc | 5284 | |
c2c087e6 DW |
5285 | if (level != LEVEL_CONTAINER) |
5286 | return 0; | |
5287 | if (!dev) | |
5288 | return 1; | |
5289 | ||
5290 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5291 | if (fd < 0) { | |
ba728be7 | 5292 | if (verbose > 0) |
e7b84f9d | 5293 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 5294 | dev, strerror(errno)); |
c2c087e6 DW |
5295 | return 0; |
5296 | } | |
5297 | if (!get_dev_size(fd, dev, &ldsize)) { | |
5298 | close(fd); | |
5299 | return 0; | |
5300 | } | |
f2f5c343 LM |
5301 | |
5302 | /* capabilities retrieve could be possible | |
5303 | * note that there is no fd for the disks in array. | |
5304 | */ | |
5305 | super = alloc_super(); | |
ba728be7 | 5306 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
5307 | if (rv != 0) { |
5308 | #if DEBUG | |
5309 | char str[256]; | |
5310 | fd2devname(fd, str); | |
5311 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
5312 | fd, str, super->orom, rv, raiddisks); | |
5313 | #endif | |
5314 | /* no orom/efi or non-intel hba of the disk */ | |
5315 | close(fd); | |
5316 | free_imsm(super); | |
5317 | return 0; | |
5318 | } | |
c2c087e6 | 5319 | close(fd); |
9126b9a8 CA |
5320 | if (super->orom) { |
5321 | if (raiddisks > super->orom->tds) { | |
5322 | if (verbose) | |
e7b84f9d | 5323 | pr_err("%d exceeds maximum number of" |
9126b9a8 CA |
5324 | " platform supported disks: %d\n", |
5325 | raiddisks, super->orom->tds); | |
5326 | free_imsm(super); | |
5327 | return 0; | |
5328 | } | |
5329 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
5330 | (ldsize >> 9) >> 32 > 0) { | |
5331 | if (verbose) | |
e7b84f9d | 5332 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
5333 | free_imsm(super); |
5334 | return 0; | |
5335 | } | |
f2f5c343 | 5336 | } |
c2c087e6 DW |
5337 | |
5338 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 5339 | free_imsm(super); |
c2c087e6 DW |
5340 | |
5341 | return 1; | |
cdddbdbc DW |
5342 | } |
5343 | ||
0dcecb2e DW |
5344 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5345 | { | |
5346 | const unsigned long long base_start = e[*idx].start; | |
5347 | unsigned long long end = base_start + e[*idx].size; | |
5348 | int i; | |
5349 | ||
5350 | if (base_start == end) | |
5351 | return 0; | |
5352 | ||
5353 | *idx = *idx + 1; | |
5354 | for (i = *idx; i < num_extents; i++) { | |
5355 | /* extend overlapping extents */ | |
5356 | if (e[i].start >= base_start && | |
5357 | e[i].start <= end) { | |
5358 | if (e[i].size == 0) | |
5359 | return 0; | |
5360 | if (e[i].start + e[i].size > end) | |
5361 | end = e[i].start + e[i].size; | |
5362 | } else if (e[i].start > end) { | |
5363 | *idx = i; | |
5364 | break; | |
5365 | } | |
5366 | } | |
5367 | ||
5368 | return end - base_start; | |
5369 | } | |
5370 | ||
5371 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5372 | { | |
5373 | /* build a composite disk with all known extents and generate a new | |
5374 | * 'maxsize' given the "all disks in an array must share a common start | |
5375 | * offset" constraint | |
5376 | */ | |
503975b9 | 5377 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
5378 | struct dl *dl; |
5379 | int i, j; | |
5380 | int start_extent; | |
5381 | unsigned long long pos; | |
b9d77223 | 5382 | unsigned long long start = 0; |
0dcecb2e DW |
5383 | unsigned long long maxsize; |
5384 | unsigned long reserve; | |
5385 | ||
0dcecb2e DW |
5386 | /* coalesce and sort all extents. also, check to see if we need to |
5387 | * reserve space between member arrays | |
5388 | */ | |
5389 | j = 0; | |
5390 | for (dl = super->disks; dl; dl = dl->next) { | |
5391 | if (!dl->e) | |
5392 | continue; | |
5393 | for (i = 0; i < dl->extent_cnt; i++) | |
5394 | e[j++] = dl->e[i]; | |
5395 | } | |
5396 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5397 | ||
5398 | /* merge extents */ | |
5399 | i = 0; | |
5400 | j = 0; | |
5401 | while (i < sum_extents) { | |
5402 | e[j].start = e[i].start; | |
5403 | e[j].size = find_size(e, &i, sum_extents); | |
5404 | j++; | |
5405 | if (e[j-1].size == 0) | |
5406 | break; | |
5407 | } | |
5408 | ||
5409 | pos = 0; | |
5410 | maxsize = 0; | |
5411 | start_extent = 0; | |
5412 | i = 0; | |
5413 | do { | |
5414 | unsigned long long esize; | |
5415 | ||
5416 | esize = e[i].start - pos; | |
5417 | if (esize >= maxsize) { | |
5418 | maxsize = esize; | |
5419 | start = pos; | |
5420 | start_extent = i; | |
5421 | } | |
5422 | pos = e[i].start + e[i].size; | |
5423 | i++; | |
5424 | } while (e[i-1].size); | |
5425 | free(e); | |
5426 | ||
a7dd165b DW |
5427 | if (maxsize == 0) |
5428 | return 0; | |
5429 | ||
5430 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5431 | * container | |
5432 | */ | |
0dcecb2e DW |
5433 | if (start_extent > 0) |
5434 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5435 | else | |
5436 | reserve = 0; | |
5437 | ||
5438 | if (maxsize < reserve) | |
a7dd165b | 5439 | return 0; |
0dcecb2e | 5440 | |
5551b113 | 5441 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 5442 | if (start + reserve > super->create_offset) |
a7dd165b | 5443 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5444 | super->create_offset = start + reserve; |
5445 | ||
5446 | return maxsize - reserve; | |
5447 | } | |
5448 | ||
88c32bb1 DW |
5449 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5450 | { | |
5451 | if (level < 0 || level == 6 || level == 4) | |
5452 | return 0; | |
5453 | ||
5454 | /* if we have an orom prevent invalid raid levels */ | |
5455 | if (orom) | |
5456 | switch (level) { | |
5457 | case 0: return imsm_orom_has_raid0(orom); | |
5458 | case 1: | |
5459 | if (raiddisks > 2) | |
5460 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5461 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5462 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5463 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5464 | } |
5465 | else | |
5466 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5467 | ||
5468 | return 0; | |
5469 | } | |
5470 | ||
ca9de185 LM |
5471 | |
5472 | static int | |
5473 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
5474 | int dpa, int verbose) | |
5475 | { | |
5476 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
5477 | struct mdstat_ent *memb = NULL; | |
5478 | int count = 0; | |
5479 | int num = 0; | |
5480 | struct md_list *dv = NULL; | |
5481 | int found; | |
5482 | ||
5483 | for (memb = mdstat ; memb ; memb = memb->next) { | |
5484 | if (memb->metadata_version && | |
5485 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
5486 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
5487 | !is_subarray(memb->metadata_version+9) && | |
5488 | memb->members) { | |
5489 | struct dev_member *dev = memb->members; | |
5490 | int fd = -1; | |
5491 | while(dev && (fd < 0)) { | |
503975b9 N |
5492 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
5493 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
5494 | if (num > 0) | |
5495 | fd = open(path, O_RDONLY, 0); | |
5496 | if ((num <= 0) || (fd < 0)) { | |
5497 | pr_vrb(": Cannot open %s: %s\n", | |
5498 | dev->name, strerror(errno)); | |
ca9de185 | 5499 | } |
503975b9 | 5500 | free(path); |
ca9de185 LM |
5501 | dev = dev->next; |
5502 | } | |
5503 | found = 0; | |
5504 | if ((fd >= 0) && disk_attached_to_hba(fd, hba)) { | |
5505 | struct mdstat_ent *vol; | |
5506 | for (vol = mdstat ; vol ; vol = vol->next) { | |
5507 | if ((vol->active > 0) && | |
5508 | vol->metadata_version && | |
5509 | is_container_member(vol, memb->dev)) { | |
5510 | found++; | |
5511 | count++; | |
5512 | } | |
5513 | } | |
5514 | if (*devlist && (found < dpa)) { | |
503975b9 N |
5515 | dv = xcalloc(1, sizeof(*dv)); |
5516 | dv->devname = xmalloc(strlen(memb->dev) + strlen("/dev/") + 1); | |
5517 | sprintf(dv->devname, "%s%s", "/dev/", memb->dev); | |
5518 | dv->found = found; | |
5519 | dv->used = 0; | |
5520 | dv->next = *devlist; | |
5521 | *devlist = dv; | |
ca9de185 LM |
5522 | } |
5523 | } | |
5524 | if (fd >= 0) | |
5525 | close(fd); | |
5526 | } | |
5527 | } | |
5528 | free_mdstat(mdstat); | |
5529 | return count; | |
5530 | } | |
5531 | ||
5532 | #ifdef DEBUG_LOOP | |
5533 | static struct md_list* | |
5534 | get_loop_devices(void) | |
5535 | { | |
5536 | int i; | |
5537 | struct md_list *devlist = NULL; | |
5538 | struct md_list *dv = NULL; | |
5539 | ||
5540 | for(i = 0; i < 12; i++) { | |
503975b9 N |
5541 | dv = xcalloc(1, sizeof(*dv)); |
5542 | dv->devname = xmalloc(40); | |
ca9de185 LM |
5543 | sprintf(dv->devname, "/dev/loop%d", i); |
5544 | dv->next = devlist; | |
5545 | devlist = dv; | |
5546 | } | |
5547 | return devlist; | |
5548 | } | |
5549 | #endif | |
5550 | ||
5551 | static struct md_list* | |
5552 | get_devices(const char *hba_path) | |
5553 | { | |
5554 | struct md_list *devlist = NULL; | |
5555 | struct md_list *dv = NULL; | |
5556 | struct dirent *ent; | |
5557 | DIR *dir; | |
5558 | int err = 0; | |
5559 | ||
5560 | #if DEBUG_LOOP | |
5561 | devlist = get_loop_devices(); | |
5562 | return devlist; | |
5563 | #endif | |
5564 | /* scroll through /sys/dev/block looking for devices attached to | |
5565 | * this hba | |
5566 | */ | |
5567 | dir = opendir("/sys/dev/block"); | |
5568 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
5569 | int fd; | |
5570 | char buf[1024]; | |
5571 | int major, minor; | |
5572 | char *path = NULL; | |
5573 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
5574 | continue; | |
5575 | path = devt_to_devpath(makedev(major, minor)); | |
5576 | if (!path) | |
5577 | continue; | |
5578 | if (!path_attached_to_hba(path, hba_path)) { | |
5579 | free(path); | |
5580 | path = NULL; | |
5581 | continue; | |
5582 | } | |
5583 | free(path); | |
5584 | path = NULL; | |
5585 | fd = dev_open(ent->d_name, O_RDONLY); | |
5586 | if (fd >= 0) { | |
5587 | fd2devname(fd, buf); | |
5588 | close(fd); | |
5589 | } else { | |
e7b84f9d | 5590 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
5591 | ent->d_name); |
5592 | continue; | |
5593 | } | |
5594 | ||
5595 | ||
503975b9 N |
5596 | dv = xcalloc(1, sizeof(*dv)); |
5597 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
5598 | dv->next = devlist; |
5599 | devlist = dv; | |
5600 | } | |
5601 | if (err) { | |
5602 | while(devlist) { | |
5603 | dv = devlist; | |
5604 | devlist = devlist->next; | |
5605 | free(dv->devname); | |
5606 | free(dv); | |
5607 | } | |
5608 | } | |
562aa102 | 5609 | closedir(dir); |
ca9de185 LM |
5610 | return devlist; |
5611 | } | |
5612 | ||
5613 | static int | |
5614 | count_volumes_list(struct md_list *devlist, char *homehost, | |
5615 | int verbose, int *found) | |
5616 | { | |
5617 | struct md_list *tmpdev; | |
5618 | int count = 0; | |
5619 | struct supertype *st = NULL; | |
5620 | ||
5621 | /* first walk the list of devices to find a consistent set | |
5622 | * that match the criterea, if that is possible. | |
5623 | * We flag the ones we like with 'used'. | |
5624 | */ | |
5625 | *found = 0; | |
5626 | st = match_metadata_desc_imsm("imsm"); | |
5627 | if (st == NULL) { | |
5628 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5629 | return 0; | |
5630 | } | |
5631 | ||
5632 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5633 | char *devname = tmpdev->devname; | |
5634 | struct stat stb; | |
5635 | struct supertype *tst; | |
5636 | int dfd; | |
5637 | if (tmpdev->used > 1) | |
5638 | continue; | |
5639 | tst = dup_super(st); | |
5640 | if (tst == NULL) { | |
5641 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5642 | goto err_1; | |
5643 | } | |
5644 | tmpdev->container = 0; | |
5645 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
5646 | if (dfd < 0) { | |
5647 | dprintf(": cannot open device %s: %s\n", | |
5648 | devname, strerror(errno)); | |
5649 | tmpdev->used = 2; | |
5650 | } else if (fstat(dfd, &stb)< 0) { | |
5651 | /* Impossible! */ | |
5652 | dprintf(": fstat failed for %s: %s\n", | |
5653 | devname, strerror(errno)); | |
5654 | tmpdev->used = 2; | |
5655 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
5656 | dprintf(": %s is not a block device.\n", | |
5657 | devname); | |
5658 | tmpdev->used = 2; | |
5659 | } else if (must_be_container(dfd)) { | |
5660 | struct supertype *cst; | |
5661 | cst = super_by_fd(dfd, NULL); | |
5662 | if (cst == NULL) { | |
5663 | dprintf(": cannot recognize container type %s\n", | |
5664 | devname); | |
5665 | tmpdev->used = 2; | |
5666 | } else if (tst->ss != st->ss) { | |
5667 | dprintf(": non-imsm container - ignore it: %s\n", | |
5668 | devname); | |
5669 | tmpdev->used = 2; | |
5670 | } else if (!tst->ss->load_container || | |
5671 | tst->ss->load_container(tst, dfd, NULL)) | |
5672 | tmpdev->used = 2; | |
5673 | else { | |
5674 | tmpdev->container = 1; | |
5675 | } | |
5676 | if (cst) | |
5677 | cst->ss->free_super(cst); | |
5678 | } else { | |
5679 | tmpdev->st_rdev = stb.st_rdev; | |
5680 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
5681 | dprintf(": no RAID superblock on %s\n", | |
5682 | devname); | |
5683 | tmpdev->used = 2; | |
5684 | } else if (tst->ss->compare_super == NULL) { | |
5685 | dprintf(": Cannot assemble %s metadata on %s\n", | |
5686 | tst->ss->name, devname); | |
5687 | tmpdev->used = 2; | |
5688 | } | |
5689 | } | |
5690 | if (dfd >= 0) | |
5691 | close(dfd); | |
5692 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
5693 | /* Ignore unrecognised devices during auto-assembly */ | |
5694 | goto loop; | |
5695 | } | |
5696 | else { | |
5697 | struct mdinfo info; | |
5698 | tst->ss->getinfo_super(tst, &info, NULL); | |
5699 | ||
5700 | if (st->minor_version == -1) | |
5701 | st->minor_version = tst->minor_version; | |
5702 | ||
5703 | if (memcmp(info.uuid, uuid_zero, | |
5704 | sizeof(int[4])) == 0) { | |
5705 | /* this is a floating spare. It cannot define | |
5706 | * an array unless there are no more arrays of | |
5707 | * this type to be found. It can be included | |
5708 | * in an array of this type though. | |
5709 | */ | |
5710 | tmpdev->used = 3; | |
5711 | goto loop; | |
5712 | } | |
5713 | ||
5714 | if (st->ss != tst->ss || | |
5715 | st->minor_version != tst->minor_version || | |
5716 | st->ss->compare_super(st, tst) != 0) { | |
5717 | /* Some mismatch. If exactly one array matches this host, | |
5718 | * we can resolve on that one. | |
5719 | * Or, if we are auto assembling, we just ignore the second | |
5720 | * for now. | |
5721 | */ | |
5722 | dprintf(": superblock on %s doesn't match others - assembly aborted\n", | |
5723 | devname); | |
5724 | goto loop; | |
5725 | } | |
5726 | tmpdev->used = 1; | |
5727 | *found = 1; | |
5728 | dprintf("found: devname: %s\n", devname); | |
5729 | } | |
5730 | loop: | |
5731 | if (tst) | |
5732 | tst->ss->free_super(tst); | |
5733 | } | |
5734 | if (*found != 0) { | |
5735 | int err; | |
5736 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
5737 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
5738 | for (iter = head; iter; iter = iter->next) { | |
5739 | dprintf("content->text_version: %s vol\n", | |
5740 | iter->text_version); | |
5741 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
5742 | /* do not assemble arrays with unsupported | |
5743 | configurations */ | |
5744 | dprintf(": Cannot activate member %s.\n", | |
5745 | iter->text_version); | |
5746 | } else | |
5747 | count++; | |
5748 | } | |
5749 | sysfs_free(head); | |
5750 | ||
5751 | } else { | |
5752 | dprintf(" no valid super block on device list: err: %d %p\n", | |
5753 | err, st->sb); | |
5754 | } | |
5755 | } else { | |
5756 | dprintf(" no more devices to examin\n"); | |
5757 | } | |
5758 | ||
5759 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5760 | if ((tmpdev->used == 1) && (tmpdev->found)) { | |
5761 | if (count) { | |
5762 | if (count < tmpdev->found) | |
5763 | count = 0; | |
5764 | else | |
5765 | count -= tmpdev->found; | |
5766 | } | |
5767 | } | |
5768 | if (tmpdev->used == 1) | |
5769 | tmpdev->used = 4; | |
5770 | } | |
5771 | err_1: | |
5772 | if (st) | |
5773 | st->ss->free_super(st); | |
5774 | return count; | |
5775 | } | |
5776 | ||
5777 | ||
5778 | static int | |
5779 | count_volumes(char *hba, int dpa, int verbose) | |
5780 | { | |
5781 | struct md_list *devlist = NULL; | |
5782 | int count = 0; | |
5783 | int found = 0;; | |
5784 | ||
5785 | devlist = get_devices(hba); | |
5786 | /* if no intel devices return zero volumes */ | |
5787 | if (devlist == NULL) | |
5788 | return 0; | |
5789 | ||
5790 | count = active_arrays_by_format("imsm", hba, &devlist, dpa, verbose); | |
5791 | dprintf(" path: %s active arrays: %d\n", hba, count); | |
5792 | if (devlist == NULL) | |
5793 | return 0; | |
5794 | do { | |
5795 | found = 0; | |
5796 | count += count_volumes_list(devlist, | |
5797 | NULL, | |
5798 | verbose, | |
5799 | &found); | |
5800 | dprintf("found %d count: %d\n", found, count); | |
5801 | } while (found); | |
5802 | ||
5803 | dprintf("path: %s total number of volumes: %d\n", hba, count); | |
5804 | ||
5805 | while(devlist) { | |
5806 | struct md_list *dv = devlist; | |
5807 | devlist = devlist->next; | |
5808 | free(dv->devname); | |
5809 | free(dv); | |
5810 | } | |
5811 | return count; | |
5812 | } | |
5813 | ||
cd9d1ac7 DW |
5814 | static int imsm_default_chunk(const struct imsm_orom *orom) |
5815 | { | |
5816 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
5817 | * 128 if no platform detected | |
5818 | */ | |
5819 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
5820 | ||
5821 | return min(512, (1 << fs)); | |
5822 | } | |
73408129 | 5823 | |
6592ce37 DW |
5824 | static int |
5825 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 5826 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 5827 | { |
660260d0 DW |
5828 | /* check/set platform and metadata limits/defaults */ |
5829 | if (super->orom && raiddisks > super->orom->dpa) { | |
5830 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
5831 | super->orom->dpa); | |
73408129 LM |
5832 | return 0; |
5833 | } | |
5834 | ||
5835 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
660260d0 | 5836 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
5837 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
5838 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
5839 | return 0; | |
5840 | } | |
cd9d1ac7 DW |
5841 | |
5842 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
5843 | *chunk = imsm_default_chunk(super->orom); | |
5844 | ||
5845 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
5846 | pr_vrb(": platform does not support a chunk size of: " | |
5847 | "%d\n", *chunk); | |
5848 | return 0; | |
6592ce37 | 5849 | } |
cd9d1ac7 | 5850 | |
6592ce37 DW |
5851 | if (layout != imsm_level_to_layout(level)) { |
5852 | if (level == 5) | |
5853 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
5854 | else if (level == 10) | |
5855 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
5856 | else | |
5857 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
5858 | layout, level); | |
5859 | return 0; | |
5860 | } | |
2cc699af CA |
5861 | |
5862 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && chunk && | |
5863 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { | |
5864 | pr_vrb(": platform does not support a volume size over 2TB\n"); | |
5865 | return 0; | |
5866 | } | |
6592ce37 DW |
5867 | return 1; |
5868 | } | |
5869 | ||
c2c087e6 DW |
5870 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
5871 | * FIX ME add ahci details | |
5872 | */ | |
8b353278 | 5873 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 5874 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 5875 | unsigned long long size, char *dev, |
2c514b71 NB |
5876 | unsigned long long *freesize, |
5877 | int verbose) | |
cdddbdbc | 5878 | { |
c2c087e6 DW |
5879 | struct stat stb; |
5880 | struct intel_super *super = st->sb; | |
b2916f25 | 5881 | struct imsm_super *mpb; |
c2c087e6 DW |
5882 | struct dl *dl; |
5883 | unsigned long long pos = 0; | |
5884 | unsigned long long maxsize; | |
5885 | struct extent *e; | |
5886 | int i; | |
cdddbdbc | 5887 | |
88c32bb1 DW |
5888 | /* We must have the container info already read in. */ |
5889 | if (!super) | |
c2c087e6 DW |
5890 | return 0; |
5891 | ||
b2916f25 JS |
5892 | mpb = super->anchor; |
5893 | ||
2cc699af | 5894 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
e7b84f9d | 5895 | pr_err("RAID gemetry validation failed. " |
d54559f0 | 5896 | "Cannot proceed with the action(s).\n"); |
c2c087e6 | 5897 | return 0; |
d54559f0 | 5898 | } |
c2c087e6 DW |
5899 | if (!dev) { |
5900 | /* General test: make sure there is space for | |
2da8544a DW |
5901 | * 'raiddisks' device extents of size 'size' at a given |
5902 | * offset | |
c2c087e6 | 5903 | */ |
e46273eb | 5904 | unsigned long long minsize = size; |
b7528a20 | 5905 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5906 | int dcnt = 0; |
5907 | if (minsize == 0) | |
5908 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5909 | for (dl = super->disks; dl ; dl = dl->next) { | |
5910 | int found = 0; | |
5911 | ||
bf5a934a | 5912 | pos = 0; |
c2c087e6 DW |
5913 | i = 0; |
5914 | e = get_extents(super, dl); | |
5915 | if (!e) continue; | |
5916 | do { | |
5917 | unsigned long long esize; | |
5918 | esize = e[i].start - pos; | |
5919 | if (esize >= minsize) | |
5920 | found = 1; | |
b7528a20 | 5921 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5922 | start_offset = pos; |
5923 | break; | |
5924 | } else if (found && pos != start_offset) { | |
5925 | found = 0; | |
5926 | break; | |
5927 | } | |
c2c087e6 DW |
5928 | pos = e[i].start + e[i].size; |
5929 | i++; | |
5930 | } while (e[i-1].size); | |
5931 | if (found) | |
5932 | dcnt++; | |
5933 | free(e); | |
5934 | } | |
5935 | if (dcnt < raiddisks) { | |
2c514b71 | 5936 | if (verbose) |
e7b84f9d | 5937 | pr_err("imsm: Not enough " |
2c514b71 NB |
5938 | "devices with space for this array " |
5939 | "(%d < %d)\n", | |
5940 | dcnt, raiddisks); | |
c2c087e6 DW |
5941 | return 0; |
5942 | } | |
5943 | return 1; | |
5944 | } | |
0dcecb2e | 5945 | |
c2c087e6 DW |
5946 | /* This device must be a member of the set */ |
5947 | if (stat(dev, &stb) < 0) | |
5948 | return 0; | |
5949 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
5950 | return 0; | |
5951 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
5952 | if (dl->major == (int)major(stb.st_rdev) && |
5953 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
5954 | break; |
5955 | } | |
5956 | if (!dl) { | |
2c514b71 | 5957 | if (verbose) |
e7b84f9d | 5958 | pr_err("%s is not in the " |
2c514b71 | 5959 | "same imsm set\n", dev); |
c2c087e6 | 5960 | return 0; |
a20d2ba5 DW |
5961 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
5962 | /* If a volume is present then the current creation attempt | |
5963 | * cannot incorporate new spares because the orom may not | |
5964 | * understand this configuration (all member disks must be | |
5965 | * members of each array in the container). | |
5966 | */ | |
e7b84f9d | 5967 | pr_err("%s is a spare and a volume" |
a20d2ba5 | 5968 | " is already defined for this container\n", dev); |
e7b84f9d | 5969 | pr_err("The option-rom requires all member" |
a20d2ba5 DW |
5970 | " disks to be a member of all volumes\n"); |
5971 | return 0; | |
5fe62b94 WD |
5972 | } else if (super->orom && mpb->num_raid_devs > 0 && |
5973 | mpb->num_disks != raiddisks) { | |
e7b84f9d | 5974 | pr_err("The option-rom requires all member" |
5fe62b94 WD |
5975 | " disks to be a member of all volumes\n"); |
5976 | return 0; | |
c2c087e6 | 5977 | } |
0dcecb2e DW |
5978 | |
5979 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
5980 | e = get_extents(super, dl); |
5981 | maxsize = 0; | |
5982 | i = 0; | |
0dcecb2e DW |
5983 | if (e) { |
5984 | do { | |
5985 | unsigned long long esize; | |
5986 | ||
5987 | esize = e[i].start - pos; | |
5988 | if (esize >= maxsize) | |
5989 | maxsize = esize; | |
5990 | pos = e[i].start + e[i].size; | |
5991 | i++; | |
5992 | } while (e[i-1].size); | |
5993 | dl->e = e; | |
5994 | dl->extent_cnt = i; | |
5995 | } else { | |
5996 | if (verbose) | |
