]>
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; | |
3c309c82 MT |
561 | } else |
562 | /* IMSM metadata disallows to attach disks to multiple | |
563 | * controllers. | |
564 | */ | |
88654014 | 565 | return 2; |
88654014 LM |
566 | } |
567 | ||
568 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
569 | { | |
9bc4ae77 | 570 | struct sys_dev *list, *elem; |
88654014 LM |
571 | char *disk_path; |
572 | ||
573 | if ((list = find_intel_devices()) == NULL) | |
574 | return 0; | |
575 | ||
576 | if (fd < 0) | |
577 | disk_path = (char *) devname; | |
578 | else | |
579 | disk_path = diskfd_to_devpath(fd); | |
580 | ||
9bc4ae77 | 581 | if (!disk_path) |
88654014 | 582 | return 0; |
88654014 | 583 | |
9bc4ae77 N |
584 | for (elem = list; elem; elem = elem->next) |
585 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 586 | return elem; |
9bc4ae77 | 587 | |
88654014 LM |
588 | if (disk_path != devname) |
589 | free(disk_path); | |
88654014 LM |
590 | |
591 | return NULL; | |
592 | } | |
593 | ||
594 | ||
d424212e N |
595 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
596 | char *devname); | |
f2f5c343 | 597 | |
cdddbdbc DW |
598 | static struct supertype *match_metadata_desc_imsm(char *arg) |
599 | { | |
600 | struct supertype *st; | |
601 | ||
602 | if (strcmp(arg, "imsm") != 0 && | |
603 | strcmp(arg, "default") != 0 | |
604 | ) | |
605 | return NULL; | |
606 | ||
503975b9 | 607 | st = xcalloc(1, sizeof(*st)); |
d1d599ea | 608 | st->container_dev = NoMdDev; |
cdddbdbc DW |
609 | st->ss = &super_imsm; |
610 | st->max_devs = IMSM_MAX_DEVICES; | |
611 | st->minor_version = 0; | |
612 | st->sb = NULL; | |
613 | return st; | |
614 | } | |
615 | ||
0e600426 | 616 | #ifndef MDASSEMBLE |
cdddbdbc DW |
617 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
618 | { | |
619 | return &mpb->sig[MPB_SIG_LEN]; | |
620 | } | |
9e2d750d | 621 | #endif |
cdddbdbc | 622 | |
949c47a0 DW |
623 | /* retrieve a disk directly from the anchor when the anchor is known to be |
624 | * up-to-date, currently only at load time | |
625 | */ | |
626 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 627 | { |
949c47a0 | 628 | if (index >= mpb->num_disks) |
cdddbdbc DW |
629 | return NULL; |
630 | return &mpb->disk[index]; | |
631 | } | |
632 | ||
95d07a2c LM |
633 | /* retrieve the disk description based on a index of the disk |
634 | * in the sub-array | |
635 | */ | |
636 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 637 | { |
b9f594fe DW |
638 | struct dl *d; |
639 | ||
640 | for (d = super->disks; d; d = d->next) | |
641 | if (d->index == index) | |
95d07a2c LM |
642 | return d; |
643 | ||
644 | return NULL; | |
645 | } | |
646 | /* retrieve a disk from the parsed metadata */ | |
647 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
648 | { | |
649 | struct dl *dl; | |
650 | ||
651 | dl = get_imsm_dl_disk(super, index); | |
652 | if (dl) | |
653 | return &dl->disk; | |
654 | ||
b9f594fe | 655 | return NULL; |
949c47a0 DW |
656 | } |
657 | ||
658 | /* generate a checksum directly from the anchor when the anchor is known to be | |
659 | * up-to-date, currently only at load or write_super after coalescing | |
660 | */ | |
661 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
662 | { |
663 | __u32 end = mpb->mpb_size / sizeof(end); | |
664 | __u32 *p = (__u32 *) mpb; | |
665 | __u32 sum = 0; | |
666 | ||
5d500228 N |
667 | while (end--) { |
668 | sum += __le32_to_cpu(*p); | |
97f734fd N |
669 | p++; |
670 | } | |
cdddbdbc | 671 | |
5d500228 | 672 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
673 | } |
674 | ||
a965f303 DW |
675 | static size_t sizeof_imsm_map(struct imsm_map *map) |
676 | { | |
677 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
678 | } | |
679 | ||
680 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 681 | { |
5e7b0330 AK |
682 | /* A device can have 2 maps if it is in the middle of a migration. |
683 | * If second_map is: | |
238c0a71 AK |
684 | * MAP_0 - we return the first map |
685 | * MAP_1 - we return the second map if it exists, else NULL | |
686 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 687 | */ |
a965f303 | 688 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 689 | struct imsm_map *map2 = NULL; |
a965f303 | 690 | |
9535fc47 AK |
691 | if (dev->vol.migr_state) |
692 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 693 | |
9535fc47 | 694 | switch (second_map) { |
3b451610 | 695 | case MAP_0: |
9535fc47 | 696 | break; |
3b451610 | 697 | case MAP_1: |
9535fc47 AK |
698 | map = map2; |
699 | break; | |
238c0a71 | 700 | case MAP_X: |
9535fc47 AK |
701 | if (map2) |
702 | map = map2; | |
703 | break; | |
9535fc47 AK |
704 | default: |
705 | map = NULL; | |
706 | } | |
707 | return map; | |
5e7b0330 | 708 | |
a965f303 | 709 | } |
cdddbdbc | 710 | |
3393c6af DW |
711 | /* return the size of the device. |
712 | * migr_state increases the returned size if map[0] were to be duplicated | |
713 | */ | |
714 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
715 | { |
716 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 717 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
718 | |
719 | /* migrating means an additional map */ | |
a965f303 | 720 | if (dev->vol.migr_state) |
238c0a71 | 721 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 722 | else if (migr_state) |
238c0a71 | 723 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
724 | |
725 | return size; | |
726 | } | |
727 | ||
54c2c1ea DW |
728 | #ifndef MDASSEMBLE |
729 | /* retrieve disk serial number list from a metadata update */ | |
730 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
731 | { | |
732 | void *u = update; | |
733 | struct disk_info *inf; | |
734 | ||
735 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
736 | sizeof_imsm_dev(&update->dev, 0); | |
737 | ||
738 | return inf; | |
739 | } | |
740 | #endif | |
741 | ||
949c47a0 | 742 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
743 | { |
744 | int offset; | |
745 | int i; | |
746 | void *_mpb = mpb; | |
747 | ||
949c47a0 | 748 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
749 | return NULL; |
750 | ||
751 | /* devices start after all disks */ | |
752 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
753 | ||
754 | for (i = 0; i <= index; i++) | |
755 | if (i == index) | |
756 | return _mpb + offset; | |
757 | else | |
3393c6af | 758 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
759 | |
760 | return NULL; | |
761 | } | |
762 | ||
949c47a0 DW |
763 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
764 | { | |
ba2de7ba DW |
765 | struct intel_dev *dv; |
766 | ||
949c47a0 DW |
767 | if (index >= super->anchor->num_raid_devs) |
768 | return NULL; | |
ba2de7ba DW |
769 | for (dv = super->devlist; dv; dv = dv->next) |
770 | if (dv->index == index) | |
771 | return dv->dev; | |
772 | return NULL; | |
949c47a0 DW |
773 | } |
774 | ||
98130f40 AK |
775 | /* |
776 | * for second_map: | |
238c0a71 AK |
777 | * == MAP_0 get first map |
778 | * == MAP_1 get second map | |
779 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
780 | */ |
781 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
782 | int slot, | |
783 | int second_map) | |
7eef0453 DW |
784 | { |
785 | struct imsm_map *map; | |
786 | ||
5e7b0330 | 787 | map = get_imsm_map(dev, second_map); |
7eef0453 | 788 | |
ff077194 DW |
789 | /* top byte identifies disk under rebuild */ |
790 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
791 | } | |
792 | ||
793 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 794 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 795 | { |
98130f40 | 796 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
797 | |
798 | return ord_to_idx(ord); | |
7eef0453 DW |
799 | } |
800 | ||
be73972f DW |
801 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
802 | { | |
803 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
804 | } | |
805 | ||
f21e18ca | 806 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
807 | { |
808 | int slot; | |
809 | __u32 ord; | |
810 | ||
811 | for (slot = 0; slot < map->num_members; slot++) { | |
812 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
813 | if (ord_to_idx(ord) == idx) | |
814 | return slot; | |
815 | } | |
816 | ||
817 | return -1; | |
818 | } | |
819 | ||
cdddbdbc DW |
820 | static int get_imsm_raid_level(struct imsm_map *map) |
821 | { | |
822 | if (map->raid_level == 1) { | |
823 | if (map->num_members == 2) | |
824 | return 1; | |
825 | else | |
826 | return 10; | |
827 | } | |
828 | ||
829 | return map->raid_level; | |
830 | } | |
831 | ||
c2c087e6 DW |
832 | static int cmp_extent(const void *av, const void *bv) |
833 | { | |
834 | const struct extent *a = av; | |
835 | const struct extent *b = bv; | |
836 | if (a->start < b->start) | |
837 | return -1; | |
838 | if (a->start > b->start) | |
839 | return 1; | |
840 | return 0; | |
841 | } | |
842 | ||
0dcecb2e | 843 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 844 | { |
c2c087e6 | 845 | int memberships = 0; |
620b1713 | 846 | int i; |
c2c087e6 | 847 | |
949c47a0 DW |
848 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
849 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 850 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 851 | |
620b1713 DW |
852 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
853 | memberships++; | |
c2c087e6 | 854 | } |
0dcecb2e DW |
855 | |
856 | return memberships; | |
857 | } | |
858 | ||
b81221b7 CA |
859 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
860 | ||
5551b113 CA |
861 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
862 | { | |
863 | if (lo == 0 || hi == 0) | |
864 | return 1; | |
865 | *lo = __le32_to_cpu((unsigned)n); | |
866 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
867 | return 0; | |
868 | } | |
869 | ||
870 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
871 | { | |
872 | return (unsigned long long)__le32_to_cpu(lo) | | |
873 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
874 | } | |
875 | ||
876 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
877 | { | |
878 | if (disk == NULL) | |
879 | return 0; | |
880 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
881 | } | |
882 | ||
883 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
884 | { | |
885 | if (map == NULL) | |
886 | return 0; | |
887 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
888 | } | |
889 | ||
890 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
891 | { | |
892 | if (map == NULL) | |
893 | return 0; | |
894 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
895 | } | |
896 | ||
ae2416e7 | 897 | #ifndef MDASSEMBLE |
5551b113 CA |
898 | static unsigned long long num_data_stripes(struct imsm_map *map) |
899 | { | |
900 | if (map == NULL) | |
901 | return 0; | |
902 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
903 | } | |
904 | ||
905 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
906 | { | |
907 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
908 | } | |
ae2416e7 | 909 | #endif |
5551b113 CA |
910 | |
911 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
912 | { | |
913 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
914 | } | |
915 | ||
916 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
917 | { | |
918 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
919 | } | |
920 | ||
921 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
922 | { | |
923 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
924 | } | |
925 | ||
0dcecb2e DW |
926 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
927 | { | |
928 | /* find a list of used extents on the given physical device */ | |
929 | struct extent *rv, *e; | |
620b1713 | 930 | int i; |
0dcecb2e | 931 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
932 | __u32 reservation; |
933 | ||
934 | /* trim the reserved area for spares, so they can join any array | |
935 | * regardless of whether the OROM has assigned sectors from the | |
936 | * IMSM_RESERVED_SECTORS region | |
937 | */ | |
938 | if (dl->index == -1) | |
b81221b7 | 939 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
940 | else |
941 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 942 | |
503975b9 | 943 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
944 | e = rv; |
945 | ||
949c47a0 DW |
946 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
947 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 948 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 949 | |
620b1713 | 950 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
951 | e->start = pba_of_lba0(map); |
952 | e->size = blocks_per_member(map); | |
620b1713 | 953 | e++; |
c2c087e6 DW |
954 | } |
955 | } | |
956 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
957 | ||
14e8215b DW |
958 | /* determine the start of the metadata |
959 | * when no raid devices are defined use the default | |
960 | * ...otherwise allow the metadata to truncate the value | |
961 | * as is the case with older versions of imsm | |
962 | */ | |
963 | if (memberships) { | |
964 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 965 | unsigned long long remainder; |
14e8215b | 966 | |
5551b113 | 967 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
968 | /* round down to 1k block to satisfy precision of the kernel |
969 | * 'size' interface | |
970 | */ | |
971 | remainder &= ~1UL; | |
972 | /* make sure remainder is still sane */ | |
f21e18ca | 973 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 974 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
975 | if (reservation > remainder) |
976 | reservation = remainder; | |
977 | } | |
5551b113 | 978 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
979 | e->size = 0; |
980 | return rv; | |
981 | } | |
982 | ||
14e8215b DW |
983 | /* try to determine how much space is reserved for metadata from |
984 | * the last get_extents() entry, otherwise fallback to the | |
985 | * default | |
986 | */ | |
987 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
988 | { | |
989 | struct extent *e; | |
990 | int i; | |
991 | __u32 rv; | |
992 | ||
993 | /* for spares just return a minimal reservation which will grow | |
994 | * once the spare is picked up by an array | |
995 | */ | |
996 | if (dl->index == -1) | |
997 | return MPB_SECTOR_CNT; | |
998 | ||
999 | e = get_extents(super, dl); | |
1000 | if (!e) | |
1001 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1002 | ||
1003 | /* scroll to last entry */ | |
1004 | for (i = 0; e[i].size; i++) | |
1005 | continue; | |
1006 | ||
5551b113 | 1007 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1008 | |
1009 | free(e); | |
1010 | ||
1011 | return rv; | |
1012 | } | |
1013 | ||
25ed7e59 DW |
1014 | static int is_spare(struct imsm_disk *disk) |
1015 | { | |
1016 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1017 | } | |
1018 | ||
1019 | static int is_configured(struct imsm_disk *disk) | |
1020 | { | |
1021 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1022 | } | |
1023 | ||
1024 | static int is_failed(struct imsm_disk *disk) | |
1025 | { | |
1026 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1027 | } | |
1028 | ||
b81221b7 CA |
1029 | /* try to determine how much space is reserved for metadata from |
1030 | * the last get_extents() entry on the smallest active disk, | |
1031 | * otherwise fallback to the default | |
1032 | */ | |
1033 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1034 | { | |
1035 | struct extent *e; | |
1036 | int i; | |
5551b113 CA |
1037 | unsigned long long min_active; |
1038 | __u32 remainder; | |
b81221b7 CA |
1039 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1040 | struct dl *dl, *dl_min = NULL; | |
1041 | ||
1042 | if (!super) | |
1043 | return rv; | |
1044 | ||
1045 | min_active = 0; | |
1046 | for (dl = super->disks; dl; dl = dl->next) { | |
1047 | if (dl->index < 0) | |
1048 | continue; | |
5551b113 CA |
1049 | unsigned long long blocks = total_blocks(&dl->disk); |
1050 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1051 | dl_min = dl; |
5551b113 | 1052 | min_active = blocks; |
b81221b7 CA |
1053 | } |
1054 | } | |
1055 | if (!dl_min) | |
1056 | return rv; | |
1057 | ||
1058 | /* find last lba used by subarrays on the smallest active disk */ | |
1059 | e = get_extents(super, dl_min); | |
1060 | if (!e) | |
1061 | return rv; | |
1062 | for (i = 0; e[i].size; i++) | |
1063 | continue; | |
1064 | ||
1065 | remainder = min_active - e[i].start; | |
1066 | free(e); | |
1067 | ||
1068 | /* to give priority to recovery we should not require full | |
1069 | IMSM_RESERVED_SECTORS from the spare */ | |
1070 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1071 | ||
1072 | /* if real reservation is smaller use that value */ | |
1073 | return (remainder < rv) ? remainder : rv; | |
1074 | } | |
1075 | ||
80e7f8c3 AC |
1076 | /* Return minimum size of a spare that can be used in this array*/ |
1077 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1078 | { | |
1079 | struct intel_super *super = st->sb; | |
1080 | struct dl *dl; | |
1081 | struct extent *e; | |
1082 | int i; | |
1083 | unsigned long long rv = 0; | |
1084 | ||
1085 | if (!super) | |
1086 | return rv; | |
1087 | /* find first active disk in array */ | |
1088 | dl = super->disks; | |
1089 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1090 | dl = dl->next; | |
1091 | if (!dl) | |
1092 | return rv; | |
1093 | /* find last lba used by subarrays */ | |
1094 | e = get_extents(super, dl); | |
1095 | if (!e) | |
1096 | return rv; | |
1097 | for (i = 0; e[i].size; i++) | |
1098 | continue; | |
1099 | if (i > 0) | |
1100 | rv = e[i-1].start + e[i-1].size; | |
1101 | free(e); | |
b81221b7 | 1102 | |
80e7f8c3 | 1103 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1104 | rv = rv + imsm_min_reserved_sectors(super); |
1105 | ||
80e7f8c3 AC |
1106 | return rv * 512; |
1107 | } | |
1108 | ||
d1e02575 AK |
1109 | static int is_gen_migration(struct imsm_dev *dev); |
1110 | ||
1799c9e8 | 1111 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1112 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1113 | struct imsm_dev *dev); | |
1e5c6983 | 1114 | |
c47b0ff6 AK |
1115 | static void print_imsm_dev(struct intel_super *super, |
1116 | struct imsm_dev *dev, | |
1117 | char *uuid, | |
1118 | int disk_idx) | |
cdddbdbc DW |
1119 | { |
1120 | __u64 sz; | |
0d80bb2f | 1121 | int slot, i; |
238c0a71 AK |
1122 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1123 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1124 | __u32 ord; |
cdddbdbc DW |
1125 | |
1126 | printf("\n"); | |
1e7bc0ed | 1127 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1128 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1129 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1130 | if (map2) | |
1131 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1132 | printf("\n"); | |
1133 | printf(" Members : %d", map->num_members); | |
1134 | if (map2) | |
1135 | printf(" <-- %d", map2->num_members); | |
1136 | printf("\n"); | |
0d80bb2f DW |
1137 | printf(" Slots : ["); |
1138 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1139 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1140 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1141 | } | |
dd8bcb3b AK |
1142 | printf("]"); |
1143 | if (map2) { | |
1144 | printf(" <-- ["); | |
1145 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1146 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1147 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1148 | } | |
1149 | printf("]"); | |
1150 | } | |
1151 | printf("\n"); | |
7095bccb AK |
1152 | printf(" Failed disk : "); |
1153 | if (map->failed_disk_num == 0xff) | |
1154 | printf("none"); | |
1155 | else | |
1156 | printf("%i", map->failed_disk_num); | |
1157 | printf("\n"); | |
620b1713 DW |
1158 | slot = get_imsm_disk_slot(map, disk_idx); |
1159 | if (slot >= 0) { | |
238c0a71 | 1160 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1161 | printf(" This Slot : %d%s\n", slot, |
1162 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1163 | } else | |
cdddbdbc DW |
1164 | printf(" This Slot : ?\n"); |
1165 | sz = __le32_to_cpu(dev->size_high); | |
1166 | sz <<= 32; | |
1167 | sz += __le32_to_cpu(dev->size_low); | |
1168 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1169 | human_size(sz * 512)); | |
5551b113 | 1170 | sz = blocks_per_member(map); |
cdddbdbc DW |
1171 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, |
1172 | human_size(sz * 512)); | |
5551b113 CA |
1173 | printf(" Sector Offset : %llu\n", |
1174 | pba_of_lba0(map)); | |
1175 | printf(" Num Stripes : %llu\n", | |
1176 | num_data_stripes(map)); | |
dd8bcb3b | 1177 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1178 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1179 | if (map2) |
1180 | printf(" <-- %u KiB", | |
1181 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1182 | printf("\n"); | |
cdddbdbc | 1183 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1184 | printf(" Migrate State : "); |
1484e727 DW |
1185 | if (dev->vol.migr_state) { |
1186 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1187 | printf("initialize\n"); |
1484e727 | 1188 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1189 | printf("rebuild\n"); |
1484e727 | 1190 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1191 | printf("check\n"); |
1484e727 | 1192 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1193 | printf("general migration\n"); |
1484e727 | 1194 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1195 | printf("state change\n"); |
1484e727 | 1196 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1197 | printf("repair\n"); |
1484e727 | 1198 | else |
8655a7b1 DW |
1199 | printf("<unknown:%d>\n", migr_type(dev)); |
1200 | } else | |
1201 | printf("idle\n"); | |
3393c6af DW |
1202 | printf(" Map State : %s", map_state_str[map->map_state]); |
1203 | if (dev->vol.migr_state) { | |
238c0a71 | 1204 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1205 | |
b10b37b8 | 1206 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1207 | printf("\n Checkpoint : %u ", |
1208 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
3136abe5 | 1209 | if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0))) |
464d40e8 LD |
1210 | printf("(N/A)"); |
1211 | else | |
1212 | printf("(%llu)", (unsigned long long) | |
1213 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1214 | } |
1215 | printf("\n"); | |
cdddbdbc | 1216 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1217 | } |
1218 | ||
0ec1f4e8 | 1219 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1220 | { |
1f24f035 | 1221 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1222 | __u64 sz; |
1223 | ||
0ec1f4e8 | 1224 | if (index < -1 || !disk) |
e9d82038 DW |
1225 | return; |
1226 | ||
cdddbdbc | 1227 | printf("\n"); |
1f24f035 | 1228 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1229 | if (index >= 0) |
1230 | printf(" Disk%02d Serial : %s\n", index, str); | |
1231 | else | |
1232 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1233 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1234 | is_configured(disk) ? " active" : "", | |
1235 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1236 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1237 | sz = total_blocks(disk) - reserved; |
cdddbdbc DW |
1238 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1239 | human_size(sz * 512)); | |
1240 | } | |
1241 | ||
520e69e2 AK |
1242 | void examine_migr_rec_imsm(struct intel_super *super) |
1243 | { | |
1244 | struct migr_record *migr_rec = super->migr_rec; | |
1245 | struct imsm_super *mpb = super->anchor; | |
1246 | int i; | |
1247 | ||
1248 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1249 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1250 | struct imsm_map *map; |
b4ab44d8 | 1251 | int slot = -1; |
3136abe5 | 1252 | |
520e69e2 AK |
1253 | if (is_gen_migration(dev) == 0) |
1254 | continue; | |
1255 | ||
1256 | printf("\nMigration Record Information:"); | |
3136abe5 | 1257 | |
44bfe6df AK |
1258 | /* first map under migration */ |
1259 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1260 | if (map) |
1261 | slot = get_imsm_disk_slot(map, super->disks->index); | |
1262 | if ((map == NULL) || (slot > 1) || (slot < 0)) { | |
520e69e2 AK |
1263 | printf(" Empty\n "); |
1264 | printf("Examine one of first two disks in array\n"); | |
1265 | break; | |
1266 | } | |
1267 | printf("\n Status : "); | |
1268 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1269 | printf("Normal\n"); | |
1270 | else | |
1271 | printf("Contains Data\n"); | |
1272 | printf(" Current Unit : %u\n", | |
1273 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1274 | printf(" Family : %u\n", | |
1275 | __le32_to_cpu(migr_rec->family_num)); | |
1276 | printf(" Ascending : %u\n", | |
1277 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1278 | printf(" Blocks Per Unit : %u\n", | |
1279 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1280 | printf(" Dest. Depth Per Unit : %u\n", | |
1281 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1282 | printf(" Checkpoint Area pba : %u\n", | |
1283 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1284 | printf(" First member lba : %u\n", | |
1285 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1286 | printf(" Total Number of Units : %u\n", | |
1287 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1288 | printf(" Size of volume : %u\n", | |
1289 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1290 | printf(" Expansion space for LBA64 : %u\n", | |
1291 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1292 | printf(" Record was read from : %u\n", | |
1293 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1294 | ||
1295 | break; | |
1296 | } | |
1297 | } | |
9e2d750d | 1298 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1299 | /******************************************************************************* |
1300 | * function: imsm_check_attributes | |
1301 | * Description: Function checks if features represented by attributes flags | |
1302 | * are supported by mdadm. | |
1303 | * Parameters: | |
1304 | * attributes - Attributes read from metadata | |
1305 | * Returns: | |
1306 | * 0 - passed attributes contains unsupported features flags | |
1307 | * 1 - all features are supported | |
1308 | ******************************************************************************/ | |
1309 | static int imsm_check_attributes(__u32 attributes) | |
1310 | { | |
1311 | int ret_val = 1; | |
418f9b36 N |
1312 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1313 | ||
1314 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1315 | |
1316 | not_supported &= attributes; | |
1317 | if (not_supported) { | |
e7b84f9d | 1318 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1319 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1320 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1321 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1322 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1323 | } | |
1324 | if (not_supported & MPB_ATTRIB_2TB) { | |
1325 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1326 | not_supported ^= MPB_ATTRIB_2TB; | |
1327 | } | |
1328 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1329 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1330 | not_supported ^= MPB_ATTRIB_RAID0; | |
1331 | } | |
1332 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1333 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1334 | not_supported ^= MPB_ATTRIB_RAID1; | |
1335 | } | |
1336 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1337 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1338 | not_supported ^= MPB_ATTRIB_RAID10; | |
1339 | } | |
1340 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1341 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1342 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1343 | } | |
1344 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1345 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1346 | not_supported ^= MPB_ATTRIB_RAID5; | |
1347 | } | |
1348 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1349 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1350 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1351 | } | |
1352 | if (not_supported & MPB_ATTRIB_BBM) { | |
1353 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1354 | not_supported ^= MPB_ATTRIB_BBM; | |
1355 | } | |
1356 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1357 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1358 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1359 | } | |
1360 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1361 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1362 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1363 | } | |
1364 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1365 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1366 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1367 | } | |
1368 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1369 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1370 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1371 | } | |
1372 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1373 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1374 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1375 | } | |
1376 | ||
1377 | if (not_supported) | |
1378 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1379 | ||
1380 | ret_val = 0; | |
1381 | } | |
1382 | ||
1383 | return ret_val; | |
1384 | } | |
1385 | ||
9e2d750d | 1386 | #ifndef MDASSEMBLE |
a5d85af7 | 1387 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1388 | |
cdddbdbc DW |
1389 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1390 | { | |
1391 | struct intel_super *super = st->sb; | |
949c47a0 | 1392 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1393 | char str[MAX_SIGNATURE_LENGTH]; |
1394 | int i; | |
27fd6274 DW |
1395 | struct mdinfo info; |
1396 | char nbuf[64]; | |
cdddbdbc | 1397 | __u32 sum; |
14e8215b | 1398 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1399 | struct dl *dl; |
27fd6274 | 1400 | |
cdddbdbc DW |
1401 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1402 | printf(" Magic : %s\n", str); | |
1403 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1404 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1405 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1406 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1407 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1408 | printf(" Attributes : "); |
1409 | if (imsm_check_attributes(mpb->attributes)) | |
1410 | printf("All supported\n"); | |
1411 | else | |
1412 | printf("not supported\n"); | |
a5d85af7 | 1413 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1414 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1415 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1416 | sum = __le32_to_cpu(mpb->check_sum); |
1417 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1418 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1419 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1420 | printf(" Disks : %d\n", mpb->num_disks); |
1421 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1422 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1423 | if (super->bbm_log) { |
1424 | struct bbm_log *log = super->bbm_log; | |
1425 | ||
1426 | printf("\n"); | |
1427 | printf("Bad Block Management Log:\n"); | |
1428 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1429 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1430 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1431 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1432 | printf(" First Spare : %llx\n", |
1433 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1434 | } |
44470971 DW |
1435 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1436 | struct mdinfo info; | |
1437 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1438 | ||
1439 | super->current_vol = i; | |
a5d85af7 | 1440 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1441 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1442 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1443 | } |
cdddbdbc DW |
1444 | for (i = 0; i < mpb->num_disks; i++) { |
1445 | if (i == super->disks->index) | |
1446 | continue; | |
0ec1f4e8 | 1447 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1448 | } |
94827db3 | 1449 | |
0ec1f4e8 DW |
1450 | for (dl = super->disks; dl; dl = dl->next) |
1451 | if (dl->index == -1) | |
1452 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1453 | |
1454 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1455 | } |
1456 | ||
061f2c6a | 1457 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1458 | { |
27fd6274 | 1459 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1460 | struct mdinfo info; |
1461 | char nbuf[64]; | |
1e7bc0ed | 1462 | struct intel_super *super = st->sb; |
1e7bc0ed | 1463 | |
0d5a423f DW |
1464 | if (!super->anchor->num_raid_devs) { |
1465 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1466 | return; |
0d5a423f | 1467 | } |
ff54de6e | 1468 | |
a5d85af7 | 1469 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1470 | fname_from_uuid(st, &info, nbuf, ':'); |
1471 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1472 | } | |
1473 | ||
1474 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1475 | { | |
1476 | /* We just write a generic IMSM ARRAY entry */ | |
1477 | struct mdinfo info; | |
1478 | char nbuf[64]; | |
1479 | char nbuf1[64]; | |
1480 | struct intel_super *super = st->sb; | |
1481 | int i; | |
1482 | ||
1483 | if (!super->anchor->num_raid_devs) | |
1484 | return; | |
1485 | ||
a5d85af7 | 1486 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1487 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1488 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1489 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1490 | ||
1491 | super->current_vol = i; | |
a5d85af7 | 1492 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1493 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1494 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1495 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1496 | } |
cdddbdbc DW |
1497 | } |
1498 | ||
9d84c8ea DW |
1499 | static void export_examine_super_imsm(struct supertype *st) |
1500 | { | |
1501 | struct intel_super *super = st->sb; | |
1502 | struct imsm_super *mpb = super->anchor; | |
1503 | struct mdinfo info; | |
1504 | char nbuf[64]; | |
1505 | ||
a5d85af7 | 1506 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1507 | fname_from_uuid(st, &info, nbuf, ':'); |
1508 | printf("MD_METADATA=imsm\n"); | |
1509 | printf("MD_LEVEL=container\n"); | |
1510 | printf("MD_UUID=%s\n", nbuf+5); | |
1511 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1512 | } | |
1513 | ||
cdddbdbc DW |
1514 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1515 | { | |
3ebe00a1 DW |
1516 | struct mdinfo info; |
1517 | char nbuf[64]; | |
1518 | ||
a5d85af7 | 1519 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1520 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1521 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1522 | } |
1523 | ||
1524 | static void brief_detail_super_imsm(struct supertype *st) | |
1525 | { | |
ff54de6e N |
1526 | struct mdinfo info; |
1527 | char nbuf[64]; | |
a5d85af7 | 1528 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1529 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1530 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1531 | } |
d665cc31 DW |
1532 | |
1533 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1534 | static void fd2devname(int fd, char *name); | |
1535 | ||
120dc887 | 1536 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1537 | { |
120dc887 LM |
1538 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1539 | * controller, as well as non-connected ports | |
d665cc31 DW |
1540 | */ |
1541 | int hba_len = strlen(hba_path) + 1; | |
1542 | struct dirent *ent; | |
1543 | DIR *dir; | |
1544 | char *path = NULL; | |
1545 | int err = 0; | |
1546 | unsigned long port_mask = (1 << port_count) - 1; | |
1547 | ||
f21e18ca | 1548 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 1549 | if (verbose > 0) |
e7b84f9d | 1550 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
1551 | return 2; |
1552 | } | |
1553 | ||
1554 | /* scroll through /sys/dev/block looking for devices attached to | |
1555 | * this hba | |
1556 | */ | |
1557 | dir = opendir("/sys/dev/block"); | |
1558 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1559 | int fd; | |
1560 | char model[64]; | |
1561 | char vendor[64]; | |
1562 | char buf[1024]; | |
1563 | int major, minor; | |
1564 | char *device; | |
1565 | char *c; | |
1566 | int port; | |
1567 | int type; | |
1568 | ||
1569 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1570 | continue; | |
1571 | path = devt_to_devpath(makedev(major, minor)); | |
1572 | if (!path) | |
1573 | continue; | |
1574 | if (!path_attached_to_hba(path, hba_path)) { | |
1575 | free(path); | |
1576 | path = NULL; | |
1577 | continue; | |
1578 | } | |
1579 | ||
1580 | /* retrieve the scsi device type */ | |
1581 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 1582 | if (verbose > 0) |
e7b84f9d | 1583 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
1584 | err = 2; |
1585 | break; | |
1586 | } | |
1587 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1588 | if (load_sys(device, buf) != 0) { | |
ba728be7 | 1589 | if (verbose > 0) |
e7b84f9d | 1590 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
1591 | path); |
1592 | err = 2; | |
1593 | free(device); | |
1594 | break; | |
1595 | } | |
1596 | type = strtoul(buf, NULL, 10); | |
1597 | ||
1598 | /* if it's not a disk print the vendor and model */ | |
1599 | if (!(type == 0 || type == 7 || type == 14)) { | |
1600 | vendor[0] = '\0'; | |
1601 | model[0] = '\0'; | |
1602 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1603 | if (load_sys(device, buf) == 0) { | |
1604 | strncpy(vendor, buf, sizeof(vendor)); | |
1605 | vendor[sizeof(vendor) - 1] = '\0'; | |
1606 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1607 | while (isspace(*c) || *c == '\0') | |
1608 | *c-- = '\0'; | |
1609 | ||
1610 | } | |
1611 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1612 | if (load_sys(device, buf) == 0) { | |
1613 | strncpy(model, buf, sizeof(model)); | |
1614 | model[sizeof(model) - 1] = '\0'; | |
1615 | c = (char *) &model[sizeof(model) - 1]; | |
1616 | while (isspace(*c) || *c == '\0') | |
1617 | *c-- = '\0'; | |
1618 | } | |
1619 | ||
1620 | if (vendor[0] && model[0]) | |
1621 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1622 | else | |
1623 | switch (type) { /* numbers from hald/linux/device.c */ | |
1624 | case 1: sprintf(buf, "tape"); break; | |
1625 | case 2: sprintf(buf, "printer"); break; | |
1626 | case 3: sprintf(buf, "processor"); break; | |
1627 | case 4: | |
1628 | case 5: sprintf(buf, "cdrom"); break; | |
1629 | case 6: sprintf(buf, "scanner"); break; | |
1630 | case 8: sprintf(buf, "media_changer"); break; | |
1631 | case 9: sprintf(buf, "comm"); break; | |
1632 | case 12: sprintf(buf, "raid"); break; | |
1633 | default: sprintf(buf, "unknown"); | |
1634 | } | |
1635 | } else | |
1636 | buf[0] = '\0'; | |
1637 | free(device); | |
1638 | ||
1639 | /* chop device path to 'host%d' and calculate the port number */ | |
1640 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 1641 | if (!c) { |
ba728be7 | 1642 | if (verbose > 0) |
e7b84f9d | 1643 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
1644 | err = 2; |
1645 | break; | |
1646 | } | |
d665cc31 DW |
1647 | *c = '\0'; |
1648 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1649 | port -= host_base; | |
1650 | else { | |
ba728be7 | 1651 | if (verbose > 0) { |
d665cc31 | 1652 | *c = '/'; /* repair the full string */ |
e7b84f9d | 1653 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
1654 | path); |
1655 | } | |
1656 | err = 2; | |
1657 | break; | |
1658 | } | |
1659 | ||
1660 | /* mark this port as used */ | |
1661 | port_mask &= ~(1 << port); | |
1662 | ||
1663 | /* print out the device information */ | |
1664 | if (buf[0]) { | |
1665 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1666 | continue; | |
1667 | } | |
1668 | ||
1669 | fd = dev_open(ent->d_name, O_RDONLY); | |
1670 | if (fd < 0) | |
1671 | printf(" Port%d : - disk info unavailable -\n", port); | |
1672 | else { | |
1673 | fd2devname(fd, buf); | |
1674 | printf(" Port%d : %s", port, buf); | |
1675 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1676 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1677 | else |
664d5325 | 1678 | printf(" ()\n"); |
4dab422a | 1679 | close(fd); |
d665cc31 | 1680 | } |
d665cc31 DW |
1681 | free(path); |
1682 | path = NULL; | |
1683 | } | |
1684 | if (path) | |
1685 | free(path); | |
1686 | if (dir) | |
1687 | closedir(dir); | |
1688 | if (err == 0) { | |
1689 | int i; | |
1690 | ||
1691 | for (i = 0; i < port_count; i++) | |
1692 | if (port_mask & (1 << i)) | |
1693 | printf(" Port%d : - no device attached -\n", i); | |
1694 | } | |
1695 | ||
1696 | return err; | |
1697 | } | |
1698 | ||
120dc887 LM |
1699 | static void print_found_intel_controllers(struct sys_dev *elem) |
1700 | { | |
1701 | for (; elem; elem = elem->next) { | |
e7b84f9d | 1702 | pr_err("found Intel(R) "); |
120dc887 LM |
1703 | if (elem->type == SYS_DEV_SATA) |
1704 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1705 | else if (elem->type == SYS_DEV_SAS) |
1706 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1707 | fprintf(stderr, "RAID controller"); |
1708 | if (elem->pci_id) | |
1709 | fprintf(stderr, " at %s", elem->pci_id); | |
1710 | fprintf(stderr, ".\n"); | |
1711 | } | |
1712 | fflush(stderr); | |
1713 | } | |
1714 | ||
120dc887 LM |
1715 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1716 | { | |
1717 | struct dirent *ent; | |
1718 | DIR *dir; | |
1719 | int host_base = -1; | |
1720 | ||
1721 | *port_count = 0; | |
1722 | if ((dir = opendir(hba_path)) == NULL) | |
1723 | return -1; | |
1724 | ||
1725 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1726 | int host; | |
1727 | ||
1728 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1729 | continue; | |
1730 | if (*port_count == 0) | |
1731 | host_base = host; | |
1732 | else if (host < host_base) | |
1733 | host_base = host; | |
1734 | ||
1735 | if (host + 1 > *port_count + host_base) | |
1736 | *port_count = host + 1 - host_base; | |
1737 | } | |
1738 | closedir(dir); | |
1739 | return host_base; | |
1740 | } | |
1741 | ||
a891a3c2 LM |
1742 | static void print_imsm_capability(const struct imsm_orom *orom) |
1743 | { | |
1744 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1745 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1746 | orom->hotfix_ver, orom->build); | |
1747 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1748 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1749 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1750 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1751 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1752 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1753 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1754 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1755 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1756 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1757 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1758 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1759 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1760 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1761 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1762 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1763 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1764 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1765 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1766 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1767 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1768 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1769 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
1770 | printf(" 2TB volumes :%s supported\n", |
1771 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
1772 | printf(" 2TB disks :%s supported\n", | |
1773 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 LM |
1774 | printf(" Max Disks : %d\n", orom->tds); |
1775 | printf(" Max Volumes : %d per array, %d per controller\n", | |
1776 | orom->vpa, orom->vphba); | |
a891a3c2 LM |
1777 | return; |
1778 | } | |
1779 | ||
e50cf220 MN |
1780 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
1781 | { | |
1782 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
1783 | printf("IMSM_VERSION=%d.%d.%d.%d\n",orom->major_ver, orom->minor_ver, | |
1784 | orom->hotfix_ver, orom->build); | |
1785 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", | |
1786 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
1787 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
1788 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
1789 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
1790 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
1791 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1792 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
1793 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
1794 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
1795 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
1796 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
1797 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
1798 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
1799 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
1800 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
1801 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
1802 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
1803 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
1804 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
1805 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
1806 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
1807 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
1808 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
1809 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
1810 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
1811 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
1812 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
1813 | } | |
1814 | ||
9eafa1de | 1815 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
1816 | { |
1817 | /* There are two components to imsm platform support, the ahci SATA | |
1818 | * controller and the option-rom. To find the SATA controller we | |
1819 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1820 | * controller with the Intel vendor id is present. This approach | |
1821 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1822 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1823 | * detect platform raid capabilities. The option-rom resides in a | |
1824 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1825 | * platform capabilities. If raid support is disabled in the BIOS the | |
1826 | * option-rom capability structure will not be available. | |
1827 | */ | |
1828 | const struct imsm_orom *orom; | |
1829 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1830 | int host_base = 0; |
1831 | int port_count = 0; | |
9eafa1de | 1832 | int result=1; |
d665cc31 | 1833 | |
5615172f | 1834 | if (enumerate_only) { |
a891a3c2 | 1835 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1836 | return 0; |
a891a3c2 LM |
1837 | list = find_intel_devices(); |
1838 | if (!list) | |
1839 | return 2; | |
1840 | for (hba = list; hba; hba = hba->next) { | |
1841 | orom = find_imsm_capability(hba->type); | |
1842 | if (!orom) { | |
1843 | result = 2; | |
1844 | break; | |
1845 | } | |
9eafa1de MN |
1846 | else |
1847 | result = 0; | |
a891a3c2 | 1848 | } |
a891a3c2 | 1849 | return result; |
5615172f DW |
1850 | } |
1851 | ||
155cbb4c LM |
1852 | list = find_intel_devices(); |
1853 | if (!list) { | |
ba728be7 | 1854 | if (verbose > 0) |
e7b84f9d | 1855 | pr_err("no active Intel(R) RAID " |
155cbb4c | 1856 | "controller found.\n"); |
d665cc31 | 1857 | return 2; |
ba728be7 | 1858 | } else if (verbose > 0) |
155cbb4c | 1859 | print_found_intel_controllers(list); |
d665cc31 | 1860 | |
a891a3c2 | 1861 | for (hba = list; hba; hba = hba->next) { |
9eafa1de MN |
1862 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
1863 | continue; | |
a891a3c2 LM |
1864 | orom = find_imsm_capability(hba->type); |
1865 | if (!orom) | |
e7b84f9d | 1866 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
a891a3c2 | 1867 | hba->path, get_sys_dev_type(hba->type)); |
9eafa1de MN |
1868 | else { |
1869 | result = 0; | |
a891a3c2 | 1870 | print_imsm_capability(orom); |
9eafa1de MN |
1871 | printf(" I/O Controller : %s (%s)\n", |
1872 | hba->path, get_sys_dev_type(hba->type)); | |
1873 | if (hba->type == SYS_DEV_SATA) { | |
1874 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1875 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1876 | if (verbose > 0) | |
1877 | pr_err("failed to enumerate " | |
1878 | "ports on SATA controller at %s.\n", hba->pci_id); | |
1879 | result |= 2; | |
1880 | } | |
120dc887 LM |
1881 | } |
1882 | } | |
d665cc31 | 1883 | } |
155cbb4c | 1884 | |
9eafa1de MN |
1885 | if (controller_path && result == 1) |
1886 | pr_err("no active Intel(R) RAID " | |
1887 | "controller found under %s\n",controller_path); | |
1888 | ||
120dc887 | 1889 | return result; |
d665cc31 | 1890 | } |
e50cf220 | 1891 | |
9eafa1de | 1892 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 MN |
1893 | { |
1894 | const struct imsm_orom *orom; | |
1895 | struct sys_dev *list, *hba; | |
1896 | int result=1; | |
1897 | ||
1898 | list = find_intel_devices(); | |
1899 | if (!list) { | |
1900 | if (verbose > 0) | |
1901 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
1902 | result = 2; | |
e50cf220 MN |
1903 | return result; |
1904 | } | |
1905 | ||
1906 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
1907 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
1908 | continue; | |
e50cf220 MN |
1909 | orom = find_imsm_capability(hba->type); |
1910 | if (!orom) { | |
1911 | if (verbose > 0) | |
1912 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",hba->path); | |
1913 | } | |
1914 | else { | |
1915 | print_imsm_capability_export(orom); | |
1916 | result = 0; | |
1917 | } | |
1918 | } | |
1919 | ||
e50cf220 MN |
1920 | return result; |
1921 | } | |
1922 | ||
cdddbdbc DW |
1923 | #endif |
1924 | ||
1925 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1926 | { | |
5115ca67 DW |
1927 | /* the imsm metadata format does not specify any host |
1928 | * identification information. We return -1 since we can never | |
1929 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1930 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1931 | * exclude member disks that do not belong, and we rely on |
1932 | * mdadm.conf to specify the arrays that should be assembled. | |
1933 | * Auto-assembly may still pick up "foreign" arrays. | |
1934 | */ | |
cdddbdbc | 1935 | |
9362c1c8 | 1936 | return -1; |
cdddbdbc DW |
1937 | } |
1938 | ||
1939 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1940 | { | |
51006d85 N |
1941 | /* The uuid returned here is used for: |
1942 | * uuid to put into bitmap file (Create, Grow) | |
1943 | * uuid for backup header when saving critical section (Grow) | |
1944 | * comparing uuids when re-adding a device into an array | |
1945 | * In these cases the uuid required is that of the data-array, | |
1946 | * not the device-set. | |
1947 | * uuid to recognise same set when adding a missing device back | |
1948 | * to an array. This is a uuid for the device-set. | |
1949 | * | |
1950 | * For each of these we can make do with a truncated | |
1951 | * or hashed uuid rather than the original, as long as | |
1952 | * everyone agrees. | |
1953 | * In each case the uuid required is that of the data-array, | |
1954 | * not the device-set. | |
43dad3d6 | 1955 | */ |
51006d85 N |
1956 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1957 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1958 | * - the orig_family_num of the container |
51006d85 N |
1959 | * - the index number of the volume |
1960 | * - the 'serial' number of the volume. | |
1961 | * Hopefully these are all constant. | |
1962 | */ | |
1963 | struct intel_super *super = st->sb; | |
43dad3d6 | 1964 | |
51006d85 N |
1965 | char buf[20]; |
1966 | struct sha1_ctx ctx; | |
1967 | struct imsm_dev *dev = NULL; | |
148acb7b | 1968 | __u32 family_num; |
51006d85 | 1969 | |
148acb7b DW |
1970 | /* some mdadm versions failed to set ->orig_family_num, in which |
1971 | * case fall back to ->family_num. orig_family_num will be | |
1972 | * fixed up with the first metadata update. | |
1973 | */ | |
1974 | family_num = super->anchor->orig_family_num; | |
1975 | if (family_num == 0) | |
1976 | family_num = super->anchor->family_num; | |
51006d85 | 1977 | sha1_init_ctx(&ctx); |
92bd8f8d | 1978 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1979 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1980 | if (super->current_vol >= 0) |
1981 | dev = get_imsm_dev(super, super->current_vol); | |
1982 | if (dev) { | |
1983 | __u32 vol = super->current_vol; | |
1984 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1985 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1986 | } | |
1987 | sha1_finish_ctx(&ctx, buf); | |
1988 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1989 | } |
1990 | ||
0d481d37 | 1991 | #if 0 |
4f5bc454 DW |
1992 | static void |
1993 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1994 | { |
cdddbdbc DW |
1995 | __u8 *v = get_imsm_version(mpb); |
1996 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1997 | char major[] = { 0, 0, 0 }; | |
1998 | char minor[] = { 0 ,0, 0 }; | |
1999 | char patch[] = { 0, 0, 0 }; | |
2000 | char *ver_parse[] = { major, minor, patch }; | |
2001 | int i, j; | |
2002 | ||
2003 | i = j = 0; | |
2004 | while (*v != '\0' && v < end) { | |
2005 | if (*v != '.' && j < 2) | |
2006 | ver_parse[i][j++] = *v; | |
2007 | else { | |
2008 | i++; | |
2009 | j = 0; | |
2010 | } | |
2011 | v++; | |
2012 | } | |
2013 | ||
4f5bc454 DW |
2014 | *m = strtol(minor, NULL, 0); |
2015 | *p = strtol(patch, NULL, 0); | |
2016 | } | |
0d481d37 | 2017 | #endif |
4f5bc454 | 2018 | |
1e5c6983 DW |
2019 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2020 | { | |
2021 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2022 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2023 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2024 | ||
2025 | switch (get_imsm_raid_level(map)) { | |
2026 | case 5: | |
2027 | case 10: | |
2028 | return chunk; | |
2029 | default: | |
2030 | return 128*1024 >> 9; | |
2031 | } | |
2032 | } | |
2033 | ||
2034 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2035 | { | |
2036 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2037 | * this is different than migr_strip_size_resync(), but it's good | |
2038 | * to be compatible | |
2039 | */ | |
238c0a71 | 2040 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2041 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2042 | ||
2043 | switch (get_imsm_raid_level(map)) { | |
2044 | case 1: | |
2045 | case 10: | |
2046 | if (map->num_members % map->num_domains == 0) | |
2047 | return 128*1024 >> 9; | |
2048 | else | |
2049 | return chunk; | |
2050 | case 5: | |
2051 | return max((__u32) 64*1024 >> 9, chunk); | |
2052 | default: | |
2053 | return 128*1024 >> 9; | |
2054 | } | |
2055 | } | |
2056 | ||
2057 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2058 | { | |
238c0a71 AK |
2059 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2060 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2061 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2062 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2063 | ||
2064 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2065 | } | |
2066 | ||
2067 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2068 | { | |
238c0a71 | 2069 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2070 | int level = get_imsm_raid_level(lo); |
2071 | ||
2072 | if (level == 1 || level == 10) { | |
238c0a71 | 2073 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2074 | |
2075 | return hi->num_domains; | |
2076 | } else | |
2077 | return num_stripes_per_unit_resync(dev); | |
2078 | } | |
2079 | ||
98130f40 | 2080 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2081 | { |
2082 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2083 | * counter-intuitively have the same number of data disks | |
2084 | */ | |
98130f40 | 2085 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2086 | |
2087 | switch (get_imsm_raid_level(map)) { | |
2088 | case 0: | |
36fd8ccc AK |
2089 | return map->num_members; |
2090 | break; | |
1e5c6983 DW |
2091 | case 1: |
2092 | case 10: | |
36fd8ccc | 2093 | return map->num_members/2; |
1e5c6983 DW |
2094 | case 5: |
2095 | return map->num_members - 1; | |
2096 | default: | |
2097 | dprintf("%s: unsupported raid level\n", __func__); | |
2098 | return 0; | |
2099 | } | |
2100 | } | |
2101 | ||
2102 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2103 | { | |
238c0a71 | 2104 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2105 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2106 | ||
2107 | switch(get_imsm_raid_level(map)) { | |
2108 | case 1: | |
2109 | case 10: | |
2110 | return chunk * map->num_domains; | |
2111 | case 5: | |
2112 | return chunk * map->num_members; | |
2113 | default: | |
2114 | return chunk; | |
2115 | } | |
2116 | } | |
2117 | ||
2118 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2119 | { | |
238c0a71 | 2120 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2121 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2122 | __u32 strip = block / chunk; | |
2123 | ||
2124 | switch (get_imsm_raid_level(map)) { | |
2125 | case 1: | |
2126 | case 10: { | |
2127 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2128 | __u32 vol_stripe = vol_strip / map->num_members; | |
2129 | ||
2130 | return vol_stripe * chunk + block % chunk; | |
2131 | } case 5: { | |
2132 | __u32 stripe = strip / (map->num_members - 1); | |
2133 | ||
2134 | return stripe * chunk + block % chunk; | |
2135 | } | |
2136 | default: | |
2137 | return 0; | |
2138 | } | |
2139 | } | |
2140 | ||
c47b0ff6 AK |
2141 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2142 | struct imsm_dev *dev) | |
1e5c6983 DW |
2143 | { |
2144 | /* calculate the conversion factor between per member 'blocks' | |
2145 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2146 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2147 | */ | |
2148 | if (!dev->vol.migr_state) | |
2149 | return 0; | |
2150 | ||
2151 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2152 | case MIGR_GEN_MIGR: { |
2153 | struct migr_record *migr_rec = super->migr_rec; | |
2154 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2155 | } | |
1e5c6983 DW |
2156 | case MIGR_VERIFY: |
2157 | case MIGR_REPAIR: | |
2158 | case MIGR_INIT: { | |
238c0a71 | 2159 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2160 | __u32 stripes_per_unit; |
2161 | __u32 blocks_per_unit; | |
2162 | __u32 parity_depth; | |
2163 | __u32 migr_chunk; | |
2164 | __u32 block_map; | |
2165 | __u32 block_rel; | |
2166 | __u32 segment; | |
2167 | __u32 stripe; | |
2168 | __u8 disks; | |
2169 | ||
2170 | /* yes, this is really the translation of migr_units to | |
2171 | * per-member blocks in the 'resync' case | |
2172 | */ | |
2173 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2174 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2175 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2176 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2177 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2178 | segment = blocks_per_unit / stripe; |
2179 | block_rel = blocks_per_unit - segment * stripe; | |
2180 | parity_depth = parity_segment_depth(dev); | |
2181 | block_map = map_migr_block(dev, block_rel); | |
2182 | return block_map + parity_depth * segment; | |
2183 | } | |
2184 | case MIGR_REBUILD: { | |
2185 | __u32 stripes_per_unit; | |
2186 | __u32 migr_chunk; | |
2187 | ||
2188 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2189 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2190 | return migr_chunk * stripes_per_unit; | |
2191 | } | |
1e5c6983 DW |
2192 | case MIGR_STATE_CHANGE: |
2193 | default: | |
2194 | return 0; | |
2195 | } | |
2196 | } | |
2197 | ||
c2c087e6 DW |
2198 | static int imsm_level_to_layout(int level) |
2199 | { | |
2200 | switch (level) { | |
2201 | case 0: | |
2202 | case 1: | |
2203 | return 0; | |
2204 | case 5: | |
2205 | case 6: | |
a380c027 | 2206 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2207 | case 10: |
c92a2527 | 2208 | return 0x102; |
c2c087e6 | 2209 | } |
a18a888e | 2210 | return UnSet; |
c2c087e6 DW |
2211 | } |
2212 | ||
8e59f3d8 AK |
2213 | /******************************************************************************* |
2214 | * Function: read_imsm_migr_rec | |
2215 | * Description: Function reads imsm migration record from last sector of disk | |
2216 | * Parameters: | |
2217 | * fd : disk descriptor | |
2218 | * super : metadata info | |
2219 | * Returns: | |
2220 | * 0 : success, | |
2221 | * -1 : fail | |
2222 | ******************************************************************************/ | |
2223 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2224 | { | |
2225 | int ret_val = -1; | |
2226 | unsigned long long dsize; | |
2227 | ||
2228 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2229 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2230 | pr_err("Cannot seek to anchor block: %s\n", |
2231 | strerror(errno)); | |
8e59f3d8 AK |
2232 | goto out; |
2233 | } | |
17a4eaf9 AK |
2234 | if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2235 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2236 | pr_err("Cannot read migr record block: %s\n", |
2237 | strerror(errno)); | |
8e59f3d8 AK |
2238 | goto out; |
2239 | } | |
2240 | ret_val = 0; | |
2241 | ||
2242 | out: | |
2243 | return ret_val; | |
2244 | } | |
2245 | ||
3136abe5 AK |
2246 | static struct imsm_dev *imsm_get_device_during_migration( |
2247 | struct intel_super *super) | |
2248 | { | |
2249 | ||
2250 | struct intel_dev *dv; | |
2251 | ||
2252 | for (dv = super->devlist; dv; dv = dv->next) { | |
2253 | if (is_gen_migration(dv->dev)) | |
2254 | return dv->dev; | |
2255 | } | |
2256 | return NULL; | |
2257 | } | |
2258 | ||
8e59f3d8 AK |
2259 | /******************************************************************************* |
2260 | * Function: load_imsm_migr_rec | |
2261 | * Description: Function reads imsm migration record (it is stored at the last | |
2262 | * sector of disk) | |
2263 | * Parameters: | |
2264 | * super : imsm internal array info | |
2265 | * info : general array info | |
2266 | * Returns: | |
2267 | * 0 : success | |
2268 | * -1 : fail | |
4c965cc9 | 2269 | * -2 : no migration in progress |
8e59f3d8 AK |
2270 | ******************************************************************************/ |
2271 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2272 | { | |
2273 | struct mdinfo *sd; | |
2274 | struct dl *dl = NULL; | |
2275 | char nm[30]; | |
2276 | int retval = -1; | |
2277 | int fd = -1; | |
3136abe5 AK |
2278 | struct imsm_dev *dev; |
2279 | struct imsm_map *map = NULL; | |
b4ab44d8 | 2280 | int slot = -1; |
3136abe5 AK |
2281 | |
2282 | /* find map under migration */ | |
2283 | dev = imsm_get_device_during_migration(super); | |
2284 | /* nothing to load,no migration in progress? | |
2285 | */ | |
2286 | if (dev == NULL) | |
4c965cc9 | 2287 | return -2; |
44bfe6df | 2288 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 AK |
2289 | |
2290 | if (info) { | |
2291 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3136abe5 AK |
2292 | /* skip spare and failed disks |
2293 | */ | |
2294 | if (sd->disk.raid_disk < 0) | |
2295 | continue; | |
8e59f3d8 | 2296 | /* read only from one of the first two slots */ |
3136abe5 AK |
2297 | if (map) |
2298 | slot = get_imsm_disk_slot(map, | |
2299 | sd->disk.raid_disk); | |
2300 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 | 2301 | continue; |
3136abe5 | 2302 | |
8e59f3d8 AK |
2303 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2304 | fd = dev_open(nm, O_RDONLY); | |
2305 | if (fd >= 0) | |
2306 | break; | |
2307 | } | |
2308 | } | |
2309 | if (fd < 0) { | |
2310 | for (dl = super->disks; dl; dl = dl->next) { | |
3136abe5 AK |
2311 | /* skip spare and failed disks |
2312 | */ | |
2313 | if (dl->index < 0) | |
2314 | continue; | |
8e59f3d8 | 2315 | /* read only from one of the first two slots */ |
3136abe5 AK |
2316 | if (map) |
2317 | slot = get_imsm_disk_slot(map, dl->index); | |
2318 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 AK |
2319 | continue; |
2320 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2321 | fd = dev_open(nm, O_RDONLY); | |
2322 | if (fd >= 0) | |
2323 | break; | |
2324 | } | |
2325 | } | |
2326 | if (fd < 0) | |
2327 | goto out; | |
2328 | retval = read_imsm_migr_rec(fd, super); | |
2329 | ||
2330 | out: | |
2331 | if (fd >= 0) | |
2332 | close(fd); | |
2333 | return retval; | |
2334 | } | |
2335 | ||
9e2d750d | 2336 | #ifndef MDASSEMBLE |
c17608ea AK |
2337 | /******************************************************************************* |
2338 | * function: imsm_create_metadata_checkpoint_update | |
2339 | * Description: It creates update for checkpoint change. | |
2340 | * Parameters: | |
2341 | * super : imsm internal array info | |
2342 | * u : pointer to prepared update | |
2343 | * Returns: | |
2344 | * Uptate length. | |
2345 | * If length is equal to 0, input pointer u contains no update | |
2346 | ******************************************************************************/ | |
2347 | static int imsm_create_metadata_checkpoint_update( | |
2348 | struct intel_super *super, | |
2349 | struct imsm_update_general_migration_checkpoint **u) | |
2350 | { | |
2351 | ||
2352 | int update_memory_size = 0; | |
2353 | ||
2354 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2355 | ||
2356 | if (u == NULL) | |
2357 | return 0; | |
2358 | *u = NULL; | |
2359 | ||
2360 | /* size of all update data without anchor */ | |
2361 | update_memory_size = | |
2362 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2363 | ||
503975b9 | 2364 | *u = xcalloc(1, update_memory_size); |
c17608ea AK |
2365 | if (*u == NULL) { |
2366 | dprintf("error: cannot get memory for " | |
2367 | "imsm_create_metadata_checkpoint_update update\n"); | |
2368 | return 0; | |
2369 | } | |
2370 | (*u)->type = update_general_migration_checkpoint; | |
2371 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2372 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2373 | (*u)->curr_migr_unit); | |
2374 | ||
2375 | return update_memory_size; | |
2376 | } | |
2377 | ||
2378 | ||
2379 | static void imsm_update_metadata_locally(struct supertype *st, | |
2380 | void *buf, int len); | |
2381 | ||
687629c2 AK |
2382 | /******************************************************************************* |
2383 | * Function: write_imsm_migr_rec | |
2384 | * Description: Function writes imsm migration record | |
2385 | * (at the last sector of disk) | |
2386 | * Parameters: | |
2387 | * super : imsm internal array info | |
2388 | * Returns: | |
2389 | * 0 : success | |
2390 | * -1 : if fail | |
2391 | ******************************************************************************/ | |
2392 | static int write_imsm_migr_rec(struct supertype *st) | |
2393 | { | |
2394 | struct intel_super *super = st->sb; | |
2395 | unsigned long long dsize; | |
2396 | char nm[30]; | |
2397 | int fd = -1; | |
2398 | int retval = -1; | |
2399 | struct dl *sd; | |
c17608ea AK |
2400 | int len; |
2401 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 AK |
2402 | struct imsm_dev *dev; |
2403 | struct imsm_map *map = NULL; | |
2404 | ||
2405 | /* find map under migration */ | |
2406 | dev = imsm_get_device_during_migration(super); | |
2407 | /* if no migration, write buffer anyway to clear migr_record | |
2408 | * on disk based on first available device | |
2409 | */ | |
2410 | if (dev == NULL) | |
2411 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2412 | super->current_vol); | |
2413 | ||
44bfe6df | 2414 | map = get_imsm_map(dev, MAP_0); |
687629c2 AK |
2415 | |
2416 | for (sd = super->disks ; sd ; sd = sd->next) { | |
b4ab44d8 | 2417 | int slot = -1; |
3136abe5 AK |
2418 | |
2419 | /* skip failed and spare devices */ | |
2420 | if (sd->index < 0) | |
2421 | continue; | |
687629c2 | 2422 | /* write to 2 first slots only */ |
3136abe5 AK |
2423 | if (map) |
2424 | slot = get_imsm_disk_slot(map, sd->index); | |
2425 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
687629c2 | 2426 | continue; |
3136abe5 | 2427 | |
687629c2 AK |
2428 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
2429 | fd = dev_open(nm, O_RDWR); | |
2430 | if (fd < 0) | |
2431 | continue; | |
2432 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2433 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2434 | pr_err("Cannot seek to anchor block: %s\n", |
2435 | strerror(errno)); | |
687629c2 AK |
2436 | goto out; |
2437 | } | |
17a4eaf9 AK |
2438 | if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2439 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2440 | pr_err("Cannot write migr record block: %s\n", |
2441 | strerror(errno)); | |
687629c2 AK |
2442 | goto out; |
2443 | } | |
2444 | close(fd); | |
2445 | fd = -1; | |
2446 | } | |
c17608ea AK |
2447 | /* update checkpoint information in metadata */ |
2448 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2449 | ||
2450 | if (len <= 0) { | |
2451 | dprintf("imsm: Cannot prepare update\n"); | |
2452 | goto out; | |
2453 | } | |
2454 | /* update metadata locally */ | |
2455 | imsm_update_metadata_locally(st, u, len); | |
2456 | /* and possibly remotely */ | |
2457 | if (st->update_tail) { | |
2458 | append_metadata_update(st, u, len); | |
2459 | /* during reshape we do all work inside metadata handler | |
2460 | * manage_reshape(), so metadata update has to be triggered | |
2461 | * insida it | |
2462 | */ | |
2463 | flush_metadata_updates(st); | |
2464 | st->update_tail = &st->updates; | |
2465 | } else | |
2466 | free(u); | |
687629c2 AK |
2467 | |
2468 | retval = 0; | |
2469 | out: | |
2470 | if (fd >= 0) | |
2471 | close(fd); | |
2472 | return retval; | |
2473 | } | |
9e2d750d | 2474 | #endif /* MDASSEMBLE */ |
687629c2 | 2475 | |
e2962bfc AK |
2476 | /* spare/missing disks activations are not allowe when |
2477 | * array/container performs reshape operation, because | |
2478 | * all arrays in container works on the same disks set | |
2479 | */ | |
2480 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2481 | { | |
2482 | int rv = 0; | |
2483 | struct intel_dev *i_dev; | |
2484 | struct imsm_dev *dev; | |
2485 | ||
2486 | /* check whole container | |
2487 | */ | |
2488 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2489 | dev = i_dev->dev; | |
3ad25638 | 2490 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2491 | /* No repair during any migration in container |
2492 | */ | |
2493 | rv = 1; | |
2494 | break; | |
2495 | } | |
2496 | } | |
2497 | return rv; | |
2498 | } | |
c41e00b2 AK |
2499 | static unsigned long long imsm_component_size_aligment_check(int level, |
2500 | int chunk_size, | |
2501 | unsigned long long component_size) | |
2502 | { | |
2503 | unsigned int component_size_alligment; | |
2504 | ||
2505 | /* check component size aligment | |
2506 | */ | |
2507 | component_size_alligment = component_size % (chunk_size/512); | |
2508 | ||
2509 | dprintf("imsm_component_size_aligment_check(Level: %i, " | |
2510 | "chunk_size = %i, component_size = %llu), " | |
2511 | "component_size_alligment = %u\n", | |
2512 | level, chunk_size, component_size, | |
2513 | component_size_alligment); | |
2514 | ||
2515 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
2516 | dprintf("imsm: reported component size alligned from %llu ", | |
2517 | component_size); | |
2518 | component_size -= component_size_alligment; | |
2519 | dprintf("to %llu (%i).\n", | |
2520 | component_size, component_size_alligment); | |
2521 | } | |
2522 | ||
2523 | return component_size; | |
2524 | } | |
e2962bfc | 2525 | |
a5d85af7 | 2526 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2527 | { |
2528 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2529 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2530 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2531 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2532 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2533 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2534 | struct dl *dl; |
e207da2f | 2535 | char *devname; |
a5d85af7 | 2536 | int map_disks = info->array.raid_disks; |
bf5a934a | 2537 | |
95eeceeb | 2538 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2539 | if (prev_map) |
2540 | map_to_analyse = prev_map; | |
2541 | ||
ca0748fa | 2542 | dl = super->current_disk; |
9894ec0d | 2543 | |
bf5a934a | 2544 | info->container_member = super->current_vol; |
cd0430a1 | 2545 | info->array.raid_disks = map->num_members; |
b335e593 | 2546 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2547 | info->array.layout = imsm_level_to_layout(info->array.level); |
2548 | info->array.md_minor = -1; | |
2549 | info->array.ctime = 0; | |
2550 | info->array.utime = 0; | |
b335e593 AK |
2551 | info->array.chunk_size = |
2552 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2553 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2554 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2555 | info->custom_array_size <<= 32; | |
2556 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2557 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2558 | ||
3f510843 | 2559 | if (is_gen_migration(dev)) { |
3f83228a | 2560 | info->reshape_active = 1; |
b335e593 AK |
2561 | info->new_level = get_imsm_raid_level(map); |
2562 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2563 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2564 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2565 | if (info->delta_disks) { |
2566 | /* this needs to be applied to every array | |
2567 | * in the container. | |
2568 | */ | |
81219e70 | 2569 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2570 | } |
3f83228a N |
2571 | /* We shape information that we give to md might have to be |
2572 | * modify to cope with md's requirement for reshaping arrays. | |
2573 | * For example, when reshaping a RAID0, md requires it to be | |
2574 | * presented as a degraded RAID4. | |
2575 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2576 | * the array as already being RAID5, but the 'before' layout | |
2577 | * is a RAID4-like layout. | |
2578 | */ | |
2579 | switch (info->array.level) { | |
2580 | case 0: | |
2581 | switch(info->new_level) { | |
2582 | case 0: | |
2583 | /* conversion is happening as RAID4 */ | |
2584 | info->array.level = 4; | |
2585 | info->array.raid_disks += 1; | |
2586 | break; | |
2587 | case 5: | |
2588 | /* conversion is happening as RAID5 */ | |
2589 | info->array.level = 5; | |
2590 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2591 | info->delta_disks -= 1; |
2592 | break; | |
2593 | default: | |
2594 | /* FIXME error message */ | |
2595 | info->array.level = UnSet; | |
2596 | break; | |
2597 | } | |
2598 | break; | |
2599 | } | |
b335e593 AK |
2600 | } else { |
2601 | info->new_level = UnSet; | |
2602 | info->new_layout = UnSet; | |
2603 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2604 | info->delta_disks = 0; |
b335e593 | 2605 | } |
ca0748fa | 2606 | |
efb30e7f DW |
2607 | if (dl) { |
2608 | info->disk.major = dl->major; | |
2609 | info->disk.minor = dl->minor; | |
ca0748fa | 2610 | info->disk.number = dl->index; |
656b6b5a N |
2611 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2612 | dl->index); | |
efb30e7f | 2613 | } |
bf5a934a | 2614 | |
5551b113 CA |
2615 | info->data_offset = pba_of_lba0(map_to_analyse); |
2616 | info->component_size = blocks_per_member(map_to_analyse); | |
139dae11 | 2617 | |
c41e00b2 AK |
2618 | info->component_size = imsm_component_size_aligment_check( |
2619 | info->array.level, | |
2620 | info->array.chunk_size, | |
2621 | info->component_size); | |
139dae11 | 2622 | |
301406c9 | 2623 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2624 | info->recovery_start = MaxSector; |
bf5a934a | 2625 | |
d2e6d5d6 | 2626 | info->reshape_progress = 0; |
b6796ce1 | 2627 | info->resync_start = MaxSector; |
b9172665 AK |
2628 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2629 | dev->vol.dirty) && | |
2630 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2631 | info->resync_start = 0; |
b6796ce1 AK |
2632 | } |
2633 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2634 | switch (migr_type(dev)) { |
2635 | case MIGR_REPAIR: | |
2636 | case MIGR_INIT: { | |
c47b0ff6 AK |
2637 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2638 | dev); | |
1e5c6983 DW |
2639 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2640 | ||
2641 | info->resync_start = blocks_per_unit * units; | |
2642 | break; | |
2643 | } | |
d2e6d5d6 | 2644 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2645 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2646 | dev); | |
2647 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2648 | unsigned long long array_blocks; |
2649 | int used_disks; | |
d2e6d5d6 | 2650 | |
befb629b AK |
2651 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2652 | (units < | |
2653 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2654 | (super->migr_rec->rec_status == | |
2655 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2656 | units++; | |
2657 | ||
d2e6d5d6 | 2658 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2659 | |
d2e6d5d6 AK |
2660 | dprintf("IMSM: General Migration checkpoint : %llu " |
2661 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2662 | (unsigned long long)units, |
2663 | (unsigned long long)blocks_per_unit, | |
2664 | info->reshape_progress); | |
75156c46 | 2665 | |
238c0a71 | 2666 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 2667 | if (used_disks > 0) { |