e7b84f9d | 5997 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
5998 | dev); |
5999 | return 0; | |
6000 | } | |
6001 | if (maxsize < size) { | |
6002 | if (verbose) | |
e7b84f9d | 6003 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
6004 | dev, maxsize, size); |
6005 | return 0; | |
6006 | } | |
6007 | ||
6008 | /* count total number of extents for merge */ | |
6009 | i = 0; | |
6010 | for (dl = super->disks; dl; dl = dl->next) | |
6011 | if (dl->e) | |
6012 | i += dl->extent_cnt; | |
6013 | ||
6014 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
6015 | |
6016 | if (!check_env("IMSM_NO_PLATFORM") && | |
6017 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
e7b84f9d | 6018 | pr_err("attempting to create a second " |
3baa56ab LO |
6019 | "volume with size less then remaining space. " |
6020 | "Aborting...\n"); | |
6021 | return 0; | |
6022 | } | |
6023 | ||
a7dd165b | 6024 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
6025 | if (verbose) { |
6026 | if (maxsize == 0) | |
e7b84f9d | 6027 | pr_err("no free space" |
b3071342 LD |
6028 | " left on device. Aborting...\n"); |
6029 | else | |
e7b84f9d | 6030 | pr_err("not enough space" |
b3071342 LD |
6031 | " to create volume of given size" |
6032 | " (%llu < %llu). Aborting...\n", | |
6033 | maxsize, size); | |
6034 | } | |
0dcecb2e | 6035 | return 0; |
0dcecb2e DW |
6036 | } |
6037 | ||
c2c087e6 DW |
6038 | *freesize = maxsize; |
6039 | ||
ca9de185 LM |
6040 | if (super->orom) { |
6041 | int count = count_volumes(super->hba->path, | |
6042 | super->orom->dpa, verbose); | |
6043 | if (super->orom->vphba <= count) { | |
40110b90 | 6044 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6045 | super->orom->vphba); |
6046 | return 0; | |
6047 | } | |
6048 | } | |
c2c087e6 | 6049 | return 1; |
cdddbdbc DW |
6050 | } |
6051 | ||
13bcac90 | 6052 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
6053 | unsigned long long size, int chunk, |
6054 | unsigned long long *freesize) | |
6055 | { | |
6056 | struct intel_super *super = st->sb; | |
6057 | struct imsm_super *mpb = super->anchor; | |
6058 | struct dl *dl; | |
6059 | int i; | |
6060 | int extent_cnt; | |
6061 | struct extent *e; | |
6062 | unsigned long long maxsize; | |
6063 | unsigned long long minsize; | |
6064 | int cnt; | |
6065 | int used; | |
6066 | ||
6067 | /* find the largest common start free region of the possible disks */ | |
6068 | used = 0; | |
6069 | extent_cnt = 0; | |
6070 | cnt = 0; | |
6071 | for (dl = super->disks; dl; dl = dl->next) { | |
6072 | dl->raiddisk = -1; | |
6073 | ||
6074 | if (dl->index >= 0) | |
6075 | used++; | |
6076 | ||
6077 | /* don't activate new spares if we are orom constrained | |
6078 | * and there is already a volume active in the container | |
6079 | */ | |
6080 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
6081 | continue; | |
6082 | ||
6083 | e = get_extents(super, dl); | |
6084 | if (!e) | |
6085 | continue; | |
6086 | for (i = 1; e[i-1].size; i++) | |
6087 | ; | |
6088 | dl->e = e; | |
6089 | dl->extent_cnt = i; | |
6090 | extent_cnt += i; | |
6091 | cnt++; | |
6092 | } | |
6093 | ||
6094 | maxsize = merge_extents(super, extent_cnt); | |
6095 | minsize = size; | |
6096 | if (size == 0) | |
612e59d8 CA |
6097 | /* chunk is in K */ |
6098 | minsize = chunk * 2; | |
efb30e7f DW |
6099 | |
6100 | if (cnt < raiddisks || | |
6101 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6102 | maxsize < minsize || |
6103 | maxsize == 0) { | |
e7b84f9d | 6104 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
6105 | return 0; /* No enough free spaces large enough */ |
6106 | } | |
6107 | ||
6108 | if (size == 0) { | |
6109 | size = maxsize; | |
6110 | if (chunk) { | |
612e59d8 CA |
6111 | size /= 2 * chunk; |
6112 | size *= 2 * chunk; | |
efb30e7f | 6113 | } |
f878b242 LM |
6114 | maxsize = size; |
6115 | } | |
6116 | if (!check_env("IMSM_NO_PLATFORM") && | |
6117 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
e7b84f9d | 6118 | pr_err("attempting to create a second " |
f878b242 LM |
6119 | "volume with size less then remaining space. " |
6120 | "Aborting...\n"); | |
6121 | return 0; | |
efb30e7f | 6122 | } |
efb30e7f DW |
6123 | cnt = 0; |
6124 | for (dl = super->disks; dl; dl = dl->next) | |
6125 | if (dl->e) | |
6126 | dl->raiddisk = cnt++; | |
6127 | ||
6128 | *freesize = size; | |
6129 | ||
13bcac90 AK |
6130 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
6131 | ||
efb30e7f DW |
6132 | return 1; |
6133 | } | |
6134 | ||
13bcac90 AK |
6135 | static int reserve_space(struct supertype *st, int raiddisks, |
6136 | unsigned long long size, int chunk, | |
6137 | unsigned long long *freesize) | |
6138 | { | |
6139 | struct intel_super *super = st->sb; | |
6140 | struct dl *dl; | |
6141 | int cnt; | |
6142 | int rv = 0; | |
6143 | ||
6144 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
6145 | if (rv) { | |
6146 | cnt = 0; | |
6147 | for (dl = super->disks; dl; dl = dl->next) | |
6148 | if (dl->e) | |
6149 | dl->raiddisk = cnt++; | |
6150 | rv = 1; | |
6151 | } | |
6152 | ||
6153 | return rv; | |
6154 | } | |
6155 | ||
bf5a934a | 6156 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6157 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
6158 | char *dev, unsigned long long *freesize, |
6159 | int verbose) | |
6160 | { | |
6161 | int fd, cfd; | |
6162 | struct mdinfo *sra; | |
20cbe8d2 | 6163 | int is_member = 0; |
bf5a934a | 6164 | |
d54559f0 LM |
6165 | /* load capability |
6166 | * if given unused devices create a container | |
bf5a934a DW |
6167 | * if given given devices in a container create a member volume |
6168 | */ | |
6169 | if (level == LEVEL_CONTAINER) { | |
6170 | /* Must be a fresh device to add to a container */ | |
6171 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
6172 | raiddisks, |
6173 | chunk?*chunk:0, size, | |
bf5a934a DW |
6174 | dev, freesize, |
6175 | verbose); | |
6176 | } | |
9587c373 | 6177 | |
8592f29d | 6178 | if (!dev) { |
e91a3bad | 6179 | if (st->sb) { |
ca9de185 | 6180 | struct intel_super *super = st->sb; |
e91a3bad | 6181 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 6182 | raiddisks, chunk, size, |
e91a3bad LM |
6183 | verbose)) |
6184 | return 0; | |
efb30e7f DW |
6185 | /* we are being asked to automatically layout a |
6186 | * new volume based on the current contents of | |
6187 | * the container. If the the parameters can be | |
6188 | * satisfied reserve_space will record the disks, | |
6189 | * start offset, and size of the volume to be | |
6190 | * created. add_to_super and getinfo_super | |
6191 | * detect when autolayout is in progress. | |
6192 | */ | |
ca9de185 LM |
6193 | /* assuming that freesize is always given when array is |
6194 | created */ | |
6195 | if (super->orom && freesize) { | |
6196 | int count; | |
6197 | count = count_volumes(super->hba->path, | |
6198 | super->orom->dpa, verbose); | |
6199 | if (super->orom->vphba <= count) { | |
6200 | pr_vrb(": platform does not support more" | |
40110b90 | 6201 | " than %d raid volumes.\n", |
ca9de185 LM |
6202 | super->orom->vphba); |
6203 | return 0; | |
6204 | } | |
6205 | } | |
e91a3bad LM |
6206 | if (freesize) |
6207 | return reserve_space(st, raiddisks, size, | |
6208 | chunk?*chunk:0, freesize); | |
8592f29d N |
6209 | } |
6210 | return 1; | |
6211 | } | |
bf5a934a DW |
6212 | if (st->sb) { |
6213 | /* creating in a given container */ | |
6214 | return validate_geometry_imsm_volume(st, level, layout, | |
6215 | raiddisks, chunk, size, | |
6216 | dev, freesize, verbose); | |
6217 | } | |
6218 | ||
bf5a934a DW |
6219 | /* This device needs to be a device in an 'imsm' container */ |
6220 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6221 | if (fd >= 0) { | |
6222 | if (verbose) | |
e7b84f9d N |
6223 | pr_err("Cannot create this array on device %s\n", |
6224 | dev); | |
bf5a934a DW |
6225 | close(fd); |
6226 | return 0; | |
6227 | } | |
6228 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6229 | if (verbose) | |
e7b84f9d | 6230 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
6231 | dev, strerror(errno)); |
6232 | return 0; | |
6233 | } | |
6234 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6235 | cfd = open_container(fd); | |
20cbe8d2 | 6236 | close(fd); |
bf5a934a | 6237 | if (cfd < 0) { |
bf5a934a | 6238 | if (verbose) |
e7b84f9d | 6239 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
6240 | dev); |
6241 | return 0; | |
6242 | } | |
6243 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 6244 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
6245 | strcmp(sra->text_version, "imsm") == 0) |
6246 | is_member = 1; | |
6247 | sysfs_free(sra); | |
6248 | if (is_member) { | |
bf5a934a DW |
6249 | /* This is a member of a imsm container. Load the container |
6250 | * and try to create a volume | |
6251 | */ | |
6252 | struct intel_super *super; | |
6253 | ||
ec50f7b6 | 6254 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a DW |
6255 | st->sb = super; |
6256 | st->container_dev = fd2devnum(cfd); | |
6257 | close(cfd); | |
6258 | return validate_geometry_imsm_volume(st, level, layout, | |
6259 | raiddisks, chunk, | |
6260 | size, dev, | |
ecbd9e81 N |
6261 | freesize, 1) |
6262 | ? 1 : -1; | |
bf5a934a | 6263 | } |
20cbe8d2 | 6264 | } |
bf5a934a | 6265 | |
20cbe8d2 | 6266 | if (verbose) |
e7b84f9d | 6267 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
6268 | |
6269 | close(cfd); | |
6270 | return 0; | |
bf5a934a | 6271 | } |
0bd16cf2 | 6272 | |
30f58b22 | 6273 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
6274 | { |
6275 | struct intel_super *super = st->sb; | |
6276 | ||
30f58b22 DW |
6277 | if (level && *level == UnSet) |
6278 | *level = LEVEL_CONTAINER; | |
6279 | ||
6280 | if (level && layout && *layout == UnSet) | |
6281 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 6282 | |
cd9d1ac7 DW |
6283 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
6284 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
6285 | } |
6286 | ||
33414a01 DW |
6287 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
6288 | ||
6289 | static int kill_subarray_imsm(struct supertype *st) | |
6290 | { | |
6291 | /* remove the subarray currently referenced by ->current_vol */ | |
6292 | __u8 i; | |
6293 | struct intel_dev **dp; | |
6294 | struct intel_super *super = st->sb; | |
6295 | __u8 current_vol = super->current_vol; | |
6296 | struct imsm_super *mpb = super->anchor; | |
6297 | ||
6298 | if (super->current_vol < 0) | |
6299 | return 2; | |
6300 | super->current_vol = -1; /* invalidate subarray cursor */ | |
6301 | ||
6302 | /* block deletions that would change the uuid of active subarrays | |
6303 | * | |
6304 | * FIXME when immutable ids are available, but note that we'll | |
6305 | * also need to fixup the invalidated/active subarray indexes in | |
6306 | * mdstat | |
6307 | */ | |
6308 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6309 | char subarray[4]; | |
6310 | ||
6311 | if (i < current_vol) | |
6312 | continue; | |
6313 | sprintf(subarray, "%u", i); | |
6314 | if (is_subarray_active(subarray, st->devname)) { | |
e7b84f9d N |
6315 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
6316 | current_vol, i); | |
33414a01 DW |
6317 | |
6318 | return 2; | |
6319 | } | |
6320 | } | |
6321 | ||
6322 | if (st->update_tail) { | |
503975b9 | 6323 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 6324 | |
33414a01 DW |
6325 | u->type = update_kill_array; |
6326 | u->dev_idx = current_vol; | |
6327 | append_metadata_update(st, u, sizeof(*u)); | |
6328 | ||
6329 | return 0; | |
6330 | } | |
6331 | ||
6332 | for (dp = &super->devlist; *dp;) | |
6333 | if ((*dp)->index == current_vol) { | |
6334 | *dp = (*dp)->next; | |
6335 | } else { | |
6336 | handle_missing(super, (*dp)->dev); | |
6337 | if ((*dp)->index > current_vol) | |
6338 | (*dp)->index--; | |
6339 | dp = &(*dp)->next; | |
6340 | } | |
6341 | ||
6342 | /* no more raid devices, all active components are now spares, | |
6343 | * but of course failed are still failed | |
6344 | */ | |
6345 | if (--mpb->num_raid_devs == 0) { | |
6346 | struct dl *d; | |
6347 | ||
6348 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
6349 | if (d->index > -2) |
6350 | mark_spare(d); | |
33414a01 DW |
6351 | } |
6352 | ||
6353 | super->updates_pending++; | |
6354 | ||
6355 | return 0; | |
6356 | } | |
aa534678 | 6357 | |
a951a4f7 | 6358 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 6359 | char *update, struct mddev_ident *ident) |
aa534678 DW |
6360 | { |
6361 | /* update the subarray currently referenced by ->current_vol */ | |
6362 | struct intel_super *super = st->sb; | |
6363 | struct imsm_super *mpb = super->anchor; | |
6364 | ||
aa534678 DW |
6365 | if (strcmp(update, "name") == 0) { |
6366 | char *name = ident->name; | |
a951a4f7 N |
6367 | char *ep; |
6368 | int vol; | |
aa534678 | 6369 | |
a951a4f7 | 6370 | if (is_subarray_active(subarray, st->devname)) { |
e7b84f9d | 6371 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
6372 | return 2; |
6373 | } | |
6374 | ||
6375 | if (!check_name(super, name, 0)) | |
6376 | return 2; | |
6377 | ||
a951a4f7 N |
6378 | vol = strtoul(subarray, &ep, 10); |
6379 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
6380 | return 2; | |
6381 | ||
aa534678 | 6382 | if (st->update_tail) { |
503975b9 | 6383 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 6384 | |
aa534678 | 6385 | u->type = update_rename_array; |
a951a4f7 | 6386 | u->dev_idx = vol; |
aa534678 DW |
6387 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
6388 | append_metadata_update(st, u, sizeof(*u)); | |
6389 | } else { | |
6390 | struct imsm_dev *dev; | |
6391 | int i; | |
6392 | ||
a951a4f7 | 6393 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
6394 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
6395 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6396 | dev = get_imsm_dev(super, i); | |
6397 | handle_missing(super, dev); | |
6398 | } | |
6399 | super->updates_pending++; | |
6400 | } | |
6401 | } else | |
6402 | return 2; | |
6403 | ||
6404 | return 0; | |
6405 | } | |
d1e02575 | 6406 | #endif /* MDASSEMBLE */ |
bf5a934a | 6407 | |
28bce06f AK |
6408 | static int is_gen_migration(struct imsm_dev *dev) |
6409 | { | |
7534230b AK |
6410 | if (dev == NULL) |
6411 | return 0; | |
6412 | ||
28bce06f AK |
6413 | if (!dev->vol.migr_state) |
6414 | return 0; | |
6415 | ||
6416 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
6417 | return 1; | |
6418 | ||
6419 | return 0; | |
6420 | } | |
6421 | ||
1e5c6983 DW |
6422 | static int is_rebuilding(struct imsm_dev *dev) |
6423 | { | |
6424 | struct imsm_map *migr_map; | |
6425 | ||
6426 | if (!dev->vol.migr_state) | |
6427 | return 0; | |
6428 | ||
6429 | if (migr_type(dev) != MIGR_REBUILD) | |
6430 | return 0; | |
6431 | ||
238c0a71 | 6432 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
6433 | |
6434 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
6435 | return 1; | |
6436 | else | |
6437 | return 0; | |
6438 | } | |
6439 | ||
b4ab44d8 | 6440 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
6441 | static int is_initializing(struct imsm_dev *dev) |
6442 | { | |
6443 | struct imsm_map *migr_map; | |
6444 | ||
6445 | if (!dev->vol.migr_state) | |
6446 | return 0; | |
6447 | ||
6448 | if (migr_type(dev) != MIGR_INIT) | |
6449 | return 0; | |
6450 | ||
238c0a71 | 6451 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
6452 | |
6453 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6454 | return 1; | |
6455 | ||
6456 | return 0; | |
6ce1fbf1 | 6457 | } |
b4ab44d8 | 6458 | #endif |
6ce1fbf1 | 6459 | |
c47b0ff6 AK |
6460 | static void update_recovery_start(struct intel_super *super, |
6461 | struct imsm_dev *dev, | |
6462 | struct mdinfo *array) | |
1e5c6983 DW |
6463 | { |
6464 | struct mdinfo *rebuild = NULL; | |
6465 | struct mdinfo *d; | |
6466 | __u32 units; | |
6467 | ||
6468 | if (!is_rebuilding(dev)) | |
6469 | return; | |
6470 | ||
6471 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
6472 | for (d = array->devs; d; d = d->next) | |
6473 | if (d->recovery_start == 0) { | |
6474 | if (rebuild) | |
6475 | return; | |
6476 | rebuild = d; | |
6477 | } | |
6478 | ||
4363fd80 DW |
6479 | if (!rebuild) { |
6480 | /* (?) none of the disks are marked with | |
6481 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
6482 | * disk_ord_tbl was not correctly updated | |
6483 | */ | |
6484 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
6485 | return; | |
6486 | } | |
6487 | ||
1e5c6983 | 6488 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 6489 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
6490 | } |
6491 | ||
9e2d750d | 6492 | #ifndef MDASSEMBLE |
276d77db | 6493 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 6494 | #endif |
1e5c6983 | 6495 | |
00bbdbda | 6496 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 6497 | { |
4f5bc454 DW |
6498 | /* Given a container loaded by load_super_imsm_all, |
6499 | * extract information about all the arrays into | |
6500 | * an mdinfo tree. | |
00bbdbda | 6501 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
6502 | * |
6503 | * For each imsm_dev create an mdinfo, fill it in, | |
6504 | * then look for matching devices in super->disks | |
6505 | * and create appropriate device mdinfo. | |
6506 | */ | |
6507 | struct intel_super *super = st->sb; | |
949c47a0 | 6508 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 6509 | struct mdinfo *rest = NULL; |
00bbdbda | 6510 | unsigned int i; |
81219e70 | 6511 | int sb_errors = 0; |
abef11a3 AK |
6512 | struct dl *d; |
6513 | int spare_disks = 0; | |
cdddbdbc | 6514 | |
19482bcc AK |
6515 | /* do not assemble arrays when not all attributes are supported */ |
6516 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 6517 | sb_errors = 1; |
e7b84f9d | 6518 | pr_err("Unsupported attributes in IMSM metadata." |
81219e70 | 6519 | "Arrays activation is blocked.\n"); |
19482bcc AK |
6520 | } |
6521 | ||
a06d022d | 6522 | /* check for bad blocks */ |
81219e70 | 6523 | if (imsm_bbm_log_size(super->anchor)) { |
e7b84f9d N |
6524 | pr_err("BBM log found in IMSM metadata." |
6525 | "Arrays activation is blocked.\n"); | |
81219e70 LM |
6526 | sb_errors = 1; |
6527 | } | |
6528 | ||
604b746f | 6529 | |
abef11a3 AK |
6530 | /* count spare devices, not used in maps |
6531 | */ | |
6532 | for (d = super->disks; d; d = d->next) | |
6533 | if (d->index == -1) | |
6534 | spare_disks++; | |
6535 | ||
4f5bc454 | 6536 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
6537 | struct imsm_dev *dev; |
6538 | struct imsm_map *map; | |
86e3692b | 6539 | struct imsm_map *map2; |
4f5bc454 | 6540 | struct mdinfo *this; |
a6482415 N |
6541 | int slot; |
6542 | #ifndef MDASSEMBLE | |
6543 | int chunk; | |
6544 | #endif | |
00bbdbda N |
6545 | char *ep; |
6546 | ||
6547 | if (subarray && | |
6548 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
6549 | continue; | |
6550 | ||
6551 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
6552 | map = get_imsm_map(dev, MAP_0); |
6553 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 6554 | |
1ce0101c DW |
6555 | /* do not publish arrays that are in the middle of an |
6556 | * unsupported migration | |
6557 | */ | |
6558 | if (dev->vol.migr_state && | |
28bce06f | 6559 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
e7b84f9d | 6560 | pr_err("cannot assemble volume '%.16s':" |
1ce0101c DW |
6561 | " unsupported migration in progress\n", |
6562 | dev->volume); | |
6563 | continue; | |
6564 | } | |
2db86302 LM |
6565 | /* do not publish arrays that are not support by controller's |
6566 | * OROM/EFI | |
6567 | */ | |
1ce0101c | 6568 | |
503975b9 | 6569 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 6570 | |
301406c9 | 6571 | super->current_vol = i; |
a5d85af7 | 6572 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 6573 | this->next = rest; |
81219e70 | 6574 | #ifndef MDASSEMBLE |
a6482415 | 6575 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
6576 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
6577 | if (!validate_geometry_imsm_orom(super, | |
6578 | get_imsm_raid_level(map), /* RAID level */ | |
6579 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
6580 | map->num_members, /* raid disks */ | |
2cc699af | 6581 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 6582 | 1 /* verbose */)) { |
e7b84f9d | 6583 | pr_err("IMSM RAID geometry validation" |
446894ea | 6584 | " failed. Array %s activation is blocked.\n", |
81219e70 LM |
6585 | dev->volume); |
6586 | this->array.state |= | |
6587 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6588 | (1<<MD_SB_BLOCK_VOLUME); | |
6589 | } | |
6590 | #endif | |
6591 | ||
6592 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
6593 | if (sb_errors) | |
6594 | this->array.state |= | |
6595 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6596 | (1<<MD_SB_BLOCK_VOLUME); | |
6597 | ||
4f5bc454 | 6598 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 6599 | unsigned long long recovery_start; |
4f5bc454 DW |
6600 | struct mdinfo *info_d; |
6601 | struct dl *d; | |
6602 | int idx; | |
9a1608e5 | 6603 | int skip; |
7eef0453 | 6604 | __u32 ord; |
4f5bc454 | 6605 | |
9a1608e5 | 6606 | skip = 0; |
238c0a71 AK |
6607 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
6608 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
6609 | for (d = super->disks; d ; d = d->next) |
6610 | if (d->index == idx) | |
0fbd635c | 6611 | break; |
4f5bc454 | 6612 | |
1e5c6983 | 6613 | recovery_start = MaxSector; |
4f5bc454 | 6614 | if (d == NULL) |
9a1608e5 | 6615 | skip = 1; |
25ed7e59 | 6616 | if (d && is_failed(&d->disk)) |
9a1608e5 | 6617 | skip = 1; |
7eef0453 | 6618 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 6619 | recovery_start = 0; |
9a1608e5 DW |
6620 | |
6621 | /* | |
6622 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
6623 | * reset resync start to avoid a dirty-degraded |
6624 | * situation when performing the intial sync | |
9a1608e5 DW |
6625 | * |
6626 | * FIXME handle dirty degraded | |
6627 | */ | |
1e5c6983 | 6628 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 6629 | this->resync_start = MaxSector; |
9a1608e5 DW |
6630 | if (skip) |
6631 | continue; | |
4f5bc454 | 6632 | |
503975b9 | 6633 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
6634 | info_d->next = this->devs; |
6635 | this->devs = info_d; | |
6636 | ||
4f5bc454 DW |
6637 | info_d->disk.number = d->index; |
6638 | info_d->disk.major = d->major; | |
6639 | info_d->disk.minor = d->minor; | |
6640 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 6641 | info_d->recovery_start = recovery_start; |
86e3692b AK |
6642 | if (map2) { |
6643 | if (slot < map2->num_members) | |
6644 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6645 | else |
6646 | this->array.spare_disks++; | |
86e3692b AK |
6647 | } else { |
6648 | if (slot < map->num_members) | |
6649 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6650 | else |
6651 | this->array.spare_disks++; | |
86e3692b | 6652 | } |
1e5c6983 DW |
6653 | if (info_d->recovery_start == MaxSector) |
6654 | this->array.working_disks++; | |
4f5bc454 DW |
6655 | |
6656 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 CA |
6657 | info_d->data_offset = pba_of_lba0(map); |
6658 | info_d->component_size = blocks_per_member(map); | |
4f5bc454 | 6659 | } |
1e5c6983 | 6660 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 6661 | update_recovery_start(super, dev, this); |
abef11a3 | 6662 | this->array.spare_disks += spare_disks; |
276d77db | 6663 | |
9e2d750d | 6664 | #ifndef MDASSEMBLE |
276d77db AK |
6665 | /* check for reshape */ |
6666 | if (this->reshape_active == 1) | |
6667 | recover_backup_imsm(st, this); | |
9e2d750d | 6668 | #endif |
9a1608e5 | 6669 | rest = this; |
4f5bc454 DW |
6670 | } |
6671 | ||
6672 | return rest; | |
cdddbdbc DW |
6673 | } |
6674 | ||
845dea95 | 6675 | |
3b451610 AK |
6676 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
6677 | int failed, int look_in_map) | |
c2a1e7da | 6678 | { |
3b451610 AK |
6679 | struct imsm_map *map; |
6680 | ||
6681 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
6682 | |
6683 | if (!failed) | |
3393c6af DW |
6684 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
6685 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
6686 | |
6687 | switch (get_imsm_raid_level(map)) { | |
6688 | case 0: | |
6689 | return IMSM_T_STATE_FAILED; | |
6690 | break; | |
6691 | case 1: | |
6692 | if (failed < map->num_members) | |
6693 | return IMSM_T_STATE_DEGRADED; | |
6694 | else | |
6695 | return IMSM_T_STATE_FAILED; | |
6696 | break; | |
6697 | case 10: | |
6698 | { | |
6699 | /** | |
c92a2527 DW |
6700 | * check to see if any mirrors have failed, otherwise we |
6701 | * are degraded. Even numbered slots are mirrored on | |
6702 | * slot+1 | |
c2a1e7da | 6703 | */ |
c2a1e7da | 6704 | int i; |
d9b420a5 N |
6705 | /* gcc -Os complains that this is unused */ |
6706 | int insync = insync; | |
c2a1e7da DW |
6707 | |
6708 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6709 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6710 | int idx = ord_to_idx(ord); |
6711 | struct imsm_disk *disk; | |
c2a1e7da | 6712 | |
c92a2527 DW |
6713 | /* reset the potential in-sync count on even-numbered |
6714 | * slots. num_copies is always 2 for imsm raid10 | |
6715 | */ | |
6716 | if ((i & 1) == 0) | |
6717 | insync = 2; | |
c2a1e7da | 6718 | |
c92a2527 | 6719 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6720 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6721 | insync--; |
c2a1e7da | 6722 | |
c92a2527 DW |
6723 | /* no in-sync disks left in this mirror the |
6724 | * array has failed | |
6725 | */ | |
6726 | if (insync == 0) | |
6727 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6728 | } |
6729 | ||
6730 | return IMSM_T_STATE_DEGRADED; | |
6731 | } | |
6732 | case 5: | |
6733 | if (failed < 2) | |
6734 | return IMSM_T_STATE_DEGRADED; | |
6735 | else | |
6736 | return IMSM_T_STATE_FAILED; | |
6737 | break; | |
6738 | default: | |
6739 | break; | |
6740 | } | |
6741 | ||
6742 | return map->map_state; | |
6743 | } | |
6744 | ||
3b451610 AK |
6745 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6746 | int look_in_map) | |
c2a1e7da DW |
6747 | { |
6748 | int i; | |
6749 | int failed = 0; | |
6750 | struct imsm_disk *disk; | |
d5985138 AK |
6751 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6752 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6753 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6754 | __u32 ord; |
6755 | int idx; | |
d5985138 | 6756 | int idx_1; |
c2a1e7da | 6757 | |
0556e1a2 DW |
6758 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6759 | * disks that are being rebuilt. New failures are recorded to | |
6760 | * map[0]. So we look through all the disks we started with and | |
6761 | * see if any failures are still present, or if any new ones | |
6762 | * have arrived | |
0556e1a2 | 6763 | */ |
d5985138 AK |
6764 | map_for_loop = map; |
6765 | if (prev && (map->num_members < prev->num_members)) | |
6766 | map_for_loop = prev; | |
68fe4598 LD |
6767 | |
6768 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6769 | idx_1 = -255; |
238c0a71 AK |
6770 | /* when MAP_X is passed both maps failures are counted |
6771 | */ | |
d5985138 | 6772 | if (prev && |
238c0a71 AK |
6773 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6774 | (i < prev->num_members)) { | |
d5985138 AK |
6775 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6776 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6777 | |
d5985138 AK |
6778 | disk = get_imsm_disk(super, idx_1); |
6779 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6780 | failed++; | |
6781 | } | |
238c0a71 AK |
6782 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6783 | (i < map->num_members)) { | |
d5985138 AK |
6784 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6785 | idx = ord_to_idx(ord); | |
6786 | ||
6787 | if (idx != idx_1) { | |
6788 | disk = get_imsm_disk(super, idx); | |
6789 | if (!disk || is_failed(disk) || | |
6790 | ord & IMSM_ORD_REBUILD) | |
6791 | failed++; | |
6792 | } | |
6793 | } | |
c2a1e7da DW |
6794 | } |
6795 | ||
6796 | return failed; | |
845dea95 NB |
6797 | } |
6798 | ||
97b4d0e9 DW |
6799 | #ifndef MDASSEMBLE |
6800 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
6801 | char *inst) | |
6802 | { | |
6803 | struct intel_super *super = c->sb; | |
6804 | struct imsm_super *mpb = super->anchor; | |
9587c373 | 6805 | |
97b4d0e9 DW |
6806 | if (atoi(inst) >= mpb->num_raid_devs) { |
6807 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
6808 | __func__, atoi(inst)); | |
6809 | return -ENODEV; | |
6810 | } | |
6811 | ||
6812 | dprintf("imsm: open_new %s\n", inst); | |
6813 | a->info.container_member = atoi(inst); | |
6814 | return 0; | |
6815 | } | |
6816 | ||
0c046afd DW |
6817 | static int is_resyncing(struct imsm_dev *dev) |
6818 | { | |
6819 | struct imsm_map *migr_map; | |
6820 | ||
6821 | if (!dev->vol.migr_state) | |
6822 | return 0; | |
6823 | ||
1484e727 DW |
6824 | if (migr_type(dev) == MIGR_INIT || |
6825 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
6826 | return 1; |
6827 | ||
4c9bc37b AK |
6828 | if (migr_type(dev) == MIGR_GEN_MIGR) |
6829 | return 0; | |
6830 | ||
238c0a71 | 6831 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 6832 | |
4c9bc37b AK |
6833 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
6834 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
6835 | return 1; |
6836 | else | |
6837 | return 0; | |
6838 | } | |
6839 | ||
0556e1a2 DW |
6840 | /* return true if we recorded new information */ |
6841 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 6842 | { |
0556e1a2 DW |
6843 | __u32 ord; |
6844 | int slot; | |
6845 | struct imsm_map *map; | |
86c54047 DW |
6846 | char buf[MAX_RAID_SERIAL_LEN+3]; |
6847 | unsigned int len, shift = 0; | |
0556e1a2 DW |
6848 | |
6849 | /* new failures are always set in map[0] */ | |
238c0a71 | 6850 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
6851 | |
6852 | slot = get_imsm_disk_slot(map, idx); | |
6853 | if (slot < 0) | |
6854 | return 0; | |
6855 | ||
6856 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 6857 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
6858 | return 0; |
6859 | ||
7d0c5e24 LD |
6860 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
6861 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
6862 | strcat(buf, ":0"); | |
86c54047 DW |
6863 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
6864 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
6865 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
6866 | ||
f2f27e63 | 6867 | disk->status |= FAILED_DISK; |
0556e1a2 | 6868 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
6869 | /* mark failures in second map if second map exists and this disk |
6870 | * in this slot. | |
6871 | * This is valid for migration, initialization and rebuild | |
6872 | */ | |
6873 | if (dev->vol.migr_state) { | |
238c0a71 | 6874 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
6875 | int slot2 = get_imsm_disk_slot(map2, idx); |
6876 | ||
6877 | if ((slot2 < map2->num_members) && | |
6878 | (slot2 >= 0)) | |
6879 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
6880 | idx | IMSM_ORD_REBUILD); |
6881 | } | |
f21e18ca | 6882 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
6883 | map->failed_disk_num = slot; |
6884 | return 1; | |
6885 | } | |
6886 | ||
6887 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
6888 | { | |
6889 | mark_failure(dev, disk, idx); | |
6890 | ||
6891 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
6892 | return; | |
6893 | ||
47ee5a45 DW |
6894 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
6895 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
6896 | } | |
6897 | ||
33414a01 DW |
6898 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
6899 | { | |
33414a01 | 6900 | struct dl *dl; |
33414a01 DW |
6901 | |
6902 | if (!super->missing) | |
6903 | return; | |
33414a01 DW |
6904 | |
6905 | dprintf("imsm: mark missing\n"); | |
3d59f0c0 AK |
6906 | /* end process for initialization and rebuild only |
6907 | */ | |
6908 | if (is_gen_migration(dev) == 0) { | |
6909 | __u8 map_state; | |
6910 | int failed; | |
6911 | ||
6912 | failed = imsm_count_failed(super, dev, MAP_0); | |
6913 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
6914 | ||
6915 | end_migration(dev, super, map_state); | |
6916 | } | |
33414a01 DW |
6917 | for (dl = super->missing; dl; dl = dl->next) |
6918 | mark_missing(dev, &dl->disk, dl->index); | |
6919 | super->updates_pending++; | |
6920 | } | |
6921 | ||
f3871fdc AK |
6922 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
6923 | long long new_size) | |
70bdf0dc | 6924 | { |
238c0a71 | 6925 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
6926 | unsigned long long array_blocks; |
6927 | struct imsm_map *map; | |
6928 | ||
6929 | if (used_disks == 0) { | |
6930 | /* when problems occures | |
6931 | * return current array_blocks value | |
6932 | */ | |
6933 | array_blocks = __le32_to_cpu(dev->size_high); | |
6934 | array_blocks = array_blocks << 32; | |
6935 | array_blocks += __le32_to_cpu(dev->size_low); | |
6936 | ||
6937 | return array_blocks; | |
6938 | } | |
6939 | ||
6940 | /* set array size in metadata | |
6941 | */ | |
f3871fdc AK |
6942 | if (new_size <= 0) { |
6943 | /* OLCE size change is caused by added disks | |
6944 | */ | |
6945 | map = get_imsm_map(dev, MAP_0); | |
6946 | array_blocks = blocks_per_member(map) * used_disks; | |
6947 | } else { | |
6948 | /* Online Volume Size Change | |
6949 | * Using available free space | |
6950 | */ | |
6951 | array_blocks = new_size; | |
6952 | } | |
70bdf0dc AK |
6953 | |
6954 | /* round array size down to closest MB | |
6955 | */ | |
6956 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
6957 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
6958 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
6959 | ||
6960 | return array_blocks; | |
6961 | } | |
6962 | ||
28bce06f AK |
6963 | static void imsm_set_disk(struct active_array *a, int n, int state); |
6964 | ||
0e2d1a4e AK |
6965 | static void imsm_progress_container_reshape(struct intel_super *super) |
6966 | { | |
6967 | /* if no device has a migr_state, but some device has a | |
6968 | * different number of members than the previous device, start | |
6969 | * changing the number of devices in this device to match | |
6970 | * previous. | |
6971 | */ | |
6972 | struct imsm_super *mpb = super->anchor; | |
6973 | int prev_disks = -1; | |
6974 | int i; | |
1dfaa380 | 6975 | int copy_map_size; |
0e2d1a4e AK |
6976 | |
6977 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6978 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 6979 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
6980 | struct imsm_map *map2; |
6981 | int prev_num_members; | |
0e2d1a4e AK |
6982 | |
6983 | if (dev->vol.migr_state) | |
6984 | return; | |
6985 | ||
6986 | if (prev_disks == -1) | |
6987 | prev_disks = map->num_members; | |
6988 | if (prev_disks == map->num_members) | |
6989 | continue; | |
6990 | ||
6991 | /* OK, this array needs to enter reshape mode. | |
6992 | * i.e it needs a migr_state | |
6993 | */ | |
6994 | ||
1dfaa380 | 6995 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
6996 | prev_num_members = map->num_members; |
6997 | map->num_members = prev_disks; | |
6998 | dev->vol.migr_state = 1; | |
6999 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7000 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
7001 | for (i = prev_num_members; |
7002 | i < map->num_members; i++) | |
7003 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 7004 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 7005 | /* Copy the current map */ |
1dfaa380 | 7006 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
7007 | map2->num_members = prev_num_members; |
7008 | ||
f3871fdc | 7009 | imsm_set_array_size(dev, -1); |
51d83f5d | 7010 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
7011 | super->updates_pending++; |
7012 | } | |
7013 | } | |
7014 | ||
aad6f216 | 7015 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
7016 | * states are handled in imsm_set_disk() with one exception, when a |
7017 | * resync is stopped due to a new failure this routine will set the | |
7018 | * 'degraded' state for the array. | |
7019 | */ | |
01f157d7 | 7020 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
7021 | { |
7022 | int inst = a->info.container_member; | |
7023 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7024 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7025 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
7026 | int failed = imsm_count_failed(super, dev, MAP_0); |
7027 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 7028 | __u32 blocks_per_unit; |
a862209d | 7029 | |
1af97990 AK |
7030 | if (dev->vol.migr_state && |
7031 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
7032 | /* array state change is blocked due to reshape action | |
aad6f216 N |
7033 | * We might need to |
7034 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
7035 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
7036 | * - update curr_migr_unit | |
1af97990 | 7037 | */ |
aad6f216 N |
7038 | if (a->curr_action == reshape) { |
7039 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 7040 | goto mark_checkpoint; |
aad6f216 N |
7041 | } else { |
7042 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
7043 | /* for some reason we aborted the reshape. | |
b66e591b AK |
7044 | * |
7045 | * disable automatic metadata rollback | |
7046 | * user action is required to recover process | |
aad6f216 | 7047 | */ |
b66e591b | 7048 | if (0) { |
238c0a71 AK |
7049 | struct imsm_map *map2 = |
7050 | get_imsm_map(dev, MAP_1); | |
7051 | dev->vol.migr_state = 0; | |
7052 | set_migr_type(dev, 0); | |
7053 | dev->vol.curr_migr_unit = 0; | |
7054 | memcpy(map, map2, | |
7055 | sizeof_imsm_map(map2)); | |
7056 | super->updates_pending++; | |
b66e591b | 7057 | } |
aad6f216 N |
7058 | } |
7059 | if (a->last_checkpoint >= a->info.component_size) { | |
7060 | unsigned long long array_blocks; | |
7061 | int used_disks; | |
e154ced3 | 7062 | struct mdinfo *mdi; |
aad6f216 | 7063 | |
238c0a71 | 7064 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
7065 | if (used_disks > 0) { |
7066 | array_blocks = | |
5551b113 | 7067 | blocks_per_member(map) * |
d55adef9 AK |
7068 | used_disks; |
7069 | /* round array size down to closest MB | |
7070 | */ | |
7071 | array_blocks = (array_blocks | |
7072 | >> SECT_PER_MB_SHIFT) | |
7073 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
7074 | a->info.custom_array_size = array_blocks; |
7075 | /* encourage manager to update array | |
7076 | * size | |
7077 | */ | |
e154ced3 | 7078 | |
d55adef9 | 7079 | a->check_reshape = 1; |
633b5610 | 7080 | } |
e154ced3 AK |
7081 | /* finalize online capacity expansion/reshape */ |
7082 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
7083 | imsm_set_disk(a, | |
7084 | mdi->disk.raid_disk, | |
7085 | mdi->curr_state); | |
7086 | ||
0e2d1a4e | 7087 | imsm_progress_container_reshape(super); |
e154ced3 | 7088 | } |
aad6f216 | 7089 | } |
1af97990 AK |
7090 | } |
7091 | ||
47ee5a45 | 7092 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
7093 | if (consistent == 2) |
7094 | handle_missing(super, dev); | |
1e5c6983 | 7095 | |
0c046afd | 7096 | if (consistent == 2 && |
b7941fd6 | 7097 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7098 | map_state != IMSM_T_STATE_NORMAL || |
7099 | dev->vol.migr_state)) | |
01f157d7 | 7100 | consistent = 0; |
272906ef | 7101 | |
b7941fd6 | 7102 | if (is_resync_complete(&a->info)) { |
0c046afd | 7103 | /* complete intialization / resync, |
0556e1a2 DW |
7104 | * recovery and interrupted recovery is completed in |
7105 | * ->set_disk | |
0c046afd DW |
7106 | */ |
7107 | if (is_resyncing(dev)) { | |
7108 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7109 | end_migration(dev, super, map_state); |
115c3803 | 7110 | super->updates_pending++; |
484240d8 | 7111 | a->last_checkpoint = 0; |
115c3803 | 7112 | } |
b9172665 AK |
7113 | } else if ((!is_resyncing(dev) && !failed) && |
7114 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7115 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7116 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7117 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7118 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7119 | else |
8e59f3d8 | 7120 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7121 | super->updates_pending++; |
115c3803 | 7122 | } |
a862209d | 7123 | |
633b5610 | 7124 | mark_checkpoint: |
5b83bacf AK |
7125 | /* skip checkpointing for general migration, |
7126 | * it is controlled in mdadm | |
7127 | */ | |
7128 | if (is_gen_migration(dev)) | |
7129 | goto skip_mark_checkpoint; | |
7130 | ||
1e5c6983 | 7131 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7132 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7133 | if (blocks_per_unit) { |
1e5c6983 DW |
7134 | __u32 units32; |
7135 | __u64 units; | |
7136 | ||
4f0a7acc | 7137 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7138 | units32 = units; |
7139 | ||
7140 | /* check that we did not overflow 32-bits, and that | |
7141 | * curr_migr_unit needs updating | |
7142 | */ | |
7143 | if (units32 == units && | |
bfd80a56 | 7144 | units32 != 0 && |
1e5c6983 DW |
7145 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7146 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7147 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7148 | super->updates_pending++; | |
7149 | } | |
7150 | } | |
f8f603f1 | 7151 | |
5b83bacf | 7152 | skip_mark_checkpoint: |
3393c6af | 7153 | /* mark dirty / clean */ |
0c046afd | 7154 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7155 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7156 | if (consistent) |
7157 | dev->vol.dirty = 0; | |
7158 | else | |
7159 | dev->vol.dirty = 1; | |
a862209d DW |
7160 | super->updates_pending++; |
7161 | } | |
28bce06f | 7162 | |
01f157d7 | 7163 | return consistent; |
a862209d DW |
7164 | } |
7165 | ||
8d45d196 | 7166 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7167 | { |
8d45d196 DW |
7168 | int inst = a->info.container_member; |
7169 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7170 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7171 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7172 | struct imsm_disk *disk; |
7ce05701 LD |
7173 | struct mdinfo *mdi; |
7174 | int recovery_not_finished = 0; | |
0c046afd | 7175 | int failed; |
b10b37b8 | 7176 | __u32 ord; |
0c046afd | 7177 | __u8 map_state; |
8d45d196 DW |
7178 | |
7179 | if (n > map->num_members) | |
7180 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
7181 | n, map->num_members - 1); | |
7182 | ||
7183 | if (n < 0) | |
7184 | return; | |
7185 | ||
4e6e574a | 7186 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 7187 | |
238c0a71 | 7188 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 7189 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7190 | |
5802a811 | 7191 | /* check for new failures */ |
0556e1a2 DW |
7192 | if (state & DS_FAULTY) { |
7193 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
7194 | super->updates_pending++; | |
8d45d196 | 7195 | } |
47ee5a45 | 7196 | |
19859edc | 7197 | /* check if in_sync */ |
0556e1a2 | 7198 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7199 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7200 | |
7201 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7202 | super->updates_pending++; |
7203 | } | |
8d45d196 | 7204 | |
3b451610 AK |
7205 | failed = imsm_count_failed(super, dev, MAP_0); |
7206 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7207 | |
0c046afd | 7208 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7209 | dprintf("imsm: Detected transition to state "); |
7210 | switch (map_state) { | |
7211 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
7212 | dprintf("normal: "); | |
7213 | if (is_rebuilding(dev)) { | |
7214 | dprintf("while rebuilding"); | |
7ce05701 LD |
7215 | /* check if recovery is really finished */ |
7216 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
7217 | if (mdi->recovery_start != MaxSector) { | |
7218 | recovery_not_finished = 1; | |
7219 | break; | |
7220 | } | |
7221 | if (recovery_not_finished) { | |
7222 | dprintf("\nimsm: Rebuild has not finished yet, " | |
7223 | "state not changed"); | |
7224 | if (a->last_checkpoint < mdi->recovery_start) { | |
7225 | a->last_checkpoint = mdi->recovery_start; | |
7226 | super->updates_pending++; | |
7227 | } | |
7228 | break; | |
7229 | } | |
94002678 | 7230 | end_migration(dev, super, map_state); |
238c0a71 | 7231 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
7232 | map->failed_disk_num = ~0; |
7233 | super->updates_pending++; | |
7234 | a->last_checkpoint = 0; | |
7235 | break; | |
7236 | } | |
7237 | if (is_gen_migration(dev)) { | |
7238 | dprintf("while general migration"); | |
bf2f0071 | 7239 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7240 | end_migration(dev, super, map_state); |
94002678 AK |
7241 | else |
7242 | map->map_state = map_state; | |
238c0a71 | 7243 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 7244 | map->failed_disk_num = ~0; |
94002678 | 7245 | super->updates_pending++; |
bf2f0071 | 7246 | break; |
94002678 AK |
7247 | } |
7248 | break; | |
7249 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
7250 | dprintf("degraded: "); | |
7251 | if ((map->map_state != map_state) && | |
7252 | !dev->vol.migr_state) { | |
7253 | dprintf("mark degraded"); | |
7254 | map->map_state = map_state; | |
7255 | super->updates_pending++; | |
7256 | a->last_checkpoint = 0; | |
7257 | break; | |
7258 | } | |
7259 | if (is_rebuilding(dev)) { | |
7260 | dprintf("while rebuilding."); | |
7261 | if (map->map_state != map_state) { | |
7262 | dprintf(" Map state change"); | |
7263 | end_migration(dev, super, map_state); | |
7264 | super->updates_pending++; | |
7265 | } | |
7266 | break; | |
7267 | } | |
7268 | if (is_gen_migration(dev)) { | |
7269 | dprintf("while general migration"); | |
bf2f0071 | 7270 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7271 | end_migration(dev, super, map_state); |
94002678 AK |
7272 | else { |
7273 | map->map_state = map_state; | |
3b451610 | 7274 | manage_second_map(super, dev); |
94002678 AK |
7275 | } |
7276 | super->updates_pending++; | |
bf2f0071 | 7277 | break; |
28bce06f | 7278 | } |
6ce1fbf1 AK |
7279 | if (is_initializing(dev)) { |
7280 | dprintf("while initialization."); | |
7281 | map->map_state = map_state; | |
7282 | super->updates_pending++; | |
7283 | break; | |
7284 | } | |
94002678 AK |
7285 | break; |
7286 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
7287 | dprintf("failed: "); | |
7288 | if (is_gen_migration(dev)) { | |
7289 | dprintf("while general migration"); | |
7290 | map->map_state = map_state; | |
7291 | super->updates_pending++; | |
7292 | break; | |
7293 | } | |
7294 | if (map->map_state != map_state) { | |
7295 | dprintf("mark failed"); | |
7296 | end_migration(dev, super, map_state); | |
7297 | super->updates_pending++; | |
7298 | a->last_checkpoint = 0; | |
7299 | break; | |
7300 | } | |
7301 | break; | |
7302 | default: | |
7303 | dprintf("state %i\n", map_state); | |
5802a811 | 7304 | } |
94002678 AK |
7305 | dprintf("\n"); |
7306 | ||
845dea95 NB |
7307 | } |
7308 | ||
f796af5d | 7309 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 7310 | { |
f796af5d | 7311 | void *buf = mpb; |
c2a1e7da DW |
7312 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
7313 | unsigned long long dsize; | |
7314 | unsigned long long sectors; | |
7315 | ||
7316 | get_dev_size(fd, NULL, &dsize); | |
7317 | ||
272f648f DW |
7318 | if (mpb_size > 512) { |
7319 | /* -1 to account for anchor */ | |
7320 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 7321 | |
272f648f DW |
7322 | /* write the extended mpb to the sectors preceeding the anchor */ |