5551b113 | 2668 | array_blocks = blocks_per_member(map) * |
75156c46 AK |
2669 | used_disks; |
2670 | /* round array size down to closest MB | |
2671 | */ | |
2672 | info->custom_array_size = (array_blocks | |
2673 | >> SECT_PER_MB_SHIFT) | |
2674 | << SECT_PER_MB_SHIFT; | |
2675 | } | |
d2e6d5d6 | 2676 | } |
1e5c6983 DW |
2677 | case MIGR_VERIFY: |
2678 | /* we could emulate the checkpointing of | |
2679 | * 'sync_action=check' migrations, but for now | |
2680 | * we just immediately complete them | |
2681 | */ | |
2682 | case MIGR_REBUILD: | |
2683 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2684 | case MIGR_STATE_CHANGE: |
2685 | /* FIXME handle other migrations */ | |
2686 | default: | |
2687 | /* we are not dirty, so... */ | |
2688 | info->resync_start = MaxSector; | |
2689 | } | |
b6796ce1 | 2690 | } |
301406c9 DW |
2691 | |
2692 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2693 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2694 | |
f35f2525 N |
2695 | info->array.major_version = -1; |
2696 | info->array.minor_version = -2; | |
e207da2f AW |
2697 | devname = devnum2devname(st->container_dev); |
2698 | *info->text_version = '\0'; | |
2699 | if (devname) | |
2700 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2701 | free(devname); | |
a67dd8cc | 2702 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2703 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2704 | |
2705 | if (dmap) { | |
2706 | int i, j; | |
2707 | for (i=0; i<map_disks; i++) { | |
2708 | dmap[i] = 0; | |
2709 | if (i < info->array.raid_disks) { | |
2710 | struct imsm_disk *dsk; | |
238c0a71 | 2711 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2712 | dsk = get_imsm_disk(super, j); |
2713 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2714 | dmap[i] = 1; | |
2715 | } | |
2716 | } | |
2717 | } | |
81ac8b4d | 2718 | } |
bf5a934a | 2719 | |
3b451610 AK |
2720 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2721 | int failed, int look_in_map); | |
2722 | ||
2723 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2724 | int look_in_map); | |
2725 | ||
b4ab44d8 AK |
2726 | |
2727 | #ifndef MDASSEMBLE | |
3b451610 AK |
2728 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
2729 | { | |
2730 | if (is_gen_migration(dev)) { | |
2731 | int failed; | |
2732 | __u8 map_state; | |
2733 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2734 | ||
2735 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2736 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2737 | if (map2->map_state != map_state) { |
2738 | map2->map_state = map_state; | |
2739 | super->updates_pending++; | |
2740 | } | |
2741 | } | |
2742 | } | |
b4ab44d8 | 2743 | #endif |
97b4d0e9 DW |
2744 | |
2745 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2746 | { | |
2747 | struct dl *d; | |
2748 | ||
2749 | for (d = super->missing; d; d = d->next) | |
2750 | if (d->index == index) | |
2751 | return &d->disk; | |
2752 | return NULL; | |
2753 | } | |
2754 | ||
a5d85af7 | 2755 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2756 | { |
2757 | struct intel_super *super = st->sb; | |
4f5bc454 | 2758 | struct imsm_disk *disk; |
a5d85af7 | 2759 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2760 | int max_enough = -1; |
2761 | int i; | |
2762 | struct imsm_super *mpb; | |
4f5bc454 | 2763 | |
bf5a934a | 2764 | if (super->current_vol >= 0) { |
a5d85af7 | 2765 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2766 | return; |
2767 | } | |
95eeceeb | 2768 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2769 | |
2770 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2771 | * spares | |
2772 | */ | |
2773 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2774 | info->array.level = LEVEL_CONTAINER; |
2775 | info->array.layout = 0; | |
2776 | info->array.md_minor = -1; | |
c2c087e6 | 2777 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2778 | info->array.utime = 0; |
2779 | info->array.chunk_size = 0; | |
2780 | ||
2781 | info->disk.major = 0; | |
2782 | info->disk.minor = 0; | |
cdddbdbc | 2783 | info->disk.raid_disk = -1; |
c2c087e6 | 2784 | info->reshape_active = 0; |
f35f2525 N |
2785 | info->array.major_version = -1; |
2786 | info->array.minor_version = -2; | |
c2c087e6 | 2787 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2788 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2789 | info->disk.number = -1; |
2790 | info->disk.state = 0; | |
c5afc314 | 2791 | info->name[0] = 0; |
921d9e16 | 2792 | info->recovery_start = MaxSector; |
3ad25638 | 2793 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2794 | |
97b4d0e9 | 2795 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2796 | mpb = super->anchor; |
97b4d0e9 | 2797 | |
ab3cb6b3 N |
2798 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2799 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2800 | int failed, enough, j, missing = 0; | |
2801 | struct imsm_map *map; | |
2802 | __u8 state; | |
97b4d0e9 | 2803 | |
3b451610 AK |
2804 | failed = imsm_count_failed(super, dev, MAP_0); |
2805 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2806 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2807 | |
2808 | /* any newly missing disks? | |
2809 | * (catches single-degraded vs double-degraded) | |
2810 | */ | |
2811 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2812 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2813 | __u32 idx = ord_to_idx(ord); |
2814 | ||
2815 | if (!(ord & IMSM_ORD_REBUILD) && | |
2816 | get_imsm_missing(super, idx)) { | |
2817 | missing = 1; | |
2818 | break; | |
2819 | } | |
97b4d0e9 | 2820 | } |
ab3cb6b3 N |
2821 | |
2822 | if (state == IMSM_T_STATE_FAILED) | |
2823 | enough = -1; | |
2824 | else if (state == IMSM_T_STATE_DEGRADED && | |
2825 | (state != map->map_state || missing)) | |
2826 | enough = 0; | |
2827 | else /* we're normal, or already degraded */ | |
2828 | enough = 1; | |
d2bde6d3 AK |
2829 | if (is_gen_migration(dev) && missing) { |
2830 | /* during general migration we need all disks | |
2831 | * that process is running on. | |
2832 | * No new missing disk is allowed. | |
2833 | */ | |
2834 | max_enough = -1; | |
2835 | enough = -1; | |
2836 | /* no more checks necessary | |
2837 | */ | |
2838 | break; | |
2839 | } | |
ab3cb6b3 N |
2840 | /* in the missing/failed disk case check to see |
2841 | * if at least one array is runnable | |
2842 | */ | |
2843 | max_enough = max(max_enough, enough); | |
2844 | } | |
2845 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2846 | info->container_enough = max_enough; | |
97b4d0e9 | 2847 | |
4a04ec6c | 2848 | if (super->disks) { |
14e8215b DW |
2849 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2850 | ||
b9f594fe | 2851 | disk = &super->disks->disk; |
5551b113 | 2852 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 2853 | info->component_size = reserved; |
25ed7e59 | 2854 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2855 | /* we don't change info->disk.raid_disk here because |
2856 | * this state will be finalized in mdmon after we have | |
2857 | * found the 'most fresh' version of the metadata | |
2858 | */ | |
25ed7e59 DW |
2859 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2860 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2861 | } |
a575e2a7 DW |
2862 | |
2863 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2864 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2865 | */ | |
2866 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2867 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2868 | else |
2869 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2870 | |
2871 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2872 | if (map) { | |
2873 | int i; | |
2874 | for (i = 0; i < map_disks; i++) | |
2875 | map[i] = 1; | |
2876 | } | |
2877 | ||
cdddbdbc DW |
2878 | } |
2879 | ||
5c4cd5da AC |
2880 | /* allocates memory and fills disk in mdinfo structure |
2881 | * for each disk in array */ | |
2882 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2883 | { | |
2884 | struct mdinfo *mddev = NULL; | |
2885 | struct intel_super *super = st->sb; | |
2886 | struct imsm_disk *disk; | |
2887 | int count = 0; | |
2888 | struct dl *dl; | |
2889 | if (!super || !super->disks) | |
2890 | return NULL; | |
2891 | dl = super->disks; | |
503975b9 | 2892 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
2893 | while (dl) { |
2894 | struct mdinfo *tmp; | |
2895 | disk = &dl->disk; | |
503975b9 | 2896 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
2897 | if (mddev->devs) |
2898 | tmp->next = mddev->devs; | |
2899 | mddev->devs = tmp; | |
2900 | tmp->disk.number = count++; | |
2901 | tmp->disk.major = dl->major; | |
2902 | tmp->disk.minor = dl->minor; | |
2903 | tmp->disk.state = is_configured(disk) ? | |
2904 | (1 << MD_DISK_ACTIVE) : 0; | |
2905 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2906 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2907 | tmp->disk.raid_disk = -1; | |
2908 | dl = dl->next; | |
2909 | } | |
2910 | return mddev; | |
2911 | } | |
2912 | ||
cdddbdbc DW |
2913 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2914 | char *update, char *devname, int verbose, | |
2915 | int uuid_set, char *homehost) | |
2916 | { | |
f352c545 DW |
2917 | /* For 'assemble' and 'force' we need to return non-zero if any |
2918 | * change was made. For others, the return value is ignored. | |
2919 | * Update options are: | |
2920 | * force-one : This device looks a bit old but needs to be included, | |
2921 | * update age info appropriately. | |
2922 | * assemble: clear any 'faulty' flag to allow this device to | |
2923 | * be assembled. | |
2924 | * force-array: Array is degraded but being forced, mark it clean | |
2925 | * if that will be needed to assemble it. | |
2926 | * | |
2927 | * newdev: not used ???? | |
2928 | * grow: Array has gained a new device - this is currently for | |
2929 | * linear only | |
2930 | * resync: mark as dirty so a resync will happen. | |
2931 | * name: update the name - preserving the homehost | |
6e46bf34 | 2932 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2933 | * |
2934 | * Following are not relevant for this imsm: | |
2935 | * sparc2.2 : update from old dodgey metadata | |
2936 | * super-minor: change the preferred_minor number | |
2937 | * summaries: update redundant counters. | |
f352c545 DW |
2938 | * homehost: update the recorded homehost |
2939 | * _reshape_progress: record new reshape_progress position. | |
2940 | */ | |
6e46bf34 DW |
2941 | int rv = 1; |
2942 | struct intel_super *super = st->sb; | |
2943 | struct imsm_super *mpb; | |
f352c545 | 2944 | |
6e46bf34 DW |
2945 | /* we can only update container info */ |
2946 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2947 | return 1; | |
2948 | ||
2949 | mpb = super->anchor; | |
2950 | ||
81a5b4f5 N |
2951 | if (strcmp(update, "uuid") == 0) { |
2952 | /* We take this to mean that the family_num should be updated. | |
2953 | * However that is much smaller than the uuid so we cannot really | |
2954 | * allow an explicit uuid to be given. And it is hard to reliably | |
2955 | * know if one was. | |
2956 | * So if !uuid_set we know the current uuid is random and just used | |
2957 | * the first 'int' and copy it to the other 3 positions. | |
2958 | * Otherwise we require the 4 'int's to be the same as would be the | |
2959 | * case if we are using a random uuid. So an explicit uuid will be | |
2960 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 2961 | */ |
81a5b4f5 N |
2962 | if (!uuid_set) { |
2963 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 2964 | rv = 0; |
81a5b4f5 N |
2965 | } else { |
2966 | if (info->uuid[0] != info->uuid[1] || | |
2967 | info->uuid[1] != info->uuid[2] || | |
2968 | info->uuid[2] != info->uuid[3]) | |
2969 | rv = -1; | |
2970 | else | |
2971 | rv = 0; | |
6e46bf34 | 2972 | } |
81a5b4f5 N |
2973 | if (rv == 0) |
2974 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
2975 | } else if (strcmp(update, "assemble") == 0) |
2976 | rv = 0; | |
2977 | else | |
1e2b2765 | 2978 | rv = -1; |
f352c545 | 2979 | |
6e46bf34 DW |
2980 | /* successful update? recompute checksum */ |
2981 | if (rv == 0) | |
2982 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2983 | |
2984 | return rv; | |
cdddbdbc DW |
2985 | } |
2986 | ||
c2c087e6 | 2987 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2988 | { |
c2c087e6 | 2989 | size_t size; |
cdddbdbc | 2990 | |
c2c087e6 DW |
2991 | size = sizeof(struct imsm_super); |
2992 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2993 | size += 2 * sizeof(struct imsm_dev); | |
2994 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2995 | size += (4 - 2) * sizeof(struct imsm_map); | |
2996 | /* 4 possible disk_ord_tbl's */ | |
2997 | size += 4 * (disks - 1) * sizeof(__u32); | |
2998 | ||
2999 | return size; | |
3000 | } | |
3001 | ||
387fcd59 N |
3002 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3003 | unsigned long long data_offset) | |
c2c087e6 DW |
3004 | { |
3005 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3006 | return 0; | |
3007 | ||
3008 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3009 | } |
3010 | ||
ba2de7ba DW |
3011 | static void free_devlist(struct intel_super *super) |
3012 | { | |
3013 | struct intel_dev *dv; | |
3014 | ||
3015 | while (super->devlist) { | |
3016 | dv = super->devlist->next; | |
3017 | free(super->devlist->dev); | |
3018 | free(super->devlist); | |
3019 | super->devlist = dv; | |
3020 | } | |
3021 | } | |
3022 | ||
3023 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3024 | { | |
3025 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3026 | } | |
3027 | ||
cdddbdbc DW |
3028 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3029 | { | |
3030 | /* | |
3031 | * return: | |
3032 | * 0 same, or first was empty, and second was copied | |
3033 | * 1 second had wrong number | |
3034 | * 2 wrong uuid | |
3035 | * 3 wrong other info | |
3036 | */ | |
3037 | struct intel_super *first = st->sb; | |
3038 | struct intel_super *sec = tst->sb; | |
3039 | ||
5d500228 N |
3040 | if (!first) { |
3041 | st->sb = tst->sb; | |
3042 | tst->sb = NULL; | |
3043 | return 0; | |
3044 | } | |
8603ea6f LM |
3045 | /* in platform dependent environment test if the disks |
3046 | * use the same Intel hba | |
cb8f6859 | 3047 | * If not on Intel hba at all, allow anything. |
8603ea6f LM |
3048 | */ |
3049 | if (!check_env("IMSM_NO_PLATFORM")) { | |
cb8f6859 | 3050 | if (first->hba && sec->hba && |
3c309c82 | 3051 | strcmp(first->hba->path, sec->hba->path) != 0) { |
8603ea6f LM |
3052 | fprintf(stderr, |
3053 | "HBAs of devices does not match %s != %s\n", | |
3c309c82 MT |
3054 | first->hba ? first->hba->path : NULL, |
3055 | sec->hba ? sec->hba->path : NULL); | |
8603ea6f LM |
3056 | return 3; |
3057 | } | |
3058 | } | |
cdddbdbc | 3059 | |
d23fe947 DW |
3060 | /* if an anchor does not have num_raid_devs set then it is a free |
3061 | * floating spare | |
3062 | */ | |
3063 | if (first->anchor->num_raid_devs > 0 && | |
3064 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3065 | /* Determine if these disks might ever have been |
3066 | * related. Further disambiguation can only take place | |
3067 | * in load_super_imsm_all | |
3068 | */ | |
3069 | __u32 first_family = first->anchor->orig_family_num; | |
3070 | __u32 sec_family = sec->anchor->orig_family_num; | |
3071 | ||
f796af5d DW |
3072 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3073 | MAX_SIGNATURE_LENGTH) != 0) | |
3074 | return 3; | |
3075 | ||
a2b97981 DW |
3076 | if (first_family == 0) |
3077 | first_family = first->anchor->family_num; | |
3078 | if (sec_family == 0) | |
3079 | sec_family = sec->anchor->family_num; | |
3080 | ||
3081 | if (first_family != sec_family) | |
d23fe947 | 3082 | return 3; |
f796af5d | 3083 | |
d23fe947 | 3084 | } |
cdddbdbc | 3085 | |
f796af5d | 3086 | |
3e372e5a DW |
3087 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3088 | * family number | |
3089 | */ | |
3090 | if (first->anchor->num_raid_devs == 0 && | |
3091 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3092 | int i; |
ba2de7ba DW |
3093 | struct intel_dev *dv; |
3094 | struct imsm_dev *dev; | |
78d30f94 DW |
3095 | |
3096 | /* we need to copy raid device info from sec if an allocation | |
3097 | * fails here we don't associate the spare | |
3098 | */ | |
3099 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3100 | dv = xmalloc(sizeof(*dv)); |
3101 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3102 | dv->dev = dev; |
3103 | dv->index = i; | |
3104 | dv->next = first->devlist; | |
3105 | first->devlist = dv; | |
78d30f94 | 3106 | } |
709743c5 | 3107 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3108 | /* allocation failure */ |
3109 | free_devlist(first); | |
3110 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
3111 | return 3; | |
78d30f94 | 3112 | } |
3e372e5a | 3113 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3114 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3115 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3116 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3117 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3118 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3119 | } |
3120 | ||
cdddbdbc DW |
3121 | return 0; |
3122 | } | |
3123 | ||
0030e8d6 DW |
3124 | static void fd2devname(int fd, char *name) |
3125 | { | |
3126 | struct stat st; | |
3127 | char path[256]; | |
33a6535d | 3128 | char dname[PATH_MAX]; |
0030e8d6 DW |
3129 | char *nm; |
3130 | int rv; | |
3131 | ||
3132 | name[0] = '\0'; | |
3133 | if (fstat(fd, &st) != 0) | |
3134 | return; | |
3135 | sprintf(path, "/sys/dev/block/%d:%d", | |
3136 | major(st.st_rdev), minor(st.st_rdev)); | |
3137 | ||
9cf014ec | 3138 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3139 | if (rv <= 0) |
3140 | return; | |
9587c373 | 3141 | |
0030e8d6 DW |
3142 | dname[rv] = '\0'; |
3143 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3144 | if (nm) { |
3145 | nm++; | |
3146 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3147 | } | |
0030e8d6 DW |
3148 | } |
3149 | ||
cdddbdbc DW |
3150 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3151 | ||
3152 | static int imsm_read_serial(int fd, char *devname, | |
3153 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3154 | { | |
3155 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
3156 | int rv; |
3157 | int rsp_len; | |
1f24f035 | 3158 | int len; |
316e2bf4 DW |
3159 | char *dest; |
3160 | char *src; | |
3161 | char *rsp_buf; | |
3162 | int i; | |
cdddbdbc DW |
3163 | |
3164 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 3165 | |
f9ba0ff1 DW |
3166 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
3167 | ||
40ebbb9c | 3168 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3169 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3170 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3171 | return 0; |
3172 | } | |
3173 | ||
cdddbdbc DW |
3174 | if (rv != 0) { |
3175 | if (devname) | |
e7b84f9d N |
3176 | pr_err("Failed to retrieve serial for %s\n", |
3177 | devname); | |
cdddbdbc DW |
3178 | return rv; |
3179 | } | |
3180 | ||
3181 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
3182 | if (!rsp_len) { |
3183 | if (devname) | |
e7b84f9d N |
3184 | pr_err("Failed to retrieve serial for %s\n", |
3185 | devname); | |
03cd4cc8 DW |
3186 | return 2; |
3187 | } | |
1f24f035 | 3188 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 3189 | |
316e2bf4 DW |
3190 | /* trim all whitespace and non-printable characters and convert |
3191 | * ':' to ';' | |
3192 | */ | |
3193 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
3194 | src = &rsp_buf[i]; | |
3195 | if (*src > 0x20) { | |
3196 | /* ':' is reserved for use in placeholder serial | |
3197 | * numbers for missing disks | |
3198 | */ | |
3199 | if (*src == ':') | |
3200 | *dest++ = ';'; | |
3201 | else | |
3202 | *dest++ = *src; | |
3203 | } | |
3204 | } | |
3205 | len = dest - rsp_buf; | |
3206 | dest = rsp_buf; | |
3207 | ||
3208 | /* truncate leading characters */ | |
3209 | if (len > MAX_RAID_SERIAL_LEN) { | |
3210 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3211 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3212 | } |
5c3db629 | 3213 | |
5c3db629 | 3214 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3215 | memcpy(serial, dest, len); |
cdddbdbc DW |
3216 | |
3217 | return 0; | |
3218 | } | |
3219 | ||
1f24f035 DW |
3220 | static int serialcmp(__u8 *s1, __u8 *s2) |
3221 | { | |
3222 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3223 | } | |
3224 | ||
3225 | static void serialcpy(__u8 *dest, __u8 *src) | |
3226 | { | |
3227 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3228 | } | |
3229 | ||
54c2c1ea DW |
3230 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3231 | { | |
3232 | struct dl *dl; | |
3233 | ||
3234 | for (dl = super->disks; dl; dl = dl->next) | |
3235 | if (serialcmp(dl->serial, serial) == 0) | |
3236 | break; | |
3237 | ||
3238 | return dl; | |
3239 | } | |
3240 | ||
a2b97981 DW |
3241 | static struct imsm_disk * |
3242 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3243 | { | |
3244 | int i; | |
3245 | ||
3246 | for (i = 0; i < mpb->num_disks; i++) { | |
3247 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3248 | ||
3249 | if (serialcmp(disk->serial, serial) == 0) { | |
3250 | if (idx) | |
3251 | *idx = i; | |
3252 | return disk; | |
3253 | } | |
3254 | } | |
3255 | ||
3256 | return NULL; | |
3257 | } | |
3258 | ||
cdddbdbc DW |
3259 | static int |
3260 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3261 | { | |
a2b97981 | 3262 | struct imsm_disk *disk; |
cdddbdbc DW |
3263 | struct dl *dl; |
3264 | struct stat stb; | |
cdddbdbc | 3265 | int rv; |
a2b97981 | 3266 | char name[40]; |
d23fe947 DW |
3267 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3268 | ||
3269 | rv = imsm_read_serial(fd, devname, serial); | |
3270 | ||
3271 | if (rv != 0) | |
3272 | return 2; | |
3273 | ||
503975b9 | 3274 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3275 | |
a2b97981 DW |
3276 | fstat(fd, &stb); |
3277 | dl->major = major(stb.st_rdev); | |
3278 | dl->minor = minor(stb.st_rdev); | |
3279 | dl->next = super->disks; | |
3280 | dl->fd = keep_fd ? fd : -1; | |
3281 | assert(super->disks == NULL); | |
3282 | super->disks = dl; | |
3283 | serialcpy(dl->serial, serial); | |
3284 | dl->index = -2; | |
3285 | dl->e = NULL; | |
3286 | fd2devname(fd, name); | |
3287 | if (devname) | |
503975b9 | 3288 | dl->devname = xstrdup(devname); |
a2b97981 | 3289 | else |
503975b9 | 3290 | dl->devname = xstrdup(name); |
cdddbdbc | 3291 | |
d23fe947 | 3292 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3293 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3294 | if (disk) { | |
3295 | dl->disk = *disk; | |
3296 | /* only set index on disks that are a member of a | |
3297 | * populated contianer, i.e. one with raid_devs | |
3298 | */ | |
3299 | if (is_failed(&dl->disk)) | |
3f6efecc | 3300 | dl->index = -2; |
a2b97981 DW |
3301 | else if (is_spare(&dl->disk)) |
3302 | dl->index = -1; | |
3f6efecc DW |
3303 | } |
3304 | ||
949c47a0 DW |
3305 | return 0; |
3306 | } | |
3307 | ||
0e600426 | 3308 | #ifndef MDASSEMBLE |
0c046afd DW |
3309 | /* When migrating map0 contains the 'destination' state while map1 |
3310 | * contains the current state. When not migrating map0 contains the | |
3311 | * current state. This routine assumes that map[0].map_state is set to | |
3312 | * the current array state before being called. | |
3313 | * | |
3314 | * Migration is indicated by one of the following states | |
3315 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3316 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3317 | * map1state=unitialized) |
1484e727 | 3318 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3319 | * map1state=normal) |
e3bba0e0 | 3320 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3321 | * map1state=degraded) |
8e59f3d8 AK |
3322 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3323 | * map1state=normal) | |
0c046afd | 3324 | */ |
8e59f3d8 AK |
3325 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3326 | __u8 to_state, int migr_type) | |
3393c6af | 3327 | { |
0c046afd | 3328 | struct imsm_map *dest; |
238c0a71 | 3329 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3330 | |
0c046afd | 3331 | dev->vol.migr_state = 1; |
1484e727 | 3332 | set_migr_type(dev, migr_type); |
f8f603f1 | 3333 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3334 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3335 | |
0556e1a2 | 3336 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3337 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3338 | if ((migr_type == MIGR_REBUILD) || |
3339 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3340 | __u32 ord; |
3341 | int i; | |
3342 | ||
3343 | for (i = 0; i < src->num_members; i++) { | |
3344 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3345 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3346 | } | |
3347 | } | |
3348 | ||
8e59f3d8 AK |
3349 | if (migr_type == MIGR_GEN_MIGR) |
3350 | /* Clear migration record */ | |
3351 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3352 | ||
0c046afd | 3353 | src->map_state = to_state; |
949c47a0 | 3354 | } |
f8f603f1 | 3355 | |
809da78e AK |
3356 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3357 | __u8 map_state) | |
f8f603f1 | 3358 | { |
238c0a71 AK |
3359 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3360 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3361 | MAP_0 : MAP_1); | |
28bce06f | 3362 | int i, j; |
0556e1a2 DW |
3363 | |
3364 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3365 | * completed in the last migration. | |
3366 | * | |
28bce06f | 3367 | * FIXME add support for raid-level-migration |
0556e1a2 | 3368 | */ |
809da78e AK |
3369 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3370 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3371 | /* when final map state is other than expected | |
3372 | * merge maps (not for migration) | |
3373 | */ | |
3374 | int failed; | |
3375 | ||
3376 | for (i = 0; i < prev->num_members; i++) | |
3377 | for (j = 0; j < map->num_members; j++) | |
3378 | /* during online capacity expansion | |
3379 | * disks position can be changed | |
3380 | * if takeover is used | |
3381 | */ | |
3382 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3383 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3384 | map->disk_ord_tbl[j] |= | |
3385 | prev->disk_ord_tbl[i]; | |
3386 | break; | |
3387 | } | |
3388 | failed = imsm_count_failed(super, dev, MAP_0); | |
3389 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3390 | } | |
f8f603f1 DW |
3391 | |
3392 | dev->vol.migr_state = 0; | |
ea672ee1 | 3393 | set_migr_type(dev, 0); |
f8f603f1 DW |
3394 | dev->vol.curr_migr_unit = 0; |
3395 | map->map_state = map_state; | |
3396 | } | |
0e600426 | 3397 | #endif |
949c47a0 DW |
3398 | |
3399 | static int parse_raid_devices(struct intel_super *super) | |
3400 | { | |
3401 | int i; | |
3402 | struct imsm_dev *dev_new; | |
4d7b1503 | 3403 | size_t len, len_migr; |
401d313b | 3404 | size_t max_len = 0; |
4d7b1503 DW |
3405 | size_t space_needed = 0; |
3406 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3407 | |
3408 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3409 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3410 | struct intel_dev *dv; |
949c47a0 | 3411 | |
4d7b1503 DW |
3412 | len = sizeof_imsm_dev(dev_iter, 0); |
3413 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3414 | if (len_migr > len) | |
3415 | space_needed += len_migr - len; | |
ca9de185 | 3416 | |
503975b9 | 3417 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
3418 | if (max_len < len_migr) |
3419 | max_len = len_migr; | |
3420 | if (max_len > len_migr) | |
3421 | space_needed += max_len - len_migr; | |
503975b9 | 3422 | dev_new = xmalloc(max_len); |
949c47a0 | 3423 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3424 | dv->dev = dev_new; |
3425 | dv->index = i; | |
3426 | dv->next = super->devlist; | |
3427 | super->devlist = dv; | |
949c47a0 | 3428 | } |
cdddbdbc | 3429 | |
4d7b1503 DW |
3430 | /* ensure that super->buf is large enough when all raid devices |
3431 | * are migrating | |
3432 | */ | |
3433 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3434 | void *buf; | |
3435 | ||
3436 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3437 | if (posix_memalign(&buf, 512, len) != 0) | |
3438 | return 1; | |
3439 | ||
1f45a8ad DW |
3440 | memcpy(buf, super->buf, super->len); |
3441 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3442 | free(super->buf); |
3443 | super->buf = buf; | |
3444 | super->len = len; | |
3445 | } | |
ca9de185 | 3446 | |
cdddbdbc DW |
3447 | return 0; |
3448 | } | |
3449 | ||
604b746f JD |
3450 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3451 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3452 | { | |
3453 | void *ptr = NULL; | |
3454 | ||
3455 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3456 | ptr = mpb; | |
3457 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3458 | } | |
3459 | ||
3460 | return ptr; | |
3461 | } | |
3462 | ||
e2f41b2c AK |
3463 | /******************************************************************************* |
3464 | * Function: check_mpb_migr_compatibility | |
3465 | * Description: Function checks for unsupported migration features: | |
3466 | * - migration optimization area (pba_of_lba0) | |
3467 | * - descending reshape (ascending_migr) | |
3468 | * Parameters: | |
3469 | * super : imsm metadata information | |
3470 | * Returns: | |
3471 | * 0 : migration is compatible | |
3472 | * -1 : migration is not compatible | |
3473 | ******************************************************************************/ | |
3474 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3475 | { | |
3476 | struct imsm_map *map0, *map1; | |
3477 | struct migr_record *migr_rec = super->migr_rec; | |
3478 | int i; | |
3479 | ||
3480 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3481 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3482 | ||
3483 | if (dev_iter && | |
3484 | dev_iter->vol.migr_state == 1 && | |
3485 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3486 | /* This device is migrating */ | |
238c0a71 AK |
3487 | map0 = get_imsm_map(dev_iter, MAP_0); |
3488 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 3489 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
3490 | /* migration optimization area was used */ |
3491 | return -1; | |
3492 | if (migr_rec->ascending_migr == 0 | |
3493 | && migr_rec->dest_depth_per_unit > 0) | |
3494 | /* descending reshape not supported yet */ | |
3495 | return -1; | |
3496 | } | |
3497 | } | |
3498 | return 0; | |
3499 | } | |
3500 | ||
d23fe947 | 3501 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3502 | |
cdddbdbc | 3503 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3504 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3505 | */ |
3506 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3507 | { | |
3508 | unsigned long long dsize; | |
cdddbdbc DW |
3509 | unsigned long long sectors; |
3510 | struct stat; | |
6416d527 | 3511 | struct imsm_super *anchor; |
cdddbdbc DW |
3512 | __u32 check_sum; |
3513 | ||
cdddbdbc | 3514 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3515 | if (dsize < 1024) { |
3516 | if (devname) | |
e7b84f9d N |
3517 | pr_err("%s: device to small for imsm\n", |
3518 | devname); | |
64436f06 N |
3519 | return 1; |
3520 | } | |
cdddbdbc DW |
3521 | |
3522 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3523 | if (devname) | |
e7b84f9d N |
3524 | pr_err("Cannot seek to anchor block on %s: %s\n", |
3525 | devname, strerror(errno)); | |
cdddbdbc DW |
3526 | return 1; |
3527 | } | |
3528 | ||
949c47a0 | 3529 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e | 3530 | if (devname) |
e7b84f9d N |
3531 | pr_err("Failed to allocate imsm anchor buffer" |
3532 | " on %s\n", devname); | |
ad97895e DW |
3533 | return 1; |
3534 | } | |
949c47a0 | 3535 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc | 3536 | if (devname) |
e7b84f9d N |
3537 | pr_err("Cannot read anchor block on %s: %s\n", |
3538 | devname, strerror(errno)); | |
6416d527 | 3539 | free(anchor); |
cdddbdbc DW |
3540 | return 1; |
3541 | } | |
3542 | ||
6416d527 | 3543 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 3544 | if (devname) |
e7b84f9d | 3545 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 3546 | free(anchor); |
cdddbdbc DW |
3547 | return 2; |
3548 | } | |
3549 | ||
d23fe947 | 3550 | __free_imsm(super, 0); |
f2f5c343 LM |
3551 | /* reload capability and hba */ |
3552 | ||
3553 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3554 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3555 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3556 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc | 3557 | if (devname) |
e7b84f9d N |
3558 | pr_err("unable to allocate %zu byte mpb buffer\n", |
3559 | super->len); | |
6416d527 | 3560 | free(anchor); |
cdddbdbc DW |
3561 | return 2; |
3562 | } | |
949c47a0 | 3563 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3564 | |
6416d527 NB |
3565 | sectors = mpb_sectors(anchor) - 1; |
3566 | free(anchor); | |
8e59f3d8 | 3567 | |
17a4eaf9 | 3568 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
e7b84f9d | 3569 | pr_err("%s could not allocate migr_rec buffer\n", __func__); |
8e59f3d8 AK |
3570 | free(super->buf); |
3571 | return 2; | |
3572 | } | |
51d83f5d | 3573 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 3574 | |
949c47a0 | 3575 | if (!sectors) { |
ecf45690 DW |
3576 | check_sum = __gen_imsm_checksum(super->anchor); |
3577 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3578 | if (devname) | |
e7b84f9d N |
3579 | pr_err("IMSM checksum %x != %x on %s\n", |
3580 | check_sum, | |
3581 | __le32_to_cpu(super->anchor->check_sum), | |
3582 | devname); | |
ecf45690 DW |
3583 | return 2; |
3584 | } | |
3585 | ||
a2b97981 | 3586 | return 0; |
949c47a0 | 3587 | } |
cdddbdbc DW |
3588 | |
3589 | /* read the extended mpb */ | |
3590 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3591 | if (devname) | |
e7b84f9d N |
3592 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
3593 | devname, strerror(errno)); | |
cdddbdbc DW |
3594 | return 1; |
3595 | } | |
3596 | ||
f21e18ca | 3597 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc | 3598 | if (devname) |
e7b84f9d N |
3599 | pr_err("Cannot read extended mpb on %s: %s\n", |
3600 | devname, strerror(errno)); | |
cdddbdbc DW |
3601 | return 2; |
3602 | } | |
3603 | ||
949c47a0 DW |
3604 | check_sum = __gen_imsm_checksum(super->anchor); |
3605 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 3606 | if (devname) |
e7b84f9d N |
3607 | pr_err("IMSM checksum %x != %x on %s\n", |
3608 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
3609 | devname); | |
db575f3b | 3610 | return 3; |
cdddbdbc DW |
3611 | } |
3612 | ||
604b746f JD |
3613 | /* FIXME the BBM log is disk specific so we cannot use this global |
3614 | * buffer for all disks. Ok for now since we only look at the global | |
3615 | * bbm_log_size parameter to gate assembly | |
3616 | */ | |
3617 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3618 | ||
a2b97981 DW |
3619 | return 0; |
3620 | } | |
3621 | ||
8e59f3d8 AK |
3622 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3623 | ||
97f81ee2 CA |
3624 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
3625 | static void clear_hi(struct intel_super *super) | |
3626 | { | |
3627 | struct imsm_super *mpb = super->anchor; | |
3628 | int i, n; | |
3629 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
3630 | return; | |
3631 | for (i = 0; i < mpb->num_disks; ++i) { | |
3632 | struct imsm_disk *disk = &mpb->disk[i]; | |
3633 | disk->total_blocks_hi = 0; | |
3634 | } | |
3635 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
3636 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3637 | if (!dev) | |
3638 | return; | |
3639 | for (n = 0; n < 2; ++n) { | |
3640 | struct imsm_map *map = get_imsm_map(dev, n); | |
3641 | if (!map) | |
3642 | continue; | |
3643 | map->pba_of_lba0_hi = 0; | |
3644 | map->blocks_per_member_hi = 0; | |
3645 | map->num_data_stripes_hi = 0; | |
3646 | } | |
3647 | } | |
3648 | } | |
3649 | ||
a2b97981 DW |
3650 | static int |
3651 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3652 | { | |
3653 | int err; | |
3654 | ||
3655 | err = load_imsm_mpb(fd, super, devname); | |
3656 | if (err) | |
3657 | return err; | |
3658 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3659 | if (err) | |
3660 | return err; | |
3661 | err = parse_raid_devices(super); | |
97f81ee2 | 3662 | clear_hi(super); |
a2b97981 | 3663 | return err; |
cdddbdbc DW |
3664 | } |
3665 | ||
ae6aad82 DW |
3666 | static void __free_imsm_disk(struct dl *d) |
3667 | { | |
3668 | if (d->fd >= 0) | |
3669 | close(d->fd); | |
3670 | if (d->devname) | |
3671 | free(d->devname); | |
0dcecb2e DW |
3672 | if (d->e) |
3673 | free(d->e); | |
ae6aad82 DW |
3674 | free(d); |
3675 | ||
3676 | } | |
1a64be56 | 3677 | |
cdddbdbc DW |
3678 | static void free_imsm_disks(struct intel_super *super) |
3679 | { | |
47ee5a45 | 3680 | struct dl *d; |
cdddbdbc | 3681 | |
47ee5a45 DW |
3682 | while (super->disks) { |
3683 | d = super->disks; | |
cdddbdbc | 3684 | super->disks = d->next; |
ae6aad82 | 3685 | __free_imsm_disk(d); |
cdddbdbc | 3686 | } |
cb82edca AK |
3687 | while (super->disk_mgmt_list) { |
3688 | d = super->disk_mgmt_list; | |
3689 | super->disk_mgmt_list = d->next; | |
3690 | __free_imsm_disk(d); | |
3691 | } | |
47ee5a45 DW |
3692 | while (super->missing) { |
3693 | d = super->missing; | |
3694 | super->missing = d->next; | |
3695 | __free_imsm_disk(d); | |
3696 | } | |
3697 | ||
cdddbdbc DW |
3698 | } |
3699 | ||
9ca2c81c | 3700 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3701 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3702 | { |
88654014 LM |
3703 | struct intel_hba *elem, *next; |
3704 | ||
9ca2c81c | 3705 | if (super->buf) { |
949c47a0 | 3706 | free(super->buf); |
9ca2c81c DW |
3707 | super->buf = NULL; |
3708 | } | |
f2f5c343 LM |
3709 | /* unlink capability description */ |
3710 | super->orom = NULL; | |
8e59f3d8 AK |
3711 | if (super->migr_rec_buf) { |
3712 | free(super->migr_rec_buf); | |
3713 | super->migr_rec_buf = NULL; | |
3714 | } | |
d23fe947 DW |
3715 | if (free_disks) |
3716 | free_imsm_disks(super); | |
ba2de7ba | 3717 | free_devlist(super); |
88654014 LM |
3718 | elem = super->hba; |
3719 | while (elem) { | |
3720 | if (elem->path) | |
3721 | free((void *)elem->path); | |
3722 | next = elem->next; | |
3723 | free(elem); | |
3724 | elem = next; | |
88c32bb1 | 3725 | } |
88654014 | 3726 | super->hba = NULL; |
cdddbdbc DW |
3727 | } |
3728 | ||
9ca2c81c DW |
3729 | static void free_imsm(struct intel_super *super) |
3730 | { | |
d23fe947 | 3731 | __free_imsm(super, 1); |
9ca2c81c DW |
3732 | free(super); |
3733 | } | |
cdddbdbc DW |
3734 | |
3735 | static void free_super_imsm(struct supertype *st) | |
3736 | { | |
3737 | struct intel_super *super = st->sb; | |
3738 | ||
3739 | if (!super) | |
3740 | return; | |
3741 | ||
3742 | free_imsm(super); | |
3743 | st->sb = NULL; | |
3744 | } | |
3745 | ||