7323 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
7324 | return 1; | |
c2a1e7da | 7325 | |
f21e18ca N |
7326 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
7327 | != 512 * sectors) | |
272f648f DW |
7328 | return 1; |
7329 | } | |
c2a1e7da | 7330 | |
272f648f DW |
7331 | /* first block is stored on second to last sector of the disk */ |
7332 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
7333 | return 1; |
7334 | ||
f796af5d | 7335 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
7336 | return 1; |
7337 | ||
c2a1e7da DW |
7338 | return 0; |
7339 | } | |
7340 | ||
2e735d19 | 7341 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 7342 | { |
2e735d19 | 7343 | struct intel_super *super = container->sb; |
c2a1e7da | 7344 | |
1a64be56 | 7345 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
7346 | if (!super->updates_pending) |
7347 | return; | |
7348 | ||
36988a3d | 7349 | write_super_imsm(container, 0); |
c2a1e7da DW |
7350 | |
7351 | super->updates_pending = 0; | |
845dea95 NB |
7352 | } |
7353 | ||
272906ef DW |
7354 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
7355 | { | |
7356 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7357 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
7358 | struct dl *dl; |
7359 | ||
7360 | for (dl = super->disks; dl; dl = dl->next) | |
7361 | if (dl->index == i) | |
7362 | break; | |
7363 | ||
25ed7e59 | 7364 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
7365 | dl = NULL; |
7366 | ||
7367 | if (dl) | |
7368 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
7369 | ||
7370 | return dl; | |
7371 | } | |
7372 | ||
a20d2ba5 | 7373 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
7374 | struct active_array *a, int activate_new, |
7375 | struct mdinfo *additional_test_list) | |
272906ef DW |
7376 | { |
7377 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7378 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
7379 | struct imsm_super *mpb = super->anchor; |
7380 | struct imsm_map *map; | |
272906ef DW |
7381 | unsigned long long pos; |
7382 | struct mdinfo *d; | |
7383 | struct extent *ex; | |
a20d2ba5 | 7384 | int i, j; |
272906ef | 7385 | int found; |
569cc43f DW |
7386 | __u32 array_start = 0; |
7387 | __u32 array_end = 0; | |
272906ef | 7388 | struct dl *dl; |
6c932028 | 7389 | struct mdinfo *test_list; |
272906ef DW |
7390 | |
7391 | for (dl = super->disks; dl; dl = dl->next) { | |
7392 | /* If in this array, skip */ | |
7393 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
7394 | if (d->state_fd >= 0 && |
7395 | d->disk.major == dl->major && | |
272906ef | 7396 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
7397 | dprintf("%x:%x already in array\n", |
7398 | dl->major, dl->minor); | |
272906ef DW |
7399 | break; |
7400 | } | |
7401 | if (d) | |
7402 | continue; | |
6c932028 AK |
7403 | test_list = additional_test_list; |
7404 | while (test_list) { | |
7405 | if (test_list->disk.major == dl->major && | |
7406 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
7407 | dprintf("%x:%x already in additional test list\n", |
7408 | dl->major, dl->minor); | |
7409 | break; | |
7410 | } | |
6c932028 | 7411 | test_list = test_list->next; |
8ba77d32 | 7412 | } |
6c932028 | 7413 | if (test_list) |
8ba77d32 | 7414 | continue; |
272906ef | 7415 | |
e553d2a4 | 7416 | /* skip in use or failed drives */ |
25ed7e59 | 7417 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
7418 | dl->index == -2) { |
7419 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 7420 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
7421 | continue; |
7422 | } | |
7423 | ||
a20d2ba5 DW |
7424 | /* skip pure spares when we are looking for partially |
7425 | * assimilated drives | |
7426 | */ | |
7427 | if (dl->index == -1 && !activate_new) | |
7428 | continue; | |
7429 | ||
272906ef | 7430 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 7431 | * It needs to be able to cover all member volumes |
272906ef DW |
7432 | */ |
7433 | ex = get_extents(super, dl); | |
7434 | if (!ex) { | |
7435 | dprintf("cannot get extents\n"); | |
7436 | continue; | |
7437 | } | |
a20d2ba5 DW |
7438 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7439 | dev = get_imsm_dev(super, i); | |
238c0a71 | 7440 | map = get_imsm_map(dev, MAP_0); |
272906ef | 7441 | |
a20d2ba5 DW |
7442 | /* check if this disk is already a member of |
7443 | * this array | |
272906ef | 7444 | */ |
620b1713 | 7445 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
7446 | continue; |
7447 | ||
7448 | found = 0; | |
7449 | j = 0; | |
7450 | pos = 0; | |
5551b113 | 7451 | array_start = pba_of_lba0(map); |
329c8278 | 7452 | array_end = array_start + |
5551b113 | 7453 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
7454 | |
7455 | do { | |
7456 | /* check that we can start at pba_of_lba0 with | |
7457 | * blocks_per_member of space | |
7458 | */ | |
329c8278 | 7459 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
7460 | found = 1; |
7461 | break; | |
7462 | } | |
7463 | pos = ex[j].start + ex[j].size; | |
7464 | j++; | |
7465 | } while (ex[j-1].size); | |
7466 | ||
7467 | if (!found) | |
272906ef | 7468 | break; |
a20d2ba5 | 7469 | } |
272906ef DW |
7470 | |
7471 | free(ex); | |
a20d2ba5 | 7472 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
7473 | dprintf("%x:%x does not have %u to %u available\n", |
7474 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
7475 | /* No room */ |
7476 | continue; | |
a20d2ba5 DW |
7477 | } |
7478 | return dl; | |
272906ef DW |
7479 | } |
7480 | ||
7481 | return dl; | |
7482 | } | |
7483 | ||
95d07a2c LM |
7484 | |
7485 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
7486 | { | |
7487 | struct imsm_dev *dev2; | |
7488 | struct imsm_map *map; | |
7489 | struct dl *idisk; | |
7490 | int slot; | |
7491 | int idx; | |
7492 | __u8 state; | |
7493 | ||
7494 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
7495 | if (dev2) { | |
238c0a71 | 7496 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 7497 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 7498 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
7499 | if (!map) |
7500 | return 1; | |
7501 | for (slot = 0; slot < map->num_members; slot++) { | |
7502 | /* | |
7503 | * Check if failed disks are deleted from intel | |
7504 | * disk list or are marked to be deleted | |
7505 | */ | |
238c0a71 | 7506 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
7507 | idisk = get_imsm_dl_disk(cont->sb, idx); |
7508 | /* | |
7509 | * Do not rebuild the array if failed disks | |
7510 | * from failed sub-array are not removed from | |
7511 | * container. | |
7512 | */ | |
7513 | if (idisk && | |
7514 | is_failed(&idisk->disk) && | |
7515 | (idisk->action != DISK_REMOVE)) | |
7516 | return 0; | |
7517 | } | |
7518 | } | |
7519 | } | |
7520 | return 1; | |
7521 | } | |
7522 | ||
88758e9d DW |
7523 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
7524 | struct metadata_update **updates) | |
7525 | { | |
7526 | /** | |
d23fe947 DW |
7527 | * Find a device with unused free space and use it to replace a |
7528 | * failed/vacant region in an array. We replace failed regions one a | |
7529 | * array at a time. The result is that a new spare disk will be added | |
7530 | * to the first failed array and after the monitor has finished | |
7531 | * propagating failures the remainder will be consumed. | |
88758e9d | 7532 | * |
d23fe947 DW |
7533 | * FIXME add a capability for mdmon to request spares from another |
7534 | * container. | |
88758e9d DW |
7535 | */ |
7536 | ||
7537 | struct intel_super *super = a->container->sb; | |
88758e9d | 7538 | int inst = a->info.container_member; |
949c47a0 | 7539 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7540 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
7541 | int failed = a->info.array.raid_disks; |
7542 | struct mdinfo *rv = NULL; | |
7543 | struct mdinfo *d; | |
7544 | struct mdinfo *di; | |
7545 | struct metadata_update *mu; | |
7546 | struct dl *dl; | |
7547 | struct imsm_update_activate_spare *u; | |
7548 | int num_spares = 0; | |
7549 | int i; | |
95d07a2c | 7550 | int allowed; |
88758e9d DW |
7551 | |
7552 | for (d = a->info.devs ; d ; d = d->next) { | |
7553 | if ((d->curr_state & DS_FAULTY) && | |
7554 | d->state_fd >= 0) | |
7555 | /* wait for Removal to happen */ | |
7556 | return NULL; | |
7557 | if (d->state_fd >= 0) | |
7558 | failed--; | |
7559 | } | |
7560 | ||
7561 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
7562 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 7563 | |
e2962bfc AK |
7564 | if (imsm_reshape_blocks_arrays_changes(super)) |
7565 | return NULL; | |
1af97990 | 7566 | |
fc8ca064 AK |
7567 | /* Cannot activate another spare if rebuild is in progress already |
7568 | */ | |
7569 | if (is_rebuilding(dev)) { | |
7570 | dprintf("imsm: No spare activation allowed. " | |
7571 | "Rebuild in progress already.\n"); | |
7572 | return NULL; | |
7573 | } | |
7574 | ||
89c67882 AK |
7575 | if (a->info.array.level == 4) |
7576 | /* No repair for takeovered array | |
7577 | * imsm doesn't support raid4 | |
7578 | */ | |
7579 | return NULL; | |
7580 | ||
3b451610 AK |
7581 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
7582 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
7583 | return NULL; |
7584 | ||
95d07a2c LM |
7585 | /* |
7586 | * If there are any failed disks check state of the other volume. | |
7587 | * Block rebuild if the another one is failed until failed disks | |
7588 | * are removed from container. | |
7589 | */ | |
7590 | if (failed) { | |
c4acd1e5 | 7591 | dprintf("found failed disks in %.*s, check if there another" |
95d07a2c | 7592 | "failed sub-array.\n", |
c4acd1e5 | 7593 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
7594 | /* check if states of the other volumes allow for rebuild */ |
7595 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
7596 | if (i != inst) { | |
7597 | allowed = imsm_rebuild_allowed(a->container, | |
7598 | i, failed); | |
7599 | if (!allowed) | |
7600 | return NULL; | |
7601 | } | |
7602 | } | |
7603 | } | |
7604 | ||
88758e9d | 7605 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
7606 | for (i = 0; i < a->info.array.raid_disks; i++) { |
7607 | for (d = a->info.devs ; d ; d = d->next) | |
7608 | if (d->disk.raid_disk == i) | |
7609 | break; | |
7610 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
7611 | if (d && (d->state_fd >= 0)) | |
7612 | continue; | |
7613 | ||
272906ef | 7614 | /* |
a20d2ba5 DW |
7615 | * OK, this device needs recovery. Try to re-add the |
7616 | * previous occupant of this slot, if this fails see if | |
7617 | * we can continue the assimilation of a spare that was | |
7618 | * partially assimilated, finally try to activate a new | |
7619 | * spare. | |
272906ef DW |
7620 | */ |
7621 | dl = imsm_readd(super, i, a); | |
7622 | if (!dl) | |
b303fe21 | 7623 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 7624 | if (!dl) |
b303fe21 | 7625 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
7626 | if (!dl) |
7627 | continue; | |
7628 | ||
7629 | /* found a usable disk with enough space */ | |
503975b9 | 7630 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
7631 | |
7632 | /* dl->index will be -1 in the case we are activating a | |
7633 | * pristine spare. imsm_process_update() will create a | |
7634 | * new index in this case. Once a disk is found to be | |
7635 | * failed in all member arrays it is kicked from the | |
7636 | * metadata | |
7637 | */ | |
7638 | di->disk.number = dl->index; | |
d23fe947 | 7639 | |
272906ef DW |
7640 | /* (ab)use di->devs to store a pointer to the device |
7641 | * we chose | |
7642 | */ | |
7643 | di->devs = (struct mdinfo *) dl; | |
7644 | ||
7645 | di->disk.raid_disk = i; | |
7646 | di->disk.major = dl->major; | |
7647 | di->disk.minor = dl->minor; | |
7648 | di->disk.state = 0; | |
d23534e4 | 7649 | di->recovery_start = 0; |
5551b113 | 7650 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
7651 | di->component_size = a->info.component_size; |
7652 | di->container_member = inst; | |
148acb7b | 7653 | super->random = random32(); |
272906ef DW |
7654 | di->next = rv; |
7655 | rv = di; | |
7656 | num_spares++; | |
7657 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
7658 | i, di->data_offset); | |
88758e9d DW |
7659 | } |
7660 | ||
7661 | if (!rv) | |
7662 | /* No spares found */ | |
7663 | return rv; | |
7664 | /* Now 'rv' has a list of devices to return. | |
7665 | * Create a metadata_update record to update the | |
7666 | * disk_ord_tbl for the array | |
7667 | */ | |
503975b9 N |
7668 | mu = xmalloc(sizeof(*mu)); |
7669 | mu->buf = xcalloc(num_spares, | |
7670 | sizeof(struct imsm_update_activate_spare)); | |
88758e9d | 7671 | mu->space = NULL; |
cb23f1f4 | 7672 | mu->space_list = NULL; |
88758e9d DW |
7673 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
7674 | mu->next = *updates; | |
7675 | u = (struct imsm_update_activate_spare *) mu->buf; | |
7676 | ||
7677 | for (di = rv ; di ; di = di->next) { | |
7678 | u->type = update_activate_spare; | |
d23fe947 DW |
7679 | u->dl = (struct dl *) di->devs; |
7680 | di->devs = NULL; | |
88758e9d DW |
7681 | u->slot = di->disk.raid_disk; |
7682 | u->array = inst; | |
7683 | u->next = u + 1; | |
7684 | u++; | |
7685 | } | |
7686 | (u-1)->next = NULL; | |
7687 | *updates = mu; | |
7688 | ||
7689 | return rv; | |
7690 | } | |
7691 | ||
54c2c1ea | 7692 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 7693 | { |
54c2c1ea | 7694 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
7695 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7696 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
7697 | struct disk_info *inf = get_disk_info(u); |
7698 | struct imsm_disk *disk; | |
8273f55e DW |
7699 | int i; |
7700 | int j; | |
8273f55e | 7701 | |
54c2c1ea | 7702 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 7703 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
7704 | for (j = 0; j < new_map->num_members; j++) |
7705 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
7706 | return 1; |
7707 | } | |
7708 | ||
7709 | return 0; | |
7710 | } | |
7711 | ||
1a64be56 LM |
7712 | |
7713 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
7714 | { | |
7715 | struct dl *dl = NULL; | |
7716 | for (dl = super->disks; dl; dl = dl->next) | |
7717 | if ((dl->major == major) && (dl->minor == minor)) | |
7718 | return dl; | |
7719 | return NULL; | |
7720 | } | |
7721 | ||
7722 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7723 | { | |
7724 | struct dl *prev = NULL; | |
7725 | struct dl *dl; | |
7726 | ||
7727 | prev = NULL; | |
7728 | for (dl = super->disks; dl; dl = dl->next) { | |
7729 | if ((dl->major == major) && (dl->minor == minor)) { | |
7730 | /* remove */ | |
7731 | if (prev) | |
7732 | prev->next = dl->next; | |
7733 | else | |
7734 | super->disks = dl->next; | |
7735 | dl->next = NULL; | |
7736 | __free_imsm_disk(dl); | |
7737 | dprintf("%s: removed %x:%x\n", | |
7738 | __func__, major, minor); | |
7739 | break; | |
7740 | } | |
7741 | prev = dl; | |
7742 | } | |
7743 | return 0; | |
7744 | } | |
7745 | ||
f21e18ca | 7746 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7747 | |
1a64be56 LM |
7748 | static int add_remove_disk_update(struct intel_super *super) |
7749 | { | |
7750 | int check_degraded = 0; | |
7751 | struct dl *disk = NULL; | |
7752 | /* add/remove some spares to/from the metadata/contrainer */ | |
7753 | while (super->disk_mgmt_list) { | |
7754 | struct dl *disk_cfg; | |
7755 | ||
7756 | disk_cfg = super->disk_mgmt_list; | |
7757 | super->disk_mgmt_list = disk_cfg->next; | |
7758 | disk_cfg->next = NULL; | |
7759 | ||
7760 | if (disk_cfg->action == DISK_ADD) { | |
7761 | disk_cfg->next = super->disks; | |
7762 | super->disks = disk_cfg; | |
7763 | check_degraded = 1; | |
7764 | dprintf("%s: added %x:%x\n", | |
7765 | __func__, disk_cfg->major, | |
7766 | disk_cfg->minor); | |
7767 | } else if (disk_cfg->action == DISK_REMOVE) { | |
7768 | dprintf("Disk remove action processed: %x.%x\n", | |
7769 | disk_cfg->major, disk_cfg->minor); | |
7770 | disk = get_disk_super(super, | |
7771 | disk_cfg->major, | |
7772 | disk_cfg->minor); | |
7773 | if (disk) { | |
7774 | /* store action status */ | |
7775 | disk->action = DISK_REMOVE; | |
7776 | /* remove spare disks only */ | |
7777 | if (disk->index == -1) { | |
7778 | remove_disk_super(super, | |
7779 | disk_cfg->major, | |
7780 | disk_cfg->minor); | |
7781 | } | |
7782 | } | |
7783 | /* release allocate disk structure */ | |
7784 | __free_imsm_disk(disk_cfg); | |
7785 | } | |
7786 | } | |
7787 | return check_degraded; | |
7788 | } | |
7789 | ||
a29911da PC |
7790 | |
7791 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
7792 | struct intel_super *super, | |
7793 | void ***space_list) | |
7794 | { | |
7795 | struct intel_dev *id; | |
7796 | void **tofree = NULL; | |
7797 | int ret_val = 0; | |
7798 | ||
7799 | dprintf("apply_reshape_migration_update()\n"); | |
7800 | if ((u->subdev < 0) || | |
7801 | (u->subdev > 1)) { | |
7802 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7803 | return ret_val; | |
7804 | } | |
7805 | if ((space_list == NULL) || (*space_list == NULL)) { | |
7806 | dprintf("imsm: Error: Memory is not allocated\n"); | |
7807 | return ret_val; | |
7808 | } | |
7809 | ||
7810 | for (id = super->devlist ; id; id = id->next) { | |
7811 | if (id->index == (unsigned)u->subdev) { | |
7812 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7813 | struct imsm_map *map; | |
7814 | struct imsm_dev *new_dev = | |
7815 | (struct imsm_dev *)*space_list; | |
238c0a71 | 7816 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
7817 | int to_state; |
7818 | struct dl *new_disk; | |
7819 | ||
7820 | if (new_dev == NULL) | |
7821 | return ret_val; | |
7822 | *space_list = **space_list; | |
7823 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 7824 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
7825 | if (migr_map) { |
7826 | dprintf("imsm: Error: migration in progress"); | |
7827 | return ret_val; | |
7828 | } | |
7829 | ||
7830 | to_state = map->map_state; | |
7831 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
7832 | map->num_members++; | |
7833 | /* this should not happen */ | |
7834 | if (u->new_disks[0] < 0) { | |
7835 | map->failed_disk_num = | |
7836 | map->num_members - 1; | |
7837 | to_state = IMSM_T_STATE_DEGRADED; | |
7838 | } else | |
7839 | to_state = IMSM_T_STATE_NORMAL; | |
7840 | } | |
8e59f3d8 | 7841 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
7842 | if (u->new_level > -1) |
7843 | map->raid_level = u->new_level; | |
238c0a71 | 7844 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
7845 | if ((u->new_level == 5) && |
7846 | (migr_map->raid_level == 0)) { | |
7847 | int ord = map->num_members - 1; | |
7848 | migr_map->num_members--; | |
7849 | if (u->new_disks[0] < 0) | |
7850 | ord |= IMSM_ORD_REBUILD; | |
7851 | set_imsm_ord_tbl_ent(map, | |
7852 | map->num_members - 1, | |
7853 | ord); | |
7854 | } | |
7855 | id->dev = new_dev; | |
7856 | tofree = (void **)dev; | |
7857 | ||
4bba0439 PC |
7858 | /* update chunk size |
7859 | */ | |
7860 | if (u->new_chunksize > 0) | |
7861 | map->blocks_per_strip = | |
7862 | __cpu_to_le16(u->new_chunksize * 2); | |
7863 | ||
a29911da PC |
7864 | /* add disk |
7865 | */ | |
7866 | if ((u->new_level != 5) || | |
7867 | (migr_map->raid_level != 0) || | |
7868 | (migr_map->raid_level == map->raid_level)) | |
7869 | goto skip_disk_add; | |
7870 | ||
7871 | if (u->new_disks[0] >= 0) { | |
7872 | /* use passes spare | |
7873 | */ | |
7874 | new_disk = get_disk_super(super, | |
7875 | major(u->new_disks[0]), | |
7876 | minor(u->new_disks[0])); | |
7877 | dprintf("imsm: new disk for reshape is: %i:%i " | |
7878 | "(%p, index = %i)\n", | |
7879 | major(u->new_disks[0]), | |
7880 | minor(u->new_disks[0]), | |
7881 | new_disk, new_disk->index); | |
7882 | if (new_disk == NULL) | |
7883 | goto error_disk_add; | |
7884 | ||
7885 | new_disk->index = map->num_members - 1; | |
7886 | /* slot to fill in autolayout | |
7887 | */ | |
7888 | new_disk->raiddisk = new_disk->index; | |
7889 | new_disk->disk.status |= CONFIGURED_DISK; | |
7890 | new_disk->disk.status &= ~SPARE_DISK; | |
7891 | } else | |
7892 | goto error_disk_add; | |
7893 | ||
7894 | skip_disk_add: | |
7895 | *tofree = *space_list; | |
7896 | /* calculate new size | |
7897 | */ | |
f3871fdc | 7898 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
7899 | |
7900 | ret_val = 1; | |
7901 | } | |
7902 | } | |
7903 | ||
7904 | if (tofree) | |
7905 | *space_list = tofree; | |
7906 | return ret_val; | |
7907 | ||
7908 | error_disk_add: | |
7909 | dprintf("Error: imsm: Cannot find disk.\n"); | |
7910 | return ret_val; | |
7911 | } | |
7912 | ||
f3871fdc AK |
7913 | static int apply_size_change_update(struct imsm_update_size_change *u, |
7914 | struct intel_super *super) | |
7915 | { | |
7916 | struct intel_dev *id; | |
7917 | int ret_val = 0; | |
7918 | ||
7919 | dprintf("apply_size_change_update()\n"); | |
7920 | if ((u->subdev < 0) || | |
7921 | (u->subdev > 1)) { | |
7922 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7923 | return ret_val; | |
7924 | } | |
7925 | ||
7926 | for (id = super->devlist ; id; id = id->next) { | |
7927 | if (id->index == (unsigned)u->subdev) { | |
7928 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7929 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
7930 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
7931 | unsigned long long blocks_per_member; | |
7932 | ||
7933 | /* calculate new size | |
7934 | */ | |
7935 | blocks_per_member = u->new_size / used_disks; | |
7936 | dprintf("imsm: apply_size_change_update(size: %llu, " | |
7937 | "blocks per member: %llu)\n", | |
7938 | u->new_size, blocks_per_member); | |
7939 | set_blocks_per_member(map, blocks_per_member); | |
7940 | imsm_set_array_size(dev, u->new_size); | |
7941 | ||
7942 | ret_val = 1; | |
7943 | break; | |
7944 | } | |
7945 | } | |
7946 | ||
7947 | return ret_val; | |
7948 | } | |
7949 | ||
7950 | ||
061d7da3 | 7951 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 7952 | struct intel_super *super, |
061d7da3 LO |
7953 | struct active_array *active_array) |
7954 | { | |
7955 | struct imsm_super *mpb = super->anchor; | |
7956 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 7957 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
7958 | struct imsm_map *migr_map; |
7959 | struct active_array *a; | |
7960 | struct imsm_disk *disk; | |
7961 | __u8 to_state; | |
7962 | struct dl *dl; | |
7963 | unsigned int found; | |
7964 | int failed; | |
5961eeec | 7965 | int victim; |
061d7da3 | 7966 | int i; |
5961eeec | 7967 | int second_map_created = 0; |
061d7da3 | 7968 | |
5961eeec | 7969 | for (; u; u = u->next) { |
238c0a71 | 7970 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 7971 | |
5961eeec | 7972 | if (victim < 0) |
7973 | return 0; | |
061d7da3 | 7974 | |
5961eeec | 7975 | for (dl = super->disks; dl; dl = dl->next) |
7976 | if (dl == u->dl) | |
7977 | break; | |
061d7da3 | 7978 | |
5961eeec | 7979 | if (!dl) { |
7980 | fprintf(stderr, "error: imsm_activate_spare passed " | |
7981 | "an unknown disk (index: %d)\n", | |
7982 | u->dl->index); | |
7983 | return 0; | |
7984 | } | |
061d7da3 | 7985 | |
5961eeec | 7986 | /* count failures (excluding rebuilds and the victim) |
7987 | * to determine map[0] state | |
7988 | */ | |
7989 | failed = 0; | |
7990 | for (i = 0; i < map->num_members; i++) { | |
7991 | if (i == u->slot) | |
7992 | continue; | |
7993 | disk = get_imsm_disk(super, | |
238c0a71 | 7994 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 7995 | if (!disk || is_failed(disk)) |
7996 | failed++; | |
7997 | } | |
061d7da3 | 7998 | |
5961eeec | 7999 | /* adding a pristine spare, assign a new index */ |
8000 | if (dl->index < 0) { | |
8001 | dl->index = super->anchor->num_disks; | |
8002 | super->anchor->num_disks++; | |
8003 | } | |
8004 | disk = &dl->disk; | |
8005 | disk->status |= CONFIGURED_DISK; | |
8006 | disk->status &= ~SPARE_DISK; | |
8007 | ||
8008 | /* mark rebuild */ | |
238c0a71 | 8009 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 8010 | if (!second_map_created) { |
8011 | second_map_created = 1; | |
8012 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8013 | migrate(dev, super, to_state, MIGR_REBUILD); | |
8014 | } else | |
8015 | map->map_state = to_state; | |
238c0a71 | 8016 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 8017 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
8018 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
8019 | dl->index | IMSM_ORD_REBUILD); | |
8020 | ||
8021 | /* update the family_num to mark a new container | |
8022 | * generation, being careful to record the existing | |
8023 | * family_num in orig_family_num to clean up after | |
8024 | * earlier mdadm versions that neglected to set it. | |
8025 | */ | |
8026 | if (mpb->orig_family_num == 0) | |
8027 | mpb->orig_family_num = mpb->family_num; | |
8028 | mpb->family_num += super->random; | |
8029 | ||
8030 | /* count arrays using the victim in the metadata */ | |
8031 | found = 0; | |
8032 | for (a = active_array; a ; a = a->next) { | |
8033 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8034 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 8035 | |
5961eeec | 8036 | if (get_imsm_disk_slot(map, victim) >= 0) |
8037 | found++; | |
8038 | } | |
061d7da3 | 8039 | |
5961eeec | 8040 | /* delete the victim if it is no longer being |
8041 | * utilized anywhere | |
061d7da3 | 8042 | */ |
5961eeec | 8043 | if (!found) { |
8044 | struct dl **dlp; | |
061d7da3 | 8045 | |
5961eeec | 8046 | /* We know that 'manager' isn't touching anything, |
8047 | * so it is safe to delete | |
8048 | */ | |
8049 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
8050 | if ((*dlp)->index == victim) |
8051 | break; | |
5961eeec | 8052 | |
8053 | /* victim may be on the missing list */ | |
8054 | if (!*dlp) | |
8055 | for (dlp = &super->missing; *dlp; | |
8056 | dlp = &(*dlp)->next) | |
8057 | if ((*dlp)->index == victim) | |
8058 | break; | |
8059 | imsm_delete(super, dlp, victim); | |
8060 | } | |
061d7da3 LO |
8061 | } |
8062 | ||
8063 | return 1; | |
8064 | } | |
a29911da | 8065 | |
2e5dc010 N |
8066 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
8067 | struct intel_super *super, | |
8068 | void ***space_list) | |
8069 | { | |
8070 | struct dl *new_disk; | |
8071 | struct intel_dev *id; | |
8072 | int i; | |
8073 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 8074 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
8075 | void **tofree = NULL; |
8076 | int devices_to_reshape = 1; | |
8077 | struct imsm_super *mpb = super->anchor; | |
8078 | int ret_val = 0; | |
d098291a | 8079 | unsigned int dev_id; |
2e5dc010 | 8080 | |
ed7333bd | 8081 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
8082 | |
8083 | /* enable spares to use in array */ | |
8084 | for (i = 0; i < delta_disks; i++) { | |
8085 | new_disk = get_disk_super(super, | |
8086 | major(u->new_disks[i]), | |
8087 | minor(u->new_disks[i])); | |
ed7333bd AK |
8088 | dprintf("imsm: new disk for reshape is: %i:%i " |
8089 | "(%p, index = %i)\n", | |
2e5dc010 N |
8090 | major(u->new_disks[i]), minor(u->new_disks[i]), |
8091 | new_disk, new_disk->index); | |
8092 | if ((new_disk == NULL) || | |
8093 | ((new_disk->index >= 0) && | |
8094 | (new_disk->index < u->old_raid_disks))) | |
8095 | goto update_reshape_exit; | |
ee4beede | 8096 | new_disk->index = disk_count++; |
2e5dc010 N |
8097 | /* slot to fill in autolayout |
8098 | */ | |
8099 | new_disk->raiddisk = new_disk->index; | |
8100 | new_disk->disk.status |= | |
8101 | CONFIGURED_DISK; | |
8102 | new_disk->disk.status &= ~SPARE_DISK; | |
8103 | } | |
8104 | ||
ed7333bd AK |
8105 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
8106 | mpb->num_raid_devs); | |
2e5dc010 N |
8107 | /* manage changes in volume |
8108 | */ | |
d098291a | 8109 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
8110 | void **sp = *space_list; |
8111 | struct imsm_dev *newdev; | |
8112 | struct imsm_map *newmap, *oldmap; | |
8113 | ||
d098291a AK |
8114 | for (id = super->devlist ; id; id = id->next) { |
8115 | if (id->index == dev_id) | |
8116 | break; | |
8117 | } | |
8118 | if (id == NULL) | |
8119 | break; | |
2e5dc010 N |
8120 | if (!sp) |
8121 | continue; | |
8122 | *space_list = *sp; | |
8123 | newdev = (void*)sp; | |
8124 | /* Copy the dev, but not (all of) the map */ | |
8125 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
8126 | oldmap = get_imsm_map(id->dev, MAP_0); |
8127 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
8128 | /* Copy the current map */ |
8129 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
8130 | /* update one device only | |
8131 | */ | |
8132 | if (devices_to_reshape) { | |
ed7333bd AK |
8133 | dprintf("imsm: modifying subdev: %i\n", |
8134 | id->index); | |
2e5dc010 N |
8135 | devices_to_reshape--; |
8136 | newdev->vol.migr_state = 1; | |
8137 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8138 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8139 | newmap->num_members = u->new_raid_disks; |
8140 | for (i = 0; i < delta_disks; i++) { | |
8141 | set_imsm_ord_tbl_ent(newmap, | |
8142 | u->old_raid_disks + i, | |
8143 | u->old_raid_disks + i); | |
8144 | } | |
8145 | /* New map is correct, now need to save old map | |
8146 | */ | |
238c0a71 | 8147 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8148 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8149 | ||
f3871fdc | 8150 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
8151 | } |
8152 | ||
8153 | sp = (void **)id->dev; | |
8154 | id->dev = newdev; | |
8155 | *sp = tofree; | |
8156 | tofree = sp; | |
8e59f3d8 AK |
8157 | |
8158 | /* Clear migration record */ | |
8159 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8160 | } |
819bc634 AK |
8161 | if (tofree) |
8162 | *space_list = tofree; | |
2e5dc010 N |
8163 | ret_val = 1; |
8164 | ||
8165 | update_reshape_exit: | |
8166 | ||
8167 | return ret_val; | |
8168 | } | |
8169 | ||
bb025c2f | 8170 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8171 | struct intel_super *super, |
8172 | void ***space_list) | |
bb025c2f KW |
8173 | { |
8174 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8175 | struct intel_dev *dv; |
8176 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8177 | struct imsm_map *map; |
8178 | struct dl *dm, *du; | |
8ca6df95 | 8179 | int i; |
bb025c2f KW |
8180 | |
8181 | for (dv = super->devlist; dv; dv = dv->next) | |
8182 | if (dv->index == (unsigned int)u->subarray) { | |
8183 | dev = dv->dev; | |
8184 | break; | |
8185 | } | |
8186 | ||
8187 | if (dev == NULL) | |
8188 | return 0; | |
8189 | ||
238c0a71 | 8190 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8191 | |
8192 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 8193 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
8194 | if (imsm_count_failed(super, dev, MAP_0) != |
8195 | (map->num_members / 2)) | |
43d5ec18 KW |
8196 | return 0; |
8197 | ||
bb025c2f KW |
8198 | /* iterate through devices to mark removed disks as spare */ |
8199 | for (dm = super->disks; dm; dm = dm->next) { | |
8200 | if (dm->disk.status & FAILED_DISK) { | |
8201 | int idx = dm->index; | |
8202 | /* update indexes on the disk list */ | |
8203 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
8204 | the index values will end up being correct.... NB */ | |
8205 | for (du = super->disks; du; du = du->next) | |
8206 | if (du->index > idx) | |
8207 | du->index--; | |
8208 | /* mark as spare disk */ | |
a8619d23 | 8209 | mark_spare(dm); |
bb025c2f KW |
8210 | } |
8211 | } | |
bb025c2f KW |
8212 | /* update map */ |
8213 | map->num_members = map->num_members / 2; | |
8214 | map->map_state = IMSM_T_STATE_NORMAL; | |
8215 | map->num_domains = 1; | |
8216 | map->raid_level = 0; | |
8217 | map->failed_disk_num = -1; | |
8218 | } | |
8219 | ||
8ca6df95 KW |
8220 | if (u->direction == R0_TO_R10) { |
8221 | void **space; | |
8222 | /* update slots in current disk list */ | |
8223 | for (dm = super->disks; dm; dm = dm->next) { | |
8224 | if (dm->index >= 0) | |
8225 | dm->index *= 2; | |
8226 | } | |
8227 | /* create new *missing* disks */ | |
8228 | for (i = 0; i < map->num_members; i++) { | |
8229 | space = *space_list; | |
8230 | if (!space) | |
8231 | continue; | |
8232 | *space_list = *space; | |
8233 | du = (void *)space; | |
8234 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
8235 | du->fd = -1; |
8236 | du->minor = 0; | |
8237 | du->major = 0; | |
8238 | du->index = (i * 2) + 1; | |
8239 | sprintf((char *)du->disk.serial, | |
8240 | " MISSING_%d", du->index); | |
8241 | sprintf((char *)du->serial, | |
8242 | "MISSING_%d", du->index); | |
8243 | du->next = super->missing; | |
8244 | super->missing = du; | |
8245 | } | |
8246 | /* create new dev and map */ | |
8247 | space = *space_list; | |
8248 | if (!space) | |
8249 | return 0; | |
8250 | *space_list = *space; | |
8251 | dev_new = (void *)space; | |
8252 | memcpy(dev_new, dev, sizeof(*dev)); | |
8253 | /* update new map */ | |
238c0a71 | 8254 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 8255 | map->num_members = map->num_members * 2; |
1a2487c2 | 8256 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
8257 | map->num_domains = 2; |
8258 | map->raid_level = 1; | |
8259 | /* replace dev<->dev_new */ | |
8260 | dv->dev = dev_new; | |
8261 | } | |
bb025c2f KW |
8262 | /* update disk order table */ |
8263 | for (du = super->disks; du; du = du->next) | |
8264 | if (du->index >= 0) | |
8265 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 8266 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
8267 | if (du->index >= 0) { |
8268 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 8269 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 8270 | } |
bb025c2f KW |
8271 | |
8272 | return 1; | |
8273 | } | |
8274 | ||
e8319a19 DW |
8275 | static void imsm_process_update(struct supertype *st, |
8276 | struct metadata_update *update) | |
8277 | { | |
8278 | /** | |
8279 | * crack open the metadata_update envelope to find the update record | |
8280 | * update can be one of: | |
d195167d AK |
8281 | * update_reshape_container_disks - all the arrays in the container |
8282 | * are being reshaped to have more devices. We need to mark | |
8283 | * the arrays for general migration and convert selected spares | |
8284 | * into active devices. | |
8285 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
8286 | * device in an array, update the disk_ord_tbl. If this disk is |
8287 | * present in all member arrays then also clear the SPARE_DISK | |
8288 | * flag | |
d195167d AK |
8289 | * update_create_array |
8290 | * update_kill_array | |
8291 | * update_rename_array | |
8292 | * update_add_remove_disk | |
e8319a19 DW |
8293 | */ |
8294 | struct intel_super *super = st->sb; | |
4d7b1503 | 8295 | struct imsm_super *mpb; |
e8319a19 DW |
8296 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
8297 | ||
4d7b1503 DW |
8298 | /* update requires a larger buf but the allocation failed */ |
8299 | if (super->next_len && !super->next_buf) { | |
8300 | super->next_len = 0; | |
8301 | return; | |
8302 | } | |
8303 | ||
8304 | if (super->next_buf) { | |
8305 | memcpy(super->next_buf, super->buf, super->len); | |
8306 | free(super->buf); | |
8307 | super->len = super->next_len; | |
8308 | super->buf = super->next_buf; | |
8309 | ||
8310 | super->next_len = 0; | |
8311 | super->next_buf = NULL; | |
8312 | } | |
8313 | ||
8314 | mpb = super->anchor; | |
8315 | ||
e8319a19 | 8316 | switch (type) { |
0ec5d470 AK |
8317 | case update_general_migration_checkpoint: { |
8318 | struct intel_dev *id; | |
8319 | struct imsm_update_general_migration_checkpoint *u = | |
8320 | (void *)update->buf; | |
8321 | ||
8322 | dprintf("imsm: process_update() " | |
8323 | "for update_general_migration_checkpoint called\n"); | |
8324 | ||
8325 | /* find device under general migration */ | |
8326 | for (id = super->devlist ; id; id = id->next) { | |
8327 | if (is_gen_migration(id->dev)) { | |
8328 | id->dev->vol.curr_migr_unit = | |
8329 | __cpu_to_le32(u->curr_migr_unit); | |
8330 | super->updates_pending++; | |
8331 | } | |
8332 | } | |
8333 | break; | |
8334 | } | |
bb025c2f KW |
8335 | case update_takeover: { |
8336 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
8337 | if (apply_takeover_update(u, super, &update->space_list)) { |
8338 | imsm_update_version_info(super); | |
bb025c2f | 8339 | super->updates_pending++; |
1a2487c2 | 8340 | } |
bb025c2f KW |
8341 | break; |
8342 | } | |
8343 | ||
78b10e66 | 8344 | case update_reshape_container_disks: { |
d195167d | 8345 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
8346 | if (apply_reshape_container_disks_update( |
8347 | u, super, &update->space_list)) | |
8348 | super->updates_pending++; | |
78b10e66 N |
8349 | break; |
8350 | } | |
48c5303a | 8351 | case update_reshape_migration: { |
a29911da PC |
8352 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
8353 | if (apply_reshape_migration_update( | |
8354 | u, super, &update->space_list)) | |
8355 | super->updates_pending++; | |
48c5303a PC |
8356 | break; |
8357 | } | |
f3871fdc AK |
8358 | case update_size_change: { |
8359 | struct imsm_update_size_change *u = (void *)update->buf; | |
8360 | if (apply_size_change_update(u, super)) | |
8361 | super->updates_pending++; | |
8362 | break; | |
8363 | } | |
e8319a19 DW |
8364 | case update_activate_spare: { |
8365 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
061d7da3 LO |
8366 | if (apply_update_activate_spare(u, super, st->arrays)) |
8367 | super->updates_pending++; | |
8273f55e DW |
8368 | break; |
8369 | } | |
8370 | case update_create_array: { | |
8371 | /* someone wants to create a new array, we need to be aware of | |
8372 | * a few races/collisions: | |
8373 | * 1/ 'Create' called by two separate instances of mdadm | |
8374 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
8375 | * devices that have since been assimilated via | |
8376 | * activate_spare. | |
8377 | * In the event this update can not be carried out mdadm will | |
8378 | * (FIX ME) notice that its update did not take hold. | |
8379 | */ | |
8380 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8381 | struct intel_dev *dv; |
8273f55e DW |
8382 | struct imsm_dev *dev; |
8383 | struct imsm_map *map, *new_map; | |
8384 | unsigned long long start, end; | |
8385 | unsigned long long new_start, new_end; | |
8386 | int i; | |
54c2c1ea DW |
8387 | struct disk_info *inf; |
8388 | struct dl *dl; | |
8273f55e DW |
8389 | |
8390 | /* handle racing creates: first come first serve */ | |
8391 | if (u->dev_idx < mpb->num_raid_devs) { | |
8392 | dprintf("%s: subarray %d already defined\n", | |
8393 | __func__, u->dev_idx); | |
ba2de7ba | 8394 | goto create_error; |
8273f55e DW |
8395 | } |
8396 | ||
8397 | /* check update is next in sequence */ | |
8398 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
8399 | dprintf("%s: can not create array %d expected index %d\n", |
8400 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 8401 | goto create_error; |
8273f55e DW |
8402 | } |
8403 | ||
238c0a71 | 8404 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
8405 | new_start = pba_of_lba0(new_map); |
8406 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 8407 | inf = get_disk_info(u); |
8273f55e DW |
8408 | |
8409 | /* handle activate_spare versus create race: | |
8410 | * check to make sure that overlapping arrays do not include | |
8411 | * overalpping disks | |
8412 | */ | |
8413 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 8414 | dev = get_imsm_dev(super, i); |
238c0a71 | 8415 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
8416 | start = pba_of_lba0(map); |
8417 | end = start + blocks_per_member(map); | |
8273f55e DW |
8418 | if ((new_start >= start && new_start <= end) || |
8419 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
8420 | /* overlap */; |
8421 | else | |
8422 | continue; | |
8423 | ||
8424 | if (disks_overlap(super, i, u)) { | |
8273f55e | 8425 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 8426 | goto create_error; |
8273f55e DW |
8427 | } |
8428 | } | |
8273f55e | 8429 | |
949c47a0 DW |
8430 | /* check that prepare update was successful */ |
8431 | if (!update->space) { | |
8432 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 8433 | goto create_error; |
949c47a0 DW |
8434 | } |
8435 | ||
54c2c1ea DW |
8436 | /* check that all disks are still active before committing |
8437 | * changes. FIXME: could we instead handle this by creating a | |
8438 | * degraded array? That's probably not what the user expects, | |
8439 | * so better to drop this update on the floor. | |
8440 | */ | |
8441 | for (i = 0; i < new_map->num_members; i++) { | |
8442 | dl = serial_to_dl(inf[i].serial, super); | |
8443 | if (!dl) { | |
8444 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 8445 | goto create_error; |
54c2c1ea | 8446 | } |
949c47a0 DW |
8447 | } |
8448 | ||
8273f55e | 8449 | super->updates_pending++; |
54c2c1ea DW |
8450 | |
8451 | /* convert spares to members and fixup ord_tbl */ | |
8452 | for (i = 0; i < new_map->num_members; i++) { | |
8453 | dl = serial_to_dl(inf[i].serial, super); | |
8454 | if (dl->index == -1) { | |
8455 | dl->index = mpb->num_disks; | |
8456 | mpb->num_disks++; | |
8457 | dl->disk.status |= CONFIGURED_DISK; | |
8458 | dl->disk.status &= ~SPARE_DISK; | |
8459 | } | |
8460 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
8461 | } | |
8462 | ||
ba2de7ba DW |
8463 | dv = update->space; |
8464 | dev = dv->dev; | |
949c47a0 DW |
8465 | update->space = NULL; |
8466 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
8467 | dv->index = u->dev_idx; |
8468 | dv->next = super->devlist; | |
8469 | super->devlist = dv; | |
8273f55e | 8470 | mpb->num_raid_devs++; |
8273f55e | 8471 | |
4d1313e9 | 8472 | imsm_update_version_info(super); |
8273f55e | 8473 | break; |
ba2de7ba DW |
8474 | create_error: |
8475 | /* mdmon knows how to release update->space, but not | |
8476 | * ((struct intel_dev *) update->space)->dev | |
8477 | */ | |
8478 | if (update->space) { | |
8479 | dv = update->space; | |
8480 | free(dv->dev); | |
8481 | } | |
8273f55e | 8482 | break; |
e8319a19 | 8483 | } |
33414a01 DW |
8484 | case update_kill_array: { |
8485 | struct imsm_update_kill_array *u = (void *) update->buf; | |
8486 | int victim = u->dev_idx; | |
8487 | struct active_array *a; | |
8488 | struct intel_dev **dp; | |
8489 | struct imsm_dev *dev; | |
8490 | ||
8491 | /* sanity check that we are not affecting the uuid of | |
8492 | * active arrays, or deleting an active array | |
8493 | * | |
8494 | * FIXME when immutable ids are available, but note that | |
8495 | * we'll also need to fixup the invalidated/active | |
8496 | * subarray indexes in mdstat | |
8497 | */ | |
8498 | for (a = st->arrays; a; a = a->next) | |
8499 | if (a->info.container_member >= victim) | |
8500 | break; | |
8501 | /* by definition if mdmon is running at least one array | |
8502 | * is active in the container, so checking | |
8503 | * mpb->num_raid_devs is just extra paranoia | |
8504 | */ | |
8505 | dev = get_imsm_dev(super, victim); | |
8506 | if (a || !dev || mpb->num_raid_devs == 1) { | |
8507 | dprintf("failed to delete subarray-%d\n", victim); | |
8508 | break; | |
8509 | } | |
8510 | ||
8511 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 8512 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
8513 | *dp = (*dp)->next; |
8514 | } else { | |
f21e18ca | 8515 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
8516 | (*dp)->index--; |
8517 | dp = &(*dp)->next; | |
8518 | } | |
8519 | mpb->num_raid_devs--; | |
8520 | super->updates_pending++; | |
8521 | break; | |
8522 | } | |
aa534678 DW |
8523 | case update_rename_array: { |
8524 | struct imsm_update_rename_array *u = (void *) update->buf; | |
8525 | char name[MAX_RAID_SERIAL_LEN+1]; | |
8526 | int target = u->dev_idx; | |
8527 | struct active_array *a; | |
8528 | struct imsm_dev *dev; | |
8529 | ||
8530 | /* sanity check that we are not affecting the uuid of | |
8531 | * an active array | |
8532 | */ | |
8533 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
8534 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
8535 | for (a = st->arrays; a; a = a->next) | |
8536 | if (a->info.container_member == target) | |
8537 | break; | |
8538 | dev = get_imsm_dev(super, u->dev_idx); | |
8539 | if (a || !dev || !check_name(super, name, 1)) { | |
8540 | dprintf("failed to rename subarray-%d\n", target); | |
8541 | break; | |
8542 | } | |
8543 | ||
cdbe98cd | 8544 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
8545 | super->updates_pending++; |
8546 | break; | |
8547 | } | |
1a64be56 | 8548 | case update_add_remove_disk: { |
43dad3d6 | 8549 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
8550 | * being added, check teh status of add_remove_disk |
8551 | * if discs has been added. | |
8552 | */ | |
8553 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 8554 | struct active_array *a; |
072b727f DW |
8555 | |
8556 | super->updates_pending++; | |
1a64be56 | 8557 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
8558 | a->check_degraded = 1; |
8559 | } | |
43dad3d6 | 8560 | break; |
e8319a19 | 8561 | } |
1a64be56 LM |
8562 | default: |
8563 | fprintf(stderr, "error: unsuported process update type:" | |
8564 | "(type: %d)\n", type); | |
8565 | } | |
e8319a19 | 8566 | } |
88758e9d | 8567 | |
bc0b9d34 PC |
8568 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
8569 | ||
8273f55e DW |
8570 | static void imsm_prepare_update(struct supertype *st, |
8571 | struct metadata_update *update) | |
8572 | { | |
949c47a0 | 8573 | /** |
4d7b1503 DW |
8574 | * Allocate space to hold new disk entries, raid-device entries or a new |
8575 | * mpb if necessary. The manager synchronously waits for updates to | |
8576 | * complete in the monitor, so new mpb buffers allocated here can be | |
8577 | * integrated by the monitor thread without worrying about live pointers | |
8578 | * in the manager thread. | |
8273f55e | 8579 | */ |
949c47a0 | 8580 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
8581 | struct intel_super *super = st->sb; |
8582 | struct imsm_super *mpb = super->anchor; | |
8583 | size_t buf_len; | |
8584 | size_t len = 0; | |
949c47a0 DW |
8585 | |
8586 | switch (type) { | |
0ec5d470 AK |
8587 | case update_general_migration_checkpoint: |
8588 | dprintf("imsm: prepare_update() " | |
8589 | "for update_general_migration_checkpoint called\n"); | |
8590 | break; | |
abedf5fc KW |
8591 | case update_takeover: { |
8592 | struct imsm_update_takeover *u = (void *)update->buf; | |
8593 | if (u->direction == R0_TO_R10) { | |
8594 | void **tail = (void **)&update->space_list; | |
8595 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 8596 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
8597 | int num_members = map->num_members; |
8598 | void *space; | |
8599 | int size, i; | |
abedf5fc KW |
8600 | /* allocate memory for added disks */ |
8601 | for (i = 0; i < num_members; i++) { | |
8602 | size = sizeof(struct dl); | |
503975b9 | 8603 | space = xmalloc(size); |
abedf5fc KW |
8604 | *tail = space; |
8605 | tail = space; | |
8606 | *tail = NULL; | |
8607 | } | |
8608 | /* allocate memory for new device */ | |
8609 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
8610 | (num_members * sizeof(__u32)); | |
503975b9 N |
8611 | space = xmalloc(size); |
8612 | *tail = space; | |
8613 | tail = space; | |
8614 | *tail = NULL; | |
8615 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
8616 | } |
8617 | ||
8618 | break; | |
8619 | } | |
78b10e66 | 8620 | case update_reshape_container_disks: { |
d195167d AK |
8621 | /* Every raid device in the container is about to |
8622 | * gain some more devices, and we will enter a | |
8623 | * reconfiguration. | |
8624 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
8625 | * will now hold 2 of them. | |
8626 | * Thus we need new 'struct imsm_dev' allocations sized | |