49133e57 | 3746 | static struct intel_super *alloc_super(void) |
c2c087e6 | 3747 | { |
503975b9 | 3748 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 3749 | |
503975b9 N |
3750 | super->current_vol = -1; |
3751 | super->create_offset = ~((unsigned long long) 0); | |
c2c087e6 DW |
3752 | return super; |
3753 | } | |
3754 | ||
f0f5a016 LM |
3755 | /* |
3756 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3757 | */ | |
d424212e | 3758 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3759 | { |
3760 | struct sys_dev *hba_name; | |
3761 | int rv = 0; | |
3762 | ||
3763 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3764 | super->orom = NULL; |
f0f5a016 LM |
3765 | super->hba = NULL; |
3766 | return 0; | |
3767 | } | |
3768 | hba_name = find_disk_attached_hba(fd, NULL); | |
3769 | if (!hba_name) { | |
d424212e | 3770 | if (devname) |
e7b84f9d N |
3771 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
3772 | devname); | |
f0f5a016 LM |
3773 | return 1; |
3774 | } | |
3775 | rv = attach_hba_to_super(super, hba_name); | |
3776 | if (rv == 2) { | |
d424212e N |
3777 | if (devname) { |
3778 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3779 | |
e7b84f9d | 3780 | pr_err("%s is attached to Intel(R) %s RAID " |
f0f5a016 LM |
3781 | "controller (%s),\n" |
3782 | " but the container is assigned to Intel(R) " | |
3783 | "%s RAID controller (", | |
d424212e | 3784 | devname, |
f0f5a016 LM |
3785 | hba_name->path, |
3786 | hba_name->pci_id ? : "Err!", | |
3787 | get_sys_dev_type(hba_name->type)); | |
3788 | ||
f0f5a016 LM |
3789 | while (hba) { |
3790 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3791 | if (hba->next) | |
3792 | fprintf(stderr, ", "); | |
3793 | hba = hba->next; | |
3794 | } | |
3795 | ||
3796 | fprintf(stderr, ").\n" | |
3c309c82 | 3797 | " Mixing devices attached to multiple controllers " |
f0f5a016 LM |
3798 | "is not allowed.\n"); |
3799 | } | |
f0f5a016 LM |
3800 | return 2; |
3801 | } | |
f2f5c343 | 3802 | super->orom = find_imsm_capability(hba_name->type); |
f2f5c343 LM |
3803 | if (!super->orom) |
3804 | return 3; | |
f0f5a016 LM |
3805 | return 0; |
3806 | } | |
3807 | ||
47ee5a45 DW |
3808 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3809 | * disks that have disappeared from the system. This routine relies on | |
3810 | * the mpb being uptodate, which it is at load time. | |
3811 | */ | |
3812 | static int find_missing(struct intel_super *super) | |
3813 | { | |
3814 | int i; | |
3815 | struct imsm_super *mpb = super->anchor; | |
3816 | struct dl *dl; | |
3817 | struct imsm_disk *disk; | |
47ee5a45 DW |
3818 | |
3819 | for (i = 0; i < mpb->num_disks; i++) { | |
3820 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3821 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3822 | if (dl) |
3823 | continue; | |
47ee5a45 | 3824 | |
503975b9 | 3825 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
3826 | dl->major = 0; |
3827 | dl->minor = 0; | |
3828 | dl->fd = -1; | |
503975b9 | 3829 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
3830 | dl->index = i; |
3831 | serialcpy(dl->serial, disk->serial); | |
3832 | dl->disk = *disk; | |
689c9bf3 | 3833 | dl->e = NULL; |
47ee5a45 DW |
3834 | dl->next = super->missing; |
3835 | super->missing = dl; | |
3836 | } | |
3837 | ||
3838 | return 0; | |
3839 | } | |
3840 | ||
3960e579 | 3841 | #ifndef MDASSEMBLE |
a2b97981 DW |
3842 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3843 | { | |
3844 | struct intel_disk *idisk = disk_list; | |
3845 | ||
3846 | while (idisk) { | |
3847 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3848 | break; | |
3849 | idisk = idisk->next; | |
3850 | } | |
3851 | ||
3852 | return idisk; | |
3853 | } | |
3854 | ||
3855 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3856 | struct intel_super *super, | |
3857 | struct intel_disk **disk_list) | |
3858 | { | |
3859 | struct imsm_disk *d = &super->disks->disk; | |
3860 | struct imsm_super *mpb = super->anchor; | |
3861 | int i, j; | |
3862 | ||
3863 | for (i = 0; i < tbl_size; i++) { | |
3864 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3865 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3866 | ||
3867 | if (tbl_mpb->family_num == mpb->family_num) { | |
3868 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3869 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3870 | __func__, super->disks->major, | |
3871 | super->disks->minor, | |
3872 | table[i]->disks->major, | |
3873 | table[i]->disks->minor); | |
3874 | break; | |
3875 | } | |
3876 | ||
3877 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3878 | is_configured(d) == is_configured(tbl_d)) && | |
3879 | tbl_mpb->generation_num < mpb->generation_num) { | |
3880 | /* current version of the mpb is a | |
3881 | * better candidate than the one in | |
3882 | * super_table, but copy over "cross | |
3883 | * generational" status | |
3884 | */ | |
3885 | struct intel_disk *idisk; | |
3886 | ||
3887 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3888 | __func__, super->disks->major, | |
3889 | super->disks->minor, | |
3890 | table[i]->disks->major, | |
3891 | table[i]->disks->minor); | |
3892 | ||
3893 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3894 | if (idisk && is_failed(&idisk->disk)) | |
3895 | tbl_d->status |= FAILED_DISK; | |
3896 | break; | |
3897 | } else { | |
3898 | struct intel_disk *idisk; | |
3899 | struct imsm_disk *disk; | |
3900 | ||
3901 | /* tbl_mpb is more up to date, but copy | |
3902 | * over cross generational status before | |
3903 | * returning | |
3904 | */ | |
3905 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3906 | if (disk && is_failed(disk)) | |
3907 | d->status |= FAILED_DISK; | |
3908 | ||
3909 | idisk = disk_list_get(d->serial, *disk_list); | |
3910 | if (idisk) { | |
3911 | idisk->owner = i; | |
3912 | if (disk && is_configured(disk)) | |
3913 | idisk->disk.status |= CONFIGURED_DISK; | |
3914 | } | |
3915 | ||
3916 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3917 | __func__, super->disks->major, | |
3918 | super->disks->minor, | |
3919 | table[i]->disks->major, | |
3920 | table[i]->disks->minor); | |
3921 | ||
3922 | return tbl_size; | |
3923 | } | |
3924 | } | |
3925 | } | |
3926 | ||
3927 | if (i >= tbl_size) | |
3928 | table[tbl_size++] = super; | |
3929 | else | |
3930 | table[i] = super; | |
3931 | ||
3932 | /* update/extend the merged list of imsm_disk records */ | |
3933 | for (j = 0; j < mpb->num_disks; j++) { | |
3934 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3935 | struct intel_disk *idisk; | |
3936 | ||
3937 | idisk = disk_list_get(disk->serial, *disk_list); | |
3938 | if (idisk) { | |
3939 | idisk->disk.status |= disk->status; | |
3940 | if (is_configured(&idisk->disk) || | |
3941 | is_failed(&idisk->disk)) | |
3942 | idisk->disk.status &= ~(SPARE_DISK); | |
3943 | } else { | |
503975b9 | 3944 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
3945 | idisk->owner = IMSM_UNKNOWN_OWNER; |
3946 | idisk->disk = *disk; | |
3947 | idisk->next = *disk_list; | |
3948 | *disk_list = idisk; | |
3949 | } | |
3950 | ||
3951 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3952 | idisk->owner = i; | |
3953 | } | |
3954 | ||
3955 | return tbl_size; | |
3956 | } | |
3957 | ||
3958 | static struct intel_super * | |
3959 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3960 | const int owner) | |
3961 | { | |
3962 | struct imsm_super *mpb = super->anchor; | |
3963 | int ok_count = 0; | |
3964 | int i; | |
3965 | ||
3966 | for (i = 0; i < mpb->num_disks; i++) { | |
3967 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3968 | struct intel_disk *idisk; | |
3969 | ||
3970 | idisk = disk_list_get(disk->serial, disk_list); | |
3971 | if (idisk) { | |
3972 | if (idisk->owner == owner || | |
3973 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3974 | ok_count++; | |
3975 | else | |
3976 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3977 | __func__, disk->serial, idisk->owner, | |
3978 | owner); | |
3979 | } else { | |
3980 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
3981 | __func__, __le32_to_cpu(mpb->family_num), i, | |
3982 | disk->serial); | |
3983 | break; | |
3984 | } | |
3985 | } | |
3986 | ||
3987 | if (ok_count == mpb->num_disks) | |
3988 | return super; | |
3989 | return NULL; | |
3990 | } | |
3991 | ||
3992 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
3993 | { | |
3994 | struct intel_super *s; | |
3995 | ||
3996 | for (s = super_list; s; s = s->next) { | |
3997 | if (family_num != s->anchor->family_num) | |
3998 | continue; | |
3999 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
4000 | __le32_to_cpu(family_num), s->disks->devname); | |
4001 | } | |
4002 | } | |
4003 | ||
4004 | static struct intel_super * | |
4005 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4006 | { | |
4007 | struct intel_super *super_table[len]; | |
4008 | struct intel_disk *disk_list = NULL; | |
4009 | struct intel_super *champion, *spare; | |
4010 | struct intel_super *s, **del; | |
4011 | int tbl_size = 0; | |
4012 | int conflict; | |
4013 | int i; | |
4014 | ||
4015 | memset(super_table, 0, sizeof(super_table)); | |
4016 | for (s = *super_list; s; s = s->next) | |
4017 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4018 | ||
4019 | for (i = 0; i < tbl_size; i++) { | |
4020 | struct imsm_disk *d; | |
4021 | struct intel_disk *idisk; | |
4022 | struct imsm_super *mpb = super_table[i]->anchor; | |
4023 | ||
4024 | s = super_table[i]; | |
4025 | d = &s->disks->disk; | |
4026 | ||
4027 | /* 'd' must appear in merged disk list for its | |
4028 | * configuration to be valid | |
4029 | */ | |
4030 | idisk = disk_list_get(d->serial, disk_list); | |
4031 | if (idisk && idisk->owner == i) | |
4032 | s = validate_members(s, disk_list, i); | |
4033 | else | |
4034 | s = NULL; | |
4035 | ||
4036 | if (!s) | |
4037 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
4038 | __func__, mpb->family_num, | |
4039 | super_table[i]->disks->major, | |
4040 | super_table[i]->disks->minor); | |
4041 | super_table[i] = s; | |
4042 | } | |
4043 | ||
4044 | /* This is where the mdadm implementation differs from the Windows | |
4045 | * driver which has no strict concept of a container. We can only | |
4046 | * assemble one family from a container, so when returning a prodigal | |
4047 | * array member to this system the code will not be able to disambiguate | |
4048 | * the container contents that should be assembled ("foreign" versus | |
4049 | * "local"). It requires user intervention to set the orig_family_num | |
4050 | * to a new value to establish a new container. The Windows driver in | |
4051 | * this situation fixes up the volume name in place and manages the | |
4052 | * foreign array as an independent entity. | |
4053 | */ | |
4054 | s = NULL; | |
4055 | spare = NULL; | |
4056 | conflict = 0; | |
4057 | for (i = 0; i < tbl_size; i++) { | |
4058 | struct intel_super *tbl_ent = super_table[i]; | |
4059 | int is_spare = 0; | |
4060 | ||
4061 | if (!tbl_ent) | |
4062 | continue; | |
4063 | ||
4064 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4065 | spare = tbl_ent; | |
4066 | is_spare = 1; | |
4067 | } | |
4068 | ||
4069 | if (s && !is_spare) { | |
4070 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4071 | conflict++; | |
4072 | } else if (!s && !is_spare) | |
4073 | s = tbl_ent; | |
4074 | } | |
4075 | ||
4076 | if (!s) | |
4077 | s = spare; | |
4078 | if (!s) { | |
4079 | champion = NULL; | |
4080 | goto out; | |
4081 | } | |
4082 | champion = s; | |
4083 | ||
4084 | if (conflict) | |
4085 | fprintf(stderr, "Chose family %#x on '%s', " | |
4086 | "assemble conflicts to new container with '--update=uuid'\n", | |
4087 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
4088 | ||
4089 | /* collect all dl's onto 'champion', and update them to | |
4090 | * champion's version of the status | |
4091 | */ | |
4092 | for (s = *super_list; s; s = s->next) { | |
4093 | struct imsm_super *mpb = champion->anchor; | |
4094 | struct dl *dl = s->disks; | |
4095 | ||
4096 | if (s == champion) | |
4097 | continue; | |
4098 | ||
5d7b407a CA |
4099 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4100 | ||
a2b97981 DW |
4101 | for (i = 0; i < mpb->num_disks; i++) { |
4102 | struct imsm_disk *disk; | |
4103 | ||
4104 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4105 | if (disk) { | |
4106 | dl->disk = *disk; | |
4107 | /* only set index on disks that are a member of | |
4108 | * a populated contianer, i.e. one with | |
4109 | * raid_devs | |
4110 | */ | |
4111 | if (is_failed(&dl->disk)) | |
4112 | dl->index = -2; | |
4113 | else if (is_spare(&dl->disk)) | |
4114 | dl->index = -1; | |
4115 | break; | |
4116 | } | |
4117 | } | |
4118 | ||
4119 | if (i >= mpb->num_disks) { | |
4120 | struct intel_disk *idisk; | |
4121 | ||
4122 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4123 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4124 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4125 | dl->index = -1; | |
4126 | else { | |
4127 | dl->index = -2; | |
4128 | continue; | |
4129 | } | |
4130 | } | |
4131 | ||
4132 | dl->next = champion->disks; | |
4133 | champion->disks = dl; | |
4134 | s->disks = NULL; | |
4135 | } | |
4136 | ||
4137 | /* delete 'champion' from super_list */ | |
4138 | for (del = super_list; *del; ) { | |
4139 | if (*del == champion) { | |
4140 | *del = (*del)->next; | |
4141 | break; | |
4142 | } else | |
4143 | del = &(*del)->next; | |
4144 | } | |
4145 | champion->next = NULL; | |
4146 | ||
4147 | out: | |
4148 | while (disk_list) { | |
4149 | struct intel_disk *idisk = disk_list; | |
4150 | ||
4151 | disk_list = disk_list->next; | |
4152 | free(idisk); | |
4153 | } | |
4154 | ||
4155 | return champion; | |
4156 | } | |
4157 | ||
9587c373 LM |
4158 | |
4159 | static int | |
4160 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
9587c373 LM |
4161 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, |
4162 | int major, int minor, int keep_fd); | |
ec50f7b6 LM |
4163 | static int |
4164 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4165 | int *max, int keep_fd); | |
4166 | ||
9587c373 | 4167 | |
cdddbdbc | 4168 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4169 | char *devname, struct md_list *devlist, |
4170 | int keep_fd) | |
cdddbdbc | 4171 | { |
a2b97981 DW |
4172 | struct intel_super *super_list = NULL; |
4173 | struct intel_super *super = NULL; | |
a2b97981 | 4174 | int err = 0; |
9587c373 | 4175 | int i = 0; |
dab4a513 | 4176 | |
9587c373 LM |
4177 | if (fd >= 0) |
4178 | /* 'fd' is an opened container */ | |
4179 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4180 | else | |
ec50f7b6 LM |
4181 | /* get super block from devlist devices */ |
4182 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4183 | if (err) |
1602d52c | 4184 | goto error; |
a2b97981 DW |
4185 | /* all mpbs enter, maybe one leaves */ |
4186 | super = imsm_thunderdome(&super_list, i); | |
4187 | if (!super) { | |
4188 | err = 1; | |
4189 | goto error; | |
cdddbdbc DW |
4190 | } |
4191 | ||
47ee5a45 DW |
4192 | if (find_missing(super) != 0) { |
4193 | free_imsm(super); | |
a2b97981 DW |
4194 | err = 2; |
4195 | goto error; | |
47ee5a45 | 4196 | } |
8e59f3d8 AK |
4197 | |
4198 | /* load migration record */ | |
4199 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4200 | if (err == -1) { |
4201 | /* migration is in progress, | |
4202 | * but migr_rec cannot be loaded, | |
4203 | */ | |
8e59f3d8 AK |
4204 | err = 4; |
4205 | goto error; | |
4206 | } | |
e2f41b2c AK |
4207 | |
4208 | /* Check migration compatibility */ | |
4c965cc9 | 4209 | if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) { |
e7b84f9d | 4210 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4211 | if (devname) |
4212 | fprintf(stderr, " on %s\n", devname); | |
4213 | else | |
4214 | fprintf(stderr, " (IMSM).\n"); | |
4215 | ||
4216 | err = 5; | |
4217 | goto error; | |
4218 | } | |
4219 | ||
a2b97981 DW |
4220 | err = 0; |
4221 | ||
4222 | error: | |
4223 | while (super_list) { | |
4224 | struct intel_super *s = super_list; | |
4225 | ||
4226 | super_list = super_list->next; | |
4227 | free_imsm(s); | |
4228 | } | |
9587c373 | 4229 | |
a2b97981 DW |
4230 | |
4231 | if (err) | |
4232 | return err; | |
f7e7067b | 4233 | |
cdddbdbc | 4234 | *sbp = super; |
9587c373 LM |
4235 | if (fd >= 0) |
4236 | st->container_dev = fd2devnum(fd); | |
4237 | else | |
4238 | st->container_dev = NoMdDev; | |
a2b97981 | 4239 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4240 | st->ss = &super_imsm; |
cdddbdbc DW |
4241 | st->minor_version = 0; |
4242 | st->max_devs = IMSM_MAX_DEVICES; | |
4243 | } | |
cdddbdbc DW |
4244 | return 0; |
4245 | } | |
2b959fbf | 4246 | |
9587c373 | 4247 | |
ec50f7b6 LM |
4248 | static int |
4249 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4250 | int *max, int keep_fd) | |
4251 | { | |
4252 | struct md_list *tmpdev; | |
4253 | int err = 0; | |
4254 | int i = 0; | |
9587c373 | 4255 | |
ec50f7b6 LM |
4256 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4257 | if (tmpdev->used != 1) | |
4258 | continue; | |
4259 | if (tmpdev->container == 1) { | |
ca9de185 | 4260 | int lmax = 0; |
ec50f7b6 LM |
4261 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4262 | if (fd < 0) { | |
e7b84f9d | 4263 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4264 | tmpdev->devname, strerror(errno)); |
4265 | err = 8; | |
4266 | goto error; | |
4267 | } | |
4268 | err = get_sra_super_block(fd, super_list, | |
4269 | tmpdev->devname, &lmax, | |
4270 | keep_fd); | |
4271 | i += lmax; | |
4272 | close(fd); | |
4273 | if (err) { | |
4274 | err = 7; | |
4275 | goto error; | |
4276 | } | |
4277 | } else { | |
4278 | int major = major(tmpdev->st_rdev); | |
4279 | int minor = minor(tmpdev->st_rdev); | |
4280 | err = get_super_block(super_list, | |
4281 | -1, | |
4282 | tmpdev->devname, | |
4283 | major, minor, | |
4284 | keep_fd); | |
4285 | i++; | |
4286 | if (err) { | |
4287 | err = 6; | |
4288 | goto error; | |
4289 | } | |
4290 | } | |
4291 | } | |
4292 | error: | |
4293 | *max = i; | |
4294 | return err; | |
4295 | } | |
9587c373 LM |
4296 | |
4297 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, | |
4298 | int major, int minor, int keep_fd) | |
4299 | { | |
4300 | struct intel_super*s = NULL; | |
4301 | char nm[32]; | |
4302 | int dfd = -1; | |
9587c373 LM |
4303 | int err = 0; |
4304 | int retry; | |
4305 | ||
4306 | s = alloc_super(); | |
4307 | if (!s) { | |
4308 | err = 1; | |
4309 | goto error; | |
4310 | } | |
4311 | ||
4312 | sprintf(nm, "%d:%d", major, minor); | |
4313 | dfd = dev_open(nm, O_RDWR); | |
4314 | if (dfd < 0) { | |
4315 | err = 2; | |
4316 | goto error; | |
4317 | } | |
4318 | ||
cb8f6859 | 4319 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
4320 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
4321 | ||
4322 | /* retry the load if we might have raced against mdmon */ | |
4323 | if (err == 3 && (devnum != -1) && mdmon_running(devnum)) | |
4324 | for (retry = 0; retry < 3; retry++) { | |
4325 | usleep(3000); | |
4326 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4327 | if (err != 3) | |
4328 | break; | |
4329 | } | |
4330 | error: | |
4331 | if (!err) { | |
4332 | s->next = *super_list; | |
4333 | *super_list = s; | |
4334 | } else { | |
4335 | if (s) | |
4336 | free(s); | |
36614e95 | 4337 | if (dfd >= 0) |
9587c373 LM |
4338 | close(dfd); |
4339 | } | |
4340 | if ((dfd >= 0) && (!keep_fd)) | |
4341 | close(dfd); | |
4342 | return err; | |
4343 | ||
4344 | } | |
4345 | ||
4346 | static int | |
4347 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4348 | { | |
4349 | struct mdinfo *sra; | |
4350 | int devnum; | |
4351 | struct mdinfo *sd; | |
4352 | int err = 0; | |
4353 | int i = 0; | |
4354 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
4355 | if (!sra) | |
4356 | return 1; | |
4357 | ||
4358 | if (sra->array.major_version != -1 || | |
4359 | sra->array.minor_version != -2 || | |
4360 | strcmp(sra->text_version, "imsm") != 0) { | |
4361 | err = 1; | |
4362 | goto error; | |
4363 | } | |
4364 | /* load all mpbs */ | |
4365 | devnum = fd2devnum(fd); | |
4366 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
4367 | if (get_super_block(super_list, devnum, devname, | |
4368 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { | |
4369 | err = 7; | |
4370 | goto error; | |
4371 | } | |
4372 | } | |
4373 | error: | |
4374 | sysfs_free(sra); | |
4375 | *max = i; | |
4376 | return err; | |
4377 | } | |
4378 | ||
2b959fbf N |
4379 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4380 | { | |
ec50f7b6 | 4381 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4382 | } |
cdddbdbc DW |
4383 | #endif |
4384 | ||
4385 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4386 | { | |
4387 | struct intel_super *super; | |
4388 | int rv; | |
4389 | ||
691c6ee1 N |
4390 | if (test_partition(fd)) |
4391 | /* IMSM not allowed on partitions */ | |
4392 | return 1; | |
4393 | ||
37424f13 DW |
4394 | free_super_imsm(st); |
4395 | ||
49133e57 | 4396 | super = alloc_super(); |
ea2bc72b LM |
4397 | /* Load hba and capabilities if they exist. |
4398 | * But do not preclude loading metadata in case capabilities or hba are | |
4399 | * non-compliant and ignore_hw_compat is set. | |
4400 | */ | |
d424212e | 4401 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4402 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4403 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 | 4404 | if (devname) |
e7b84f9d | 4405 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
4406 | free_imsm(super); |
4407 | return 2; | |
4408 | } | |
a2b97981 | 4409 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
4410 | |
4411 | if (rv) { | |
4412 | if (devname) | |
e7b84f9d N |
4413 | pr_err("Failed to load all information " |
4414 | "sections on %s\n", devname); | |
cdddbdbc DW |
4415 | free_imsm(super); |
4416 | return rv; | |
4417 | } | |
4418 | ||
4419 | st->sb = super; | |
4420 | if (st->ss == NULL) { | |
4421 | st->ss = &super_imsm; | |
4422 | st->minor_version = 0; | |
4423 | st->max_devs = IMSM_MAX_DEVICES; | |
4424 | } | |
8e59f3d8 AK |
4425 | |
4426 | /* load migration record */ | |
2e062e82 AK |
4427 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4428 | /* Check for unsupported migration features */ | |
4429 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 4430 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
4431 | if (devname) |
4432 | fprintf(stderr, " on %s\n", devname); | |
4433 | else | |
4434 | fprintf(stderr, " (IMSM).\n"); | |
4435 | return 3; | |
4436 | } | |
e2f41b2c AK |
4437 | } |
4438 | ||
cdddbdbc DW |
4439 | return 0; |
4440 | } | |
4441 | ||
ef6ffade DW |
4442 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4443 | { | |
4444 | if (info->level == 1) | |
4445 | return 128; | |
4446 | return info->chunk_size >> 9; | |
4447 | } | |
4448 | ||
5551b113 CA |
4449 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
4450 | unsigned long long size) | |
fcfd9599 | 4451 | { |
4025c288 | 4452 | if (info->level == 1) |
5551b113 | 4453 | return size * 2; |
4025c288 | 4454 | else |
5551b113 | 4455 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
4456 | } |
4457 | ||
4d1313e9 DW |
4458 | static void imsm_update_version_info(struct intel_super *super) |
4459 | { | |
4460 | /* update the version and attributes */ | |
4461 | struct imsm_super *mpb = super->anchor; | |
4462 | char *version; | |
4463 | struct imsm_dev *dev; | |
4464 | struct imsm_map *map; | |
4465 | int i; | |
4466 | ||
4467 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4468 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4469 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4470 | if (__le32_to_cpu(dev->size_high) > 0) |
4471 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4472 | ||
4473 | /* FIXME detect when an array spans a port multiplier */ | |
4474 | #if 0 | |
4475 | mpb->attributes |= MPB_ATTRIB_PM; | |
4476 | #endif | |
4477 | ||
4478 | if (mpb->num_raid_devs > 1 || | |
4479 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4480 | version = MPB_VERSION_ATTRIBS; | |
4481 | switch (get_imsm_raid_level(map)) { | |
4482 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4483 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4484 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4485 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4486 | } | |
4487 | } else { | |
4488 | if (map->num_members >= 5) | |
4489 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4490 | else if (dev->status == DEV_CLONE_N_GO) | |
4491 | version = MPB_VERSION_CNG; | |
4492 | else if (get_imsm_raid_level(map) == 5) | |
4493 | version = MPB_VERSION_RAID5; | |
4494 | else if (map->num_members >= 3) | |
4495 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4496 | else if (get_imsm_raid_level(map) == 1) | |
4497 | version = MPB_VERSION_RAID1; | |
4498 | else | |
4499 | version = MPB_VERSION_RAID0; | |
4500 | } | |
4501 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4502 | } | |
4503 | } | |
4504 | ||
aa534678 DW |
4505 | static int check_name(struct intel_super *super, char *name, int quiet) |
4506 | { | |
4507 | struct imsm_super *mpb = super->anchor; | |
4508 | char *reason = NULL; | |
4509 | int i; | |
4510 | ||
4511 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4512 | reason = "must be 16 characters or less"; | |
4513 | ||
4514 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4515 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4516 | ||
4517 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4518 | reason = "already exists"; | |
4519 | break; | |
4520 | } | |
4521 | } | |
4522 | ||
4523 | if (reason && !quiet) | |
e7b84f9d | 4524 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
4525 | |
4526 | return !reason; | |
4527 | } | |
4528 | ||
8b353278 DW |
4529 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4530 | unsigned long long size, char *name, | |
83cd1e97 N |
4531 | char *homehost, int *uuid, |
4532 | long long data_offset) | |
cdddbdbc | 4533 | { |
c2c087e6 DW |
4534 | /* We are creating a volume inside a pre-existing container. |
4535 | * so st->sb is already set. | |
4536 | */ | |
4537 | struct intel_super *super = st->sb; | |
949c47a0 | 4538 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4539 | struct intel_dev *dv; |
c2c087e6 DW |
4540 | struct imsm_dev *dev; |
4541 | struct imsm_vol *vol; | |
4542 | struct imsm_map *map; | |
4543 | int idx = mpb->num_raid_devs; | |
4544 | int i; | |
4545 | unsigned long long array_blocks; | |
2c092cad | 4546 | size_t size_old, size_new; |
5551b113 | 4547 | unsigned long long num_data_stripes; |
cdddbdbc | 4548 | |
88c32bb1 | 4549 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
e7b84f9d | 4550 | pr_err("This imsm-container already has the " |
88c32bb1 | 4551 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4552 | return 0; |
4553 | } | |
4554 | ||
2c092cad DW |
4555 | /* ensure the mpb is large enough for the new data */ |
4556 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4557 | size_new = disks_to_mpb_size(info->nr_disks); | |
4558 | if (size_new > size_old) { | |
4559 | void *mpb_new; | |
4560 | size_t size_round = ROUND_UP(size_new, 512); | |
4561 | ||
4562 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
e7b84f9d | 4563 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
4564 | return 0; |
4565 | } | |
17a4eaf9 AK |
4566 | if (posix_memalign(&super->migr_rec_buf, 512, |
4567 | MIGR_REC_BUF_SIZE) != 0) { | |
e7b84f9d N |
4568 | pr_err("%s could not allocate migr_rec buffer\n", |
4569 | __func__); | |
8e59f3d8 AK |
4570 | free(super->buf); |
4571 | free(super); | |
ea944c8f | 4572 | free(mpb_new); |
8e59f3d8 AK |
4573 | return 0; |
4574 | } | |
2c092cad DW |
4575 | memcpy(mpb_new, mpb, size_old); |
4576 | free(mpb); | |
4577 | mpb = mpb_new; | |
949c47a0 | 4578 | super->anchor = mpb_new; |
2c092cad DW |
4579 | mpb->mpb_size = __cpu_to_le32(size_new); |
4580 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4581 | } | |
bf5a934a | 4582 | super->current_vol = idx; |
3960e579 DW |
4583 | |
4584 | /* handle 'failed_disks' by either: | |
4585 | * a) create dummy disk entries in the table if this the first | |
4586 | * volume in the array. We add them here as this is the only | |
4587 | * opportunity to add them. add_to_super_imsm_volume() | |
4588 | * handles the non-failed disks and continues incrementing | |
4589 | * mpb->num_disks. | |
4590 | * b) validate that 'failed_disks' matches the current number | |
4591 | * of missing disks if the container is populated | |
d23fe947 | 4592 | */ |
3960e579 | 4593 | if (super->current_vol == 0) { |
d23fe947 | 4594 | mpb->num_disks = 0; |
3960e579 DW |
4595 | for (i = 0; i < info->failed_disks; i++) { |
4596 | struct imsm_disk *disk; | |
4597 | ||
4598 | mpb->num_disks++; | |
4599 | disk = __get_imsm_disk(mpb, i); | |
4600 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4601 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4602 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4603 | "missing:%d", i); | |
4604 | } | |
4605 | find_missing(super); | |
4606 | } else { | |
4607 | int missing = 0; | |
4608 | struct dl *d; | |
4609 | ||
4610 | for (d = super->missing; d; d = d->next) | |
4611 | missing++; | |
4612 | if (info->failed_disks > missing) { | |
e7b84f9d | 4613 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
4614 | return 0; |
4615 | } | |
4616 | } | |
5a038140 | 4617 | |
aa534678 DW |
4618 | if (!check_name(super, name, 0)) |
4619 | return 0; | |
503975b9 N |
4620 | dv = xmalloc(sizeof(*dv)); |
4621 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 4622 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 4623 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 4624 | info->layout, info->chunk_size, |
5551b113 | 4625 | size * 2); |
979d38be DW |
4626 | /* round array size down to closest MB */ |
4627 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4628 | ||
c2c087e6 DW |
4629 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4630 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4631 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4632 | vol = &dev->vol; |
4633 | vol->migr_state = 0; | |
1484e727 | 4634 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4635 | vol->dirty = !info->state; |
f8f603f1 | 4636 | vol->curr_migr_unit = 0; |
238c0a71 | 4637 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
4638 | set_pba_of_lba0(map, super->create_offset); |
4639 | set_blocks_per_member(map, info_to_blocks_per_member(info, size)); | |
ef6ffade | 4640 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4641 | map->failed_disk_num = ~0; |
bf4442ab | 4642 | if (info->level > 0) |
fffaf1ff N |
4643 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
4644 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
4645 | else |
4646 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4647 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4648 | map->ddf = 1; |
ef6ffade DW |
4649 | |
4650 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4651 | free(dev); |
4652 | free(dv); | |
e7b84f9d | 4653 | pr_err("imsm does not support more than 2 disks" |
ef6ffade DW |
4654 | "in a raid1 volume\n"); |
4655 | return 0; | |
4656 | } | |
81062a36 DW |
4657 | |
4658 | map->raid_level = info->level; | |
4d1313e9 | 4659 | if (info->level == 10) { |
c2c087e6 | 4660 | map->raid_level = 1; |
4d1313e9 | 4661 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4662 | } else if (info->level == 1) |
4663 | map->num_domains = info->raid_disks; | |
4664 | else | |
ff596308 | 4665 | map->num_domains = 1; |
81062a36 | 4666 | |
5551b113 CA |
4667 | /* info->size is only int so use the 'size' parameter instead */ |
4668 | num_data_stripes = (size * 2) / info_to_blocks_per_strip(info); | |
4669 | num_data_stripes /= map->num_domains; | |
4670 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 4671 | |
c2c087e6 DW |
4672 | map->num_members = info->raid_disks; |
4673 | for (i = 0; i < map->num_members; i++) { | |
4674 | /* initialized in add_to_super */ | |
4eb26970 | 4675 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4676 | } |
949c47a0 | 4677 | mpb->num_raid_devs++; |
ba2de7ba DW |
4678 | |
4679 | dv->dev = dev; | |
4680 | dv->index = super->current_vol; | |
4681 | dv->next = super->devlist; | |
4682 | super->devlist = dv; | |
c2c087e6 | 4683 | |
4d1313e9 DW |
4684 | imsm_update_version_info(super); |
4685 | ||
c2c087e6 | 4686 | return 1; |
cdddbdbc DW |
4687 | } |
4688 | ||
bf5a934a DW |
4689 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4690 | unsigned long long size, char *name, | |
83cd1e97 N |
4691 | char *homehost, int *uuid, |
4692 | unsigned long long data_offset) | |
bf5a934a DW |
4693 | { |
4694 | /* This is primarily called by Create when creating a new array. | |
4695 | * We will then get add_to_super called for each component, and then | |
4696 | * write_init_super called to write it out to each device. | |
4697 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4698 | * array. | |
4699 | * To create on a pre-existing array a different method will be called. | |
4700 | * This one is just for fresh drives. | |
4701 | */ | |
4702 | struct intel_super *super; | |
4703 | struct imsm_super *mpb; | |
4704 | size_t mpb_size; | |
4d1313e9 | 4705 | char *version; |
bf5a934a | 4706 | |
83cd1e97 N |
4707 | if (data_offset != INVALID_SECTORS) { |
4708 | fprintf(stderr, Name ": data-offset not supported by imsm\n"); | |
4709 | return 0; | |
4710 | } | |
4711 | ||
bf5a934a | 4712 | if (st->sb) |
83cd1e97 N |
4713 | return init_super_imsm_volume(st, info, size, name, homehost, uuid, |
4714 | data_offset); | |
e683ca88 DW |
4715 | |
4716 | if (info) | |
4717 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4718 | else | |
4719 | mpb_size = 512; | |
bf5a934a | 4720 | |
49133e57 | 4721 | super = alloc_super(); |
e683ca88 | 4722 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4723 | free(super); |
e683ca88 DW |
4724 | super = NULL; |
4725 | } | |
4726 | if (!super) { | |
e7b84f9d | 4727 | pr_err("%s could not allocate superblock\n", __func__); |
bf5a934a DW |
4728 | return 0; |
4729 | } | |
17a4eaf9 | 4730 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
e7b84f9d | 4731 | pr_err("%s could not allocate migr_rec buffer\n", __func__); |
8e59f3d8 AK |
4732 | free(super->buf); |
4733 | free(super); | |
4734 | return 0; | |
4735 | } | |
e683ca88 | 4736 | memset(super->buf, 0, mpb_size); |
ef649044 | 4737 | mpb = super->buf; |
e683ca88 DW |
4738 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4739 | st->sb = super; | |
4740 | ||
4741 | if (info == NULL) { | |
4742 | /* zeroing superblock */ | |
4743 | return 0; | |
4744 | } | |
bf5a934a | 4745 | |
4d1313e9 DW |
4746 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4747 | ||
4748 | version = (char *) mpb->sig; | |
4749 | strcpy(version, MPB_SIGNATURE); | |
4750 | version += strlen(MPB_SIGNATURE); | |
4751 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4752 | |
bf5a934a DW |
4753 | return 1; |
4754 | } | |
4755 | ||
0e600426 | 4756 | #ifndef MDASSEMBLE |
f20c3968 | 4757 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4758 | int fd, char *devname) |
4759 | { | |
4760 | struct intel_super *super = st->sb; | |
d23fe947 | 4761 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4762 | struct imsm_disk *_disk; |
bf5a934a DW |
4763 | struct imsm_dev *dev; |
4764 | struct imsm_map *map; | |
3960e579 | 4765 | struct dl *dl, *df; |
4eb26970 | 4766 | int slot; |
bf5a934a | 4767 | |
949c47a0 | 4768 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4769 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4770 | |
208933a7 | 4771 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 4772 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
4773 | devname); |
4774 | return 1; | |
4775 | } | |
4776 | ||
efb30e7f DW |
4777 | if (fd == -1) { |
4778 | /* we're doing autolayout so grab the pre-marked (in | |
4779 | * validate_geometry) raid_disk | |
4780 | */ | |
4781 | for (dl = super->disks; dl; dl = dl->next) | |
4782 | if (dl->raiddisk == dk->raid_disk) | |
4783 | break; | |
4784 | } else { | |
4785 | for (dl = super->disks; dl ; dl = dl->next) | |
4786 | if (dl->major == dk->major && | |
4787 | dl->minor == dk->minor) | |
4788 | break; | |
4789 | } | |
d23fe947 | 4790 | |
208933a7 | 4791 | if (!dl) { |
e7b84f9d | 4792 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 4793 | return 1; |
208933a7 | 4794 | } |
bf5a934a | 4795 | |
d23fe947 DW |
4796 | /* add a pristine spare to the metadata */ |
4797 | if (dl->index < 0) { | |
4798 | dl->index = super->anchor->num_disks; | |
4799 | super->anchor->num_disks++; | |
4800 | } | |
4eb26970 DW |
4801 | /* Check the device has not already been added */ |
4802 | slot = get_imsm_disk_slot(map, dl->index); | |
4803 | if (slot >= 0 && | |
238c0a71 | 4804 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 4805 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
4806 | devname); |
4807 | return 1; | |
4808 | } | |
656b6b5a | 4809 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4810 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4811 | |
3960e579 DW |
4812 | /* update size of 'missing' disks to be at least as large as the |
4813 | * largest acitve member (we only have dummy missing disks when | |
4814 | * creating the first volume) | |
4815 | */ | |
4816 | if (super->current_vol == 0) { | |
4817 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
4818 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
4819 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
4820 | _disk = __get_imsm_disk(mpb, df->index); |
4821 | *_disk = df->disk; | |
4822 | } | |
4823 | } | |
4824 | ||
4825 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
4826 | for (df = super->missing; df; df = df->next) | |
4827 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 4828 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
4829 | |
4830 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
4831 | continue; | |
4832 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 4833 | if (is_gen_migration(dev)) { |
238c0a71 AK |
4834 | struct imsm_map *map2 = get_imsm_map(dev, |
4835 | MAP_1); | |
0a108d63 AK |
4836 | int slot2 = get_imsm_disk_slot(map2, df->index); |
4837 | if ((slot2 < map2->num_members) && | |
4838 | (slot2 >= 0)) { | |
1ace8403 | 4839 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
4840 | slot2, |
4841 | MAP_1); | |
1ace8403 AK |
4842 | if ((unsigned)df->index == |
4843 | ord_to_idx(ord2)) | |
4844 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 4845 | slot2, |
1ace8403 AK |
4846 | df->index | |
4847 | IMSM_ORD_REBUILD); | |
4848 | } | |
4849 | } | |
3960e579 DW |
4850 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
4851 | break; | |
4852 | } | |
4853 | ||
d23fe947 DW |
4854 | /* if we are creating the first raid device update the family number */ |
4855 | if (super->current_vol == 0) { | |
4856 | __u32 sum; | |
4857 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 4858 | |
3960e579 | 4859 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 4860 | if (!_dev || !_disk) { |
e7b84f9d | 4861 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
4862 | return 1; |
4863 | } | |
d23fe947 DW |
4864 | *_dev = *dev; |