8627 | * as sizeof_imsm_dev but with more devices in both maps. | |
8628 | */ | |
8629 | struct imsm_update_reshape *u = (void *)update->buf; | |
8630 | struct intel_dev *dl; | |
8631 | void **space_tail = (void**)&update->space_list; | |
8632 | ||
8633 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8634 | ||
8635 | for (dl = super->devlist; dl; dl = dl->next) { | |
8636 | int size = sizeof_imsm_dev(dl->dev, 1); | |
8637 | void *s; | |
d677e0b8 AK |
8638 | if (u->new_raid_disks > u->old_raid_disks) |
8639 | size += sizeof(__u32)*2* | |
8640 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8641 | s = xmalloc(size); |
d195167d AK |
8642 | *space_tail = s; |
8643 | space_tail = s; | |
8644 | *space_tail = NULL; | |
8645 | } | |
8646 | ||
8647 | len = disks_to_mpb_size(u->new_raid_disks); | |
8648 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
8649 | break; |
8650 | } | |
48c5303a | 8651 | case update_reshape_migration: { |
bc0b9d34 PC |
8652 | /* for migration level 0->5 we need to add disks |
8653 | * so the same as for container operation we will copy | |
8654 | * device to the bigger location. | |
8655 | * in memory prepared device and new disk area are prepared | |
8656 | * for usage in process update | |
8657 | */ | |
8658 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
8659 | struct intel_dev *id; | |
8660 | void **space_tail = (void **)&update->space_list; | |
8661 | int size; | |
8662 | void *s; | |
8663 | int current_level = -1; | |
8664 | ||
8665 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8666 | ||
8667 | /* add space for bigger array in update | |
8668 | */ | |
8669 | for (id = super->devlist; id; id = id->next) { | |
8670 | if (id->index == (unsigned)u->subdev) { | |
8671 | size = sizeof_imsm_dev(id->dev, 1); | |
8672 | if (u->new_raid_disks > u->old_raid_disks) | |
8673 | size += sizeof(__u32)*2* | |
8674 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8675 | s = xmalloc(size); |
bc0b9d34 PC |
8676 | *space_tail = s; |
8677 | space_tail = s; | |
8678 | *space_tail = NULL; | |
8679 | break; | |
8680 | } | |
8681 | } | |
8682 | if (update->space_list == NULL) | |
8683 | break; | |
8684 | ||
8685 | /* add space for disk in update | |
8686 | */ | |
8687 | size = sizeof(struct dl); | |
503975b9 | 8688 | s = xmalloc(size); |
bc0b9d34 PC |
8689 | *space_tail = s; |
8690 | space_tail = s; | |
8691 | *space_tail = NULL; | |
8692 | ||
8693 | /* add spare device to update | |
8694 | */ | |
8695 | for (id = super->devlist ; id; id = id->next) | |
8696 | if (id->index == (unsigned)u->subdev) { | |
8697 | struct imsm_dev *dev; | |
8698 | struct imsm_map *map; | |
8699 | ||
8700 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 8701 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
8702 | current_level = map->raid_level; |
8703 | break; | |
8704 | } | |
8705 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
8706 | struct mdinfo *spares; | |
8707 | ||
8708 | spares = get_spares_for_grow(st); | |
8709 | if (spares) { | |
8710 | struct dl *dl; | |
8711 | struct mdinfo *dev; | |
8712 | ||
8713 | dev = spares->devs; | |
8714 | if (dev) { | |
8715 | u->new_disks[0] = | |
8716 | makedev(dev->disk.major, | |
8717 | dev->disk.minor); | |
8718 | dl = get_disk_super(super, | |
8719 | dev->disk.major, | |
8720 | dev->disk.minor); | |
8721 | dl->index = u->old_raid_disks; | |
8722 | dev = dev->next; | |
8723 | } | |
8724 | sysfs_free(spares); | |
8725 | } | |
8726 | } | |
8727 | len = disks_to_mpb_size(u->new_raid_disks); | |
8728 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8729 | break; |
8730 | } | |
f3871fdc AK |
8731 | case update_size_change: { |
8732 | break; | |
8733 | } | |
949c47a0 DW |
8734 | case update_create_array: { |
8735 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8736 | struct intel_dev *dv; |
54c2c1ea | 8737 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8738 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8739 | struct dl *dl; |
8740 | struct disk_info *inf; | |
8741 | int i; | |
8742 | int activate = 0; | |
949c47a0 | 8743 | |
54c2c1ea DW |
8744 | inf = get_disk_info(u); |
8745 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 8746 | /* allocate a new super->devlist entry */ |
503975b9 N |
8747 | dv = xmalloc(sizeof(*dv)); |
8748 | dv->dev = xmalloc(len); | |
8749 | update->space = dv; | |
949c47a0 | 8750 | |
54c2c1ea DW |
8751 | /* count how many spares will be converted to members */ |
8752 | for (i = 0; i < map->num_members; i++) { | |
8753 | dl = serial_to_dl(inf[i].serial, super); | |
8754 | if (!dl) { | |
8755 | /* hmm maybe it failed?, nothing we can do about | |
8756 | * it here | |
8757 | */ | |
8758 | continue; | |
8759 | } | |
8760 | if (count_memberships(dl, super) == 0) | |
8761 | activate++; | |
8762 | } | |
8763 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
8764 | break; |
8765 | default: | |
8766 | break; | |
8767 | } | |
8768 | } | |
8273f55e | 8769 | |
4d7b1503 DW |
8770 | /* check if we need a larger metadata buffer */ |
8771 | if (super->next_buf) | |
8772 | buf_len = super->next_len; | |
8773 | else | |
8774 | buf_len = super->len; | |
8775 | ||
8776 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
8777 | /* ok we need a larger buf than what is currently allocated | |
8778 | * if this allocation fails process_update will notice that | |
8779 | * ->next_len is set and ->next_buf is NULL | |
8780 | */ | |
8781 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
8782 | if (super->next_buf) | |
8783 | free(super->next_buf); | |
8784 | ||
8785 | super->next_len = buf_len; | |
1f45a8ad DW |
8786 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
8787 | memset(super->next_buf, 0, buf_len); | |
8788 | else | |
4d7b1503 DW |
8789 | super->next_buf = NULL; |
8790 | } | |
8273f55e DW |
8791 | } |
8792 | ||
ae6aad82 | 8793 | /* must be called while manager is quiesced */ |
f21e18ca | 8794 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
8795 | { |
8796 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
8797 | struct dl *iter; |
8798 | struct imsm_dev *dev; | |
8799 | struct imsm_map *map; | |
24565c9a DW |
8800 | int i, j, num_members; |
8801 | __u32 ord; | |
ae6aad82 | 8802 | |
24565c9a DW |
8803 | dprintf("%s: deleting device[%d] from imsm_super\n", |
8804 | __func__, index); | |
ae6aad82 DW |
8805 | |
8806 | /* shift all indexes down one */ | |
8807 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 8808 | if (iter->index > (int)index) |
ae6aad82 | 8809 | iter->index--; |
47ee5a45 | 8810 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 8811 | if (iter->index > (int)index) |
47ee5a45 | 8812 | iter->index--; |
ae6aad82 DW |
8813 | |
8814 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8815 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8816 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
8817 | num_members = map->num_members; |
8818 | for (j = 0; j < num_members; j++) { | |
8819 | /* update ord entries being careful not to propagate | |
8820 | * ord-flags to the first map | |
8821 | */ | |
238c0a71 | 8822 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 8823 | |
24565c9a DW |
8824 | if (ord_to_idx(ord) <= index) |
8825 | continue; | |
ae6aad82 | 8826 | |
238c0a71 | 8827 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 8828 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 8829 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
8830 | if (map) |
8831 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
8832 | } |
8833 | } | |
8834 | ||
8835 | mpb->num_disks--; | |
8836 | super->updates_pending++; | |
24565c9a DW |
8837 | if (*dlp) { |
8838 | struct dl *dl = *dlp; | |
8839 | ||
8840 | *dlp = (*dlp)->next; | |
8841 | __free_imsm_disk(dl); | |
8842 | } | |
ae6aad82 | 8843 | } |
9e2d750d | 8844 | #endif /* MDASSEMBLE */ |
9a717282 AK |
8845 | |
8846 | static void close_targets(int *targets, int new_disks) | |
8847 | { | |
8848 | int i; | |
8849 | ||
8850 | if (!targets) | |
8851 | return; | |
8852 | ||
8853 | for (i = 0; i < new_disks; i++) { | |
8854 | if (targets[i] >= 0) { | |
8855 | close(targets[i]); | |
8856 | targets[i] = -1; | |
8857 | } | |
8858 | } | |
8859 | } | |
8860 | ||
8861 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
8862 | struct intel_super *super, | |
8863 | struct imsm_dev *dev) | |
8864 | { | |
8865 | switch (level) { | |
bf5cf7c7 | 8866 | case 1: |
9a717282 AK |
8867 | case 10:{ |
8868 | int ret_val = 0; | |
8869 | struct imsm_map *map; | |
8870 | int i; | |
8871 | ||
8872 | ret_val = raid_disks/2; | |
8873 | /* check map if all disks pairs not failed | |
8874 | * in both maps | |
8875 | */ | |
238c0a71 | 8876 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
8877 | for (i = 0; i < ret_val; i++) { |
8878 | int degradation = 0; | |
8879 | if (get_imsm_disk(super, i) == NULL) | |
8880 | degradation++; | |
8881 | if (get_imsm_disk(super, i + 1) == NULL) | |
8882 | degradation++; | |
8883 | if (degradation == 2) | |
8884 | return 0; | |
8885 | } | |
238c0a71 | 8886 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
8887 | /* if there is no second map |
8888 | * result can be returned | |
8889 | */ | |
8890 | if (map == NULL) | |
8891 | return ret_val; | |
8892 | /* check degradation in second map | |
8893 | */ | |
8894 | for (i = 0; i < ret_val; i++) { | |
8895 | int degradation = 0; | |
8896 | if (get_imsm_disk(super, i) == NULL) | |
8897 | degradation++; | |
8898 | if (get_imsm_disk(super, i + 1) == NULL) | |
8899 | degradation++; | |
8900 | if (degradation == 2) | |
8901 | return 0; | |
8902 | } | |
8903 | return ret_val; | |
8904 | } | |
8905 | case 5: | |
8906 | return 1; | |
8907 | case 6: | |
8908 | return 2; | |
8909 | default: | |
8910 | return 0; | |
8911 | } | |
8912 | } | |
8913 | ||
8914 | ||
687629c2 AK |
8915 | /******************************************************************************* |
8916 | * Function: open_backup_targets | |
8917 | * Description: Function opens file descriptors for all devices given in | |
8918 | * info->devs | |
8919 | * Parameters: | |
8920 | * info : general array info | |
8921 | * raid_disks : number of disks | |
8922 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
8923 | * super : intel super for raid10 degradation check |
8924 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
8925 | * Returns: |
8926 | * 0 : success | |
8927 | * -1 : fail | |
8928 | ******************************************************************************/ | |
9a717282 AK |
8929 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
8930 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
8931 | { |
8932 | struct mdinfo *sd; | |
f627f5ad | 8933 | int i; |
9a717282 | 8934 | int opened = 0; |
f627f5ad AK |
8935 | |
8936 | for (i = 0; i < raid_disks; i++) | |
8937 | raid_fds[i] = -1; | |
687629c2 AK |
8938 | |
8939 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8940 | char *dn; | |
8941 | ||
8942 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
8943 | dprintf("disk is faulty!!\n"); | |
8944 | continue; | |
8945 | } | |
8946 | ||
8947 | if ((sd->disk.raid_disk >= raid_disks) || | |
8948 | (sd->disk.raid_disk < 0)) | |
8949 | continue; | |
8950 | ||
8951 | dn = map_dev(sd->disk.major, | |
8952 | sd->disk.minor, 1); | |
8953 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
8954 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
8955 | fprintf(stderr, "cannot open component\n"); | |
9a717282 | 8956 | continue; |
687629c2 | 8957 | } |
9a717282 AK |
8958 | opened++; |
8959 | } | |
8960 | /* check if maximum array degradation level is not exceeded | |
8961 | */ | |
8962 | if ((raid_disks - opened) > | |
8963 | imsm_get_allowed_degradation(info->new_level, | |
8964 | raid_disks, | |
8965 | super, dev)) { | |
8966 | fprintf(stderr, "Not enough disks can be opened.\n"); | |
8967 | close_targets(raid_fds, raid_disks); | |
8968 | return -2; | |
687629c2 AK |
8969 | } |
8970 | return 0; | |
8971 | } | |
8972 | ||
9e2d750d | 8973 | #ifndef MDASSEMBLE |
687629c2 AK |
8974 | /******************************************************************************* |
8975 | * Function: init_migr_record_imsm | |
8976 | * Description: Function inits imsm migration record | |
8977 | * Parameters: | |
8978 | * super : imsm internal array info | |
8979 | * dev : device under migration | |
8980 | * info : general array info to find the smallest device | |
8981 | * Returns: | |
8982 | * none | |
8983 | ******************************************************************************/ | |
8984 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
8985 | struct mdinfo *info) | |
8986 | { | |
8987 | struct intel_super *super = st->sb; | |
8988 | struct migr_record *migr_rec = super->migr_rec; | |
8989 | int new_data_disks; | |
8990 | unsigned long long dsize, dev_sectors; | |
8991 | long long unsigned min_dev_sectors = -1LLU; | |
8992 | struct mdinfo *sd; | |
8993 | char nm[30]; | |
8994 | int fd; | |
238c0a71 AK |
8995 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
8996 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 8997 | unsigned long long num_migr_units; |
3ef4403c | 8998 | unsigned long long array_blocks; |
687629c2 AK |
8999 | |
9000 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
9001 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
9002 | ||
9003 | /* only ascending reshape supported now */ | |
9004 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
9005 | ||
9006 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
9007 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
9008 | migr_rec->dest_depth_per_unit *= |
9009 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 9010 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
9011 | migr_rec->blocks_per_unit = |
9012 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
9013 | migr_rec->dest_depth_per_unit = | |
9014 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 9015 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
9016 | num_migr_units = |
9017 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
9018 | ||
9019 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
9020 | num_migr_units++; | |
9021 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
9022 | ||
9023 | migr_rec->post_migr_vol_cap = dev->size_low; | |
9024 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
9025 | ||
9026 | ||
9027 | /* Find the smallest dev */ | |
9028 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9029 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
9030 | fd = dev_open(nm, O_RDONLY); | |
9031 | if (fd < 0) | |
9032 | continue; | |
9033 | get_dev_size(fd, NULL, &dsize); | |
9034 | dev_sectors = dsize / 512; | |
9035 | if (dev_sectors < min_dev_sectors) | |
9036 | min_dev_sectors = dev_sectors; | |
9037 | close(fd); | |
9038 | } | |
9039 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
9040 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
9041 | ||
9042 | write_imsm_migr_rec(st); | |
9043 | ||
9044 | return; | |
9045 | } | |
9046 | ||
9047 | /******************************************************************************* | |
9048 | * Function: save_backup_imsm | |
9049 | * Description: Function saves critical data stripes to Migration Copy Area | |
9050 | * and updates the current migration unit status. | |
9051 | * Use restore_stripes() to form a destination stripe, | |
9052 | * and to write it to the Copy Area. | |
9053 | * Parameters: | |
9054 | * st : supertype information | |
aea93171 | 9055 | * dev : imsm device that backup is saved for |
687629c2 AK |
9056 | * info : general array info |
9057 | * buf : input buffer | |
687629c2 AK |
9058 | * length : length of data to backup (blocks_per_unit) |
9059 | * Returns: | |
9060 | * 0 : success | |
9061 | *, -1 : fail | |
9062 | ******************************************************************************/ | |
9063 | int save_backup_imsm(struct supertype *st, | |
9064 | struct imsm_dev *dev, | |
9065 | struct mdinfo *info, | |
9066 | void *buf, | |
687629c2 AK |
9067 | int length) |
9068 | { | |
9069 | int rv = -1; | |
9070 | struct intel_super *super = st->sb; | |
9071 | unsigned long long *target_offsets = NULL; | |
9072 | int *targets = NULL; | |
9073 | int i; | |
238c0a71 | 9074 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 9075 | int new_disks = map_dest->num_members; |
ab724b98 AK |
9076 | int dest_layout = 0; |
9077 | int dest_chunk; | |
d1877f69 | 9078 | unsigned long long start; |
238c0a71 | 9079 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 9080 | |
503975b9 | 9081 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 9082 | |
7e45b550 AK |
9083 | for (i = 0; i < new_disks; i++) |
9084 | targets[i] = -1; | |
9085 | ||
503975b9 | 9086 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 9087 | |
d1877f69 | 9088 | start = info->reshape_progress * 512; |
687629c2 | 9089 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
9090 | target_offsets[i] = (unsigned long long) |
9091 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
9092 | /* move back copy area adderss, it will be moved forward |
9093 | * in restore_stripes() using start input variable | |
9094 | */ | |
9095 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
9096 | } |
9097 | ||
9a717282 AK |
9098 | if (open_backup_targets(info, new_disks, targets, |
9099 | super, dev)) | |
687629c2 AK |
9100 | goto abort; |
9101 | ||
68eb8bc6 | 9102 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
9103 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
9104 | ||
687629c2 AK |
9105 | if (restore_stripes(targets, /* list of dest devices */ |
9106 | target_offsets, /* migration record offsets */ | |
9107 | new_disks, | |
ab724b98 AK |
9108 | dest_chunk, |
9109 | map_dest->raid_level, | |
9110 | dest_layout, | |
9111 | -1, /* source backup file descriptor */ | |
9112 | 0, /* input buf offset | |
9113 | * always 0 buf is already offseted */ | |
d1877f69 | 9114 | start, |
687629c2 AK |
9115 | length, |
9116 | buf) != 0) { | |
e7b84f9d | 9117 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
9118 | goto abort; |
9119 | } | |
9120 | ||
9121 | rv = 0; | |
9122 | ||
9123 | abort: | |
9124 | if (targets) { | |
9a717282 | 9125 | close_targets(targets, new_disks); |
687629c2 AK |
9126 | free(targets); |
9127 | } | |
9128 | free(target_offsets); | |
9129 | ||
9130 | return rv; | |
9131 | } | |
9132 | ||
9133 | /******************************************************************************* | |
9134 | * Function: save_checkpoint_imsm | |
9135 | * Description: Function called for current unit status update | |
9136 | * in the migration record. It writes it to disk. | |
9137 | * Parameters: | |
9138 | * super : imsm internal array info | |
9139 | * info : general array info | |
9140 | * Returns: | |
9141 | * 0: success | |
9142 | * 1: failure | |
0228d92c AK |
9143 | * 2: failure, means no valid migration record |
9144 | * / no general migration in progress / | |
687629c2 AK |
9145 | ******************************************************************************/ |
9146 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
9147 | { | |
9148 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
9149 | unsigned long long blocks_per_unit; |
9150 | unsigned long long curr_migr_unit; | |
9151 | ||
2e062e82 AK |
9152 | if (load_imsm_migr_rec(super, info) != 0) { |
9153 | dprintf("imsm: ERROR: Cannot read migration record " | |
9154 | "for checkpoint save.\n"); | |
9155 | return 1; | |
9156 | } | |
9157 | ||
f8b72ef5 AK |
9158 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
9159 | if (blocks_per_unit == 0) { | |
0228d92c AK |
9160 | dprintf("imsm: no migration in progress.\n"); |
9161 | return 2; | |
687629c2 | 9162 | } |
f8b72ef5 AK |
9163 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
9164 | /* check if array is alligned to copy area | |
9165 | * if it is not alligned, add one to current migration unit value | |
9166 | * this can happend on array reshape finish only | |
9167 | */ | |
9168 | if (info->reshape_progress % blocks_per_unit) | |
9169 | curr_migr_unit++; | |
687629c2 AK |
9170 | |
9171 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 9172 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
9173 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9174 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
9175 | __cpu_to_le32(curr_migr_unit * |
9176 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
9177 | if (write_imsm_migr_rec(st) < 0) { |
9178 | dprintf("imsm: Cannot write migration record " | |
9179 | "outside backup area\n"); | |
9180 | return 1; | |
9181 | } | |
9182 | ||
9183 | return 0; | |
9184 | } | |
9185 | ||
276d77db AK |
9186 | /******************************************************************************* |
9187 | * Function: recover_backup_imsm | |
9188 | * Description: Function recovers critical data from the Migration Copy Area | |
9189 | * while assembling an array. | |
9190 | * Parameters: | |
9191 | * super : imsm internal array info | |
9192 | * info : general array info | |
9193 | * Returns: | |
9194 | * 0 : success (or there is no data to recover) | |
9195 | * 1 : fail | |
9196 | ******************************************************************************/ | |
9197 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
9198 | { | |
9199 | struct intel_super *super = st->sb; | |
9200 | struct migr_record *migr_rec = super->migr_rec; | |
9201 | struct imsm_map *map_dest = NULL; | |
9202 | struct intel_dev *id = NULL; | |
9203 | unsigned long long read_offset; | |
9204 | unsigned long long write_offset; | |
9205 | unsigned unit_len; | |
9206 | int *targets = NULL; | |
9207 | int new_disks, i, err; | |
9208 | char *buf = NULL; | |
9209 | int retval = 1; | |
9210 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
9211 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 9212 | char buffer[20]; |
6c3560c0 | 9213 | int skipped_disks = 0; |
276d77db AK |
9214 | |
9215 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
9216 | if (err < 1) | |
9217 | return 1; | |
9218 | ||
9219 | /* recover data only during assemblation */ | |
9220 | if (strncmp(buffer, "inactive", 8) != 0) | |
9221 | return 0; | |
9222 | /* no data to recover */ | |
9223 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
9224 | return 0; | |
9225 | if (curr_migr_unit >= num_migr_units) | |
9226 | return 1; | |
9227 | ||
9228 | /* find device during reshape */ | |
9229 | for (id = super->devlist; id; id = id->next) | |
9230 | if (is_gen_migration(id->dev)) | |
9231 | break; | |
9232 | if (id == NULL) | |
9233 | return 1; | |
9234 | ||
238c0a71 | 9235 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
9236 | new_disks = map_dest->num_members; |
9237 | ||
9238 | read_offset = (unsigned long long) | |
9239 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
9240 | ||
9241 | write_offset = ((unsigned long long) | |
9242 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 9243 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
9244 | |
9245 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9246 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
9247 | goto abort; | |
503975b9 | 9248 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 9249 | |
9a717282 | 9250 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 9251 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
9252 | goto abort; |
9253 | } | |
276d77db AK |
9254 | |
9255 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
9256 | if (targets[i] < 0) { |
9257 | skipped_disks++; | |
9258 | continue; | |
9259 | } | |
276d77db | 9260 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
9261 | pr_err("Cannot seek to block: %s\n", |
9262 | strerror(errno)); | |
137debce AK |
9263 | skipped_disks++; |
9264 | continue; | |
276d77db | 9265 | } |
9ec11d1a | 9266 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9267 | pr_err("Cannot read copy area block: %s\n", |
9268 | strerror(errno)); | |
137debce AK |
9269 | skipped_disks++; |
9270 | continue; | |
276d77db AK |
9271 | } |
9272 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
9273 | pr_err("Cannot seek to block: %s\n", |
9274 | strerror(errno)); | |
137debce AK |
9275 | skipped_disks++; |
9276 | continue; | |
276d77db | 9277 | } |
9ec11d1a | 9278 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9279 | pr_err("Cannot restore block: %s\n", |
9280 | strerror(errno)); | |
137debce AK |
9281 | skipped_disks++; |
9282 | continue; | |
276d77db AK |
9283 | } |
9284 | } | |
9285 | ||
137debce AK |
9286 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
9287 | new_disks, | |
9288 | super, | |
9289 | id->dev)) { | |
e7b84f9d N |
9290 | pr_err("Cannot restore data from backup." |
9291 | " Too many failed disks\n"); | |
6c3560c0 AK |
9292 | goto abort; |
9293 | } | |
9294 | ||
befb629b AK |
9295 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
9296 | /* ignore error == 2, this can mean end of reshape here | |
9297 | */ | |
9298 | dprintf("imsm: Cannot write checkpoint to " | |
9299 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
9300 | } else | |
276d77db | 9301 | retval = 0; |
276d77db AK |
9302 | |
9303 | abort: | |
9304 | if (targets) { | |
9305 | for (i = 0; i < new_disks; i++) | |
9306 | if (targets[i]) | |
9307 | close(targets[i]); | |
9308 | free(targets); | |
9309 | } | |
9310 | free(buf); | |
9311 | return retval; | |
9312 | } | |
9313 | ||
2cda7640 ML |
9314 | static char disk_by_path[] = "/dev/disk/by-path/"; |
9315 | ||
9316 | static const char *imsm_get_disk_controller_domain(const char *path) | |
9317 | { | |
2cda7640 | 9318 | char disk_path[PATH_MAX]; |
96234762 LM |
9319 | char *drv=NULL; |
9320 | struct stat st; | |
2cda7640 | 9321 | |
96234762 LM |
9322 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
9323 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
9324 | if (stat(disk_path, &st) == 0) { | |
9325 | struct sys_dev* hba; | |
9326 | char *path=NULL; | |
9327 | ||
9328 | path = devt_to_devpath(st.st_rdev); | |
9329 | if (path == NULL) | |
9330 | return "unknown"; | |
9331 | hba = find_disk_attached_hba(-1, path); | |
9332 | if (hba && hba->type == SYS_DEV_SAS) | |
9333 | drv = "isci"; | |
9334 | else if (hba && hba->type == SYS_DEV_SATA) | |
9335 | drv = "ahci"; | |
9336 | else | |
9337 | drv = "unknown"; | |
9338 | dprintf("path: %s hba: %s attached: %s\n", | |
9339 | path, (hba) ? hba->path : "NULL", drv); | |
9340 | free(path); | |
9341 | if (hba) | |
9342 | free_sys_dev(&hba); | |
2cda7640 | 9343 | } |
96234762 | 9344 | return drv; |
2cda7640 ML |
9345 | } |
9346 | ||
78b10e66 N |
9347 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
9348 | { | |
9349 | char subdev_name[20]; | |
9350 | struct mdstat_ent *mdstat; | |
9351 | ||
9352 | sprintf(subdev_name, "%d", subdev); | |
9353 | mdstat = mdstat_by_subdev(subdev_name, container); | |
9354 | if (!mdstat) | |
9355 | return -1; | |
9356 | ||
9357 | *minor = mdstat->devnum; | |
9358 | free_mdstat(mdstat); | |
9359 | return 0; | |
9360 | } | |
9361 | ||
9362 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
9363 | struct geo_params *geo, | |
fbf3d202 AK |
9364 | int *old_raid_disks, |
9365 | int direction) | |
78b10e66 | 9366 | { |
694575e7 KW |
9367 | /* currently we only support increasing the number of devices |
9368 | * for a container. This increases the number of device for each | |
9369 | * member array. They must all be RAID0 or RAID5. | |
9370 | */ | |
78b10e66 N |
9371 | int ret_val = 0; |
9372 | struct mdinfo *info, *member; | |
9373 | int devices_that_can_grow = 0; | |
9374 | ||
9375 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
9376 | "st->devnum = (%i)\n", | |
9377 | st->devnum); | |
9378 | ||
d04f65f4 | 9379 | if (geo->size > 0 || |
78b10e66 N |
9380 | geo->level != UnSet || |
9381 | geo->layout != UnSet || | |
9382 | geo->chunksize != 0 || | |
9383 | geo->raid_disks == UnSet) { | |
9384 | dprintf("imsm: Container operation is allowed for " | |
9385 | "raid disks number change only.\n"); | |
9386 | return ret_val; | |
9387 | } | |
9388 | ||
fbf3d202 AK |
9389 | if (direction == ROLLBACK_METADATA_CHANGES) { |
9390 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9391 | "container operation.\n"); | |
9392 | return ret_val; | |
9393 | } | |
9394 | ||
78b10e66 N |
9395 | info = container_content_imsm(st, NULL); |
9396 | for (member = info; member; member = member->next) { | |
9397 | int result; | |
9398 | int minor; | |
9399 | ||
9400 | dprintf("imsm: checking device_num: %i\n", | |
9401 | member->container_member); | |
9402 | ||
d7d205bd | 9403 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
9404 | /* we work on container for Online Capacity Expansion |
9405 | * only so raid_disks has to grow | |
9406 | */ | |
9407 | dprintf("imsm: for container operation raid disks " | |
9408 | "increase is required\n"); | |
9409 | break; | |
9410 | } | |
9411 | ||
9412 | if ((info->array.level != 0) && | |
9413 | (info->array.level != 5)) { | |
9414 | /* we cannot use this container with other raid level | |
9415 | */ | |
690aae1a | 9416 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
9417 | " raid level (%i) detected\n", |
9418 | info->array.level); | |
9419 | break; | |
9420 | } else { | |
9421 | /* check for platform support | |
9422 | * for this raid level configuration | |
9423 | */ | |
9424 | struct intel_super *super = st->sb; | |
9425 | if (!is_raid_level_supported(super->orom, | |
9426 | member->array.level, | |
9427 | geo->raid_disks)) { | |
690aae1a | 9428 | dprintf("platform does not support raid%d with" |
78b10e66 N |
9429 | " %d disk%s\n", |
9430 | info->array.level, | |
9431 | geo->raid_disks, | |
9432 | geo->raid_disks > 1 ? "s" : ""); | |
9433 | break; | |
9434 | } | |
2a4a08e7 AK |
9435 | /* check if component size is aligned to chunk size |
9436 | */ | |
9437 | if (info->component_size % | |
9438 | (info->array.chunk_size/512)) { | |
9439 | dprintf("Component size is not aligned to " | |
9440 | "chunk size\n"); | |
9441 | break; | |
9442 | } | |
78b10e66 N |
9443 | } |
9444 | ||
9445 | if (*old_raid_disks && | |
9446 | info->array.raid_disks != *old_raid_disks) | |
9447 | break; | |
9448 | *old_raid_disks = info->array.raid_disks; | |
9449 | ||
9450 | /* All raid5 and raid0 volumes in container | |
9451 | * have to be ready for Online Capacity Expansion | |
9452 | * so they need to be assembled. We have already | |
9453 | * checked that no recovery etc is happening. | |
9454 | */ | |
9455 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
9456 | st->container_dev, | |
9457 | &minor); | |
9458 | if (result < 0) { | |
9459 | dprintf("imsm: cannot find array\n"); | |
9460 | break; | |
9461 | } | |
9462 | devices_that_can_grow++; | |
9463 | } | |
9464 | sysfs_free(info); | |
9465 | if (!member && devices_that_can_grow) | |
9466 | ret_val = 1; | |
9467 | ||
9468 | if (ret_val) | |
9469 | dprintf("\tContainer operation allowed\n"); | |
9470 | else | |
9471 | dprintf("\tError: %i\n", ret_val); | |
9472 | ||
9473 | return ret_val; | |
9474 | } | |
9475 | ||
9476 | /* Function: get_spares_for_grow | |
9477 | * Description: Allocates memory and creates list of spare devices | |
9478 | * avaliable in container. Checks if spare drive size is acceptable. | |
9479 | * Parameters: Pointer to the supertype structure | |
9480 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
9481 | * NULL if fail | |
9482 | */ | |
9483 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
9484 | { | |
78b10e66 | 9485 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 9486 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
9487 | } |
9488 | ||
9489 | /****************************************************************************** | |
9490 | * function: imsm_create_metadata_update_for_reshape | |
9491 | * Function creates update for whole IMSM container. | |
9492 | * | |
9493 | ******************************************************************************/ | |
9494 | static int imsm_create_metadata_update_for_reshape( | |
9495 | struct supertype *st, | |
9496 | struct geo_params *geo, | |
9497 | int old_raid_disks, | |
9498 | struct imsm_update_reshape **updatep) | |
9499 | { | |
9500 | struct intel_super *super = st->sb; | |
9501 | struct imsm_super *mpb = super->anchor; | |
9502 | int update_memory_size = 0; | |
9503 | struct imsm_update_reshape *u = NULL; | |
9504 | struct mdinfo *spares = NULL; | |
9505 | int i; | |
9506 | int delta_disks = 0; | |
bbd24d86 | 9507 | struct mdinfo *dev; |
78b10e66 N |
9508 | |
9509 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
9510 | geo->raid_disks); | |
9511 | ||
9512 | delta_disks = geo->raid_disks - old_raid_disks; | |
9513 | ||
9514 | /* size of all update data without anchor */ | |
9515 | update_memory_size = sizeof(struct imsm_update_reshape); | |
9516 | ||
9517 | /* now add space for spare disks that we need to add. */ | |
9518 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
9519 | ||
503975b9 | 9520 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
9521 | u->type = update_reshape_container_disks; |
9522 | u->old_raid_disks = old_raid_disks; | |
9523 | u->new_raid_disks = geo->raid_disks; | |
9524 | ||
9525 | /* now get spare disks list | |
9526 | */ | |
9527 | spares = get_spares_for_grow(st); | |
9528 | ||
9529 | if (spares == NULL | |
9530 | || delta_disks > spares->array.spare_disks) { | |
e7b84f9d | 9531 | pr_err("imsm: ERROR: Cannot get spare devices " |
e14e5960 | 9532 | "for %s.\n", geo->dev_name); |
e4c72d1d | 9533 | i = -1; |
78b10e66 N |
9534 | goto abort; |
9535 | } | |
9536 | ||
9537 | /* we have got spares | |
9538 | * update disk list in imsm_disk list table in anchor | |
9539 | */ | |
9540 | dprintf("imsm: %i spares are available.\n\n", | |
9541 | spares->array.spare_disks); | |
9542 | ||
bbd24d86 | 9543 | dev = spares->devs; |
78b10e66 | 9544 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
9545 | struct dl *dl; |
9546 | ||
bbd24d86 AK |
9547 | if (dev == NULL) |
9548 | break; | |
78b10e66 N |
9549 | u->new_disks[i] = makedev(dev->disk.major, |
9550 | dev->disk.minor); | |
9551 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
9552 | dl->index = mpb->num_disks; |
9553 | mpb->num_disks++; | |
bbd24d86 | 9554 | dev = dev->next; |
78b10e66 | 9555 | } |
78b10e66 N |
9556 | |
9557 | abort: | |
9558 | /* free spares | |
9559 | */ | |
9560 | sysfs_free(spares); | |
9561 | ||
d677e0b8 | 9562 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 9563 | if (i == delta_disks) { |
d677e0b8 | 9564 | dprintf(" OK\n"); |
78b10e66 N |
9565 | *updatep = u; |
9566 | return update_memory_size; | |
9567 | } | |
9568 | free(u); | |
d677e0b8 | 9569 | dprintf(" Error\n"); |
78b10e66 N |
9570 | |
9571 | return 0; | |
9572 | } | |
9573 | ||
f3871fdc AK |
9574 | |
9575 | /****************************************************************************** | |
9576 | * function: imsm_create_metadata_update_for_size_change() | |
9577 | * Creates update for IMSM array for array size change. | |
9578 | * | |
9579 | ******************************************************************************/ | |
9580 | static int imsm_create_metadata_update_for_size_change( | |
9581 | struct supertype *st, | |
9582 | struct geo_params *geo, | |
9583 | struct imsm_update_size_change **updatep) | |
9584 | { | |
9585 | struct intel_super *super = st->sb; | |
9586 | int update_memory_size = 0; | |
9587 | struct imsm_update_size_change *u = NULL; | |
9588 | ||
9589 | dprintf("imsm_create_metadata_update_for_size_change(enter)" | |
9590 | " New size = %llu\n", geo->size); | |
9591 | ||
9592 | /* size of all update data without anchor */ | |
9593 | update_memory_size = sizeof(struct imsm_update_size_change); | |
9594 | ||
503975b9 | 9595 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
9596 | u->type = update_size_change; |
9597 | u->subdev = super->current_vol; | |
9598 | u->new_size = geo->size; | |
9599 | ||
9600 | dprintf("imsm: reshape update preparation : OK\n"); | |
9601 | *updatep = u; | |
9602 | ||
9603 | return update_memory_size; | |
9604 | } | |
9605 | ||
48c5303a PC |
9606 | /****************************************************************************** |
9607 | * function: imsm_create_metadata_update_for_migration() | |
9608 | * Creates update for IMSM array. | |
9609 | * | |
9610 | ******************************************************************************/ | |
9611 | static int imsm_create_metadata_update_for_migration( | |
9612 | struct supertype *st, | |
9613 | struct geo_params *geo, | |
9614 | struct imsm_update_reshape_migration **updatep) | |
9615 | { | |
9616 | struct intel_super *super = st->sb; | |
9617 | int update_memory_size = 0; | |
9618 | struct imsm_update_reshape_migration *u = NULL; | |
9619 | struct imsm_dev *dev; | |
9620 | int previous_level = -1; | |
9621 | ||
9622 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
9623 | " New Level = %i\n", geo->level); | |
9624 | ||
9625 | /* size of all update data without anchor */ | |
9626 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
9627 | ||
503975b9 | 9628 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
9629 | u->type = update_reshape_migration; |
9630 | u->subdev = super->current_vol; | |
9631 | u->new_level = geo->level; | |
9632 | u->new_layout = geo->layout; | |
9633 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
9634 | u->new_disks[0] = -1; | |
4bba0439 | 9635 | u->new_chunksize = -1; |
48c5303a PC |
9636 | |
9637 | dev = get_imsm_dev(super, u->subdev); | |
9638 | if (dev) { | |
9639 | struct imsm_map *map; | |
9640 | ||
238c0a71 | 9641 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
9642 | if (map) { |
9643 | int current_chunk_size = | |
9644 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
9645 | ||
9646 | if (geo->chunksize != current_chunk_size) { | |
9647 | u->new_chunksize = geo->chunksize / 1024; | |
9648 | dprintf("imsm: " | |
9649 | "chunk size change from %i to %i\n", | |
9650 | current_chunk_size, u->new_chunksize); | |
9651 | } | |
48c5303a | 9652 | previous_level = map->raid_level; |
4bba0439 | 9653 | } |
48c5303a PC |
9654 | } |
9655 | if ((geo->level == 5) && (previous_level == 0)) { | |
9656 | struct mdinfo *spares = NULL; | |
9657 | ||
9658 | u->new_raid_disks++; | |
9659 | spares = get_spares_for_grow(st); | |
9660 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
9661 | free(u); | |
9662 | sysfs_free(spares); | |
9663 | update_memory_size = 0; | |
9664 | dprintf("error: cannot get spare device " | |
9665 | "for requested migration"); | |
9666 | return 0; | |
9667 | } | |
9668 | sysfs_free(spares); | |
9669 | } | |
9670 | dprintf("imsm: reshape update preparation : OK\n"); | |
9671 | *updatep = u; | |
9672 | ||
9673 | return update_memory_size; | |
9674 | } | |
9675 | ||
8dd70bce AK |
9676 | static void imsm_update_metadata_locally(struct supertype *st, |
9677 | void *buf, int len) | |
9678 | { | |
9679 | struct metadata_update mu; | |
9680 | ||
9681 | mu.buf = buf; | |
9682 | mu.len = len; | |
9683 | mu.space = NULL; | |
9684 | mu.space_list = NULL; | |
9685 | mu.next = NULL; | |
9686 | imsm_prepare_update(st, &mu); | |
9687 | imsm_process_update(st, &mu); | |
9688 | ||
9689 | while (mu.space_list) { | |
9690 | void **space = mu.space_list; | |
9691 | mu.space_list = *space; | |
9692 | free(space); | |
9693 | } | |
9694 | } | |
78b10e66 | 9695 | |
471bceb6 | 9696 | /*************************************************************************** |
694575e7 | 9697 | * Function: imsm_analyze_change |
471bceb6 KW |
9698 | * Description: Function analyze change for single volume |
9699 | * and validate if transition is supported | |
fbf3d202 AK |
9700 | * Parameters: Geometry parameters, supertype structure, |
9701 | * metadata change direction (apply/rollback) | |
694575e7 | 9702 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
9703 | ****************************************************************************/ |
9704 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
9705 | struct geo_params *geo, |
9706 | int direction) | |
694575e7 | 9707 | { |
471bceb6 KW |
9708 | struct mdinfo info; |
9709 | int change = -1; | |
9710 | int check_devs = 0; | |
c21e737b | 9711 | int chunk; |
67a2db32 AK |
9712 | /* number of added/removed disks in operation result */ |
9713 | int devNumChange = 0; | |
9714 | /* imsm compatible layout value for array geometry verification */ | |
9715 | int imsm_layout = -1; | |
7abc9871 AK |
9716 | int data_disks; |
9717 | struct imsm_dev *dev; | |
9718 | struct intel_super *super; | |
d04f65f4 | 9719 | unsigned long long current_size; |
65d38cca | 9720 | unsigned long long free_size; |
d04f65f4 | 9721 | unsigned long long max_size; |
65d38cca | 9722 | int rv; |
471bceb6 KW |
9723 | |
9724 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
9725 | if ((geo->level != info.array.level) && |
9726 | (geo->level >= 0) && | |
9727 | (geo->level != UnSet)) { | |
9728 | switch (info.array.level) { | |
9729 | case 0: | |
9730 | if (geo->level == 5) { | |
b5347799 | 9731 | change = CH_MIGRATION; |
e13ce846 | 9732 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
e7b84f9d N |
9733 | pr_err("Error. Requested Layout " |
9734 | "not supported (left-asymmetric layout " | |
9735 | "is supported only)!\n"); | |
e13ce846 AK |
9736 | change = -1; |
9737 | goto analyse_change_exit; | |
9738 | } | |
67a2db32 | 9739 | imsm_layout = geo->layout; |
471bceb6 | 9740 | check_devs = 1; |
e91a3bad LM |
9741 | devNumChange = 1; /* parity disk added */ |
9742 | } else if (geo->level == 10) { | |
471bceb6 KW |
9743 | change = CH_TAKEOVER; |
9744 | check_devs = 1; | |
e91a3bad | 9745 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 9746 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 9747 | } |
dfe77a9e KW |
9748 | break; |
9749 | case 1: | |
471bceb6 KW |
9750 | case 10: |
9751 | if (geo->level == 0) { | |
9752 | change = CH_TAKEOVER; | |
9753 | check_devs = 1; | |
e91a3bad | 9754 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 9755 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
9756 | } |
9757 | break; | |
9758 | } | |
9759 | if (change == -1) { | |
e7b84f9d N |
9760 | pr_err("Error. Level Migration from %d to %d " |
9761 | "not supported!\n", | |
9762 | info.array.level, geo->level); | |
471bceb6 KW |
9763 | goto analyse_change_exit; |
9764 | } | |
9765 | } else | |
9766 | geo->level = info.array.level; | |
9767 | ||
9768 | if ((geo->layout != info.array.layout) | |
9769 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 9770 | change = CH_MIGRATION; |
471bceb6 KW |
9771 | if ((info.array.layout == 0) |
9772 | && (info.array.level == 5) | |
9773 | && (geo->layout == 5)) { | |
9774 | /* reshape 5 -> 4 */ | |
9775 | } else if ((info.array.layout == 5) | |
9776 | && (info.array.level == 5) | |
9777 | && (geo->layout == 0)) { | |
9778 | /* reshape 4 -> 5 */ | |
9779 | geo->layout = 0; | |
9780 | geo->level = 5; | |
9781 | } else { | |
e7b84f9d N |
9782 | pr_err("Error. Layout Migration from %d to %d " |
9783 | "not supported!\n", | |
9784 | info.array.layout, geo->layout); | |
471bceb6 KW |
9785 | change = -1; |
9786 | goto analyse_change_exit; | |
9787 | } | |
67a2db32 | 9788 | } else { |
471bceb6 | 9789 | geo->layout = info.array.layout; |
67a2db32 AK |
9790 | if (imsm_layout == -1) |
9791 | imsm_layout = info.array.layout; | |
9792 | } | |
471bceb6 KW |
9793 | |
9794 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
9795 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 9796 | change = CH_MIGRATION; |
471bceb6 KW |
9797 | else |
9798 | geo->chunksize = info.array.chunk_size; | |
9799 | ||
c21e737b | 9800 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
9801 | |
9802 | super = st->sb; | |
9803 | dev = get_imsm_dev(super, super->current_vol); | |
9804 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 9805 | /* compute current size per disk member |
7abc9871 | 9806 | */ |
c41e00b2 AK |
9807 | current_size = info.custom_array_size / data_disks; |
9808 | ||
d5d2c614 | 9809 | if ((geo->size > 0) && (geo->size != MAX_SIZE)) { |
c41e00b2 AK |
9810 | /* align component size |
9811 | */ | |
9812 | geo->size = imsm_component_size_aligment_check( | |
9813 | get_imsm_raid_level(dev->vol.map), | |
9814 | chunk * 1024, | |
9815 | geo->size * 2); | |
65d0b4ce LD |
9816 | if (geo->size == 0) { |
9817 | pr_err("Error. Size expansion is " \ | |
9818 | "supported only (current size is %llu, " \ | |
9819 | "requested size /rounded/ is 0).\n", | |
9820 | current_size); | |
9821 | goto analyse_change_exit; | |
9822 | } | |
c41e00b2 | 9823 | } |
7abc9871 | 9824 | |
d04f65f4 | 9825 | if ((current_size != geo->size) && (geo->size > 0)) { |
7abc9871 | 9826 | if (change != -1) { |
e7b84f9d | 9827 | pr_err("Error. Size change should be the only " |
7abc9871 AK |
9828 | "one at a time.\n"); |
9829 | change = -1; | |
9830 | goto analyse_change_exit; | |
9831 | } | |
9832 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
e7b84f9d N |
9833 | pr_err("Error. The last volume in container " |
9834 | "can be expanded only (%i/%i).\n", | |
9835 | super->current_vol, st->devnum); | |
7abc9871 AK |
9836 | goto analyse_change_exit; |
9837 | } | |
65d38cca LD |
9838 | /* check the maximum available size |
9839 | */ | |
9840 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
9841 | 0, chunk, &free_size); | |
9842 | if (rv == 0) | |
9843 | /* Cannot find maximum available space | |
9844 | */ | |
9845 | max_size = 0; | |
9846 | else { | |
9847 | max_size = free_size + current_size; | |
9848 | /* align component size | |
9849 | */ | |
9850 | max_size = imsm_component_size_aligment_check( | |
9851 | get_imsm_raid_level(dev->vol.map), | |
9852 | chunk * 1024, | |
9853 | max_size); | |
9854 | } | |
d04f65f4 | 9855 | if (geo->size == MAX_SIZE) { |
b130333f AK |
9856 | /* requested size change to the maximum available size |
9857 | */ | |
65d38cca | 9858 | if (max_size == 0) { |
e7b84f9d | 9859 | pr_err("Error. Cannot find " |
b130333f AK |
9860 | "maximum available space.\n"); |
9861 | change = -1; | |
9862 | goto analyse_change_exit; | |
65d38cca LD |
9863 | } else |
9864 | geo->size = max_size; | |
c41e00b2 | 9865 | } |
b130333f | 9866 | |
fbf3d202 AK |
9867 | if ((direction == ROLLBACK_METADATA_CHANGES)) { |
9868 | /* accept size for rollback only | |
9869 | */ | |
9870 | } else { | |
9871 | /* round size due to metadata compatibility | |
9872 | */ | |
9873 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
9874 | << SECT_PER_MB_SHIFT; | |
9875 | dprintf("Prepare update for size change to %llu\n", | |
9876 | geo->size ); | |
9877 | if (current_size >= geo->size) { | |
e7b84f9d N |
9878 | pr_err("Error. Size expansion is " |
9879 | "supported only (current size is %llu, " | |
9880 | "requested size /rounded/ is %llu).\n", | |
9881 | current_size, geo->size); | |
fbf3d202 AK |
9882 | goto analyse_change_exit; |
9883 | } | |
65d38cca | 9884 | if (max_size && geo->size > max_size) { |
e7b84f9d N |
9885 | pr_err("Error. Requested size is larger " |
9886 | "than maximum available size (maximum " | |
9887 | "available size is %llu, " | |
9888 | "requested size /rounded/ is %llu).\n", | |
9889 | max_size, geo->size); | |
65d38cca LD |
9890 | goto analyse_change_exit; |
9891 | } | |
7abc9871 AK |
9892 | } |
9893 | geo->size *= data_disks; | |
9894 | geo->raid_disks = dev->vol.map->num_members; | |
9895 | change = CH_ARRAY_SIZE; | |
9896 | } | |
471bceb6 KW |
9897 | if (!validate_geometry_imsm(st, |
9898 | geo->level, | |
67a2db32 | 9899 | imsm_layout, |
e91a3bad | 9900 | geo->raid_disks + devNumChange, |
c21e737b | 9901 | &chunk, |
471bceb6 KW |
9902 | geo->size, |
9903 | 0, 0, 1)) | |
9904 | change = -1; | |
9905 | ||
9906 | if (check_devs) { | |
9907 | struct intel_super *super = st->sb; | |
9908 | struct imsm_super *mpb = super->anchor; | |
9909 | ||
9910 | if (mpb->num_raid_devs > 1) { | |
e7b84f9d N |
9911 | pr_err("Error. Cannot perform operation on %s" |
9912 | "- for this operation it MUST be single " | |
9913 | "array in container\n", | |
9914 | geo->dev_name); | |
471bceb6 KW |
9915 | change = -1; |
9916 | } | |