4865 | *_disk = dl->disk; | |
148acb7b DW |
4866 | sum = random32(); |
4867 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4868 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4869 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4870 | } |
ca0748fa | 4871 | super->current_disk = dl; |
f20c3968 | 4872 | return 0; |
bf5a934a DW |
4873 | } |
4874 | ||
a8619d23 AK |
4875 | /* mark_spare() |
4876 | * Function marks disk as spare and restores disk serial | |
4877 | * in case it was previously marked as failed by takeover operation | |
4878 | * reruns: | |
4879 | * -1 : critical error | |
4880 | * 0 : disk is marked as spare but serial is not set | |
4881 | * 1 : success | |
4882 | */ | |
4883 | int mark_spare(struct dl *disk) | |
4884 | { | |
4885 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
4886 | int ret_val = -1; | |
4887 | ||
4888 | if (!disk) | |
4889 | return ret_val; | |
4890 | ||
4891 | ret_val = 0; | |
4892 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
4893 | /* Restore disk serial number, because takeover marks disk | |
4894 | * as failed and adds to serial ':0' before it becomes | |
4895 | * a spare disk. | |
4896 | */ | |
4897 | serialcpy(disk->serial, serial); | |
4898 | serialcpy(disk->disk.serial, serial); | |
4899 | ret_val = 1; | |
4900 | } | |
4901 | disk->disk.status = SPARE_DISK; | |
4902 | disk->index = -1; | |
4903 | ||
4904 | return ret_val; | |
4905 | } | |
88654014 | 4906 | |
f20c3968 | 4907 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
4908 | int fd, char *devname, |
4909 | unsigned long long data_offset) | |
cdddbdbc | 4910 | { |
c2c087e6 | 4911 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4912 | struct dl *dd; |
4913 | unsigned long long size; | |
f2f27e63 | 4914 | __u32 id; |
c2c087e6 DW |
4915 | int rv; |
4916 | struct stat stb; | |
4917 | ||
88654014 LM |
4918 | /* If we are on an RAID enabled platform check that the disk is |
4919 | * attached to the raid controller. | |
4920 | * We do not need to test disks attachment for container based additions, | |
4921 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4922 | */ |
d424212e | 4923 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4924 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4925 | if (rv != 0) { |
4926 | dprintf("capability: %p fd: %d ret: %d\n", | |
4927 | super->orom, fd, rv); | |
4928 | return 1; | |
88c32bb1 DW |
4929 | } |
4930 | ||
f20c3968 DW |
4931 | if (super->current_vol >= 0) |
4932 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4933 | |
c2c087e6 | 4934 | fstat(fd, &stb); |
503975b9 | 4935 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
4936 | dd->major = major(stb.st_rdev); |
4937 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 4938 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 4939 | dd->fd = fd; |
689c9bf3 | 4940 | dd->e = NULL; |
1a64be56 | 4941 | dd->action = DISK_ADD; |
c2c087e6 | 4942 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4943 | if (rv) { |
e7b84f9d | 4944 | pr_err("failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4945 | free(dd); |
0030e8d6 | 4946 | abort(); |
c2c087e6 DW |
4947 | } |
4948 | ||
c2c087e6 DW |
4949 | get_dev_size(fd, NULL, &size); |
4950 | size /= 512; | |
1f24f035 | 4951 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
4952 | set_total_blocks(&dd->disk, size); |
4953 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
4954 | struct imsm_super *mpb = super->anchor; | |
4955 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
4956 | } | |
a8619d23 | 4957 | mark_spare(dd); |
c2c087e6 | 4958 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4959 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4960 | else |
b9f594fe | 4961 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4962 | |
4963 | if (st->update_tail) { | |
1a64be56 LM |
4964 | dd->next = super->disk_mgmt_list; |
4965 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4966 | } else { |
4967 | dd->next = super->disks; | |
4968 | super->disks = dd; | |
ceaf0ee1 | 4969 | super->updates_pending++; |
43dad3d6 | 4970 | } |
f20c3968 DW |
4971 | |
4972 | return 0; | |
cdddbdbc DW |
4973 | } |
4974 | ||
1a64be56 LM |
4975 | |
4976 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4977 | { | |
4978 | struct intel_super *super = st->sb; | |
4979 | struct dl *dd; | |
4980 | ||
4981 | /* remove from super works only in mdmon - for communication | |
4982 | * manager - monitor. Check if communication memory buffer | |
4983 | * is prepared. | |
4984 | */ | |
4985 | if (!st->update_tail) { | |
e7b84f9d N |
4986 | pr_err("%s shall be used in mdmon context only" |
4987 | "(line %d).\n", __func__, __LINE__); | |
1a64be56 LM |
4988 | return 1; |
4989 | } | |
503975b9 | 4990 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
4991 | dd->major = dk->major; |
4992 | dd->minor = dk->minor; | |
1a64be56 | 4993 | dd->fd = -1; |
a8619d23 | 4994 | mark_spare(dd); |
1a64be56 LM |
4995 | dd->action = DISK_REMOVE; |
4996 | ||
4997 | dd->next = super->disk_mgmt_list; | |
4998 | super->disk_mgmt_list = dd; | |
4999 | ||
5000 | ||
5001 | return 0; | |
5002 | } | |
5003 | ||
f796af5d DW |
5004 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5005 | ||
5006 | static union { | |
5007 | char buf[512]; | |
5008 | struct imsm_super anchor; | |
5009 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 5010 | |
d23fe947 DW |
5011 | /* spare records have their own family number and do not have any defined raid |
5012 | * devices | |
5013 | */ | |
5014 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5015 | { | |
d23fe947 | 5016 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5017 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5018 | __u32 sum; |
5019 | struct dl *d; | |
5020 | ||
f796af5d DW |
5021 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
5022 | spare->generation_num = __cpu_to_le32(1UL), | |
5023 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
5024 | spare->num_disks = 1, | |
5025 | spare->num_raid_devs = 0, | |
5026 | spare->cache_size = mpb->cache_size, | |
5027 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
5028 | ||
5029 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5030 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5031 | |
5032 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5033 | if (d->index != -1) |
d23fe947 DW |
5034 | continue; |
5035 | ||
f796af5d | 5036 | spare->disk[0] = d->disk; |
027c374f CA |
5037 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5038 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5039 | ||
f796af5d DW |
5040 | sum = __gen_imsm_checksum(spare); |
5041 | spare->family_num = __cpu_to_le32(sum); | |
5042 | spare->orig_family_num = 0; | |
5043 | sum = __gen_imsm_checksum(spare); | |
5044 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5045 | |
f796af5d | 5046 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
5047 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
5048 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 5049 | return 1; |
d23fe947 DW |
5050 | } |
5051 | if (doclose) { | |
5052 | close(d->fd); | |
5053 | d->fd = -1; | |
5054 | } | |
5055 | } | |
5056 | ||
e74255d9 | 5057 | return 0; |
d23fe947 DW |
5058 | } |
5059 | ||
36988a3d | 5060 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5061 | { |
36988a3d | 5062 | struct intel_super *super = st->sb; |
949c47a0 | 5063 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5064 | struct dl *d; |
5065 | __u32 generation; | |
5066 | __u32 sum; | |
d23fe947 | 5067 | int spares = 0; |
949c47a0 | 5068 | int i; |
a48ac0a8 | 5069 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5070 | int num_disks = 0; |
146c6260 | 5071 | int clear_migration_record = 1; |
cdddbdbc | 5072 | |
c2c087e6 DW |
5073 | /* 'generation' is incremented everytime the metadata is written */ |
5074 | generation = __le32_to_cpu(mpb->generation_num); | |
5075 | generation++; | |
5076 | mpb->generation_num = __cpu_to_le32(generation); | |
5077 | ||
148acb7b DW |
5078 | /* fix up cases where previous mdadm releases failed to set |
5079 | * orig_family_num | |
5080 | */ | |
5081 | if (mpb->orig_family_num == 0) | |
5082 | mpb->orig_family_num = mpb->family_num; | |
5083 | ||
d23fe947 | 5084 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5085 | if (d->index == -1) |
d23fe947 | 5086 | spares++; |
36988a3d | 5087 | else { |
d23fe947 | 5088 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5089 | num_disks++; |
5090 | } | |
d23fe947 | 5091 | } |
36988a3d | 5092 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5093 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5094 | num_disks++; |
5095 | } | |
5096 | mpb->num_disks = num_disks; | |
5097 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5098 | |
949c47a0 DW |
5099 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5100 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5101 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5102 | if (dev && dev2) { | |
5103 | imsm_copy_dev(dev, dev2); | |
5104 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5105 | } | |
146c6260 AK |
5106 | if (is_gen_migration(dev2)) |
5107 | clear_migration_record = 0; | |
949c47a0 | 5108 | } |
a48ac0a8 DW |
5109 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
5110 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 5111 | |
c2c087e6 | 5112 | /* recalculate checksum */ |
949c47a0 | 5113 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5114 | mpb->check_sum = __cpu_to_le32(sum); |
5115 | ||
51d83f5d AK |
5116 | if (super->clean_migration_record_by_mdmon) { |
5117 | clear_migration_record = 1; | |
5118 | super->clean_migration_record_by_mdmon = 0; | |
5119 | } | |
146c6260 | 5120 | if (clear_migration_record) |
17a4eaf9 | 5121 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); |
146c6260 | 5122 | |
d23fe947 | 5123 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5124 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5125 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5126 | continue; |
30602f53 | 5127 | |
146c6260 AK |
5128 | if (clear_migration_record) { |
5129 | unsigned long long dsize; | |
5130 | ||
5131 | get_dev_size(d->fd, NULL, &dsize); | |
5132 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
17a4eaf9 AK |
5133 | if (write(d->fd, super->migr_rec_buf, |
5134 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
9e2d750d | 5135 | perror("Write migr_rec failed"); |
146c6260 AK |
5136 | } |
5137 | } | |
51d83f5d AK |
5138 | |
5139 | if (store_imsm_mpb(d->fd, mpb)) | |
5140 | fprintf(stderr, | |
5141 | "%s: failed for device %d:%d (fd: %d)%s\n", | |
5142 | __func__, d->major, d->minor, | |
5143 | d->fd, strerror(errno)); | |
5144 | ||
c2c087e6 DW |
5145 | if (doclose) { |
5146 | close(d->fd); | |
5147 | d->fd = -1; | |
5148 | } | |
5149 | } | |
5150 | ||
d23fe947 DW |
5151 | if (spares) |
5152 | return write_super_imsm_spares(super, doclose); | |
5153 | ||
e74255d9 | 5154 | return 0; |
c2c087e6 DW |
5155 | } |
5156 | ||
0e600426 | 5157 | |
9b1fb677 | 5158 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5159 | { |
5160 | size_t len; | |
5161 | struct imsm_update_create_array *u; | |
5162 | struct intel_super *super = st->sb; | |
9b1fb677 | 5163 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5164 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5165 | struct disk_info *inf; |
5166 | struct imsm_disk *disk; | |
5167 | int i; | |
43dad3d6 | 5168 | |
54c2c1ea DW |
5169 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5170 | sizeof(*inf) * map->num_members; | |
503975b9 | 5171 | u = xmalloc(len); |
43dad3d6 | 5172 | u->type = update_create_array; |
9b1fb677 | 5173 | u->dev_idx = dev_idx; |
43dad3d6 | 5174 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5175 | inf = get_disk_info(u); |
5176 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5177 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5178 | |
54c2c1ea DW |
5179 | disk = get_imsm_disk(super, idx); |
5180 | serialcpy(inf[i].serial, disk->serial); | |
5181 | } | |
43dad3d6 DW |
5182 | append_metadata_update(st, u, len); |
5183 | ||
5184 | return 0; | |
5185 | } | |
5186 | ||
1a64be56 | 5187 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5188 | { |
5189 | struct intel_super *super = st->sb; | |
5190 | size_t len; | |
1a64be56 | 5191 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5192 | |
1a64be56 | 5193 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5194 | return 0; |
5195 | ||
5196 | len = sizeof(*u); | |
503975b9 | 5197 | u = xmalloc(len); |
1a64be56 | 5198 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5199 | append_metadata_update(st, u, len); |
5200 | ||
5201 | return 0; | |
5202 | } | |
5203 | ||
c2c087e6 DW |
5204 | static int write_init_super_imsm(struct supertype *st) |
5205 | { | |
9b1fb677 DW |
5206 | struct intel_super *super = st->sb; |
5207 | int current_vol = super->current_vol; | |
5208 | ||
5209 | /* we are done with current_vol reset it to point st at the container */ | |
5210 | super->current_vol = -1; | |
5211 | ||
8273f55e | 5212 | if (st->update_tail) { |
43dad3d6 DW |
5213 | /* queue the recently created array / added disk |
5214 | * as a metadata update */ | |
43dad3d6 | 5215 | int rv; |
8273f55e | 5216 | |
43dad3d6 | 5217 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5218 | if (current_vol < 0) { |
1a64be56 LM |
5219 | /* in the mgmt (add/remove) disk case we are running |
5220 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5221 | */ |
1a64be56 | 5222 | return mgmt_disk(st); |
43dad3d6 | 5223 | } else |
9b1fb677 | 5224 | rv = create_array(st, current_vol); |
8273f55e | 5225 | |
43dad3d6 | 5226 | return rv; |
d682f344 N |
5227 | } else { |
5228 | struct dl *d; | |
5229 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 5230 | Kill(d->devname, NULL, 0, -1, 1); |
36988a3d | 5231 | return write_super_imsm(st, 1); |
d682f344 | 5232 | } |
cdddbdbc | 5233 | } |
0e600426 | 5234 | #endif |
cdddbdbc | 5235 | |
e683ca88 | 5236 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5237 | { |
e683ca88 DW |
5238 | struct intel_super *super = st->sb; |
5239 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5240 | |
e683ca88 | 5241 | if (!mpb) |
ad97895e DW |
5242 | return 1; |
5243 | ||
1799c9e8 | 5244 | #ifndef MDASSEMBLE |
e683ca88 | 5245 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5246 | #else |
5247 | return 1; | |
5248 | #endif | |
cdddbdbc DW |
5249 | } |
5250 | ||
0e600426 N |
5251 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
5252 | { | |
5253 | return __le32_to_cpu(mpb->bbm_log_size); | |
5254 | } | |
5255 | ||
5256 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
5257 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5258 | int layout, int raiddisks, int chunk, | |
af4348dd N |
5259 | unsigned long long size, |
5260 | unsigned long long data_offset, | |
5261 | char *dev, | |
2c514b71 NB |
5262 | unsigned long long *freesize, |
5263 | int verbose) | |
cdddbdbc | 5264 | { |
c2c087e6 DW |
5265 | int fd; |
5266 | unsigned long long ldsize; | |
f2f5c343 LM |
5267 | struct intel_super *super=NULL; |
5268 | int rv = 0; | |
cdddbdbc | 5269 | |
c2c087e6 DW |
5270 | if (level != LEVEL_CONTAINER) |
5271 | return 0; | |
5272 | if (!dev) | |
5273 | return 1; | |
5274 | ||
5275 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5276 | if (fd < 0) { | |
ba728be7 | 5277 | if (verbose > 0) |
e7b84f9d | 5278 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 5279 | dev, strerror(errno)); |
c2c087e6 DW |
5280 | return 0; |
5281 | } | |
5282 | if (!get_dev_size(fd, dev, &ldsize)) { | |
5283 | close(fd); | |
5284 | return 0; | |
5285 | } | |
f2f5c343 LM |
5286 | |
5287 | /* capabilities retrieve could be possible | |
5288 | * note that there is no fd for the disks in array. | |
5289 | */ | |
5290 | super = alloc_super(); | |
ba728be7 | 5291 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
5292 | if (rv != 0) { |
5293 | #if DEBUG | |
5294 | char str[256]; | |
5295 | fd2devname(fd, str); | |
5296 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
5297 | fd, str, super->orom, rv, raiddisks); | |
5298 | #endif | |
5299 | /* no orom/efi or non-intel hba of the disk */ | |
5300 | close(fd); | |
5301 | free_imsm(super); | |
5302 | return 0; | |
5303 | } | |
c2c087e6 | 5304 | close(fd); |
9126b9a8 CA |
5305 | if (super->orom) { |
5306 | if (raiddisks > super->orom->tds) { | |
5307 | if (verbose) | |
e7b84f9d | 5308 | pr_err("%d exceeds maximum number of" |
9126b9a8 CA |
5309 | " platform supported disks: %d\n", |
5310 | raiddisks, super->orom->tds); | |
5311 | free_imsm(super); | |
5312 | return 0; | |
5313 | } | |
5314 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
5315 | (ldsize >> 9) >> 32 > 0) { | |
5316 | if (verbose) | |
e7b84f9d | 5317 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
5318 | free_imsm(super); |
5319 | return 0; | |
5320 | } | |
f2f5c343 | 5321 | } |
c2c087e6 | 5322 | |
af4348dd | 5323 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 5324 | free_imsm(super); |
c2c087e6 DW |
5325 | |
5326 | return 1; | |
cdddbdbc DW |
5327 | } |
5328 | ||
0dcecb2e DW |
5329 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5330 | { | |
5331 | const unsigned long long base_start = e[*idx].start; | |
5332 | unsigned long long end = base_start + e[*idx].size; | |
5333 | int i; | |
5334 | ||
5335 | if (base_start == end) | |
5336 | return 0; | |
5337 | ||
5338 | *idx = *idx + 1; | |
5339 | for (i = *idx; i < num_extents; i++) { | |
5340 | /* extend overlapping extents */ | |
5341 | if (e[i].start >= base_start && | |
5342 | e[i].start <= end) { | |
5343 | if (e[i].size == 0) | |
5344 | return 0; | |
5345 | if (e[i].start + e[i].size > end) | |
5346 | end = e[i].start + e[i].size; | |
5347 | } else if (e[i].start > end) { | |
5348 | *idx = i; | |
5349 | break; | |
5350 | } | |
5351 | } | |
5352 | ||
5353 | return end - base_start; | |
5354 | } | |
5355 | ||
5356 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5357 | { | |
5358 | /* build a composite disk with all known extents and generate a new | |
5359 | * 'maxsize' given the "all disks in an array must share a common start | |
5360 | * offset" constraint | |
5361 | */ | |
503975b9 | 5362 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
5363 | struct dl *dl; |
5364 | int i, j; | |
5365 | int start_extent; | |
5366 | unsigned long long pos; | |
b9d77223 | 5367 | unsigned long long start = 0; |
0dcecb2e DW |
5368 | unsigned long long maxsize; |
5369 | unsigned long reserve; | |
5370 | ||
0dcecb2e DW |
5371 | /* coalesce and sort all extents. also, check to see if we need to |
5372 | * reserve space between member arrays | |
5373 | */ | |
5374 | j = 0; | |
5375 | for (dl = super->disks; dl; dl = dl->next) { | |
5376 | if (!dl->e) | |
5377 | continue; | |
5378 | for (i = 0; i < dl->extent_cnt; i++) | |
5379 | e[j++] = dl->e[i]; | |
5380 | } | |
5381 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5382 | ||
5383 | /* merge extents */ | |
5384 | i = 0; | |
5385 | j = 0; | |
5386 | while (i < sum_extents) { | |
5387 | e[j].start = e[i].start; | |
5388 | e[j].size = find_size(e, &i, sum_extents); | |
5389 | j++; | |
5390 | if (e[j-1].size == 0) | |
5391 | break; | |
5392 | } | |
5393 | ||
5394 | pos = 0; | |
5395 | maxsize = 0; | |
5396 | start_extent = 0; | |
5397 | i = 0; | |
5398 | do { | |
5399 | unsigned long long esize; | |
5400 | ||
5401 | esize = e[i].start - pos; | |
5402 | if (esize >= maxsize) { | |
5403 | maxsize = esize; | |
5404 | start = pos; | |
5405 | start_extent = i; | |
5406 | } | |
5407 | pos = e[i].start + e[i].size; | |
5408 | i++; | |
5409 | } while (e[i-1].size); | |
5410 | free(e); | |
5411 | ||
a7dd165b DW |
5412 | if (maxsize == 0) |
5413 | return 0; | |
5414 | ||
5415 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5416 | * container | |
5417 | */ | |
0dcecb2e DW |
5418 | if (start_extent > 0) |
5419 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5420 | else | |
5421 | reserve = 0; | |
5422 | ||
5423 | if (maxsize < reserve) | |
a7dd165b | 5424 | return 0; |
0dcecb2e | 5425 | |
5551b113 | 5426 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 5427 | if (start + reserve > super->create_offset) |
a7dd165b | 5428 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5429 | super->create_offset = start + reserve; |
5430 | ||
5431 | return maxsize - reserve; | |
5432 | } | |
5433 | ||
88c32bb1 DW |
5434 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5435 | { | |
5436 | if (level < 0 || level == 6 || level == 4) | |
5437 | return 0; | |
5438 | ||
5439 | /* if we have an orom prevent invalid raid levels */ | |
5440 | if (orom) | |
5441 | switch (level) { | |
5442 | case 0: return imsm_orom_has_raid0(orom); | |
5443 | case 1: | |
5444 | if (raiddisks > 2) | |
5445 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5446 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5447 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5448 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5449 | } |
5450 | else | |
5451 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5452 | ||
5453 | return 0; | |
5454 | } | |
5455 | ||
ca9de185 LM |
5456 | |
5457 | static int | |
5458 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
5459 | int dpa, int verbose) | |
5460 | { | |
5461 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
5462 | struct mdstat_ent *memb = NULL; | |
5463 | int count = 0; | |
5464 | int num = 0; | |
5465 | struct md_list *dv = NULL; | |
5466 | int found; | |
5467 | ||
5468 | for (memb = mdstat ; memb ; memb = memb->next) { | |
5469 | if (memb->metadata_version && | |
5470 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
5471 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
5472 | !is_subarray(memb->metadata_version+9) && | |
5473 | memb->members) { | |
5474 | struct dev_member *dev = memb->members; | |
5475 | int fd = -1; | |
5476 | while(dev && (fd < 0)) { | |
503975b9 N |
5477 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
5478 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
5479 | if (num > 0) | |
5480 | fd = open(path, O_RDONLY, 0); | |
5481 | if ((num <= 0) || (fd < 0)) { | |
5482 | pr_vrb(": Cannot open %s: %s\n", | |
5483 | dev->name, strerror(errno)); | |
ca9de185 | 5484 | } |
503975b9 | 5485 | free(path); |
ca9de185 LM |
5486 | dev = dev->next; |
5487 | } | |
5488 | found = 0; | |
5489 | if ((fd >= 0) && disk_attached_to_hba(fd, hba)) { | |
5490 | struct mdstat_ent *vol; | |
5491 | for (vol = mdstat ; vol ; vol = vol->next) { | |
5492 | if ((vol->active > 0) && | |
5493 | vol->metadata_version && | |
5494 | is_container_member(vol, memb->dev)) { | |
5495 | found++; | |
5496 | count++; | |
5497 | } | |
5498 | } | |
5499 | if (*devlist && (found < dpa)) { | |
503975b9 N |
5500 | dv = xcalloc(1, sizeof(*dv)); |
5501 | dv->devname = xmalloc(strlen(memb->dev) + strlen("/dev/") + 1); | |
5502 | sprintf(dv->devname, "%s%s", "/dev/", memb->dev); | |
5503 | dv->found = found; | |
5504 | dv->used = 0; | |
5505 | dv->next = *devlist; | |
5506 | *devlist = dv; | |
ca9de185 LM |
5507 | } |
5508 | } | |
5509 | if (fd >= 0) | |
5510 | close(fd); | |
5511 | } | |
5512 | } | |
5513 | free_mdstat(mdstat); | |
5514 | return count; | |
5515 | } | |
5516 | ||
5517 | #ifdef DEBUG_LOOP | |
5518 | static struct md_list* | |
5519 | get_loop_devices(void) | |
5520 | { | |
5521 | int i; | |
5522 | struct md_list *devlist = NULL; | |
5523 | struct md_list *dv = NULL; | |
5524 | ||
5525 | for(i = 0; i < 12; i++) { | |
503975b9 N |
5526 | dv = xcalloc(1, sizeof(*dv)); |
5527 | dv->devname = xmalloc(40); | |
ca9de185 LM |
5528 | sprintf(dv->devname, "/dev/loop%d", i); |
5529 | dv->next = devlist; | |
5530 | devlist = dv; | |
5531 | } | |
5532 | return devlist; | |
5533 | } | |
5534 | #endif | |
5535 | ||
5536 | static struct md_list* | |
5537 | get_devices(const char *hba_path) | |
5538 | { | |
5539 | struct md_list *devlist = NULL; | |
5540 | struct md_list *dv = NULL; | |
5541 | struct dirent *ent; | |
5542 | DIR *dir; | |
5543 | int err = 0; | |
5544 | ||
5545 | #if DEBUG_LOOP | |
5546 | devlist = get_loop_devices(); | |
5547 | return devlist; | |
5548 | #endif | |
5549 | /* scroll through /sys/dev/block looking for devices attached to | |
5550 | * this hba | |
5551 | */ | |
5552 | dir = opendir("/sys/dev/block"); | |
5553 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
5554 | int fd; | |
5555 | char buf[1024]; | |
5556 | int major, minor; | |
5557 | char *path = NULL; | |
5558 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
5559 | continue; | |
5560 | path = devt_to_devpath(makedev(major, minor)); | |
5561 | if (!path) | |
5562 | continue; | |
5563 | if (!path_attached_to_hba(path, hba_path)) { | |
5564 | free(path); | |
5565 | path = NULL; | |
5566 | continue; | |
5567 | } | |
5568 | free(path); | |
5569 | path = NULL; | |
5570 | fd = dev_open(ent->d_name, O_RDONLY); | |
5571 | if (fd >= 0) { | |
5572 | fd2devname(fd, buf); | |
5573 | close(fd); | |
5574 | } else { | |
e7b84f9d | 5575 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
5576 | ent->d_name); |
5577 | continue; | |
5578 | } | |
5579 | ||
5580 | ||
503975b9 N |
5581 | dv = xcalloc(1, sizeof(*dv)); |
5582 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
5583 | dv->next = devlist; |
5584 | devlist = dv; | |
5585 | } | |
5586 | if (err) { | |
5587 | while(devlist) { | |
5588 | dv = devlist; | |
5589 | devlist = devlist->next; | |
5590 | free(dv->devname); | |
5591 | free(dv); | |
5592 | } | |
5593 | } | |
562aa102 | 5594 | closedir(dir); |
ca9de185 LM |
5595 | return devlist; |
5596 | } | |
5597 | ||
5598 | static int | |
5599 | count_volumes_list(struct md_list *devlist, char *homehost, | |
5600 | int verbose, int *found) | |
5601 | { | |
5602 | struct md_list *tmpdev; | |
5603 | int count = 0; | |
5604 | struct supertype *st = NULL; | |
5605 | ||
5606 | /* first walk the list of devices to find a consistent set | |
5607 | * that match the criterea, if that is possible. | |
5608 | * We flag the ones we like with 'used'. | |
5609 | */ | |
5610 | *found = 0; | |
5611 | st = match_metadata_desc_imsm("imsm"); | |
5612 | if (st == NULL) { | |
5613 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5614 | return 0; | |
5615 | } | |
5616 | ||
5617 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5618 | char *devname = tmpdev->devname; | |
5619 | struct stat stb; | |
5620 | struct supertype *tst; | |
5621 | int dfd; | |
5622 | if (tmpdev->used > 1) | |
5623 | continue; | |
5624 | tst = dup_super(st); | |
5625 | if (tst == NULL) { | |
5626 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5627 | goto err_1; | |
5628 | } | |
5629 | tmpdev->container = 0; | |
5630 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
5631 | if (dfd < 0) { | |
5632 | dprintf(": cannot open device %s: %s\n", | |
5633 | devname, strerror(errno)); | |
5634 | tmpdev->used = 2; | |
5635 | } else if (fstat(dfd, &stb)< 0) { | |
5636 | /* Impossible! */ | |
5637 | dprintf(": fstat failed for %s: %s\n", | |
5638 | devname, strerror(errno)); | |
5639 | tmpdev->used = 2; | |
5640 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
5641 | dprintf(": %s is not a block device.\n", | |
5642 | devname); | |
5643 | tmpdev->used = 2; | |
5644 | } else if (must_be_container(dfd)) { | |
5645 | struct supertype *cst; | |
5646 | cst = super_by_fd(dfd, NULL); | |
5647 | if (cst == NULL) { | |
5648 | dprintf(": cannot recognize container type %s\n", | |
5649 | devname); | |
5650 | tmpdev->used = 2; | |
5651 | } else if (tst->ss != st->ss) { | |
5652 | dprintf(": non-imsm container - ignore it: %s\n", | |
5653 | devname); | |
5654 | tmpdev->used = 2; | |
5655 | } else if (!tst->ss->load_container || | |
5656 | tst->ss->load_container(tst, dfd, NULL)) | |
5657 | tmpdev->used = 2; | |
5658 | else { | |
5659 | tmpdev->container = 1; | |
5660 | } | |
5661 | if (cst) | |
5662 | cst->ss->free_super(cst); | |
5663 | } else { | |
5664 | tmpdev->st_rdev = stb.st_rdev; | |
5665 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
5666 | dprintf(": no RAID superblock on %s\n", | |
5667 | devname); | |
5668 | tmpdev->used = 2; | |
5669 | } else if (tst->ss->compare_super == NULL) { | |
5670 | dprintf(": Cannot assemble %s metadata on %s\n", | |
5671 | tst->ss->name, devname); | |
5672 | tmpdev->used = 2; | |
5673 | } | |
5674 | } | |
5675 | if (dfd >= 0) | |
5676 | close(dfd); | |
5677 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
5678 | /* Ignore unrecognised devices during auto-assembly */ | |
5679 | goto loop; | |
5680 | } | |
5681 | else { | |
5682 | struct mdinfo info; | |
5683 | tst->ss->getinfo_super(tst, &info, NULL); | |
5684 | ||
5685 | if (st->minor_version == -1) | |
5686 | st->minor_version = tst->minor_version; | |
5687 | ||
5688 | if (memcmp(info.uuid, uuid_zero, | |
5689 | sizeof(int[4])) == 0) { | |
5690 | /* this is a floating spare. It cannot define | |
5691 | * an array unless there are no more arrays of | |
5692 | * this type to be found. It can be included | |
5693 | * in an array of this type though. | |
5694 | */ | |
5695 | tmpdev->used = 3; | |
5696 | goto loop; | |
5697 | } | |
5698 | ||
5699 | if (st->ss != tst->ss || | |
5700 | st->minor_version != tst->minor_version || | |
5701 | st->ss->compare_super(st, tst) != 0) { | |
5702 | /* Some mismatch. If exactly one array matches this host, | |
5703 | * we can resolve on that one. | |
5704 | * Or, if we are auto assembling, we just ignore the second | |
5705 | * for now. | |
5706 | */ | |
5707 | dprintf(": superblock on %s doesn't match others - assembly aborted\n", | |
5708 | devname); | |
5709 | goto loop; | |
5710 | } | |
5711 | tmpdev->used = 1; | |
5712 | *found = 1; | |
5713 | dprintf("found: devname: %s\n", devname); | |
5714 | } | |
5715 | loop: | |
5716 | if (tst) | |
5717 | tst->ss->free_super(tst); | |
5718 | } | |
5719 | if (*found != 0) { | |
5720 | int err; | |
5721 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
5722 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
5723 | for (iter = head; iter; iter = iter->next) { | |
5724 | dprintf("content->text_version: %s vol\n", | |
5725 | iter->text_version); | |
5726 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
5727 | /* do not assemble arrays with unsupported | |
5728 | configurations */ | |
5729 | dprintf(": Cannot activate member %s.\n", | |
5730 | iter->text_version); | |
5731 | } else | |
5732 | count++; | |
5733 | } | |
5734 | sysfs_free(head); | |
5735 | ||
5736 | } else { | |
5737 | dprintf(" no valid super block on device list: err: %d %p\n", | |
5738 | err, st->sb); | |
5739 | } | |
5740 | } else { | |
5741 | dprintf(" no more devices to examin\n"); | |
5742 | } | |
5743 | ||
5744 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5745 | if ((tmpdev->used == 1) && (tmpdev->found)) { | |
5746 | if (count) { | |
5747 | if (count < tmpdev->found) | |
5748 | count = 0; | |
5749 | else | |
5750 | count -= tmpdev->found; | |
5751 | } | |
5752 | } | |
5753 | if (tmpdev->used == 1) | |
5754 | tmpdev->used = 4; | |
5755 | } | |
5756 | err_1: | |
5757 | if (st) | |
5758 | st->ss->free_super(st); | |
5759 | return count; | |
5760 | } | |
5761 | ||
5762 | ||
5763 | static int | |
5764 | count_volumes(char *hba, int dpa, int verbose) | |
5765 | { | |
5766 | struct md_list *devlist = NULL; | |
5767 | int count = 0; | |
5768 | int found = 0;; | |
5769 | ||
5770 | devlist = get_devices(hba); | |
5771 | /* if no intel devices return zero volumes */ | |
5772 | if (devlist == NULL) | |
5773 | return 0; | |
5774 | ||
5775 | count = active_arrays_by_format("imsm", hba, &devlist, dpa, verbose); | |
5776 | dprintf(" path: %s active arrays: %d\n", hba, count); | |
5777 | if (devlist == NULL) | |
5778 | return 0; | |
5779 | do { | |
5780 | found = 0; | |
5781 | count += count_volumes_list(devlist, | |
5782 | NULL, | |
5783 | verbose, | |
5784 | &found); | |
5785 | dprintf("found %d count: %d\n", found, count); | |
5786 | } while (found); | |
5787 | ||
5788 | dprintf("path: %s total number of volumes: %d\n", hba, count); | |
5789 | ||
5790 | while(devlist) { | |
5791 | struct md_list *dv = devlist; | |
5792 | devlist = devlist->next; | |
5793 | free(dv->devname); | |
5794 | free(dv); | |
5795 | } | |
5796 | return count; | |
5797 | } | |
5798 | ||
cd9d1ac7 DW |
5799 | static int imsm_default_chunk(const struct imsm_orom *orom) |
5800 | { | |
5801 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
5802 | * 128 if no platform detected | |
5803 | */ | |
5804 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
5805 | ||
5806 | return min(512, (1 << fs)); | |
5807 | } | |
73408129 | 5808 | |
6592ce37 DW |
5809 | static int |
5810 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 5811 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 5812 | { |
660260d0 DW |
5813 | /* check/set platform and metadata limits/defaults */ |
5814 | if (super->orom && raiddisks > super->orom->dpa) { | |
5815 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
5816 | super->orom->dpa); | |
73408129 LM |
5817 | return 0; |
5818 | } | |
5819 | ||
5d500228 | 5820 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 5821 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
5822 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
5823 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
5824 | return 0; | |
5825 | } | |
cd9d1ac7 DW |
5826 | |
5827 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
5828 | *chunk = imsm_default_chunk(super->orom); | |
5829 | ||
5830 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
5831 | pr_vrb(": platform does not support a chunk size of: " | |
5832 | "%d\n", *chunk); | |
5833 | return 0; | |
6592ce37 | 5834 | } |
cd9d1ac7 | 5835 | |
6592ce37 DW |
5836 | if (layout != imsm_level_to_layout(level)) { |
5837 | if (level == 5) | |
5838 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
5839 | else if (level == 10) | |
5840 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
5841 | else | |
5842 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
5843 | layout, level); | |
5844 | return 0; | |
5845 | } | |
2cc699af CA |
5846 | |
5847 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && chunk && | |
5848 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { | |
5849 | pr_vrb(": platform does not support a volume size over 2TB\n"); | |
5850 | return 0; | |
5851 | } | |
6592ce37 DW |
5852 | return 1; |
5853 | } | |
5854 | ||
c2c087e6 DW |
5855 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
5856 | * FIX ME add ahci details | |
5857 | */ | |
8b353278 | 5858 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 5859 | int layout, int raiddisks, int *chunk, |
af4348dd N |
5860 | unsigned long long size, |
5861 | unsigned long long data_offset, | |
5862 | char *dev, | |
2c514b71 NB |
5863 | unsigned long long *freesize, |
5864 | int verbose) | |
cdddbdbc | 5865 | { |
c2c087e6 DW |
5866 | struct stat stb; |
5867 | struct intel_super *super = st->sb; | |
b2916f25 | 5868 | struct imsm_super *mpb; |
c2c087e6 DW |
5869 | struct dl *dl; |
5870 | unsigned long long pos = 0; | |
5871 | unsigned long long maxsize; | |
5872 | struct extent *e; | |
5873 | int i; | |
cdddbdbc | 5874 | |
88c32bb1 DW |
5875 | /* We must have the container info already read in. */ |
5876 | if (!super) | |
c2c087e6 DW |
5877 | return 0; |
5878 | ||
b2916f25 JS |
5879 | mpb = super->anchor; |
5880 | ||
2cc699af | 5881 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
e7b84f9d | 5882 | pr_err("RAID gemetry validation failed. " |
d54559f0 | 5883 | "Cannot proceed with the action(s).\n"); |
c2c087e6 | 5884 | return 0; |
d54559f0 | 5885 | } |
c2c087e6 DW |
5886 | if (!dev) { |
5887 | /* General test: make sure there is space for | |
2da8544a DW |
5888 | * 'raiddisks' device extents of size 'size' at a given |
5889 | * offset | |
c2c087e6 | 5890 | */ |
e46273eb | 5891 | unsigned long long minsize = size; |
b7528a20 | 5892 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5893 | int dcnt = 0; |
5894 | if (minsize == 0) | |
5895 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5896 | for (dl = super->disks; dl ; dl = dl->next) { | |
5897 | int found = 0; | |
5898 | ||
bf5a934a | 5899 | pos = 0; |
c2c087e6 DW |
5900 | i = 0; |
5901 | e = get_extents(super, dl); | |
5902 | if (!e) continue; | |
5903 | do { | |
5904 | unsigned long long esize; | |
5905 | esize = e[i].start - pos; | |
5906 | if (esize >= minsize) | |
5907 | found = 1; | |
b7528a20 | 5908 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5909 | start_offset = pos; |
5910 | break; | |
5911 | } else if (found && pos != start_offset) { | |
5912 | found = 0; | |
5913 | break; | |
5914 | } | |
c2c087e6 DW |
5915 | pos = e[i].start + e[i].size; |
5916 | i++; | |
5917 | } while (e[i-1].size); | |
5918 | if (found) | |
5919 | dcnt++; | |
5920 | free(e); | |
5921 | } | |
5922 | if (dcnt < raiddisks) { | |
2c514b71 | 5923 | if (verbose) |
e7b84f9d | 5924 | pr_err("imsm: Not enough " |
2c514b71 NB |
5925 | "devices with space for this array " |
5926 | "(%d < %d)\n", | |
5927 | dcnt, raiddisks); | |
c2c087e6 DW |
5928 | return 0; |
5929 | } | |
5930 | return 1; | |
5931 | } | |
0dcecb2e | 5932 | |
c2c087e6 DW |
5933 | /* This device must be a member of the set */ |
5934 | if (stat(dev, &stb) < 0) | |
5935 | return 0; | |
5936 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
5937 | return 0; | |
5938 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
5939 | if (dl->major == (int)major(stb.st_rdev) && |
5940 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
5941 | break; |
5942 | } | |
5943 | if (!dl) { | |
2c514b71 | 5944 | if (verbose) |
e7b84f9d | 5945 | pr_err("%s is not in the " |
2c514b71 | 5946 | "same imsm set\n", dev); |
c2c087e6 | 5947 | return 0; |
a20d2ba5 DW |
5948 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
5949 | /* If a volume is present then the current creation attempt | |
5950 | * cannot incorporate new spares because the orom may not | |
5951 | * understand this configuration (all member disks must be | |
5952 | * members of each array in the container). | |
5953 | */ | |
e7b84f9d | 5954 | pr_err("%s is a spare and a volume" |
a20d2ba5 | 5955 | " is already defined for this container\n", dev); |
e7b84f9d | 5956 | pr_err("The option-rom requires all member" |
a20d2ba5 DW |
5957 | " disks to be a member of all volumes\n"); |
5958 | return 0; | |
5fe62b94 WD |
5959 | } else if (super->orom && mpb->num_raid_devs > 0 && |
5960 | mpb->num_disks != raiddisks) { | |
e7b84f9d | 5961 | pr_err("The option-rom requires all member" |
5fe62b94 WD |
5962 | " disks to be a member of all volumes\n"); |
5963 | return 0; | |
c2c087e6 | 5964 | } |
0dcecb2e DW |
5965 | |
5966 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
5967 | e = get_extents(super, dl); |
5968 | maxsize = 0; | |
5969 | i = 0; | |
0dcecb2e DW |
5970 | if (e) { |
5971 | do { | |
5972 | unsigned long long esize; | |
5973 | ||
5974 | esize = e[i].start - pos; | |
5975 | if (esize >= maxsize) | |
5976 | maxsize = esize; | |
5977 | pos = e[i].start + e[i].size; | |
5978 | i++; | |
5979 | } while (e[i-1].size); | |
5980 | dl->e = e; | |
5981 | dl->extent_cnt = i; | |
5982 | } else { | |
5983 | if (verbose) | |
e7b84f9d | 5984 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
5985 | dev); |
5986 | return 0; | |
5987 | } | |
5988 | if (maxsize < size) { | |
5989 | if (verbose) | |
e7b84f9d | 5990 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
5991 | dev, maxsize, size); |
5992 | return 0; | |
5993 | } | |
5994 | ||
5995 | /* count total number of extents for merge */ | |
5996 | i = 0; | |
5997 | for (dl = super->disks; dl; dl = dl->next) | |
5998 | if (dl->e) | |
5999 | i += dl->extent_cnt; | |
6000 | ||
6001 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
6002 | |
6003 | if (!check_env("IMSM_NO_PLATFORM") && | |
6004 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
e7b84f9d | 6005 | pr_err("attempting to create a second " |
3baa56ab LO |
6006 | "volume with size less then remaining space. " |
6007 | "Aborting...\n"); | |
6008 | return 0; | |
6009 | } | |
6010 | ||
a7dd165b | 6011 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
6012 | if (verbose) { |
6013 | if (maxsize == 0) | |
e7b84f9d | 6014 | pr_err("no free space" |
b3071342 LD |
6015 | " left on device. Aborting...\n"); |
6016 | else | |
e7b84f9d | 6017 | pr_err("not enough space" |
b3071342 LD |
6018 | " to create volume of given size" |
6019 | " (%llu < %llu). Aborting...\n", | |
6020 | maxsize, size); | |
6021 | } | |
0dcecb2e | 6022 | return 0; |
0dcecb2e DW |
6023 | } |
6024 | ||
c2c087e6 DW |
6025 | *freesize = maxsize; |
6026 | ||
ca9de185 LM |
6027 | if (super->orom) { |
6028 | int count = count_volumes(super->hba->path, | |
6029 | super->orom->dpa, verbose); | |
6030 | if (super->orom->vphba <= count) { | |
40110b90 | 6031 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6032 | super->orom->vphba); |
6033 | return 0; | |
6034 | } | |
6035 | } | |
c2c087e6 | 6036 | return 1; |
cdddbdbc DW |
6037 | } |
6038 | ||
13bcac90 | 6039 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
6040 | unsigned long long size, int chunk, |
6041 | unsigned long long *freesize) | |
6042 | { | |
6043 | struct intel_super *super = st->sb; | |
6044 | struct imsm_super *mpb = super->anchor; | |
6045 | struct dl *dl; | |
6046 | int i; | |
6047 | int extent_cnt; | |
6048 | struct extent *e; | |
6049 | unsigned long long maxsize; | |
6050 | unsigned long long minsize; | |
6051 | int cnt; | |
6052 | int used; | |
6053 | ||
6054 | /* find the largest common start free region of the possible disks */ | |
6055 | used = 0; | |
6056 | extent_cnt = 0; | |
6057 | cnt = 0; | |
6058 | for (dl = super->disks; dl; dl = dl->next) { | |
6059 | dl->raiddisk = -1; | |
6060 | ||
6061 | if (dl->index >= 0) | |
6062 | used++; | |
6063 | ||
6064 | /* don't activate new spares if we are orom constrained | |
6065 | * and there is already a volume active in the container | |
6066 | */ | |
6067 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
6068 | continue; | |
6069 | ||
6070 | e = get_extents(super, dl); | |
6071 | if (!e) | |
6072 | continue; | |
6073 | for (i = 1; e[i-1].size; i++) | |
6074 | ; | |
6075 | dl->e = e; | |
6076 | dl->extent_cnt = i; | |
6077 | extent_cnt += i; | |
6078 | cnt++; | |
6079 | } | |
6080 | ||
6081 | maxsize = merge_extents(super, extent_cnt); | |
6082 | minsize = size; | |
6083 | if (size == 0) | |
612e59d8 CA |
6084 | /* chunk is in K */ |
6085 | minsize = chunk * 2; | |
efb30e7f DW |
6086 | |
6087 | if (cnt < raiddisks || | |
6088 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6089 | maxsize < minsize || |
6090 | maxsize == 0) { | |
e7b84f9d | 6091 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
6092 | return 0; /* No enough free spaces large enough */ |
6093 | } | |
6094 | ||
6095 | if (size == 0) { | |
6096 | size = maxsize; | |
6097 | if (chunk) { | |
612e59d8 CA |
6098 | size /= 2 * chunk; |
6099 | size *= 2 * chunk; | |
efb30e7f | 6100 | } |
f878b242 LM |
6101 | maxsize = size; |
6102 | } | |
6103 | if (!check_env("IMSM_NO_PLATFORM") && | |
6104 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
e7b84f9d | 6105 | pr_err("attempting to create a second " |
f878b242 LM |
6106 | "volume with size less then remaining space. " |
6107 | "Aborting...\n"); | |
6108 | return 0; | |
efb30e7f | 6109 | } |
efb30e7f DW |
6110 | cnt = 0; |
6111 | for (dl = super->disks; dl; dl = dl->next) | |
6112 | if (dl->e) | |
6113 | dl->raiddisk = cnt++; | |
6114 | ||
6115 | *freesize = size; | |
6116 | ||
13bcac90 AK |
6117 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
6118 | ||
efb30e7f DW |
6119 | return 1; |
6120 | } | |
6121 | ||
13bcac90 AK |
6122 | static int reserve_space(struct supertype *st, int raiddisks, |
6123 | unsigned long long size, int chunk, | |
6124 | unsigned long long *freesize) | |
6125 | { | |
6126 | struct intel_super *super = st->sb; | |
6127 | struct dl *dl; | |
6128 | int cnt; | |
6129 | int rv = 0; | |
6130 | ||
6131 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
6132 | if (rv) { | |
6133 | cnt = 0; | |
6134 | for (dl = super->disks; dl; dl = dl->next) | |
6135 | if (dl->e) | |
6136 | dl->raiddisk = cnt++; | |
6137 | rv = 1; | |
6138 | } | |
6139 | ||
6140 | return rv; | |
6141 | } | |
6142 | ||
bf5a934a | 6143 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6144 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 6145 | unsigned long long data_offset, |
bf5a934a DW |
6146 | char *dev, unsigned long long *freesize, |
6147 | int verbose) | |
6148 | { | |
6149 | int fd, cfd; | |
6150 | struct mdinfo *sra; | |
20cbe8d2 | 6151 | int is_member = 0; |
bf5a934a | 6152 | |
d54559f0 LM |
6153 | /* load capability |
6154 | * if given unused devices create a container | |
bf5a934a DW |
6155 | * if given given devices in a container create a member volume |
6156 | */ | |
6157 | if (level == LEVEL_CONTAINER) { | |
6158 | /* Must be a fresh device to add to a container */ | |
6159 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 6160 | raiddisks, |
af4348dd N |
6161 | chunk?*chunk:0, |
6162 | size, data_offset, | |
bf5a934a DW |
6163 | dev, freesize, |
6164 | verbose); | |
6165 | } | |
9587c373 | 6166 | |
8592f29d | 6167 | if (!dev) { |
e91a3bad | 6168 | if (st->sb) { |
ca9de185 | 6169 | struct intel_super *super = st->sb; |
e91a3bad | 6170 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 6171 | raiddisks, chunk, size, |
e91a3bad LM |
6172 | verbose)) |
6173 | return 0; | |
efb30e7f DW |
6174 | /* we are being asked to automatically layout a |
6175 | * new volume based on the current contents of | |
6176 | * the container. If the the parameters can be | |
6177 | * satisfied reserve_space will record the disks, | |
6178 | * start offset, and size of the volume to be | |
6179 | * created. add_to_super and getinfo_super | |
6180 | * detect when autolayout is in progress. | |
6181 | */ | |
ca9de185 LM |
6182 | /* assuming that freesize is always given when array is |
6183 | created */ | |
6184 | if (super->orom && freesize) { | |
6185 | int count; | |
6186 | count = count_volumes(super->hba->path, | |
6187 | super->orom->dpa, verbose); | |
6188 | if (super->orom->vphba <= count) { | |
6189 | pr_vrb(": platform does not support more" | |
40110b90 | 6190 | " than %d raid volumes.\n", |
ca9de185 LM |
6191 | super->orom->vphba); |
6192 | return 0; | |
6193 | } | |
6194 | } | |
e91a3bad LM |
6195 | if (freesize) |
6196 | return reserve_space(st, raiddisks, size, | |
6197 | chunk?*chunk:0, freesize); | |
8592f29d N |
6198 | } |
6199 | return 1; | |
6200 | } | |
bf5a934a DW |
6201 | if (st->sb) { |
6202 | /* creating in a given container */ | |
6203 | return validate_geometry_imsm_volume(st, level, layout, | |
6204 | raiddisks, chunk, size, | |
af4348dd | 6205 | data_offset, |
bf5a934a DW |
6206 | dev, freesize, verbose); |
6207 | } | |
6208 | ||
bf5a934a DW |
6209 | /* This device needs to be a device in an 'imsm' container */ |
6210 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6211 | if (fd >= 0) { | |
6212 | if (verbose) | |
e7b84f9d N |
6213 | pr_err("Cannot create this array on device %s\n", |
6214 | dev); | |
bf5a934a DW |
6215 | close(fd); |
6216 | return 0; | |
6217 | } | |
6218 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6219 | if (verbose) | |
e7b84f9d | 6220 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
6221 | dev, strerror(errno)); |
6222 | return 0; | |
6223 | } | |
6224 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6225 | cfd = open_container(fd); | |
20cbe8d2 | 6226 | close(fd); |
bf5a934a | 6227 | if (cfd < 0) { |
bf5a934a | 6228 | if (verbose) |
e7b84f9d | 6229 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
6230 | dev); |
6231 | return 0; | |
6232 | } | |
6233 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 6234 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
6235 | strcmp(sra->text_version, "imsm") == 0) |
6236 | is_member = 1; | |
6237 | sysfs_free(sra); | |
6238 | if (is_member) { | |
bf5a934a DW |
6239 | /* This is a member of a imsm container. Load the container |
6240 | * and try to create a volume | |
6241 | */ | |
6242 | struct intel_super *super; | |
6243 | ||
ec50f7b6 | 6244 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a DW |
6245 | st->sb = super; |
6246 | st->container_dev = fd2devnum(cfd); | |
6247 | close(cfd); | |
6248 | return validate_geometry_imsm_volume(st, level, layout, | |
6249 | raiddisks, chunk, | |
af4348dd | 6250 | size, data_offset, dev, |
ecbd9e81 N |
6251 | freesize, 1) |
6252 | ? 1 : -1; | |
bf5a934a | 6253 | } |
20cbe8d2 | 6254 | } |
bf5a934a | 6255 | |
20cbe8d2 | 6256 | if (verbose) |
e7b84f9d | 6257 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
6258 | |
6259 | close(cfd); | |
6260 | return 0; | |
bf5a934a | 6261 | } |
0bd16cf2 | 6262 | |
30f58b22 | 6263 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
6264 | { |
6265 | struct intel_super *super = st->sb; | |
6266 | ||
30f58b22 DW |
6267 | if (level && *level == UnSet) |
6268 | *level = LEVEL_CONTAINER; | |
6269 | ||
6270 | if (level && layout && *layout == UnSet) | |
6271 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 6272 | |
cd9d1ac7 DW |
6273 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
6274 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
6275 | } |
6276 | ||
33414a01 DW |
6277 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
6278 | ||
6279 | static int kill_subarray_imsm(struct supertype *st) | |
6280 | { | |
6281 | /* remove the subarray currently referenced by ->current_vol */ | |
6282 | __u8 i; | |
6283 | struct intel_dev **dp; | |
6284 | struct intel_super *super = st->sb; | |
6285 | __u8 current_vol = super->current_vol; | |
6286 | struct imsm_super *mpb = super->anchor; | |
6287 | ||
6288 | if (super->current_vol < 0) | |
6289 | return 2; | |
6290 | super->current_vol = -1; /* invalidate subarray cursor */ | |
6291 | ||
6292 | /* block deletions that would change the uuid of active subarrays | |
6293 | * | |
6294 | * FIXME when immutable ids are available, but note that we'll | |
6295 | * also need to fixup the invalidated/active subarray indexes in | |
6296 | * mdstat | |
6297 | */ | |
6298 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6299 | char subarray[4]; | |
6300 | ||
6301 | if (i < current_vol) | |
6302 | continue; | |
6303 | sprintf(subarray, "%u", i); | |
6304 | if (is_subarray_active(subarray, st->devname)) { | |
e7b84f9d N |
6305 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
6306 | current_vol, i); | |
33414a01 DW |
6307 | |
6308 | return 2; | |
6309 | } | |
6310 | } | |
6311 | ||
6312 | if (st->update_tail) { | |
503975b9 | 6313 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 6314 | |
33414a01 DW |
6315 | u->type = update_kill_array; |
6316 | u->dev_idx = current_vol; | |
6317 | append_metadata_update(st, u, sizeof(*u)); | |
6318 | ||
6319 | return 0; | |
6320 | } | |
6321 | ||
6322 | for (dp = &super->devlist; *dp;) | |
6323 | if ((*dp)->index == current_vol) { | |
6324 | *dp = (*dp)->next; | |
6325 | } else { | |
6326 | handle_missing(super, (*dp)->dev); | |
6327 | if ((*dp)->index > current_vol) | |
6328 | (*dp)->index--; | |
6329 | dp = &(*dp)->next; | |
6330 | } | |
6331 | ||
6332 | /* no more raid devices, all active components are now spares, | |
6333 | * but of course failed are still failed | |
6334 | */ | |
6335 | if (--mpb->num_raid_devs == 0) { | |
6336 | struct dl *d; | |
6337 | ||
6338 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
6339 | if (d->index > -2) |
6340 | mark_spare(d); | |
33414a01 DW |
6341 | } |
6342 | ||
6343 | super->updates_pending++; | |
6344 | ||
6345 | return 0; | |
6346 | } | |
aa534678 | 6347 | |
a951a4f7 | 6348 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 6349 | char *update, struct mddev_ident *ident) |
aa534678 DW |
6350 | { |
6351 | /* update the subarray currently referenced by ->current_vol */ | |
6352 | struct intel_super *super = st->sb; | |
6353 | struct imsm_super *mpb = super->anchor; | |
6354 | ||
aa534678 DW |
6355 | if (strcmp(update, "name") == 0) { |
6356 | char *name = ident->name; | |
a951a4f7 N |
6357 | char *ep; |
6358 | int vol; | |
aa534678 | 6359 | |
a951a4f7 | 6360 | if (is_subarray_active(subarray, st->devname)) { |
e7b84f9d | 6361 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
6362 | return 2; |
6363 | } | |
6364 | ||
6365 | if (!check_name(super, name, 0)) | |
6366 | return 2; | |
6367 | ||
a951a4f7 N |
6368 | vol = strtoul(subarray, &ep, 10); |
6369 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
6370 | return 2; | |
6371 | ||
aa534678 | 6372 | if (st->update_tail) { |
503975b9 | 6373 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 6374 | |
aa534678 | 6375 | u->type = update_rename_array; |
a951a4f7 | 6376 | u->dev_idx = vol; |
aa534678 DW |
6377 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
6378 | append_metadata_update(st, u, sizeof(*u)); | |
6379 | } else { | |
6380 | struct imsm_dev *dev; | |
6381 | int i; | |
6382 | ||
a951a4f7 | 6383 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
6384 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
6385 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6386 | dev = get_imsm_dev(super, i); | |
6387 | handle_missing(super, dev); | |
6388 | } | |
6389 | super->updates_pending++; | |
6390 | } | |
6391 | } else | |
6392 | return 2; | |
6393 | ||
6394 | return 0; | |
6395 | } | |
d1e02575 | 6396 | #endif /* MDASSEMBLE */ |
bf5a934a | 6397 | |
28bce06f AK |
6398 | static int is_gen_migration(struct imsm_dev *dev) |
6399 | { | |
7534230b AK |
6400 | if (dev == NULL) |
6401 | return 0; | |
6402 | ||
28bce06f AK |
6403 | if (!dev->vol.migr_state) |
6404 | return 0; | |
6405 | ||
6406 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
6407 | return 1; | |
6408 | ||
6409 | return 0; | |
6410 | } | |
6411 | ||
1e5c6983 DW |
6412 | static int is_rebuilding(struct imsm_dev *dev) |
6413 | { | |
6414 | struct imsm_map *migr_map; | |
6415 | ||
6416 | if (!dev->vol.migr_state) | |
6417 | return 0; | |
6418 | ||
6419 | if (migr_type(dev) != MIGR_REBUILD) | |
6420 | return 0; | |
6421 | ||
238c0a71 | 6422 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
6423 | |
6424 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
6425 | return 1; | |
6426 | else | |
6427 | return 0; | |
6428 | } | |
6429 | ||
b4ab44d8 | 6430 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
6431 | static int is_initializing(struct imsm_dev *dev) |
6432 | { | |
6433 | struct imsm_map *migr_map; | |
6434 | ||
6435 | if (!dev->vol.migr_state) | |
6436 | return 0; | |
6437 | ||
6438 | if (migr_type(dev) != MIGR_INIT) | |
6439 | return 0; | |
6440 | ||
238c0a71 | 6441 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
6442 | |
6443 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6444 | return 1; | |
6445 | ||
6446 | return 0; | |
6ce1fbf1 | 6447 | } |
b4ab44d8 | 6448 | #endif |
6ce1fbf1 | 6449 | |
c47b0ff6 AK |
6450 | static void update_recovery_start(struct intel_super *super, |
6451 | struct imsm_dev *dev, | |
6452 | struct mdinfo *array) | |
1e5c6983 DW |
6453 | { |
6454 | struct mdinfo *rebuild = NULL; | |
6455 | struct mdinfo *d; | |
6456 | __u32 units; | |
6457 | ||
6458 | if (!is_rebuilding(dev)) | |
6459 | return; | |
6460 | ||
6461 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
6462 | for (d = array->devs; d; d = d->next) | |
6463 | if (d->recovery_start == 0) { | |
6464 | if (rebuild) | |
6465 | return; | |
6466 | rebuild = d; | |
6467 | } | |
6468 | ||
4363fd80 DW |
6469 | if (!rebuild) { |
6470 | /* (?) none of the disks are marked with | |
6471 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
6472 | * disk_ord_tbl was not correctly updated | |
6473 | */ | |
6474 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
6475 | return; | |
6476 | } | |
6477 | ||
1e5c6983 | 6478 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 6479 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
6480 | } |
6481 | ||
9e2d750d | 6482 | #ifndef MDASSEMBLE |
276d77db | 6483 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 6484 | #endif |
1e5c6983 | 6485 | |
00bbdbda | 6486 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 6487 | { |
4f5bc454 DW |
6488 | /* Given a container loaded by load_super_imsm_all, |
6489 | * extract information about all the arrays into | |
6490 | * an mdinfo tree. | |
00bbdbda | 6491 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
6492 | * |
6493 | * For each imsm_dev create an mdinfo, fill it in, | |
6494 | * then look for matching devices in super->disks | |
6495 | * and create appropriate device mdinfo. | |
6496 | */ | |
6497 | struct intel_super *super = st->sb; | |
949c47a0 | 6498 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 6499 | struct mdinfo *rest = NULL; |
00bbdbda | 6500 | unsigned int i; |
81219e70 | 6501 | int sb_errors = 0; |
abef11a3 AK |
6502 | struct dl *d; |
6503 | int spare_disks = 0; | |
cdddbdbc | 6504 | |
19482bcc AK |
6505 | /* do not assemble arrays when not all attributes are supported */ |
6506 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 6507 | sb_errors = 1; |
e7b84f9d | 6508 | pr_err("Unsupported attributes in IMSM metadata." |
81219e70 | 6509 | "Arrays activation is blocked.\n"); |
19482bcc AK |
6510 | } |
6511 | ||
a06d022d | 6512 | /* check for bad blocks */ |
81219e70 | 6513 | if (imsm_bbm_log_size(super->anchor)) { |
e7b84f9d N |
6514 | pr_err("BBM log found in IMSM metadata." |
6515 | "Arrays activation is blocked.\n"); | |
81219e70 LM |
6516 | sb_errors = 1; |
6517 | } | |
6518 | ||
604b746f | 6519 | |
abef11a3 AK |
6520 | /* count spare devices, not used in maps |
6521 | */ | |
6522 | for (d = super->disks; d; d = d->next) | |
6523 | if (d->index == -1) | |
6524 | spare_disks++; | |
6525 | ||
4f5bc454 | 6526 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
6527 | struct imsm_dev *dev; |
6528 | struct imsm_map *map; | |
86e3692b | 6529 | struct imsm_map *map2; |
4f5bc454 | 6530 | struct mdinfo *this; |
a6482415 N |
6531 | int slot; |
6532 | #ifndef MDASSEMBLE | |
6533 | int chunk; | |
6534 | #endif | |
00bbdbda N |
6535 | char *ep; |
6536 | ||
6537 | if (subarray && | |
6538 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
6539 | continue; | |
6540 | ||
6541 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
6542 | map = get_imsm_map(dev, MAP_0); |
6543 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 6544 | |
1ce0101c DW |
6545 | /* do not publish arrays that are in the middle of an |
6546 | * unsupported migration | |
6547 | */ | |
6548 | if (dev->vol.migr_state && | |
28bce06f | 6549 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
e7b84f9d | 6550 | pr_err("cannot assemble volume '%.16s':" |
1ce0101c DW |
6551 | " unsupported migration in progress\n", |
6552 | dev->volume); | |
6553 | continue; | |
6554 | } | |
2db86302 LM |
6555 | /* do not publish arrays that are not support by controller's |
6556 | * OROM/EFI | |
6557 | */ | |
1ce0101c | 6558 | |
503975b9 | 6559 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 6560 | |
301406c9 | 6561 | super->current_vol = i; |
a5d85af7 | 6562 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 6563 | this->next = rest; |
81219e70 | 6564 | #ifndef MDASSEMBLE |
a6482415 | 6565 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
6566 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
6567 | if (!validate_geometry_imsm_orom(super, | |
6568 | get_imsm_raid_level(map), /* RAID level */ | |
6569 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
6570 | map->num_members, /* raid disks */ | |
2cc699af | 6571 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 6572 | 1 /* verbose */)) { |
e7b84f9d | 6573 | pr_err("IMSM RAID geometry validation" |
446894ea | 6574 | " failed. Array %s activation is blocked.\n", |
81219e70 LM |
6575 | dev->volume); |
6576 | this->array.state |= | |
6577 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6578 | (1<<MD_SB_BLOCK_VOLUME); | |
6579 | } | |
6580 | #endif | |
6581 | ||
6582 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
6583 | if (sb_errors) | |
6584 | this->array.state |= | |
6585 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6586 | (1<<MD_SB_BLOCK_VOLUME); | |
6587 | ||
4f5bc454 | 6588 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 6589 | unsigned long long recovery_start; |
4f5bc454 DW |
6590 | struct mdinfo *info_d; |
6591 | struct dl *d; | |
6592 | int idx; | |
9a1608e5 | 6593 | int skip; |
7eef0453 | 6594 | __u32 ord; |
4f5bc454 | 6595 | |
9a1608e5 | 6596 | skip = 0; |
238c0a71 AK |
6597 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
6598 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
6599 | for (d = super->disks; d ; d = d->next) |
6600 | if (d->index == idx) | |
0fbd635c | 6601 | break; |
4f5bc454 | 6602 | |
1e5c6983 | 6603 | recovery_start = MaxSector; |
4f5bc454 | 6604 | if (d == NULL) |
9a1608e5 | 6605 | skip = 1; |
25ed7e59 | 6606 | if (d && is_failed(&d->disk)) |
9a1608e5 | 6607 | skip = 1; |
7eef0453 | 6608 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 6609 | recovery_start = 0; |
9a1608e5 DW |
6610 | |
6611 | /* | |
6612 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
6613 | * reset resync start to avoid a dirty-degraded |
6614 | * situation when performing the intial sync | |
9a1608e5 DW |
6615 | * |
6616 | * FIXME handle dirty degraded | |
6617 | */ | |
1e5c6983 | 6618 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 6619 | this->resync_start = MaxSector; |
9a1608e5 DW |
6620 | if (skip) |
6621 | continue; | |
4f5bc454 | 6622 | |
503975b9 | 6623 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
6624 | info_d->next = this->devs; |
6625 | this->devs = info_d; | |
6626 | ||
4f5bc454 DW |
6627 | info_d->disk.number = d->index; |
6628 | info_d->disk.major = d->major; | |
6629 | info_d->disk.minor = d->minor; | |
6630 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 6631 | info_d->recovery_start = recovery_start; |
86e3692b AK |
6632 | if (map2) { |
6633 | if (slot < map2->num_members) | |
6634 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6635 | else |
6636 | this->array.spare_disks++; | |
86e3692b AK |
6637 | } else { |
6638 | if (slot < map->num_members) | |
6639 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6640 | else |
6641 | this->array.spare_disks++; | |
86e3692b | 6642 | } |
1e5c6983 DW |
6643 | if (info_d->recovery_start == MaxSector) |
6644 | this->array.working_disks++; | |
4f5bc454 DW |
6645 | |
6646 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 CA |
6647 | info_d->data_offset = pba_of_lba0(map); |
6648 | info_d->component_size = blocks_per_member(map); | |
4f5bc454 | 6649 | } |
1e5c6983 | 6650 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 6651 | update_recovery_start(super, dev, this); |
abef11a3 | 6652 | this->array.spare_disks += spare_disks; |
276d77db | 6653 | |
9e2d750d | 6654 | #ifndef MDASSEMBLE |
276d77db AK |
6655 | /* check for reshape */ |
6656 | if (this->reshape_active == 1) | |
6657 | recover_backup_imsm(st, this); | |
9e2d750d | 6658 | #endif |
9a1608e5 | 6659 | rest = this; |
4f5bc454 DW |
6660 | } |
6661 | ||
6662 | return rest; | |
cdddbdbc DW |
6663 | } |
6664 | ||
845dea95 | 6665 | |
3b451610 AK |
6666 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
6667 | int failed, int look_in_map) | |
c2a1e7da | 6668 | { |
3b451610 AK |
6669 | struct imsm_map *map; |
6670 | ||
6671 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
6672 | |
6673 | if (!failed) | |
3393c6af DW |
6674 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
6675 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
6676 | |
6677 | switch (get_imsm_raid_level(map)) { | |
6678 | case 0: | |
6679 | return IMSM_T_STATE_FAILED; | |
6680 | break; | |
6681 | case 1: | |
6682 | if (failed < map->num_members) | |
6683 | return IMSM_T_STATE_DEGRADED; | |
6684 | else | |
6685 | return IMSM_T_STATE_FAILED; | |
6686 | break; | |
6687 | case 10: | |
6688 | { | |
6689 | /** | |
c92a2527 DW |
6690 | * check to see if any mirrors have failed, otherwise we |
6691 | * are degraded. Even numbered slots are mirrored on | |
6692 | * slot+1 | |
c2a1e7da | 6693 | */ |
c2a1e7da | 6694 | int i; |
d9b420a5 N |
6695 | /* gcc -Os complains that this is unused */ |
6696 | int insync = insync; | |
c2a1e7da DW |
6697 | |
6698 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6699 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6700 | int idx = ord_to_idx(ord); |
6701 | struct imsm_disk *disk; | |
c2a1e7da | 6702 | |
c92a2527 DW |
6703 | /* reset the potential in-sync count on even-numbered |
6704 | * slots. num_copies is always 2 for imsm raid10 | |
6705 | */ | |
6706 | if ((i & 1) == 0) | |
6707 | insync = 2; | |
c2a1e7da | 6708 | |
c92a2527 | 6709 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6710 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6711 | insync--; |
c2a1e7da | 6712 | |
c92a2527 DW |
6713 | /* no in-sync disks left in this mirror the |
6714 | * array has failed | |
6715 | */ | |
6716 | if (insync == 0) | |
6717 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6718 | } |
6719 | ||
6720 | return IMSM_T_STATE_DEGRADED; | |
6721 | } | |
6722 | case 5: | |
6723 | if (failed < 2) | |
6724 | return IMSM_T_STATE_DEGRADED; | |
6725 | else | |
6726 | return IMSM_T_STATE_FAILED; | |
6727 | break; | |
6728 | default: | |
6729 | break; | |
6730 | } | |
6731 | ||
6732 | return map->map_state; | |
6733 | } | |
6734 | ||
3b451610 AK |
6735 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6736 | int look_in_map) | |
c2a1e7da DW |
6737 | { |
6738 | int i; | |
6739 | int failed = 0; | |
6740 | struct imsm_disk *disk; | |
d5985138 AK |
6741 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6742 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6743 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6744 | __u32 ord; |
6745 | int idx; | |
d5985138 | 6746 | int idx_1; |
c2a1e7da | 6747 | |
0556e1a2 DW |
6748 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6749 | * disks that are being rebuilt. New failures are recorded to | |
6750 | * map[0]. So we look through all the disks we started with and | |
6751 | * see if any failures are still present, or if any new ones | |
6752 | * have arrived | |
0556e1a2 | 6753 | */ |
d5985138 AK |
6754 | map_for_loop = map; |
6755 | if (prev && (map->num_members < prev->num_members)) | |
6756 | map_for_loop = prev; | |
68fe4598 LD |
6757 | |
6758 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6759 | idx_1 = -255; |
238c0a71 AK |
6760 | /* when MAP_X is passed both maps failures are counted |
6761 | */ | |
d5985138 | 6762 | if (prev && |
238c0a71 AK |
6763 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6764 | (i < prev->num_members)) { | |
d5985138 AK |
6765 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6766 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6767 | |
d5985138 AK |
6768 | disk = get_imsm_disk(super, idx_1); |
6769 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6770 | failed++; | |
6771 | } | |
238c0a71 AK |
6772 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6773 | (i < map->num_members)) { | |
d5985138 AK |
6774 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6775 | idx = ord_to_idx(ord); | |
6776 | ||
6777 | if (idx != idx_1) { | |
6778 | disk = get_imsm_disk(super, idx); | |
6779 | if (!disk || is_failed(disk) || | |
6780 | ord & IMSM_ORD_REBUILD) | |
6781 | failed++; | |
6782 | } | |
6783 | } | |
c2a1e7da DW |
6784 | } |
6785 | ||
6786 | return failed; | |
845dea95 NB |
6787 | } |
6788 | ||
97b4d0e9 DW |
6789 | #ifndef MDASSEMBLE |
6790 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
6791 | char *inst) | |
6792 | { | |
6793 | struct intel_super *super = c->sb; | |
6794 | struct imsm_super *mpb = super->anchor; | |
9587c373 | 6795 | |
97b4d0e9 DW |
6796 | if (atoi(inst) >= mpb->num_raid_devs) { |
6797 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
6798 | __func__, atoi(inst)); | |
6799 | return -ENODEV; | |
6800 | } | |
6801 | ||
6802 | dprintf("imsm: open_new %s\n", inst); | |
6803 | a->info.container_member = atoi(inst); | |
6804 | return 0; | |
6805 | } | |
6806 | ||
0c046afd DW |
6807 | static int is_resyncing(struct imsm_dev *dev) |
6808 | { | |
6809 | struct imsm_map *migr_map; | |
6810 | ||
6811 | if (!dev->vol.migr_state) | |
6812 | return 0; | |
6813 | ||
1484e727 DW |
6814 | if (migr_type(dev) == MIGR_INIT || |
6815 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
6816 | return 1; |
6817 | ||
4c9bc37b AK |
6818 | if (migr_type(dev) == MIGR_GEN_MIGR) |
6819 | return 0; | |
6820 | ||
238c0a71 | 6821 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 6822 | |
4c9bc37b AK |
6823 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
6824 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
6825 | return 1; |
6826 | else | |
6827 | return 0; | |
6828 | } | |
6829 | ||
0556e1a2 DW |
6830 | /* return true if we recorded new information */ |
6831 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 6832 | { |
0556e1a2 DW |
6833 | __u32 ord; |
6834 | int slot; | |
6835 | struct imsm_map *map; | |
86c54047 DW |
6836 | char buf[MAX_RAID_SERIAL_LEN+3]; |
6837 | unsigned int len, shift = 0; | |
0556e1a2 DW |
6838 | |
6839 | /* new failures are always set in map[0] */ | |
238c0a71 | 6840 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
6841 | |
6842 | slot = get_imsm_disk_slot(map, idx); | |
6843 | if (slot < 0) | |
6844 | return 0; | |
6845 | ||
6846 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 6847 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
6848 | return 0; |
6849 | ||
7d0c5e24 LD |
6850 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
6851 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
6852 | strcat(buf, ":0"); | |
86c54047 DW |
6853 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
6854 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
6855 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
6856 | ||
f2f27e63 | 6857 | disk->status |= FAILED_DISK; |
0556e1a2 | 6858 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
6859 | /* mark failures in second map if second map exists and this disk |
6860 | * in this slot. | |
6861 | * This is valid for migration, initialization and rebuild | |
6862 | */ | |
6863 | if (dev->vol.migr_state) { | |
238c0a71 | 6864 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
6865 | int slot2 = get_imsm_disk_slot(map2, idx); |
6866 | ||
6867 | if ((slot2 < map2->num_members) && | |
6868 | (slot2 >= 0)) | |
6869 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
6870 | idx | IMSM_ORD_REBUILD); |
6871 | } | |
f21e18ca | 6872 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
6873 | map->failed_disk_num = slot; |
6874 | return 1; | |
6875 | } | |
6876 | ||
6877 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
6878 | { | |
6879 | mark_failure(dev, disk, idx); | |
6880 | ||
6881 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
6882 | return; | |
6883 | ||
47ee5a45 DW |
6884 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
6885 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
6886 | } | |
6887 | ||
33414a01 DW |
6888 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
6889 | { | |
33414a01 | 6890 | struct dl *dl; |
33414a01 DW |
6891 | |
6892 | if (!super->missing) | |
6893 | return; | |
33414a01 DW |
6894 | |
6895 | dprintf("imsm: mark missing\n"); | |
3d59f0c0 AK |
6896 | /* end process for initialization and rebuild only |
6897 | */ | |
6898 | if (is_gen_migration(dev) == 0) { | |
6899 | __u8 map_state; | |
6900 | int failed; | |
6901 | ||
6902 | failed = imsm_count_failed(super, dev, MAP_0); | |
6903 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
6904 | ||
6905 | end_migration(dev, super, map_state); | |
6906 | } | |
33414a01 DW |
6907 | for (dl = super->missing; dl; dl = dl->next) |
6908 | mark_missing(dev, &dl->disk, dl->index); | |
6909 | super->updates_pending++; | |
6910 | } | |
6911 | ||
f3871fdc AK |
6912 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
6913 | long long new_size) | |
70bdf0dc | 6914 | { |
238c0a71 | 6915 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
6916 | unsigned long long array_blocks; |
6917 | struct imsm_map *map; | |
6918 | ||
6919 | if (used_disks == 0) { | |
6920 | /* when problems occures | |
6921 | * return current array_blocks value | |
6922 | */ | |
6923 | array_blocks = __le32_to_cpu(dev->size_high); | |
6924 | array_blocks = array_blocks << 32; | |
6925 | array_blocks += __le32_to_cpu(dev->size_low); | |
6926 | ||
6927 | return array_blocks; | |
6928 | } | |
6929 | ||
6930 | /* set array size in metadata | |
6931 | */ | |
f3871fdc AK |
6932 | if (new_size <= 0) { |
6933 | /* OLCE size change is caused by added disks | |
6934 | */ | |
6935 | map = get_imsm_map(dev, MAP_0); | |
6936 | array_blocks = blocks_per_member(map) * used_disks; | |
6937 | } else { | |
6938 | /* Online Volume Size Change | |
6939 | * Using available free space | |
6940 | */ | |
6941 | array_blocks = new_size; | |
6942 | } | |
70bdf0dc AK |
6943 | |
6944 | /* round array size down to closest MB | |
6945 | */ | |
6946 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
6947 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
6948 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
6949 | ||
6950 | return array_blocks; | |
6951 | } | |
6952 | ||
28bce06f AK |
6953 | static void imsm_set_disk(struct active_array *a, int n, int state); |
6954 | ||
0e2d1a4e AK |
6955 | static void imsm_progress_container_reshape(struct intel_super *super) |
6956 | { | |
6957 | /* if no device has a migr_state, but some device has a | |
6958 | * different number of members than the previous device, start | |
6959 | * changing the number of devices in this device to match | |
6960 | * previous. | |
6961 | */ | |
6962 | struct imsm_super *mpb = super->anchor; | |
6963 | int prev_disks = -1; | |
6964 | int i; | |
1dfaa380 | 6965 | int copy_map_size; |
0e2d1a4e AK |
6966 | |
6967 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6968 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 6969 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
6970 | struct imsm_map *map2; |
6971 | int prev_num_members; | |
0e2d1a4e AK |
6972 | |
6973 | if (dev->vol.migr_state) | |
6974 | return; | |
6975 | ||
6976 | if (prev_disks == -1) | |
6977 | prev_disks = map->num_members; | |
6978 | if (prev_disks == map->num_members) | |
6979 | continue; | |
6980 | ||
6981 | /* OK, this array needs to enter reshape mode. | |
6982 | * i.e it needs a migr_state | |
6983 | */ | |
6984 | ||
1dfaa380 | 6985 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
6986 | prev_num_members = map->num_members; |
6987 | map->num_members = prev_disks; | |
6988 | dev->vol.migr_state = 1; | |
6989 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 6990 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
6991 | for (i = prev_num_members; |
6992 | i < map->num_members; i++) | |
6993 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 6994 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 6995 | /* Copy the current map */ |
1dfaa380 | 6996 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
6997 | map2->num_members = prev_num_members; |
6998 | ||
f3871fdc | 6999 | imsm_set_array_size(dev, -1); |
51d83f5d | 7000 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
7001 | super->updates_pending++; |
7002 | } | |
7003 | } | |
7004 | ||
aad6f216 | 7005 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
7006 | * states are handled in imsm_set_disk() with one exception, when a |
7007 | * resync is stopped due to a new failure this routine will set the | |
7008 | * 'degraded' state for the array. | |
7009 | */ | |
01f157d7 | 7010 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
7011 | { |
7012 | int inst = a->info.container_member; | |
7013 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7014 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7015 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
7016 | int failed = imsm_count_failed(super, dev, MAP_0); |
7017 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 7018 | __u32 blocks_per_unit; |
a862209d | 7019 | |
1af97990 AK |
7020 | if (dev->vol.migr_state && |
7021 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
7022 | /* array state change is blocked due to reshape action | |
aad6f216 N |
7023 | * We might need to |
7024 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
7025 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
7026 | * - update curr_migr_unit | |
1af97990 | 7027 | */ |
aad6f216 N |
7028 | if (a->curr_action == reshape) { |
7029 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 7030 | goto mark_checkpoint; |
aad6f216 N |
7031 | } else { |
7032 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
7033 | /* for some reason we aborted the reshape. | |
b66e591b AK |
7034 | * |
7035 | * disable automatic metadata rollback | |
7036 | * user action is required to recover process | |
aad6f216 | 7037 | */ |
b66e591b | 7038 | if (0) { |
238c0a71 AK |
7039 | struct imsm_map *map2 = |
7040 | get_imsm_map(dev, MAP_1); | |
7041 | dev->vol.migr_state = 0; | |
7042 | set_migr_type(dev, 0); | |
7043 | dev->vol.curr_migr_unit = 0; | |
7044 | memcpy(map, map2, | |
7045 | sizeof_imsm_map(map2)); | |
7046 | super->updates_pending++; | |
b66e591b | 7047 | } |
aad6f216 N |
7048 | } |
7049 | if (a->last_checkpoint >= a->info.component_size) { | |
7050 | unsigned long long array_blocks; | |
7051 | int used_disks; | |
e154ced3 | 7052 | struct mdinfo *mdi; |
aad6f216 | 7053 | |
238c0a71 | 7054 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
7055 | if (used_disks > 0) { |
7056 | array_blocks = | |
5551b113 | 7057 | blocks_per_member(map) * |
d55adef9 AK |
7058 | used_disks; |
7059 | /* round array size down to closest MB | |
7060 | */ | |
7061 | array_blocks = (array_blocks | |
7062 | >> SECT_PER_MB_SHIFT) | |
7063 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
7064 | a->info.custom_array_size = array_blocks; |
7065 | /* encourage manager to update array | |
7066 | * size | |
7067 | */ | |
e154ced3 | 7068 | |
d55adef9 | 7069 | a->check_reshape = 1; |
633b5610 | 7070 | } |
e154ced3 AK |
7071 | /* finalize online capacity expansion/reshape */ |
7072 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
7073 | imsm_set_disk(a, | |
7074 | mdi->disk.raid_disk, | |
7075 | mdi->curr_state); | |
7076 | ||
0e2d1a4e | 7077 | imsm_progress_container_reshape(super); |
e154ced3 | 7078 | } |
aad6f216 | 7079 | } |
1af97990 AK |
7080 | } |
7081 | ||
47ee5a45 | 7082 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