9917 | } | |
9918 | ||
9919 | analyse_change_exit: | |
fbf3d202 AK |
9920 | if ((direction == ROLLBACK_METADATA_CHANGES) && |
9921 | ((change == CH_MIGRATION) || (change == CH_TAKEOVER))) { | |
9922 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9923 | "migration and takeover operations.\n"); | |
9924 | change = -1; | |
9925 | } | |
471bceb6 | 9926 | return change; |
694575e7 KW |
9927 | } |
9928 | ||
bb025c2f KW |
9929 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
9930 | { | |
9931 | struct intel_super *super = st->sb; | |
9932 | struct imsm_update_takeover *u; | |
9933 | ||
503975b9 | 9934 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
9935 | |
9936 | u->type = update_takeover; | |
9937 | u->subarray = super->current_vol; | |
9938 | ||
9939 | /* 10->0 transition */ | |
9940 | if (geo->level == 0) | |
9941 | u->direction = R10_TO_R0; | |
9942 | ||
0529c688 KW |
9943 | /* 0->10 transition */ |
9944 | if (geo->level == 10) | |
9945 | u->direction = R0_TO_R10; | |
9946 | ||
bb025c2f KW |
9947 | /* update metadata locally */ |
9948 | imsm_update_metadata_locally(st, u, | |
9949 | sizeof(struct imsm_update_takeover)); | |
9950 | /* and possibly remotely */ | |
9951 | if (st->update_tail) | |
9952 | append_metadata_update(st, u, | |
9953 | sizeof(struct imsm_update_takeover)); | |
9954 | else | |
9955 | free(u); | |
9956 | ||
9957 | return 0; | |
9958 | } | |
9959 | ||
d04f65f4 N |
9960 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
9961 | int level, | |
78b10e66 | 9962 | int layout, int chunksize, int raid_disks, |
41784c88 | 9963 | int delta_disks, char *backup, char *dev, |
016e00f5 | 9964 | int direction, int verbose) |
78b10e66 | 9965 | { |
78b10e66 N |
9966 | int ret_val = 1; |
9967 | struct geo_params geo; | |
9968 | ||
9969 | dprintf("imsm: reshape_super called.\n"); | |
9970 | ||
71204a50 | 9971 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
9972 | |
9973 | geo.dev_name = dev; | |
694575e7 | 9974 | geo.dev_id = st->devnum; |
78b10e66 N |
9975 | geo.size = size; |
9976 | geo.level = level; | |
9977 | geo.layout = layout; | |
9978 | geo.chunksize = chunksize; | |
9979 | geo.raid_disks = raid_disks; | |
41784c88 AK |
9980 | if (delta_disks != UnSet) |
9981 | geo.raid_disks += delta_disks; | |
78b10e66 N |
9982 | |
9983 | dprintf("\tfor level : %i\n", geo.level); | |
9984 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
9985 | ||
9986 | if (experimental() == 0) | |
9987 | return ret_val; | |
9988 | ||
78b10e66 | 9989 | if (st->container_dev == st->devnum) { |
694575e7 KW |
9990 | /* On container level we can only increase number of devices. */ |
9991 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 9992 | int old_raid_disks = 0; |
6dc0be30 | 9993 | |
78b10e66 | 9994 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 9995 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
9996 | struct imsm_update_reshape *u = NULL; |
9997 | int len; | |
9998 | ||
9999 | len = imsm_create_metadata_update_for_reshape( | |
10000 | st, &geo, old_raid_disks, &u); | |
10001 | ||
ed08d51c AK |
10002 | if (len <= 0) { |
10003 | dprintf("imsm: Cannot prepare update\n"); | |
10004 | goto exit_imsm_reshape_super; | |
10005 | } | |
10006 | ||
8dd70bce AK |
10007 | ret_val = 0; |
10008 | /* update metadata locally */ | |
10009 | imsm_update_metadata_locally(st, u, len); | |
10010 | /* and possibly remotely */ | |
10011 | if (st->update_tail) | |
10012 | append_metadata_update(st, u, len); | |
10013 | else | |
ed08d51c | 10014 | free(u); |
8dd70bce | 10015 | |
694575e7 | 10016 | } else { |
e7b84f9d | 10017 | pr_err("(imsm) Operation " |
e7ff7e40 | 10018 | "is not allowed on this container\n"); |
694575e7 KW |
10019 | } |
10020 | } else { | |
10021 | /* On volume level we support following operations | |
471bceb6 KW |
10022 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
10023 | * - chunk size migration | |
10024 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
10025 | */ | |
10026 | struct intel_super *super = st->sb; | |
10027 | struct intel_dev *dev = super->devlist; | |
10028 | int change, devnum; | |
694575e7 | 10029 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
10030 | /* find requested device */ |
10031 | while (dev) { | |
19986c72 MB |
10032 | if (imsm_find_array_minor_by_subdev( |
10033 | dev->index, st->container_dev, &devnum) == 0 | |
10034 | && devnum == geo.dev_id) | |
471bceb6 KW |
10035 | break; |
10036 | dev = dev->next; | |
10037 | } | |
10038 | if (dev == NULL) { | |
e7b84f9d | 10039 | pr_err("Cannot find %s (%i) subarray\n", |
471bceb6 KW |
10040 | geo.dev_name, geo.dev_id); |
10041 | goto exit_imsm_reshape_super; | |
10042 | } | |
10043 | super->current_vol = dev->index; | |
fbf3d202 | 10044 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 10045 | switch (change) { |
471bceb6 | 10046 | case CH_TAKEOVER: |
bb025c2f | 10047 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 10048 | break; |
48c5303a PC |
10049 | case CH_MIGRATION: { |
10050 | struct imsm_update_reshape_migration *u = NULL; | |
10051 | int len = | |
10052 | imsm_create_metadata_update_for_migration( | |
10053 | st, &geo, &u); | |
10054 | if (len < 1) { | |
10055 | dprintf("imsm: " | |
10056 | "Cannot prepare update\n"); | |
10057 | break; | |
10058 | } | |
471bceb6 | 10059 | ret_val = 0; |
48c5303a PC |
10060 | /* update metadata locally */ |
10061 | imsm_update_metadata_locally(st, u, len); | |
10062 | /* and possibly remotely */ | |
10063 | if (st->update_tail) | |
10064 | append_metadata_update(st, u, len); | |
10065 | else | |
10066 | free(u); | |
10067 | } | |
10068 | break; | |
7abc9871 | 10069 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
10070 | struct imsm_update_size_change *u = NULL; |
10071 | int len = | |
10072 | imsm_create_metadata_update_for_size_change( | |
10073 | st, &geo, &u); | |
10074 | if (len < 1) { | |
10075 | dprintf("imsm: " | |
10076 | "Cannot prepare update\n"); | |
10077 | break; | |
10078 | } | |
10079 | ret_val = 0; | |
10080 | /* update metadata locally */ | |
10081 | imsm_update_metadata_locally(st, u, len); | |
10082 | /* and possibly remotely */ | |
10083 | if (st->update_tail) | |
10084 | append_metadata_update(st, u, len); | |
10085 | else | |
10086 | free(u); | |
7abc9871 AK |
10087 | } |
10088 | break; | |
471bceb6 KW |
10089 | default: |
10090 | ret_val = 1; | |
694575e7 | 10091 | } |
694575e7 | 10092 | } |
78b10e66 | 10093 | |
ed08d51c | 10094 | exit_imsm_reshape_super: |
78b10e66 N |
10095 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
10096 | return ret_val; | |
10097 | } | |
2cda7640 | 10098 | |
eee67a47 AK |
10099 | /******************************************************************************* |
10100 | * Function: wait_for_reshape_imsm | |
10101 | * Description: Function writes new sync_max value and waits until | |
10102 | * reshape process reach new position | |
10103 | * Parameters: | |
10104 | * sra : general array info | |
eee67a47 AK |
10105 | * ndata : number of disks in new array's layout |
10106 | * Returns: | |
10107 | * 0 : success, | |
10108 | * 1 : there is no reshape in progress, | |
10109 | * -1 : fail | |
10110 | ******************************************************************************/ | |
ae9f01f8 | 10111 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
10112 | { |
10113 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
10114 | unsigned long long completed; | |
ae9f01f8 AK |
10115 | /* to_complete : new sync_max position */ |
10116 | unsigned long long to_complete = sra->reshape_progress; | |
10117 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 10118 | |
ae9f01f8 AK |
10119 | if (fd < 0) { |
10120 | dprintf("imsm: wait_for_reshape_imsm() " | |
10121 | "cannot open reshape_position\n"); | |
eee67a47 | 10122 | return 1; |
ae9f01f8 | 10123 | } |
eee67a47 | 10124 | |
ae9f01f8 AK |
10125 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
10126 | dprintf("imsm: wait_for_reshape_imsm() " | |
10127 | "cannot read reshape_position (no reshape in progres)\n"); | |
10128 | close(fd); | |
10129 | return 0; | |
10130 | } | |
eee67a47 | 10131 | |
ae9f01f8 AK |
10132 | if (completed > to_complete) { |
10133 | dprintf("imsm: wait_for_reshape_imsm() " | |
10134 | "wrong next position to set %llu (%llu)\n", | |
10135 | to_complete, completed); | |
10136 | close(fd); | |
10137 | return -1; | |
10138 | } | |
10139 | dprintf("Position set: %llu\n", position_to_set); | |
10140 | if (sysfs_set_num(sra, NULL, "sync_max", | |
10141 | position_to_set) != 0) { | |
10142 | dprintf("imsm: wait_for_reshape_imsm() " | |
10143 | "cannot set reshape position to %llu\n", | |
10144 | position_to_set); | |
10145 | close(fd); | |
10146 | return -1; | |
eee67a47 AK |
10147 | } |
10148 | ||
eee67a47 AK |
10149 | do { |
10150 | char action[20]; | |
10151 | fd_set rfds; | |
10152 | FD_ZERO(&rfds); | |
10153 | FD_SET(fd, &rfds); | |
a47e44fb AK |
10154 | select(fd+1, &rfds, NULL, NULL, NULL); |
10155 | if (sysfs_get_str(sra, NULL, "sync_action", | |
10156 | action, 20) > 0 && | |
10157 | strncmp(action, "reshape", 7) != 0) | |
10158 | break; | |
eee67a47 | 10159 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
10160 | dprintf("imsm: wait_for_reshape_imsm() " |
10161 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
10162 | close(fd); |
10163 | return 1; | |
10164 | } | |
eee67a47 AK |
10165 | } while (completed < to_complete); |
10166 | close(fd); | |
10167 | return 0; | |
10168 | ||
10169 | } | |
10170 | ||
b915c95f AK |
10171 | /******************************************************************************* |
10172 | * Function: check_degradation_change | |
10173 | * Description: Check that array hasn't become failed. | |
10174 | * Parameters: | |
10175 | * info : for sysfs access | |
10176 | * sources : source disks descriptors | |
10177 | * degraded: previous degradation level | |
10178 | * Returns: | |
10179 | * degradation level | |
10180 | ******************************************************************************/ | |
10181 | int check_degradation_change(struct mdinfo *info, | |
10182 | int *sources, | |
10183 | int degraded) | |
10184 | { | |
10185 | unsigned long long new_degraded; | |
e1993023 LD |
10186 | int rv; |
10187 | ||
10188 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
10189 | if ((rv == -1) || (new_degraded != (unsigned long long)degraded)) { | |
b915c95f AK |
10190 | /* check each device to ensure it is still working */ |
10191 | struct mdinfo *sd; | |
10192 | new_degraded = 0; | |
10193 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10194 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
10195 | continue; | |
10196 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
10197 | char sbuf[20]; | |
10198 | if (sysfs_get_str(info, | |
10199 | sd, "state", sbuf, 20) < 0 || | |
10200 | strstr(sbuf, "faulty") || | |
10201 | strstr(sbuf, "in_sync") == NULL) { | |
10202 | /* this device is dead */ | |
10203 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
10204 | if (sd->disk.raid_disk >= 0 && | |
10205 | sources[sd->disk.raid_disk] >= 0) { | |
10206 | close(sources[ | |
10207 | sd->disk.raid_disk]); | |
10208 | sources[sd->disk.raid_disk] = | |
10209 | -1; | |
10210 | } | |
10211 | new_degraded++; | |
10212 | } | |
10213 | } | |
10214 | } | |
10215 | } | |
10216 | ||
10217 | return new_degraded; | |
10218 | } | |
10219 | ||
10f22854 AK |
10220 | /******************************************************************************* |
10221 | * Function: imsm_manage_reshape | |
10222 | * Description: Function finds array under reshape and it manages reshape | |
10223 | * process. It creates stripes backups (if required) and sets | |
10224 | * checheckpoits. | |
10225 | * Parameters: | |
10226 | * afd : Backup handle (nattive) - not used | |
10227 | * sra : general array info | |
10228 | * reshape : reshape parameters - not used | |
10229 | * st : supertype structure | |
10230 | * blocks : size of critical section [blocks] | |
10231 | * fds : table of source device descriptor | |
10232 | * offsets : start of array (offest per devices) | |
10233 | * dests : not used | |
10234 | * destfd : table of destination device descriptor | |
10235 | * destoffsets : table of destination offsets (per device) | |
10236 | * Returns: | |
10237 | * 1 : success, reshape is done | |
10238 | * 0 : fail | |
10239 | ******************************************************************************/ | |
999b4972 N |
10240 | static int imsm_manage_reshape( |
10241 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 10242 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
10243 | int *fds, unsigned long long *offsets, |
10244 | int dests, int *destfd, unsigned long long *destoffsets) | |
10245 | { | |
10f22854 AK |
10246 | int ret_val = 0; |
10247 | struct intel_super *super = st->sb; | |
10248 | struct intel_dev *dv = NULL; | |
10249 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 10250 | struct imsm_map *map_src; |
10f22854 AK |
10251 | int migr_vol_qan = 0; |
10252 | int ndata, odata; /* [bytes] */ | |
10253 | int chunk; /* [bytes] */ | |
10254 | struct migr_record *migr_rec; | |
10255 | char *buf = NULL; | |
10256 | unsigned int buf_size; /* [bytes] */ | |
10257 | unsigned long long max_position; /* array size [bytes] */ | |
10258 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
10259 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
10260 | unsigned long long start_src; /* [bytes] */ |
10261 | unsigned long long start; /* [bytes] */ | |
10262 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 10263 | int degraded = 0; |
ab724b98 | 10264 | int source_layout = 0; |
10f22854 | 10265 | |
1ab242d8 | 10266 | if (!fds || !offsets || !sra) |
10f22854 AK |
10267 | goto abort; |
10268 | ||
10269 | /* Find volume during the reshape */ | |
10270 | for (dv = super->devlist; dv; dv = dv->next) { | |
10271 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
10272 | && dv->dev->vol.migr_state == 1) { | |
10273 | dev = dv->dev; | |
10274 | migr_vol_qan++; | |
10275 | } | |
10276 | } | |
10277 | /* Only one volume can migrate at the same time */ | |
10278 | if (migr_vol_qan != 1) { | |
e7b84f9d | 10279 | pr_err(": %s", migr_vol_qan ? |
10f22854 AK |
10280 | "Number of migrating volumes greater than 1\n" : |
10281 | "There is no volume during migrationg\n"); | |
10282 | goto abort; | |
10283 | } | |
10284 | ||
238c0a71 | 10285 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
10286 | if (map_src == NULL) |
10287 | goto abort; | |
10f22854 | 10288 | |
238c0a71 AK |
10289 | ndata = imsm_num_data_members(dev, MAP_0); |
10290 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 10291 | |
7b1ab482 | 10292 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
10293 | old_data_stripe_length = odata * chunk; |
10294 | ||
10295 | migr_rec = super->migr_rec; | |
10296 | ||
10f22854 AK |
10297 | /* initialize migration record for start condition */ |
10298 | if (sra->reshape_progress == 0) | |
10299 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
10300 | else { |
10301 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
10302 | dprintf("imsm: cannot restart migration when data " | |
10303 | "are present in copy area.\n"); | |
10304 | goto abort; | |
10305 | } | |
6a75c8ca AK |
10306 | /* Save checkpoint to update migration record for current |
10307 | * reshape position (in md). It can be farther than current | |
10308 | * reshape position in metadata. | |
10309 | */ | |
10310 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
10311 | /* ignore error == 2, this can mean end of reshape here | |
10312 | */ | |
10313 | dprintf("imsm: Cannot write checkpoint to " | |
10314 | "migration record (UNIT_SRC_NORMAL, " | |
10315 | "initial save)\n"); | |
10316 | goto abort; | |
10317 | } | |
b2c59438 | 10318 | } |
10f22854 AK |
10319 | |
10320 | /* size for data */ | |
10321 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
10322 | /* extend buffer size for parity disk */ | |
10323 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
10324 | /* add space for stripe aligment */ | |
10325 | buf_size += old_data_stripe_length; | |
10326 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
10327 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
10328 | goto abort; | |
10329 | } | |
10330 | ||
3ef4403c | 10331 | max_position = sra->component_size * ndata; |
68eb8bc6 | 10332 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
10333 | |
10334 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
10335 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
10336 | /* current reshape position [blocks] */ | |
10337 | unsigned long long current_position = | |
10338 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
10339 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
10340 | unsigned long long border; | |
10341 | ||
b915c95f AK |
10342 | /* Check that array hasn't become failed. |
10343 | */ | |
10344 | degraded = check_degradation_change(sra, fds, degraded); | |
10345 | if (degraded > 1) { | |
10346 | dprintf("imsm: Abort reshape due to degradation" | |
10347 | " level (%i)\n", degraded); | |
10348 | goto abort; | |
10349 | } | |
10350 | ||
10f22854 AK |
10351 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
10352 | ||
10353 | if ((current_position + next_step) > max_position) | |
10354 | next_step = max_position - current_position; | |
10355 | ||
92144abf | 10356 | start = current_position * 512; |
10f22854 AK |
10357 | |
10358 | /* allign reading start to old geometry */ | |
10359 | start_buf_shift = start % old_data_stripe_length; | |
10360 | start_src = start - start_buf_shift; | |
10361 | ||
10362 | border = (start_src / odata) - (start / ndata); | |
10363 | border /= 512; | |
10364 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
10365 | /* save critical stripes to buf | |
10366 | * start - start address of current unit | |
10367 | * to backup [bytes] | |
10368 | * start_src - start address of current unit | |
10369 | * to backup alligned to source array | |
10370 | * [bytes] | |
10371 | */ | |
10372 | unsigned long long next_step_filler = 0; | |
10373 | unsigned long long copy_length = next_step * 512; | |
10374 | ||
10375 | /* allign copy area length to stripe in old geometry */ | |
10376 | next_step_filler = ((copy_length + start_buf_shift) | |
10377 | % old_data_stripe_length); | |
10378 | if (next_step_filler) | |
10379 | next_step_filler = (old_data_stripe_length | |
10380 | - next_step_filler); | |
10381 | dprintf("save_stripes() parameters: start = %llu," | |
10382 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
10383 | "\tstart_in_buf_shift = %llu," | |
10384 | "\tnext_step_filler = %llu\n", | |
10385 | start, start_src, copy_length, | |
10386 | start_buf_shift, next_step_filler); | |
10387 | ||
10388 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
10389 | chunk, map_src->raid_level, |
10390 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
10391 | copy_length + |
10392 | next_step_filler + start_buf_shift, | |
10393 | buf)) { | |
10394 | dprintf("imsm: Cannot save stripes" | |
10395 | " to buffer\n"); | |
10396 | goto abort; | |
10397 | } | |
10398 | /* Convert data to destination format and store it | |
10399 | * in backup general migration area | |
10400 | */ | |
10401 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 10402 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
10403 | dprintf("imsm: Cannot save stripes to " |
10404 | "target devices\n"); | |
10405 | goto abort; | |
10406 | } | |
10407 | if (save_checkpoint_imsm(st, sra, | |
10408 | UNIT_SRC_IN_CP_AREA)) { | |
10409 | dprintf("imsm: Cannot write checkpoint to " | |
10410 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
10411 | goto abort; | |
10412 | } | |
8016a6d4 AK |
10413 | } else { |
10414 | /* set next step to use whole border area */ | |
10415 | border /= next_step; | |
10416 | if (border > 1) | |
10417 | next_step *= border; | |
10f22854 AK |
10418 | } |
10419 | /* When data backed up, checkpoint stored, | |
10420 | * kick the kernel to reshape unit of data | |
10421 | */ | |
10422 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
10423 | /* limit next step to array max position */ |
10424 | if (next_step > max_position) | |
10425 | next_step = max_position; | |
10f22854 AK |
10426 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
10427 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 10428 | sra->reshape_progress = next_step; |
10f22854 AK |
10429 | |
10430 | /* wait until reshape finish */ | |
ae9f01f8 | 10431 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
10432 | dprintf("wait_for_reshape_imsm returned error!\n"); |
10433 | goto abort; | |
10434 | } | |
10f22854 | 10435 | |
0228d92c AK |
10436 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
10437 | /* ignore error == 2, this can mean end of reshape here | |
10438 | */ | |
10f22854 AK |
10439 | dprintf("imsm: Cannot write checkpoint to " |
10440 | "migration record (UNIT_SRC_NORMAL)\n"); | |
10441 | goto abort; | |
10442 | } | |
10443 | ||
10444 | } | |
10445 | ||
10446 | /* return '1' if done */ | |
10447 | ret_val = 1; | |
10448 | abort: | |
10449 | free(buf); | |
10450 | abort_reshape(sra); | |
10451 | ||
10452 | return ret_val; | |
999b4972 | 10453 | } |
71204a50 | 10454 | #endif /* MDASSEMBLE */ |
999b4972 | 10455 | |
cdddbdbc DW |
10456 | struct superswitch super_imsm = { |
10457 | #ifndef MDASSEMBLE | |
10458 | .examine_super = examine_super_imsm, | |
10459 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 10460 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 10461 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
10462 | .detail_super = detail_super_imsm, |
10463 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 10464 | .write_init_super = write_init_super_imsm, |
0e600426 N |
10465 | .validate_geometry = validate_geometry_imsm, |
10466 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 10467 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 10468 | .detail_platform = detail_platform_imsm, |
e50cf220 | 10469 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 10470 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 10471 | .update_subarray = update_subarray_imsm, |
2b959fbf | 10472 | .load_container = load_container_imsm, |
71204a50 N |
10473 | .default_geometry = default_geometry_imsm, |
10474 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
10475 | .reshape_super = imsm_reshape_super, | |
10476 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 10477 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
10478 | #endif |
10479 | .match_home = match_home_imsm, | |
10480 | .uuid_from_super= uuid_from_super_imsm, | |
10481 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 10482 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
10483 | .update_super = update_super_imsm, |
10484 | ||
10485 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 10486 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
10487 | |
10488 | .compare_super = compare_super_imsm, | |
10489 | ||
10490 | .load_super = load_super_imsm, | |
bf5a934a | 10491 | .init_super = init_super_imsm, |
e683ca88 | 10492 | .store_super = store_super_imsm, |
cdddbdbc DW |
10493 | .free_super = free_super_imsm, |
10494 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 10495 | .container_content = container_content_imsm, |
cdddbdbc | 10496 | |
276d77db | 10497 | |
cdddbdbc | 10498 | .external = 1, |
4cce4069 | 10499 | .name = "imsm", |
845dea95 | 10500 | |
0e600426 | 10501 | #ifndef MDASSEMBLE |
845dea95 NB |
10502 | /* for mdmon */ |
10503 | .open_new = imsm_open_new, | |
ed9d66aa | 10504 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
10505 | .set_disk = imsm_set_disk, |
10506 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 10507 | .activate_spare = imsm_activate_spare, |
e8319a19 | 10508 | .process_update = imsm_process_update, |
8273f55e | 10509 | .prepare_update = imsm_prepare_update, |
0e600426 | 10510 | #endif /* MDASSEMBLE */ |
cdddbdbc | 10511 | }; |