7083 | if (consistent == 2) |
7084 | handle_missing(super, dev); | |
1e5c6983 | 7085 | |
0c046afd | 7086 | if (consistent == 2 && |
b7941fd6 | 7087 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7088 | map_state != IMSM_T_STATE_NORMAL || |
7089 | dev->vol.migr_state)) | |
01f157d7 | 7090 | consistent = 0; |
272906ef | 7091 | |
b7941fd6 | 7092 | if (is_resync_complete(&a->info)) { |
0c046afd | 7093 | /* complete intialization / resync, |
0556e1a2 DW |
7094 | * recovery and interrupted recovery is completed in |
7095 | * ->set_disk | |
0c046afd DW |
7096 | */ |
7097 | if (is_resyncing(dev)) { | |
7098 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7099 | end_migration(dev, super, map_state); |
115c3803 | 7100 | super->updates_pending++; |
484240d8 | 7101 | a->last_checkpoint = 0; |
115c3803 | 7102 | } |
b9172665 AK |
7103 | } else if ((!is_resyncing(dev) && !failed) && |
7104 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7105 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7106 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7107 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7108 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7109 | else |
8e59f3d8 | 7110 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7111 | super->updates_pending++; |
115c3803 | 7112 | } |
a862209d | 7113 | |
633b5610 | 7114 | mark_checkpoint: |
5b83bacf AK |
7115 | /* skip checkpointing for general migration, |
7116 | * it is controlled in mdadm | |
7117 | */ | |
7118 | if (is_gen_migration(dev)) | |
7119 | goto skip_mark_checkpoint; | |
7120 | ||
1e5c6983 | 7121 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7122 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7123 | if (blocks_per_unit) { |
1e5c6983 DW |
7124 | __u32 units32; |
7125 | __u64 units; | |
7126 | ||
4f0a7acc | 7127 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7128 | units32 = units; |
7129 | ||
7130 | /* check that we did not overflow 32-bits, and that | |
7131 | * curr_migr_unit needs updating | |
7132 | */ | |
7133 | if (units32 == units && | |
bfd80a56 | 7134 | units32 != 0 && |
1e5c6983 DW |
7135 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7136 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7137 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7138 | super->updates_pending++; | |
7139 | } | |
7140 | } | |
f8f603f1 | 7141 | |
5b83bacf | 7142 | skip_mark_checkpoint: |
3393c6af | 7143 | /* mark dirty / clean */ |
0c046afd | 7144 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7145 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7146 | if (consistent) |
7147 | dev->vol.dirty = 0; | |
7148 | else | |
7149 | dev->vol.dirty = 1; | |
a862209d DW |
7150 | super->updates_pending++; |
7151 | } | |
28bce06f | 7152 | |
01f157d7 | 7153 | return consistent; |
a862209d DW |
7154 | } |
7155 | ||
8d45d196 | 7156 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7157 | { |
8d45d196 DW |
7158 | int inst = a->info.container_member; |
7159 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7160 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7161 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7162 | struct imsm_disk *disk; |
7ce05701 LD |
7163 | struct mdinfo *mdi; |
7164 | int recovery_not_finished = 0; | |
0c046afd | 7165 | int failed; |
b10b37b8 | 7166 | __u32 ord; |
0c046afd | 7167 | __u8 map_state; |
8d45d196 DW |
7168 | |
7169 | if (n > map->num_members) | |
7170 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
7171 | n, map->num_members - 1); | |
7172 | ||
7173 | if (n < 0) | |
7174 | return; | |
7175 | ||
4e6e574a | 7176 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 7177 | |
238c0a71 | 7178 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 7179 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7180 | |
5802a811 | 7181 | /* check for new failures */ |
0556e1a2 DW |
7182 | if (state & DS_FAULTY) { |
7183 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
7184 | super->updates_pending++; | |
8d45d196 | 7185 | } |
47ee5a45 | 7186 | |
19859edc | 7187 | /* check if in_sync */ |
0556e1a2 | 7188 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7189 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7190 | |
7191 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7192 | super->updates_pending++; |
7193 | } | |
8d45d196 | 7194 | |
3b451610 AK |
7195 | failed = imsm_count_failed(super, dev, MAP_0); |
7196 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7197 | |
0c046afd | 7198 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7199 | dprintf("imsm: Detected transition to state "); |
7200 | switch (map_state) { | |
7201 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
7202 | dprintf("normal: "); | |
7203 | if (is_rebuilding(dev)) { | |
7204 | dprintf("while rebuilding"); | |
7ce05701 LD |
7205 | /* check if recovery is really finished */ |
7206 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
7207 | if (mdi->recovery_start != MaxSector) { | |
7208 | recovery_not_finished = 1; | |
7209 | break; | |
7210 | } | |
7211 | if (recovery_not_finished) { | |
7212 | dprintf("\nimsm: Rebuild has not finished yet, " | |
7213 | "state not changed"); | |
7214 | if (a->last_checkpoint < mdi->recovery_start) { | |
7215 | a->last_checkpoint = mdi->recovery_start; | |
7216 | super->updates_pending++; | |
7217 | } | |
7218 | break; | |
7219 | } | |
94002678 | 7220 | end_migration(dev, super, map_state); |
238c0a71 | 7221 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
7222 | map->failed_disk_num = ~0; |
7223 | super->updates_pending++; | |
7224 | a->last_checkpoint = 0; | |
7225 | break; | |
7226 | } | |
7227 | if (is_gen_migration(dev)) { | |
7228 | dprintf("while general migration"); | |
bf2f0071 | 7229 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7230 | end_migration(dev, super, map_state); |
94002678 AK |
7231 | else |
7232 | map->map_state = map_state; | |
238c0a71 | 7233 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 7234 | map->failed_disk_num = ~0; |
94002678 | 7235 | super->updates_pending++; |
bf2f0071 | 7236 | break; |
94002678 AK |
7237 | } |
7238 | break; | |
7239 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
7240 | dprintf("degraded: "); | |
7241 | if ((map->map_state != map_state) && | |
7242 | !dev->vol.migr_state) { | |
7243 | dprintf("mark degraded"); | |
7244 | map->map_state = map_state; | |
7245 | super->updates_pending++; | |
7246 | a->last_checkpoint = 0; | |
7247 | break; | |
7248 | } | |
7249 | if (is_rebuilding(dev)) { | |
7250 | dprintf("while rebuilding."); | |
7251 | if (map->map_state != map_state) { | |
7252 | dprintf(" Map state change"); | |
7253 | end_migration(dev, super, map_state); | |
7254 | super->updates_pending++; | |
7255 | } | |
7256 | break; | |
7257 | } | |
7258 | if (is_gen_migration(dev)) { | |
7259 | dprintf("while general migration"); | |
bf2f0071 | 7260 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7261 | end_migration(dev, super, map_state); |
94002678 AK |
7262 | else { |
7263 | map->map_state = map_state; | |
3b451610 | 7264 | manage_second_map(super, dev); |
94002678 AK |
7265 | } |
7266 | super->updates_pending++; | |
bf2f0071 | 7267 | break; |
28bce06f | 7268 | } |
6ce1fbf1 AK |
7269 | if (is_initializing(dev)) { |
7270 | dprintf("while initialization."); | |
7271 | map->map_state = map_state; | |
7272 | super->updates_pending++; | |
7273 | break; | |
7274 | } | |
94002678 AK |
7275 | break; |
7276 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
7277 | dprintf("failed: "); | |
7278 | if (is_gen_migration(dev)) { | |
7279 | dprintf("while general migration"); | |
7280 | map->map_state = map_state; | |
7281 | super->updates_pending++; | |
7282 | break; | |
7283 | } | |
7284 | if (map->map_state != map_state) { | |
7285 | dprintf("mark failed"); | |
7286 | end_migration(dev, super, map_state); | |
7287 | super->updates_pending++; | |
7288 | a->last_checkpoint = 0; | |
7289 | break; | |
7290 | } | |
7291 | break; | |
7292 | default: | |
7293 | dprintf("state %i\n", map_state); | |
5802a811 | 7294 | } |
94002678 AK |
7295 | dprintf("\n"); |
7296 | ||
845dea95 NB |
7297 | } |
7298 | ||
f796af5d | 7299 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 7300 | { |
f796af5d | 7301 | void *buf = mpb; |
c2a1e7da DW |
7302 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
7303 | unsigned long long dsize; | |
7304 | unsigned long long sectors; | |
7305 | ||
7306 | get_dev_size(fd, NULL, &dsize); | |
7307 | ||
272f648f DW |
7308 | if (mpb_size > 512) { |
7309 | /* -1 to account for anchor */ | |
7310 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 7311 | |
272f648f DW |
7312 | /* write the extended mpb to the sectors preceeding the anchor */ |
7313 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
7314 | return 1; | |
c2a1e7da | 7315 | |
f21e18ca N |
7316 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
7317 | != 512 * sectors) | |
272f648f DW |
7318 | return 1; |
7319 | } | |
c2a1e7da | 7320 | |
272f648f DW |
7321 | /* first block is stored on second to last sector of the disk */ |
7322 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
7323 | return 1; |
7324 | ||
f796af5d | 7325 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
7326 | return 1; |
7327 | ||
c2a1e7da DW |
7328 | return 0; |
7329 | } | |
7330 | ||
2e735d19 | 7331 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 7332 | { |
2e735d19 | 7333 | struct intel_super *super = container->sb; |
c2a1e7da | 7334 | |
1a64be56 | 7335 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
7336 | if (!super->updates_pending) |
7337 | return; | |
7338 | ||
36988a3d | 7339 | write_super_imsm(container, 0); |
c2a1e7da DW |
7340 | |
7341 | super->updates_pending = 0; | |
845dea95 NB |
7342 | } |
7343 | ||
272906ef DW |
7344 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
7345 | { | |
7346 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7347 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
7348 | struct dl *dl; |
7349 | ||
7350 | for (dl = super->disks; dl; dl = dl->next) | |
7351 | if (dl->index == i) | |
7352 | break; | |
7353 | ||
25ed7e59 | 7354 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
7355 | dl = NULL; |
7356 | ||
7357 | if (dl) | |
7358 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
7359 | ||
7360 | return dl; | |
7361 | } | |
7362 | ||
a20d2ba5 | 7363 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
7364 | struct active_array *a, int activate_new, |
7365 | struct mdinfo *additional_test_list) | |
272906ef DW |
7366 | { |
7367 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7368 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
7369 | struct imsm_super *mpb = super->anchor; |
7370 | struct imsm_map *map; | |
272906ef DW |
7371 | unsigned long long pos; |
7372 | struct mdinfo *d; | |
7373 | struct extent *ex; | |
a20d2ba5 | 7374 | int i, j; |
272906ef | 7375 | int found; |
569cc43f DW |
7376 | __u32 array_start = 0; |
7377 | __u32 array_end = 0; | |
272906ef | 7378 | struct dl *dl; |
6c932028 | 7379 | struct mdinfo *test_list; |
272906ef DW |
7380 | |
7381 | for (dl = super->disks; dl; dl = dl->next) { | |
7382 | /* If in this array, skip */ | |
7383 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
7384 | if (d->state_fd >= 0 && |
7385 | d->disk.major == dl->major && | |
272906ef | 7386 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
7387 | dprintf("%x:%x already in array\n", |
7388 | dl->major, dl->minor); | |
272906ef DW |
7389 | break; |
7390 | } | |
7391 | if (d) | |
7392 | continue; | |
6c932028 AK |
7393 | test_list = additional_test_list; |
7394 | while (test_list) { | |
7395 | if (test_list->disk.major == dl->major && | |
7396 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
7397 | dprintf("%x:%x already in additional test list\n", |
7398 | dl->major, dl->minor); | |
7399 | break; | |
7400 | } | |
6c932028 | 7401 | test_list = test_list->next; |
8ba77d32 | 7402 | } |
6c932028 | 7403 | if (test_list) |
8ba77d32 | 7404 | continue; |
272906ef | 7405 | |
e553d2a4 | 7406 | /* skip in use or failed drives */ |
25ed7e59 | 7407 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
7408 | dl->index == -2) { |
7409 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 7410 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
7411 | continue; |
7412 | } | |
7413 | ||
a20d2ba5 DW |
7414 | /* skip pure spares when we are looking for partially |
7415 | * assimilated drives | |
7416 | */ | |
7417 | if (dl->index == -1 && !activate_new) | |
7418 | continue; | |
7419 | ||
272906ef | 7420 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 7421 | * It needs to be able to cover all member volumes |
272906ef DW |
7422 | */ |
7423 | ex = get_extents(super, dl); | |
7424 | if (!ex) { | |
7425 | dprintf("cannot get extents\n"); | |
7426 | continue; | |
7427 | } | |
a20d2ba5 DW |
7428 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7429 | dev = get_imsm_dev(super, i); | |
238c0a71 | 7430 | map = get_imsm_map(dev, MAP_0); |
272906ef | 7431 | |
a20d2ba5 DW |
7432 | /* check if this disk is already a member of |
7433 | * this array | |
272906ef | 7434 | */ |
620b1713 | 7435 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
7436 | continue; |
7437 | ||
7438 | found = 0; | |
7439 | j = 0; | |
7440 | pos = 0; | |
5551b113 | 7441 | array_start = pba_of_lba0(map); |
329c8278 | 7442 | array_end = array_start + |
5551b113 | 7443 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
7444 | |
7445 | do { | |
7446 | /* check that we can start at pba_of_lba0 with | |
7447 | * blocks_per_member of space | |
7448 | */ | |
329c8278 | 7449 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
7450 | found = 1; |
7451 | break; | |
7452 | } | |
7453 | pos = ex[j].start + ex[j].size; | |
7454 | j++; | |
7455 | } while (ex[j-1].size); | |
7456 | ||
7457 | if (!found) | |
272906ef | 7458 | break; |
a20d2ba5 | 7459 | } |
272906ef DW |
7460 | |
7461 | free(ex); | |
a20d2ba5 | 7462 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
7463 | dprintf("%x:%x does not have %u to %u available\n", |
7464 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
7465 | /* No room */ |
7466 | continue; | |
a20d2ba5 DW |
7467 | } |
7468 | return dl; | |
272906ef DW |
7469 | } |
7470 | ||
7471 | return dl; | |
7472 | } | |
7473 | ||
95d07a2c LM |
7474 | |
7475 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
7476 | { | |
7477 | struct imsm_dev *dev2; | |
7478 | struct imsm_map *map; | |
7479 | struct dl *idisk; | |
7480 | int slot; | |
7481 | int idx; | |
7482 | __u8 state; | |
7483 | ||
7484 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
7485 | if (dev2) { | |
238c0a71 | 7486 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 7487 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 7488 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
7489 | if (!map) |
7490 | return 1; | |
7491 | for (slot = 0; slot < map->num_members; slot++) { | |
7492 | /* | |
7493 | * Check if failed disks are deleted from intel | |
7494 | * disk list or are marked to be deleted | |
7495 | */ | |
238c0a71 | 7496 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
7497 | idisk = get_imsm_dl_disk(cont->sb, idx); |
7498 | /* | |
7499 | * Do not rebuild the array if failed disks | |
7500 | * from failed sub-array are not removed from | |
7501 | * container. | |
7502 | */ | |
7503 | if (idisk && | |
7504 | is_failed(&idisk->disk) && | |
7505 | (idisk->action != DISK_REMOVE)) | |
7506 | return 0; | |
7507 | } | |
7508 | } | |
7509 | } | |
7510 | return 1; | |
7511 | } | |
7512 | ||
88758e9d DW |
7513 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
7514 | struct metadata_update **updates) | |
7515 | { | |
7516 | /** | |
d23fe947 DW |
7517 | * Find a device with unused free space and use it to replace a |
7518 | * failed/vacant region in an array. We replace failed regions one a | |
7519 | * array at a time. The result is that a new spare disk will be added | |
7520 | * to the first failed array and after the monitor has finished | |
7521 | * propagating failures the remainder will be consumed. | |
88758e9d | 7522 | * |
d23fe947 DW |
7523 | * FIXME add a capability for mdmon to request spares from another |
7524 | * container. | |
88758e9d DW |
7525 | */ |
7526 | ||
7527 | struct intel_super *super = a->container->sb; | |
88758e9d | 7528 | int inst = a->info.container_member; |
949c47a0 | 7529 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7530 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
7531 | int failed = a->info.array.raid_disks; |
7532 | struct mdinfo *rv = NULL; | |
7533 | struct mdinfo *d; | |
7534 | struct mdinfo *di; | |
7535 | struct metadata_update *mu; | |
7536 | struct dl *dl; | |
7537 | struct imsm_update_activate_spare *u; | |
7538 | int num_spares = 0; | |
7539 | int i; | |
95d07a2c | 7540 | int allowed; |
88758e9d DW |
7541 | |
7542 | for (d = a->info.devs ; d ; d = d->next) { | |
7543 | if ((d->curr_state & DS_FAULTY) && | |
7544 | d->state_fd >= 0) | |
7545 | /* wait for Removal to happen */ | |
7546 | return NULL; | |
7547 | if (d->state_fd >= 0) | |
7548 | failed--; | |
7549 | } | |
7550 | ||
7551 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
7552 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 7553 | |
e2962bfc AK |
7554 | if (imsm_reshape_blocks_arrays_changes(super)) |
7555 | return NULL; | |
1af97990 | 7556 | |
fc8ca064 AK |
7557 | /* Cannot activate another spare if rebuild is in progress already |
7558 | */ | |
7559 | if (is_rebuilding(dev)) { | |
7560 | dprintf("imsm: No spare activation allowed. " | |
7561 | "Rebuild in progress already.\n"); | |
7562 | return NULL; | |
7563 | } | |
7564 | ||
89c67882 AK |
7565 | if (a->info.array.level == 4) |
7566 | /* No repair for takeovered array | |
7567 | * imsm doesn't support raid4 | |
7568 | */ | |
7569 | return NULL; | |
7570 | ||
3b451610 AK |
7571 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
7572 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
7573 | return NULL; |
7574 | ||
95d07a2c LM |
7575 | /* |
7576 | * If there are any failed disks check state of the other volume. | |
7577 | * Block rebuild if the another one is failed until failed disks | |
7578 | * are removed from container. | |
7579 | */ | |
7580 | if (failed) { | |
c4acd1e5 | 7581 | dprintf("found failed disks in %.*s, check if there another" |
95d07a2c | 7582 | "failed sub-array.\n", |
c4acd1e5 | 7583 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
7584 | /* check if states of the other volumes allow for rebuild */ |
7585 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
7586 | if (i != inst) { | |
7587 | allowed = imsm_rebuild_allowed(a->container, | |
7588 | i, failed); | |
7589 | if (!allowed) | |
7590 | return NULL; | |
7591 | } | |
7592 | } | |
7593 | } | |
7594 | ||
88758e9d | 7595 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
7596 | for (i = 0; i < a->info.array.raid_disks; i++) { |
7597 | for (d = a->info.devs ; d ; d = d->next) | |
7598 | if (d->disk.raid_disk == i) | |
7599 | break; | |
7600 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
7601 | if (d && (d->state_fd >= 0)) | |
7602 | continue; | |
7603 | ||
272906ef | 7604 | /* |
a20d2ba5 DW |
7605 | * OK, this device needs recovery. Try to re-add the |
7606 | * previous occupant of this slot, if this fails see if | |
7607 | * we can continue the assimilation of a spare that was | |
7608 | * partially assimilated, finally try to activate a new | |
7609 | * spare. | |
272906ef DW |
7610 | */ |
7611 | dl = imsm_readd(super, i, a); | |
7612 | if (!dl) | |
b303fe21 | 7613 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 7614 | if (!dl) |
b303fe21 | 7615 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
7616 | if (!dl) |
7617 | continue; | |
7618 | ||
7619 | /* found a usable disk with enough space */ | |
503975b9 | 7620 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
7621 | |
7622 | /* dl->index will be -1 in the case we are activating a | |
7623 | * pristine spare. imsm_process_update() will create a | |
7624 | * new index in this case. Once a disk is found to be | |
7625 | * failed in all member arrays it is kicked from the | |
7626 | * metadata | |
7627 | */ | |
7628 | di->disk.number = dl->index; | |
d23fe947 | 7629 | |
272906ef DW |
7630 | /* (ab)use di->devs to store a pointer to the device |
7631 | * we chose | |
7632 | */ | |
7633 | di->devs = (struct mdinfo *) dl; | |
7634 | ||
7635 | di->disk.raid_disk = i; | |
7636 | di->disk.major = dl->major; | |
7637 | di->disk.minor = dl->minor; | |
7638 | di->disk.state = 0; | |
d23534e4 | 7639 | di->recovery_start = 0; |
5551b113 | 7640 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
7641 | di->component_size = a->info.component_size; |
7642 | di->container_member = inst; | |
148acb7b | 7643 | super->random = random32(); |
272906ef DW |
7644 | di->next = rv; |
7645 | rv = di; | |
7646 | num_spares++; | |
7647 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
7648 | i, di->data_offset); | |
88758e9d DW |
7649 | } |
7650 | ||
7651 | if (!rv) | |
7652 | /* No spares found */ | |
7653 | return rv; | |
7654 | /* Now 'rv' has a list of devices to return. | |
7655 | * Create a metadata_update record to update the | |
7656 | * disk_ord_tbl for the array | |
7657 | */ | |
503975b9 N |
7658 | mu = xmalloc(sizeof(*mu)); |
7659 | mu->buf = xcalloc(num_spares, | |
7660 | sizeof(struct imsm_update_activate_spare)); | |
88758e9d | 7661 | mu->space = NULL; |
cb23f1f4 | 7662 | mu->space_list = NULL; |
88758e9d DW |
7663 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
7664 | mu->next = *updates; | |
7665 | u = (struct imsm_update_activate_spare *) mu->buf; | |
7666 | ||
7667 | for (di = rv ; di ; di = di->next) { | |
7668 | u->type = update_activate_spare; | |
d23fe947 DW |
7669 | u->dl = (struct dl *) di->devs; |
7670 | di->devs = NULL; | |
88758e9d DW |
7671 | u->slot = di->disk.raid_disk; |
7672 | u->array = inst; | |
7673 | u->next = u + 1; | |
7674 | u++; | |
7675 | } | |
7676 | (u-1)->next = NULL; | |
7677 | *updates = mu; | |
7678 | ||
7679 | return rv; | |
7680 | } | |
7681 | ||
54c2c1ea | 7682 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 7683 | { |
54c2c1ea | 7684 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
7685 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7686 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
7687 | struct disk_info *inf = get_disk_info(u); |
7688 | struct imsm_disk *disk; | |
8273f55e DW |
7689 | int i; |
7690 | int j; | |
8273f55e | 7691 | |
54c2c1ea | 7692 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 7693 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
7694 | for (j = 0; j < new_map->num_members; j++) |
7695 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
7696 | return 1; |
7697 | } | |
7698 | ||
7699 | return 0; | |
7700 | } | |
7701 | ||
1a64be56 LM |
7702 | |
7703 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
7704 | { | |
7705 | struct dl *dl = NULL; | |
7706 | for (dl = super->disks; dl; dl = dl->next) | |
7707 | if ((dl->major == major) && (dl->minor == minor)) | |
7708 | return dl; | |
7709 | return NULL; | |
7710 | } | |
7711 | ||
7712 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7713 | { | |
7714 | struct dl *prev = NULL; | |
7715 | struct dl *dl; | |
7716 | ||
7717 | prev = NULL; | |
7718 | for (dl = super->disks; dl; dl = dl->next) { | |
7719 | if ((dl->major == major) && (dl->minor == minor)) { | |
7720 | /* remove */ | |
7721 | if (prev) | |
7722 | prev->next = dl->next; | |
7723 | else | |
7724 | super->disks = dl->next; | |
7725 | dl->next = NULL; | |
7726 | __free_imsm_disk(dl); | |
7727 | dprintf("%s: removed %x:%x\n", | |
7728 | __func__, major, minor); | |
7729 | break; | |
7730 | } | |
7731 | prev = dl; | |
7732 | } | |
7733 | return 0; | |
7734 | } | |
7735 | ||
f21e18ca | 7736 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7737 | |
1a64be56 LM |
7738 | static int add_remove_disk_update(struct intel_super *super) |
7739 | { | |
7740 | int check_degraded = 0; | |
7741 | struct dl *disk = NULL; | |
7742 | /* add/remove some spares to/from the metadata/contrainer */ | |
7743 | while (super->disk_mgmt_list) { | |
7744 | struct dl *disk_cfg; | |
7745 | ||
7746 | disk_cfg = super->disk_mgmt_list; | |
7747 | super->disk_mgmt_list = disk_cfg->next; | |
7748 | disk_cfg->next = NULL; | |
7749 | ||
7750 | if (disk_cfg->action == DISK_ADD) { | |
7751 | disk_cfg->next = super->disks; | |
7752 | super->disks = disk_cfg; | |
7753 | check_degraded = 1; | |
7754 | dprintf("%s: added %x:%x\n", | |
7755 | __func__, disk_cfg->major, | |
7756 | disk_cfg->minor); | |
7757 | } else if (disk_cfg->action == DISK_REMOVE) { | |
7758 | dprintf("Disk remove action processed: %x.%x\n", | |
7759 | disk_cfg->major, disk_cfg->minor); | |
7760 | disk = get_disk_super(super, | |
7761 | disk_cfg->major, | |
7762 | disk_cfg->minor); | |
7763 | if (disk) { | |
7764 | /* store action status */ | |
7765 | disk->action = DISK_REMOVE; | |
7766 | /* remove spare disks only */ | |
7767 | if (disk->index == -1) { | |
7768 | remove_disk_super(super, | |
7769 | disk_cfg->major, | |
7770 | disk_cfg->minor); | |
7771 | } | |
7772 | } | |
7773 | /* release allocate disk structure */ | |
7774 | __free_imsm_disk(disk_cfg); | |
7775 | } | |
7776 | } | |
7777 | return check_degraded; | |
7778 | } | |
7779 | ||
a29911da PC |
7780 | |
7781 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
7782 | struct intel_super *super, | |
7783 | void ***space_list) | |
7784 | { | |
7785 | struct intel_dev *id; | |
7786 | void **tofree = NULL; | |
7787 | int ret_val = 0; | |
7788 | ||
7789 | dprintf("apply_reshape_migration_update()\n"); | |
7790 | if ((u->subdev < 0) || | |
7791 | (u->subdev > 1)) { | |
7792 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7793 | return ret_val; | |
7794 | } | |
7795 | if ((space_list == NULL) || (*space_list == NULL)) { | |
7796 | dprintf("imsm: Error: Memory is not allocated\n"); | |
7797 | return ret_val; | |
7798 | } | |
7799 | ||
7800 | for (id = super->devlist ; id; id = id->next) { | |
7801 | if (id->index == (unsigned)u->subdev) { | |
7802 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7803 | struct imsm_map *map; | |
7804 | struct imsm_dev *new_dev = | |
7805 | (struct imsm_dev *)*space_list; | |
238c0a71 | 7806 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
7807 | int to_state; |
7808 | struct dl *new_disk; | |
7809 | ||
7810 | if (new_dev == NULL) | |
7811 | return ret_val; | |
7812 | *space_list = **space_list; | |
7813 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 7814 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
7815 | if (migr_map) { |
7816 | dprintf("imsm: Error: migration in progress"); | |
7817 | return ret_val; | |
7818 | } | |
7819 | ||
7820 | to_state = map->map_state; | |
7821 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
7822 | map->num_members++; | |
7823 | /* this should not happen */ | |
7824 | if (u->new_disks[0] < 0) { | |
7825 | map->failed_disk_num = | |
7826 | map->num_members - 1; | |
7827 | to_state = IMSM_T_STATE_DEGRADED; | |
7828 | } else | |
7829 | to_state = IMSM_T_STATE_NORMAL; | |
7830 | } | |
8e59f3d8 | 7831 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
7832 | if (u->new_level > -1) |
7833 | map->raid_level = u->new_level; | |
238c0a71 | 7834 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
7835 | if ((u->new_level == 5) && |
7836 | (migr_map->raid_level == 0)) { | |
7837 | int ord = map->num_members - 1; | |
7838 | migr_map->num_members--; | |
7839 | if (u->new_disks[0] < 0) | |
7840 | ord |= IMSM_ORD_REBUILD; | |
7841 | set_imsm_ord_tbl_ent(map, | |
7842 | map->num_members - 1, | |
7843 | ord); | |
7844 | } | |
7845 | id->dev = new_dev; | |
7846 | tofree = (void **)dev; | |
7847 | ||
4bba0439 PC |
7848 | /* update chunk size |
7849 | */ | |
7850 | if (u->new_chunksize > 0) | |
7851 | map->blocks_per_strip = | |
7852 | __cpu_to_le16(u->new_chunksize * 2); | |
7853 | ||
a29911da PC |
7854 | /* add disk |
7855 | */ | |
7856 | if ((u->new_level != 5) || | |
7857 | (migr_map->raid_level != 0) || | |
7858 | (migr_map->raid_level == map->raid_level)) | |
7859 | goto skip_disk_add; | |
7860 | ||
7861 | if (u->new_disks[0] >= 0) { | |
7862 | /* use passes spare | |
7863 | */ | |
7864 | new_disk = get_disk_super(super, | |
7865 | major(u->new_disks[0]), | |
7866 | minor(u->new_disks[0])); | |
7867 | dprintf("imsm: new disk for reshape is: %i:%i " | |
7868 | "(%p, index = %i)\n", | |
7869 | major(u->new_disks[0]), | |
7870 | minor(u->new_disks[0]), | |
7871 | new_disk, new_disk->index); | |
7872 | if (new_disk == NULL) | |
7873 | goto error_disk_add; | |
7874 | ||
7875 | new_disk->index = map->num_members - 1; | |
7876 | /* slot to fill in autolayout | |
7877 | */ | |
7878 | new_disk->raiddisk = new_disk->index; | |
7879 | new_disk->disk.status |= CONFIGURED_DISK; | |
7880 | new_disk->disk.status &= ~SPARE_DISK; | |
7881 | } else | |
7882 | goto error_disk_add; | |
7883 | ||
7884 | skip_disk_add: | |
7885 | *tofree = *space_list; | |
7886 | /* calculate new size | |
7887 | */ | |
f3871fdc | 7888 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
7889 | |
7890 | ret_val = 1; | |
7891 | } | |
7892 | } | |
7893 | ||
7894 | if (tofree) | |
7895 | *space_list = tofree; | |
7896 | return ret_val; | |
7897 | ||
7898 | error_disk_add: | |
7899 | dprintf("Error: imsm: Cannot find disk.\n"); | |
7900 | return ret_val; | |
7901 | } | |
7902 | ||
f3871fdc AK |
7903 | static int apply_size_change_update(struct imsm_update_size_change *u, |
7904 | struct intel_super *super) | |
7905 | { | |
7906 | struct intel_dev *id; | |
7907 | int ret_val = 0; | |
7908 | ||
7909 | dprintf("apply_size_change_update()\n"); | |
7910 | if ((u->subdev < 0) || | |
7911 | (u->subdev > 1)) { | |
7912 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7913 | return ret_val; | |
7914 | } | |
7915 | ||
7916 | for (id = super->devlist ; id; id = id->next) { | |
7917 | if (id->index == (unsigned)u->subdev) { | |
7918 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7919 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
7920 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
7921 | unsigned long long blocks_per_member; | |
7922 | ||
7923 | /* calculate new size | |
7924 | */ | |
7925 | blocks_per_member = u->new_size / used_disks; | |
7926 | dprintf("imsm: apply_size_change_update(size: %llu, " | |
7927 | "blocks per member: %llu)\n", | |
7928 | u->new_size, blocks_per_member); | |
7929 | set_blocks_per_member(map, blocks_per_member); | |
7930 | imsm_set_array_size(dev, u->new_size); | |
7931 | ||
7932 | ret_val = 1; | |
7933 | break; | |
7934 | } | |
7935 | } | |
7936 | ||
7937 | return ret_val; | |
7938 | } | |
7939 | ||
7940 | ||
061d7da3 | 7941 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 7942 | struct intel_super *super, |
061d7da3 LO |
7943 | struct active_array *active_array) |
7944 | { | |
7945 | struct imsm_super *mpb = super->anchor; | |
7946 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 7947 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
7948 | struct imsm_map *migr_map; |
7949 | struct active_array *a; | |
7950 | struct imsm_disk *disk; | |
7951 | __u8 to_state; | |
7952 | struct dl *dl; | |
7953 | unsigned int found; | |
7954 | int failed; | |
5961eeec | 7955 | int victim; |
061d7da3 | 7956 | int i; |
5961eeec | 7957 | int second_map_created = 0; |
061d7da3 | 7958 | |
5961eeec | 7959 | for (; u; u = u->next) { |
238c0a71 | 7960 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 7961 | |
5961eeec | 7962 | if (victim < 0) |
7963 | return 0; | |
061d7da3 | 7964 | |
5961eeec | 7965 | for (dl = super->disks; dl; dl = dl->next) |
7966 | if (dl == u->dl) | |
7967 | break; | |
061d7da3 | 7968 | |
5961eeec | 7969 | if (!dl) { |
7970 | fprintf(stderr, "error: imsm_activate_spare passed " | |
7971 | "an unknown disk (index: %d)\n", | |
7972 | u->dl->index); | |
7973 | return 0; | |
7974 | } | |
061d7da3 | 7975 | |
5961eeec | 7976 | /* count failures (excluding rebuilds and the victim) |
7977 | * to determine map[0] state | |
7978 | */ | |
7979 | failed = 0; | |
7980 | for (i = 0; i < map->num_members; i++) { | |
7981 | if (i == u->slot) | |
7982 | continue; | |
7983 | disk = get_imsm_disk(super, | |
238c0a71 | 7984 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 7985 | if (!disk || is_failed(disk)) |
7986 | failed++; | |
7987 | } | |
061d7da3 | 7988 | |
5961eeec | 7989 | /* adding a pristine spare, assign a new index */ |
7990 | if (dl->index < 0) { | |
7991 | dl->index = super->anchor->num_disks; | |
7992 | super->anchor->num_disks++; | |
7993 | } | |
7994 | disk = &dl->disk; | |
7995 | disk->status |= CONFIGURED_DISK; | |
7996 | disk->status &= ~SPARE_DISK; | |
7997 | ||
7998 | /* mark rebuild */ | |
238c0a71 | 7999 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 8000 | if (!second_map_created) { |
8001 | second_map_created = 1; | |
8002 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8003 | migrate(dev, super, to_state, MIGR_REBUILD); | |
8004 | } else | |
8005 | map->map_state = to_state; | |
238c0a71 | 8006 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 8007 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
8008 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
8009 | dl->index | IMSM_ORD_REBUILD); | |
8010 | ||
8011 | /* update the family_num to mark a new container | |
8012 | * generation, being careful to record the existing | |
8013 | * family_num in orig_family_num to clean up after | |
8014 | * earlier mdadm versions that neglected to set it. | |
8015 | */ | |
8016 | if (mpb->orig_family_num == 0) | |
8017 | mpb->orig_family_num = mpb->family_num; | |
8018 | mpb->family_num += super->random; | |
8019 | ||
8020 | /* count arrays using the victim in the metadata */ | |
8021 | found = 0; | |
8022 | for (a = active_array; a ; a = a->next) { | |
8023 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8024 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 8025 | |
5961eeec | 8026 | if (get_imsm_disk_slot(map, victim) >= 0) |
8027 | found++; | |
8028 | } | |
061d7da3 | 8029 | |
5961eeec | 8030 | /* delete the victim if it is no longer being |
8031 | * utilized anywhere | |
061d7da3 | 8032 | */ |
5961eeec | 8033 | if (!found) { |
8034 | struct dl **dlp; | |
061d7da3 | 8035 | |
5961eeec | 8036 | /* We know that 'manager' isn't touching anything, |
8037 | * so it is safe to delete | |
8038 | */ | |
8039 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
8040 | if ((*dlp)->index == victim) |
8041 | break; | |
5961eeec | 8042 | |
8043 | /* victim may be on the missing list */ | |
8044 | if (!*dlp) | |
8045 | for (dlp = &super->missing; *dlp; | |
8046 | dlp = &(*dlp)->next) | |
8047 | if ((*dlp)->index == victim) | |
8048 | break; | |
8049 | imsm_delete(super, dlp, victim); | |
8050 | } | |
061d7da3 LO |
8051 | } |
8052 | ||
8053 | return 1; | |
8054 | } | |
a29911da | 8055 | |
2e5dc010 N |
8056 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
8057 | struct intel_super *super, | |
8058 | void ***space_list) | |
8059 | { | |
8060 | struct dl *new_disk; | |
8061 | struct intel_dev *id; | |
8062 | int i; | |
8063 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 8064 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
8065 | void **tofree = NULL; |
8066 | int devices_to_reshape = 1; | |
8067 | struct imsm_super *mpb = super->anchor; | |
8068 | int ret_val = 0; | |
d098291a | 8069 | unsigned int dev_id; |
2e5dc010 | 8070 | |
ed7333bd | 8071 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
8072 | |
8073 | /* enable spares to use in array */ | |
8074 | for (i = 0; i < delta_disks; i++) { | |
8075 | new_disk = get_disk_super(super, | |
8076 | major(u->new_disks[i]), | |
8077 | minor(u->new_disks[i])); | |
ed7333bd AK |
8078 | dprintf("imsm: new disk for reshape is: %i:%i " |
8079 | "(%p, index = %i)\n", | |
2e5dc010 N |
8080 | major(u->new_disks[i]), minor(u->new_disks[i]), |
8081 | new_disk, new_disk->index); | |
8082 | if ((new_disk == NULL) || | |
8083 | ((new_disk->index >= 0) && | |
8084 | (new_disk->index < u->old_raid_disks))) | |
8085 | goto update_reshape_exit; | |
ee4beede | 8086 | new_disk->index = disk_count++; |
2e5dc010 N |
8087 | /* slot to fill in autolayout |
8088 | */ | |
8089 | new_disk->raiddisk = new_disk->index; | |
8090 | new_disk->disk.status |= | |
8091 | CONFIGURED_DISK; | |
8092 | new_disk->disk.status &= ~SPARE_DISK; | |
8093 | } | |
8094 | ||
ed7333bd AK |
8095 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
8096 | mpb->num_raid_devs); | |
2e5dc010 N |
8097 | /* manage changes in volume |
8098 | */ | |
d098291a | 8099 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
8100 | void **sp = *space_list; |
8101 | struct imsm_dev *newdev; | |
8102 | struct imsm_map *newmap, *oldmap; | |
8103 | ||
d098291a AK |
8104 | for (id = super->devlist ; id; id = id->next) { |
8105 | if (id->index == dev_id) | |
8106 | break; | |
8107 | } | |
8108 | if (id == NULL) | |
8109 | break; | |
2e5dc010 N |
8110 | if (!sp) |
8111 | continue; | |
8112 | *space_list = *sp; | |
8113 | newdev = (void*)sp; | |
8114 | /* Copy the dev, but not (all of) the map */ | |
8115 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
8116 | oldmap = get_imsm_map(id->dev, MAP_0); |
8117 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
8118 | /* Copy the current map */ |
8119 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
8120 | /* update one device only | |
8121 | */ | |
8122 | if (devices_to_reshape) { | |
ed7333bd AK |
8123 | dprintf("imsm: modifying subdev: %i\n", |
8124 | id->index); | |
2e5dc010 N |
8125 | devices_to_reshape--; |
8126 | newdev->vol.migr_state = 1; | |
8127 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8128 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8129 | newmap->num_members = u->new_raid_disks; |
8130 | for (i = 0; i < delta_disks; i++) { | |
8131 | set_imsm_ord_tbl_ent(newmap, | |
8132 | u->old_raid_disks + i, | |
8133 | u->old_raid_disks + i); | |
8134 | } | |
8135 | /* New map is correct, now need to save old map | |
8136 | */ | |
238c0a71 | 8137 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8138 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8139 | ||
f3871fdc | 8140 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
8141 | } |
8142 | ||
8143 | sp = (void **)id->dev; | |
8144 | id->dev = newdev; | |
8145 | *sp = tofree; | |
8146 | tofree = sp; | |
8e59f3d8 AK |
8147 | |
8148 | /* Clear migration record */ | |
8149 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8150 | } |
819bc634 AK |
8151 | if (tofree) |
8152 | *space_list = tofree; | |
2e5dc010 N |
8153 | ret_val = 1; |
8154 | ||
8155 | update_reshape_exit: | |
8156 | ||
8157 | return ret_val; | |
8158 | } | |
8159 | ||
bb025c2f | 8160 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8161 | struct intel_super *super, |
8162 | void ***space_list) | |
bb025c2f KW |
8163 | { |
8164 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8165 | struct intel_dev *dv; |
8166 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8167 | struct imsm_map *map; |
8168 | struct dl *dm, *du; | |
8ca6df95 | 8169 | int i; |
bb025c2f KW |
8170 | |
8171 | for (dv = super->devlist; dv; dv = dv->next) | |
8172 | if (dv->index == (unsigned int)u->subarray) { | |
8173 | dev = dv->dev; | |
8174 | break; | |
8175 | } | |
8176 | ||
8177 | if (dev == NULL) | |
8178 | return 0; | |
8179 | ||
238c0a71 | 8180 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8181 | |
8182 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 8183 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
8184 | if (imsm_count_failed(super, dev, MAP_0) != |
8185 | (map->num_members / 2)) | |
43d5ec18 KW |
8186 | return 0; |
8187 | ||
bb025c2f KW |
8188 | /* iterate through devices to mark removed disks as spare */ |
8189 | for (dm = super->disks; dm; dm = dm->next) { | |
8190 | if (dm->disk.status & FAILED_DISK) { | |
8191 | int idx = dm->index; | |
8192 | /* update indexes on the disk list */ | |
8193 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
8194 | the index values will end up being correct.... NB */ | |
8195 | for (du = super->disks; du; du = du->next) | |
8196 | if (du->index > idx) | |
8197 | du->index--; | |
8198 | /* mark as spare disk */ | |
a8619d23 | 8199 | mark_spare(dm); |
bb025c2f KW |
8200 | } |
8201 | } | |
bb025c2f KW |
8202 | /* update map */ |
8203 | map->num_members = map->num_members / 2; | |
8204 | map->map_state = IMSM_T_STATE_NORMAL; | |
8205 | map->num_domains = 1; | |
8206 | map->raid_level = 0; | |
8207 | map->failed_disk_num = -1; | |
8208 | } | |
8209 | ||
8ca6df95 KW |
8210 | if (u->direction == R0_TO_R10) { |
8211 | void **space; | |
8212 | /* update slots in current disk list */ | |
8213 | for (dm = super->disks; dm; dm = dm->next) { | |
8214 | if (dm->index >= 0) | |
8215 | dm->index *= 2; | |
8216 | } | |
8217 | /* create new *missing* disks */ | |
8218 | for (i = 0; i < map->num_members; i++) { | |
8219 | space = *space_list; | |
8220 | if (!space) | |
8221 | continue; | |
8222 | *space_list = *space; | |
8223 | du = (void *)space; | |
8224 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
8225 | du->fd = -1; |
8226 | du->minor = 0; | |
8227 | du->major = 0; | |
8228 | du->index = (i * 2) + 1; | |
8229 | sprintf((char *)du->disk.serial, | |
8230 | " MISSING_%d", du->index); | |
8231 | sprintf((char *)du->serial, | |
8232 | "MISSING_%d", du->index); | |
8233 | du->next = super->missing; | |
8234 | super->missing = du; | |
8235 | } | |
8236 | /* create new dev and map */ | |
8237 | space = *space_list; | |
8238 | if (!space) | |
8239 | return 0; | |
8240 | *space_list = *space; | |
8241 | dev_new = (void *)space; | |
8242 | memcpy(dev_new, dev, sizeof(*dev)); | |
8243 | /* update new map */ | |
238c0a71 | 8244 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 8245 | map->num_members = map->num_members * 2; |
1a2487c2 | 8246 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
8247 | map->num_domains = 2; |
8248 | map->raid_level = 1; | |
8249 | /* replace dev<->dev_new */ | |
8250 | dv->dev = dev_new; | |
8251 | } | |
bb025c2f KW |
8252 | /* update disk order table */ |
8253 | for (du = super->disks; du; du = du->next) | |
8254 | if (du->index >= 0) | |
8255 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 8256 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
8257 | if (du->index >= 0) { |
8258 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 8259 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 8260 | } |
bb025c2f KW |
8261 | |
8262 | return 1; | |
8263 | } | |
8264 | ||
e8319a19 DW |
8265 | static void imsm_process_update(struct supertype *st, |
8266 | struct metadata_update *update) | |
8267 | { | |
8268 | /** | |
8269 | * crack open the metadata_update envelope to find the update record | |
8270 | * update can be one of: | |
d195167d AK |
8271 | * update_reshape_container_disks - all the arrays in the container |
8272 | * are being reshaped to have more devices. We need to mark | |
8273 | * the arrays for general migration and convert selected spares | |
8274 | * into active devices. | |
8275 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
8276 | * device in an array, update the disk_ord_tbl. If this disk is |
8277 | * present in all member arrays then also clear the SPARE_DISK | |
8278 | * flag | |
d195167d AK |
8279 | * update_create_array |
8280 | * update_kill_array | |
8281 | * update_rename_array | |
8282 | * update_add_remove_disk | |
e8319a19 DW |
8283 | */ |
8284 | struct intel_super *super = st->sb; | |
4d7b1503 | 8285 | struct imsm_super *mpb; |
e8319a19 DW |
8286 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
8287 | ||
4d7b1503 DW |
8288 | /* update requires a larger buf but the allocation failed */ |
8289 | if (super->next_len && !super->next_buf) { | |
8290 | super->next_len = 0; | |
8291 | return; | |
8292 | } | |
8293 | ||
8294 | if (super->next_buf) { | |
8295 | memcpy(super->next_buf, super->buf, super->len); | |
8296 | free(super->buf); | |
8297 | super->len = super->next_len; | |
8298 | super->buf = super->next_buf; | |
8299 | ||
8300 | super->next_len = 0; | |
8301 | super->next_buf = NULL; | |
8302 | } | |
8303 | ||
8304 | mpb = super->anchor; | |
8305 | ||
e8319a19 | 8306 | switch (type) { |
0ec5d470 AK |
8307 | case update_general_migration_checkpoint: { |
8308 | struct intel_dev *id; | |
8309 | struct imsm_update_general_migration_checkpoint *u = | |
8310 | (void *)update->buf; | |
8311 | ||
8312 | dprintf("imsm: process_update() " | |
8313 | "for update_general_migration_checkpoint called\n"); | |
8314 | ||
8315 | /* find device under general migration */ | |
8316 | for (id = super->devlist ; id; id = id->next) { | |
8317 | if (is_gen_migration(id->dev)) { | |
8318 | id->dev->vol.curr_migr_unit = | |
8319 | __cpu_to_le32(u->curr_migr_unit); | |
8320 | super->updates_pending++; | |
8321 | } | |
8322 | } | |
8323 | break; | |
8324 | } | |
bb025c2f KW |
8325 | case update_takeover: { |
8326 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
8327 | if (apply_takeover_update(u, super, &update->space_list)) { |
8328 | imsm_update_version_info(super); | |
bb025c2f | 8329 | super->updates_pending++; |
1a2487c2 | 8330 | } |
bb025c2f KW |
8331 | break; |
8332 | } | |
8333 | ||
78b10e66 | 8334 | case update_reshape_container_disks: { |
d195167d | 8335 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
8336 | if (apply_reshape_container_disks_update( |
8337 | u, super, &update->space_list)) | |
8338 | super->updates_pending++; | |
78b10e66 N |
8339 | break; |
8340 | } | |
48c5303a | 8341 | case update_reshape_migration: { |
a29911da PC |
8342 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
8343 | if (apply_reshape_migration_update( | |
8344 | u, super, &update->space_list)) | |
8345 | super->updates_pending++; | |
48c5303a PC |
8346 | break; |
8347 | } | |
f3871fdc AK |
8348 | case update_size_change: { |
8349 | struct imsm_update_size_change *u = (void *)update->buf; | |
8350 | if (apply_size_change_update(u, super)) | |
8351 | super->updates_pending++; | |
8352 | break; | |
8353 | } | |
e8319a19 DW |
8354 | case update_activate_spare: { |
8355 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
061d7da3 LO |
8356 | if (apply_update_activate_spare(u, super, st->arrays)) |
8357 | super->updates_pending++; | |
8273f55e DW |
8358 | break; |
8359 | } | |
8360 | case update_create_array: { | |
8361 | /* someone wants to create a new array, we need to be aware of | |
8362 | * a few races/collisions: | |
8363 | * 1/ 'Create' called by two separate instances of mdadm | |
8364 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
8365 | * devices that have since been assimilated via | |
8366 | * activate_spare. | |
8367 | * In the event this update can not be carried out mdadm will | |
8368 | * (FIX ME) notice that its update did not take hold. | |
8369 | */ | |
8370 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8371 | struct intel_dev *dv; |
8273f55e DW |
8372 | struct imsm_dev *dev; |
8373 | struct imsm_map *map, *new_map; | |
8374 | unsigned long long start, end; | |
8375 | unsigned long long new_start, new_end; | |
8376 | int i; | |
54c2c1ea DW |
8377 | struct disk_info *inf; |
8378 | struct dl *dl; | |
8273f55e DW |
8379 | |
8380 | /* handle racing creates: first come first serve */ | |
8381 | if (u->dev_idx < mpb->num_raid_devs) { | |
8382 | dprintf("%s: subarray %d already defined\n", | |
8383 | __func__, u->dev_idx); | |
ba2de7ba | 8384 | goto create_error; |
8273f55e DW |
8385 | } |
8386 | ||
8387 | /* check update is next in sequence */ | |
8388 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
8389 | dprintf("%s: can not create array %d expected index %d\n", |
8390 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 8391 | goto create_error; |
8273f55e DW |
8392 | } |
8393 | ||
238c0a71 | 8394 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
8395 | new_start = pba_of_lba0(new_map); |
8396 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 8397 | inf = get_disk_info(u); |
8273f55e DW |
8398 | |
8399 | /* handle activate_spare versus create race: | |
8400 | * check to make sure that overlapping arrays do not include | |
8401 | * overalpping disks | |
8402 | */ | |
8403 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 8404 | dev = get_imsm_dev(super, i); |
238c0a71 | 8405 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
8406 | start = pba_of_lba0(map); |
8407 | end = start + blocks_per_member(map); | |
8273f55e DW |
8408 | if ((new_start >= start && new_start <= end) || |
8409 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
8410 | /* overlap */; |
8411 | else | |
8412 | continue; | |
8413 | ||
8414 | if (disks_overlap(super, i, u)) { | |
8273f55e | 8415 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 8416 | goto create_error; |
8273f55e DW |
8417 | } |
8418 | } | |
8273f55e | 8419 | |
949c47a0 DW |
8420 | /* check that prepare update was successful */ |
8421 | if (!update->space) { | |
8422 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 8423 | goto create_error; |
949c47a0 DW |
8424 | } |
8425 | ||
54c2c1ea DW |
8426 | /* check that all disks are still active before committing |
8427 | * changes. FIXME: could we instead handle this by creating a | |
8428 | * degraded array? That's probably not what the user expects, | |
8429 | * so better to drop this update on the floor. | |
8430 | */ | |
8431 | for (i = 0; i < new_map->num_members; i++) { | |
8432 | dl = serial_to_dl(inf[i].serial, super); | |
8433 | if (!dl) { | |
8434 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 8435 | goto create_error; |
54c2c1ea | 8436 | } |
949c47a0 DW |
8437 | } |
8438 | ||
8273f55e | 8439 | super->updates_pending++; |
54c2c1ea DW |
8440 | |
8441 | /* convert spares to members and fixup ord_tbl */ | |
8442 | for (i = 0; i < new_map->num_members; i++) { | |
8443 | dl = serial_to_dl(inf[i].serial, super); | |
8444 | if (dl->index == -1) { | |
8445 | dl->index = mpb->num_disks; | |
8446 | mpb->num_disks++; | |
8447 | dl->disk.status |= CONFIGURED_DISK; | |
8448 | dl->disk.status &= ~SPARE_DISK; | |
8449 | } | |
8450 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
8451 | } | |
8452 | ||
ba2de7ba DW |
8453 | dv = update->space; |
8454 | dev = dv->dev; | |
949c47a0 DW |
8455 | update->space = NULL; |
8456 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
8457 | dv->index = u->dev_idx; |
8458 | dv->next = super->devlist; | |
8459 | super->devlist = dv; | |
8273f55e | 8460 | mpb->num_raid_devs++; |
8273f55e | 8461 | |
4d1313e9 | 8462 | imsm_update_version_info(super); |
8273f55e | 8463 | break; |
ba2de7ba DW |
8464 | create_error: |
8465 | /* mdmon knows how to release update->space, but not | |
8466 | * ((struct intel_dev *) update->space)->dev | |
8467 | */ | |
8468 | if (update->space) { | |
8469 | dv = update->space; | |
8470 | free(dv->dev); | |
8471 | } | |
8273f55e | 8472 | break; |
e8319a19 | 8473 | } |
33414a01 DW |
8474 | case update_kill_array: { |
8475 | struct imsm_update_kill_array *u = (void *) update->buf; | |
8476 | int victim = u->dev_idx; | |
8477 | struct active_array *a; | |
8478 | struct intel_dev **dp; | |
8479 | struct imsm_dev *dev; | |
8480 | ||
8481 | /* sanity check that we are not affecting the uuid of | |
8482 | * active arrays, or deleting an active array | |
8483 | * | |
8484 | * FIXME when immutable ids are available, but note that | |
8485 | * we'll also need to fixup the invalidated/active | |
8486 | * subarray indexes in mdstat | |
8487 | */ | |
8488 | for (a = st->arrays; a; a = a->next) | |
8489 | if (a->info.container_member >= victim) | |
8490 | break; | |
8491 | /* by definition if mdmon is running at least one array | |
8492 | * is active in the container, so checking | |
8493 | * mpb->num_raid_devs is just extra paranoia | |
8494 | */ | |
8495 | dev = get_imsm_dev(super, victim); | |
8496 | if (a || !dev || mpb->num_raid_devs == 1) { | |
8497 | dprintf("failed to delete subarray-%d\n", victim); | |
8498 | break; | |
8499 | } | |
8500 | ||
8501 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 8502 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
8503 | *dp = (*dp)->next; |
8504 | } else { | |
f21e18ca | 8505 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
8506 | (*dp)->index--; |
8507 | dp = &(*dp)->next; | |
8508 | } | |
8509 | mpb->num_raid_devs--; | |
8510 | super->updates_pending++; | |
8511 | break; | |
8512 | } | |
aa534678 DW |
8513 | case update_rename_array: { |
8514 | struct imsm_update_rename_array *u = (void *) update->buf; | |
8515 | char name[MAX_RAID_SERIAL_LEN+1]; | |
8516 | int target = u->dev_idx; | |
8517 | struct active_array *a; | |
8518 | struct imsm_dev *dev; | |
8519 | ||
8520 | /* sanity check that we are not affecting the uuid of | |
8521 | * an active array | |
8522 | */ | |
8523 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
8524 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
8525 | for (a = st->arrays; a; a = a->next) | |
8526 | if (a->info.container_member == target) | |
8527 | break; | |
8528 | dev = get_imsm_dev(super, u->dev_idx); | |
8529 | if (a || !dev || !check_name(super, name, 1)) { | |
8530 | dprintf("failed to rename subarray-%d\n", target); | |
8531 | break; | |
8532 | } | |
8533 | ||
cdbe98cd | 8534 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
8535 | super->updates_pending++; |
8536 | break; | |
8537 | } | |
1a64be56 | 8538 | case update_add_remove_disk: { |
43dad3d6 | 8539 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
8540 | * being added, check teh status of add_remove_disk |
8541 | * if discs has been added. | |
8542 | */ | |
8543 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 8544 | struct active_array *a; |
072b727f DW |
8545 | |
8546 | super->updates_pending++; | |
1a64be56 | 8547 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
8548 | a->check_degraded = 1; |
8549 | } | |
43dad3d6 | 8550 | break; |
e8319a19 | 8551 | } |
1a64be56 LM |
8552 | default: |
8553 | fprintf(stderr, "error: unsuported process update type:" | |
8554 | "(type: %d)\n", type); | |
8555 | } | |
e8319a19 | 8556 | } |
88758e9d | 8557 | |
bc0b9d34 PC |
8558 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
8559 | ||
8273f55e DW |
8560 | static void imsm_prepare_update(struct supertype *st, |
8561 | struct metadata_update *update) | |
8562 | { | |
949c47a0 | 8563 | /** |
4d7b1503 DW |
8564 | * Allocate space to hold new disk entries, raid-device entries or a new |
8565 | * mpb if necessary. The manager synchronously waits for updates to | |
8566 | * complete in the monitor, so new mpb buffers allocated here can be | |
8567 | * integrated by the monitor thread without worrying about live pointers | |
8568 | * in the manager thread. | |
8273f55e | 8569 | */ |
949c47a0 | 8570 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
8571 | struct intel_super *super = st->sb; |
8572 | struct imsm_super *mpb = super->anchor; | |
8573 | size_t buf_len; | |
8574 | size_t len = 0; | |
949c47a0 DW |
8575 | |
8576 | switch (type) { | |
0ec5d470 AK |
8577 | case update_general_migration_checkpoint: |
8578 | dprintf("imsm: prepare_update() " | |
8579 | "for update_general_migration_checkpoint called\n"); | |
8580 | break; | |
abedf5fc KW |
8581 | case update_takeover: { |
8582 | struct imsm_update_takeover *u = (void *)update->buf; | |
8583 | if (u->direction == R0_TO_R10) { | |
8584 | void **tail = (void **)&update->space_list; | |
8585 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 8586 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
8587 | int num_members = map->num_members; |
8588 | void *space; | |
8589 | int size, i; | |
abedf5fc KW |
8590 | /* allocate memory for added disks */ |
8591 | for (i = 0; i < num_members; i++) { | |
8592 | size = sizeof(struct dl); | |
503975b9 | 8593 | space = xmalloc(size); |
abedf5fc KW |
8594 | *tail = space; |
8595 | tail = space; | |
8596 | *tail = NULL; | |
8597 | } | |
8598 | /* allocate memory for new device */ | |
8599 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
8600 | (num_members * sizeof(__u32)); | |
503975b9 N |
8601 | space = xmalloc(size); |
8602 | *tail = space; | |
8603 | tail = space; | |
8604 | *tail = NULL; | |
8605 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
8606 | } |
8607 | ||
8608 | break; | |
8609 | } | |
78b10e66 | 8610 | case update_reshape_container_disks: { |
d195167d AK |
8611 | /* Every raid device in the container is about to |
8612 | * gain some more devices, and we will enter a | |
8613 | * reconfiguration. | |
8614 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
8615 | * will now hold 2 of them. | |
8616 | * Thus we need new 'struct imsm_dev' allocations sized | |
8617 | * as sizeof_imsm_dev but with more devices in both maps. | |
8618 | */ | |
8619 | struct imsm_update_reshape *u = (void *)update->buf; | |
8620 | struct intel_dev *dl; | |
8621 | void **space_tail = (void**)&update->space_list; | |
8622 | ||
8623 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8624 | ||
8625 | for (dl = super->devlist; dl; dl = dl->next) { | |
8626 | int size = sizeof_imsm_dev(dl->dev, 1); | |
8627 | void *s; | |
d677e0b8 AK |
8628 | if (u->new_raid_disks > u->old_raid_disks) |
8629 | size += sizeof(__u32)*2* | |
8630 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8631 | s = xmalloc(size); |
d195167d AK |
8632 | *space_tail = s; |
8633 | space_tail = s; | |
8634 | *space_tail = NULL; | |
8635 | } | |
8636 | ||
8637 | len = disks_to_mpb_size(u->new_raid_disks); | |
8638 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
8639 | break; |
8640 | } | |
48c5303a | 8641 | case update_reshape_migration: { |
bc0b9d34 PC |
8642 | /* for migration level 0->5 we need to add disks |
8643 | * so the same as for container operation we will copy | |
8644 | * device to the bigger location. | |
8645 | * in memory prepared device and new disk area are prepared | |
8646 | * for usage in process update | |
8647 | */ | |
8648 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
8649 | struct intel_dev *id; | |
8650 | void **space_tail = (void **)&update->space_list; | |
8651 | int size; | |
8652 | void *s; | |
8653 | int current_level = -1; | |
8654 | ||
8655 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8656 | ||
8657 | /* add space for bigger array in update | |
8658 | */ | |
8659 | for (id = super->devlist; id; id = id->next) { | |
8660 | if (id->index == (unsigned)u->subdev) { | |
8661 | size = sizeof_imsm_dev(id->dev, 1); | |
8662 | if (u->new_raid_disks > u->old_raid_disks) | |
8663 | size += sizeof(__u32)*2* | |
8664 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8665 | s = xmalloc(size); |
bc0b9d34 PC |
8666 | *space_tail = s; |
8667 | space_tail = s; | |
8668 | *space_tail = NULL; | |
8669 | break; | |
8670 | } | |
8671 | } | |
8672 | if (update->space_list == NULL) | |
8673 | break; | |
8674 | ||
8675 | /* add space for disk in update | |
8676 | */ | |
8677 | size = sizeof(struct dl); | |
503975b9 | 8678 | s = xmalloc(size); |
bc0b9d34 PC |
8679 | *space_tail = s; |
8680 | space_tail = s; | |
8681 | *space_tail = NULL; | |
8682 | ||
8683 | /* add spare device to update | |
8684 | */ | |
8685 | for (id = super->devlist ; id; id = id->next) | |
8686 | if (id->index == (unsigned)u->subdev) { | |
8687 | struct imsm_dev *dev; | |
8688 | struct imsm_map *map; | |
8689 | ||
8690 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 8691 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
8692 | current_level = map->raid_level; |
8693 | break; | |
8694 | } | |
8695 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
8696 | struct mdinfo *spares; | |
8697 | ||
8698 | spares = get_spares_for_grow(st); | |
8699 | if (spares) { | |
8700 | struct dl *dl; | |
8701 | struct mdinfo *dev; | |
8702 | ||
8703 | dev = spares->devs; | |
8704 | if (dev) { | |
8705 | u->new_disks[0] = | |
8706 | makedev(dev->disk.major, | |
8707 | dev->disk.minor); | |
8708 | dl = get_disk_super(super, | |
8709 | dev->disk.major, | |
8710 | dev->disk.minor); | |
8711 | dl->index = u->old_raid_disks; | |
8712 | dev = dev->next; | |
8713 | } | |
8714 | sysfs_free(spares); | |
8715 | } | |
8716 | } | |
8717 | len = disks_to_mpb_size(u->new_raid_disks); | |
8718 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8719 | break; |
8720 | } | |
f3871fdc AK |
8721 | case update_size_change: { |
8722 | break; | |
8723 | } | |
949c47a0 DW |
8724 | case update_create_array: { |
8725 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8726 | struct intel_dev *dv; |
54c2c1ea | 8727 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8728 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8729 | struct dl *dl; |
8730 | struct disk_info *inf; | |
8731 | int i; | |
8732 | int activate = 0; | |
949c47a0 | 8733 | |
54c2c1ea DW |
8734 | inf = get_disk_info(u); |
8735 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 8736 | /* allocate a new super->devlist entry */ |
503975b9 N |
8737 | dv = xmalloc(sizeof(*dv)); |
8738 | dv->dev = xmalloc(len); | |
8739 | update->space = dv; | |
949c47a0 | 8740 | |
54c2c1ea DW |
8741 | /* count how many spares will be converted to members */ |
8742 | for (i = 0; i < map->num_members; i++) { | |
8743 | dl = serial_to_dl(inf[i].serial, super); | |
8744 | if (!dl) { | |
8745 | /* hmm maybe it failed?, nothing we can do about | |
8746 | * it here | |
8747 | */ | |
8748 | continue; | |
8749 | } | |
8750 | if (count_memberships(dl, super) == 0) | |
8751 | activate++; | |
8752 | } | |
8753 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
8754 | break; |
8755 | default: | |
8756 | break; | |
8757 | } | |
8758 | } | |
8273f55e | 8759 | |
4d7b1503 DW |
8760 | /* check if we need a larger metadata buffer */ |
8761 | if (super->next_buf) | |
8762 | buf_len = super->next_len; | |
8763 | else | |
8764 | buf_len = super->len; | |
8765 | ||
8766 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
8767 | /* ok we need a larger buf than what is currently allocated | |
8768 | * if this allocation fails process_update will notice that | |
8769 | * ->next_len is set and ->next_buf is NULL | |
8770 | */ | |
8771 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
8772 | if (super->next_buf) | |
8773 | free(super->next_buf); | |
8774 | ||
8775 | super->next_len = buf_len; | |
1f45a8ad DW |
8776 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
8777 | memset(super->next_buf, 0, buf_len); | |
8778 | else | |
4d7b1503 DW |
8779 | super->next_buf = NULL; |
8780 | } | |
8273f55e DW |
8781 | } |
8782 | ||
ae6aad82 | 8783 | /* must be called while manager is quiesced */ |
f21e18ca | 8784 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
8785 | { |
8786 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
8787 | struct dl *iter; |
8788 | struct imsm_dev *dev; | |
8789 | struct imsm_map *map; | |
24565c9a DW |
8790 | int i, j, num_members; |
8791 | __u32 ord; | |
ae6aad82 | 8792 | |
24565c9a DW |
8793 | dprintf("%s: deleting device[%d] from imsm_super\n", |
8794 | __func__, index); | |
ae6aad82 DW |
8795 | |
8796 | /* shift all indexes down one */ | |
8797 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 8798 | if (iter->index > (int)index) |
ae6aad82 | 8799 | iter->index--; |
47ee5a45 | 8800 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 8801 | if (iter->index > (int)index) |
47ee5a45 | 8802 | iter->index--; |
ae6aad82 DW |
8803 | |
8804 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8805 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8806 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
8807 | num_members = map->num_members; |
8808 | for (j = 0; j < num_members; j++) { | |
8809 | /* update ord entries being careful not to propagate | |
8810 | * ord-flags to the first map | |
8811 | */ | |
238c0a71 | 8812 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 8813 | |
24565c9a DW |
8814 | if (ord_to_idx(ord) <= index) |
8815 | continue; | |
ae6aad82 | 8816 | |
238c0a71 | 8817 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 8818 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 8819 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
8820 | if (map) |
8821 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
8822 | } |
8823 | } | |
8824 | ||
8825 | mpb->num_disks--; | |
8826 | super->updates_pending++; | |
24565c9a DW |
8827 | if (*dlp) { |
8828 | struct dl *dl = *dlp; | |
8829 | ||
8830 | *dlp = (*dlp)->next; | |
8831 | __free_imsm_disk(dl); | |
8832 | } | |
ae6aad82 | 8833 | } |
9e2d750d | 8834 | #endif /* MDASSEMBLE */ |
9a717282 AK |
8835 | |
8836 | static void close_targets(int *targets, int new_disks) | |
8837 | { | |
8838 | int i; | |
8839 | ||
8840 | if (!targets) | |
8841 | return; | |
8842 | ||
8843 | for (i = 0; i < new_disks; i++) { | |
8844 | if (targets[i] >= 0) { | |
8845 | close(targets[i]); | |
8846 | targets[i] = -1; | |
8847 | } | |
8848 | } | |
8849 | } | |
8850 | ||
8851 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
8852 | struct intel_super *super, | |
8853 | struct imsm_dev *dev) | |
8854 | { | |
8855 | switch (level) { | |
bf5cf7c7 | 8856 | case 1: |
9a717282 AK |
8857 | case 10:{ |
8858 | int ret_val = 0; | |
8859 | struct imsm_map *map; | |
8860 | int i; | |
8861 | ||
8862 | ret_val = raid_disks/2; | |
8863 | /* check map if all disks pairs not failed | |
8864 | * in both maps | |
8865 | */ | |
238c0a71 | 8866 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
8867 | for (i = 0; i < ret_val; i++) { |
8868 | int degradation = 0; | |
8869 | if (get_imsm_disk(super, i) == NULL) | |
8870 | degradation++; | |
8871 | if (get_imsm_disk(super, i + 1) == NULL) | |
8872 | degradation++; | |
8873 | if (degradation == 2) | |
8874 | return 0; | |
8875 | } | |
238c0a71 | 8876 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
8877 | /* if there is no second map |
8878 | * result can be returned | |
8879 | */ | |
8880 | if (map == NULL) | |
8881 | return ret_val; | |
8882 | /* check degradation in second map | |
8883 | */ | |
8884 | for (i = 0; i < ret_val; i++) { | |
8885 | int degradation = 0; | |
8886 | if (get_imsm_disk(super, i) == NULL) | |
8887 | degradation++; | |
8888 | if (get_imsm_disk(super, i + 1) == NULL) | |
8889 | degradation++; | |
8890 | if (degradation == 2) | |
8891 | return 0; | |
8892 | } | |
8893 | return ret_val; | |
8894 | } | |
8895 | case 5: | |
8896 | return 1; | |
8897 | case 6: | |
8898 | return 2; | |
8899 | default: | |
8900 | return 0; | |
8901 | } | |
8902 | } | |
8903 | ||
8904 | ||
687629c2 AK |
8905 | /******************************************************************************* |
8906 | * Function: open_backup_targets | |
8907 | * Description: Function opens file descriptors for all devices given in | |
8908 | * info->devs | |
8909 | * Parameters: | |
8910 | * info : general array info | |
8911 | * raid_disks : number of disks | |
8912 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
8913 | * super : intel super for raid10 degradation check |
8914 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
8915 | * Returns: |
8916 | * 0 : success | |
8917 | * -1 : fail | |
8918 | ******************************************************************************/ | |
9a717282 AK |
8919 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
8920 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
8921 | { |
8922 | struct mdinfo *sd; | |
f627f5ad | 8923 | int i; |
9a717282 | 8924 | int opened = 0; |
f627f5ad AK |
8925 | |
8926 | for (i = 0; i < raid_disks; i++) | |
8927 | raid_fds[i] = -1; | |
687629c2 AK |
8928 | |
8929 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8930 | char *dn; | |
8931 | ||
8932 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
8933 | dprintf("disk is faulty!!\n"); | |
8934 | continue; | |
8935 | } | |
8936 | ||
8937 | if ((sd->disk.raid_disk >= raid_disks) || | |
8938 | (sd->disk.raid_disk < 0)) | |
8939 | continue; | |
8940 | ||
8941 | dn = map_dev(sd->disk.major, | |
8942 | sd->disk.minor, 1); | |
8943 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
8944 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
8945 | fprintf(stderr, "cannot open component\n"); | |
9a717282 | 8946 | continue; |
687629c2 | 8947 | } |
9a717282 AK |
8948 | opened++; |
8949 | } | |
8950 | /* check if maximum array degradation level is not exceeded | |
8951 | */ | |
8952 | if ((raid_disks - opened) > | |
8953 | imsm_get_allowed_degradation(info->new_level, | |
8954 | raid_disks, | |
8955 | super, dev)) { | |
8956 | fprintf(stderr, "Not enough disks can be opened.\n"); | |
8957 | close_targets(raid_fds, raid_disks); | |
8958 | return -2; | |
687629c2 AK |
8959 | } |
8960 | return 0; | |
8961 | } | |
8962 | ||
9e2d750d | 8963 | #ifndef MDASSEMBLE |
687629c2 AK |
8964 | /******************************************************************************* |
8965 | * Function: init_migr_record_imsm | |
8966 | * Description: Function inits imsm migration record | |
8967 | * Parameters: | |
8968 | * super : imsm internal array info | |
8969 | * dev : device under migration | |
8970 | * info : general array info to find the smallest device | |
8971 | * Returns: | |
8972 | * none | |
8973 | ******************************************************************************/ | |
8974 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
8975 | struct mdinfo *info) | |
8976 | { | |
8977 | struct intel_super *super = st->sb; | |
8978 | struct migr_record *migr_rec = super->migr_rec; | |
8979 | int new_data_disks; | |
8980 | unsigned long long dsize, dev_sectors; | |
8981 | long long unsigned min_dev_sectors = -1LLU; | |
8982 | struct mdinfo *sd; | |
8983 | char nm[30]; | |
8984 | int fd; | |
238c0a71 AK |
8985 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
8986 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 8987 | unsigned long long num_migr_units; |
3ef4403c | 8988 | unsigned long long array_blocks; |
687629c2 AK |
8989 | |
8990 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
8991 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
8992 | ||
8993 | /* only ascending reshape supported now */ | |
8994 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
8995 | ||
8996 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
8997 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
8998 | migr_rec->dest_depth_per_unit *= |
8999 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 9000 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
9001 | migr_rec->blocks_per_unit = |
9002 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
9003 | migr_rec->dest_depth_per_unit = | |
9004 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 9005 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
9006 | num_migr_units = |
9007 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
9008 | ||
9009 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
9010 | num_migr_units++; | |
9011 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
9012 | ||
9013 | migr_rec->post_migr_vol_cap = dev->size_low; | |
9014 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
9015 | ||
9016 | ||
9017 | /* Find the smallest dev */ | |
9018 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9019 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
9020 | fd = dev_open(nm, O_RDONLY); | |
9021 | if (fd < 0) | |
9022 | continue; | |
9023 | get_dev_size(fd, NULL, &dsize); | |
9024 | dev_sectors = dsize / 512; | |
9025 | if (dev_sectors < min_dev_sectors) | |
9026 | min_dev_sectors = dev_sectors; | |
9027 | close(fd); | |
9028 | } | |
9029 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
9030 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
9031 | ||
9032 | write_imsm_migr_rec(st); | |
9033 | ||
9034 | return; | |
9035 | } | |
9036 | ||
9037 | /******************************************************************************* | |
9038 | * Function: save_backup_imsm | |
9039 | * Description: Function saves critical data stripes to Migration Copy Area | |
9040 | * and updates the current migration unit status. | |
9041 | * Use restore_stripes() to form a destination stripe, | |
9042 | * and to write it to the Copy Area. | |
9043 | * Parameters: | |
9044 | * st : supertype information | |
aea93171 | 9045 | * dev : imsm device that backup is saved for |
687629c2 AK |
9046 | * info : general array info |
9047 | * buf : input buffer | |
687629c2 AK |
9048 | * length : length of data to backup (blocks_per_unit) |
9049 | * Returns: | |
9050 | * 0 : success | |
9051 | *, -1 : fail | |
9052 | ******************************************************************************/ | |
9053 | int save_backup_imsm(struct supertype *st, | |
9054 | struct imsm_dev *dev, | |
9055 | struct mdinfo *info, | |
9056 | void *buf, | |
687629c2 AK |
9057 | int length) |
9058 | { | |
9059 | int rv = -1; | |
9060 | struct intel_super *super = st->sb; | |
9061 | unsigned long long *target_offsets = NULL; | |
9062 | int *targets = NULL; | |
9063 | int i; | |
238c0a71 | 9064 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 9065 | int new_disks = map_dest->num_members; |
ab724b98 AK |
9066 | int dest_layout = 0; |
9067 | int dest_chunk; | |
d1877f69 | 9068 | unsigned long long start; |
238c0a71 | 9069 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 9070 | |
503975b9 | 9071 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 9072 | |
7e45b550 AK |
9073 | for (i = 0; i < new_disks; i++) |
9074 | targets[i] = -1; | |
9075 | ||
503975b9 | 9076 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 9077 | |
d1877f69 | 9078 | start = info->reshape_progress * 512; |
687629c2 | 9079 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
9080 | target_offsets[i] = (unsigned long long) |
9081 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
9082 | /* move back copy area adderss, it will be moved forward |
9083 | * in restore_stripes() using start input variable | |
9084 | */ | |
9085 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
9086 | } |
9087 | ||
9a717282 AK |
9088 | if (open_backup_targets(info, new_disks, targets, |
9089 | super, dev)) | |
687629c2 AK |
9090 | goto abort; |
9091 | ||
68eb8bc6 | 9092 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
9093 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
9094 | ||
687629c2 AK |
9095 | if (restore_stripes(targets, /* list of dest devices */ |
9096 | target_offsets, /* migration record offsets */ | |
9097 | new_disks, | |
ab724b98 AK |
9098 | dest_chunk, |
9099 | map_dest->raid_level, | |
9100 | dest_layout, | |
9101 | -1, /* source backup file descriptor */ | |
9102 | 0, /* input buf offset | |
9103 | * always 0 buf is already offseted */ | |
d1877f69 | 9104 | start, |
687629c2 AK |
9105 | length, |
9106 | buf) != 0) { | |
e7b84f9d | 9107 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
9108 | goto abort; |
9109 | } | |
9110 | ||
9111 | rv = 0; | |
9112 | ||
9113 | abort: | |
9114 | if (targets) { | |
9a717282 | 9115 | close_targets(targets, new_disks); |
687629c2 AK |
9116 | free(targets); |
9117 | } | |
9118 | free(target_offsets); | |
9119 | ||
9120 | return rv; | |
9121 | } | |
9122 | ||
9123 | /******************************************************************************* | |
9124 | * Function: save_checkpoint_imsm | |
9125 | * Description: Function called for current unit status update | |
9126 | * in the migration record. It writes it to disk. | |
9127 | * Parameters: | |
9128 | * super : imsm internal array info | |
9129 | * info : general array info | |
9130 | * Returns: | |
9131 | * 0: success | |
9132 | * 1: failure | |
0228d92c AK |
9133 | * 2: failure, means no valid migration record |
9134 | * / no general migration in progress / | |
687629c2 AK |
9135 | ******************************************************************************/ |
9136 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
9137 | { | |
9138 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
9139 | unsigned long long blocks_per_unit; |
9140 | unsigned long long curr_migr_unit; | |
9141 | ||
2e062e82 AK |
9142 | if (load_imsm_migr_rec(super, info) != 0) { |
9143 | dprintf("imsm: ERROR: Cannot read migration record " | |
9144 | "for checkpoint save.\n"); | |
9145 | return 1; | |
9146 | } | |
9147 | ||
f8b72ef5 AK |
9148 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
9149 | if (blocks_per_unit == 0) { | |
0228d92c AK |
9150 | dprintf("imsm: no migration in progress.\n"); |
9151 | return 2; | |
687629c2 | 9152 | } |
f8b72ef5 AK |
9153 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
9154 | /* check if array is alligned to copy area | |
9155 | * if it is not alligned, add one to current migration unit value | |
9156 | * this can happend on array reshape finish only | |
9157 | */ | |
9158 | if (info->reshape_progress % blocks_per_unit) | |
9159 | curr_migr_unit++; | |
687629c2 AK |
9160 | |
9161 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 9162 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
9163 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9164 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
9165 | __cpu_to_le32(curr_migr_unit * |
9166 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
9167 | if (write_imsm_migr_rec(st) < 0) { |
9168 | dprintf("imsm: Cannot write migration record " | |
9169 | "outside backup area\n"); | |
9170 | return 1; | |
9171 | } | |
9172 | ||
9173 | return 0; | |
9174 | } | |
9175 | ||
276d77db AK |
9176 | /******************************************************************************* |
9177 | * Function: recover_backup_imsm | |
9178 | * Description: Function recovers critical data from the Migration Copy Area | |
9179 | * while assembling an array. | |
9180 | * Parameters: | |
9181 | * super : imsm internal array info | |
9182 | * info : general array info | |
9183 | * Returns: | |
9184 | * 0 : success (or there is no data to recover) | |
9185 | * 1 : fail | |
9186 | ******************************************************************************/ | |
9187 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
9188 | { | |
9189 | struct intel_super *super = st->sb; | |
9190 | struct migr_record *migr_rec = super->migr_rec; | |
9191 | struct imsm_map *map_dest = NULL; | |
9192 | struct intel_dev *id = NULL; | |
9193 | unsigned long long read_offset; | |
9194 | unsigned long long write_offset; | |
9195 | unsigned unit_len; | |
9196 | int *targets = NULL; | |
9197 | int new_disks, i, err; | |
9198 | char *buf = NULL; | |
9199 | int retval = 1; | |
9200 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
9201 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 9202 | char buffer[20]; |
6c3560c0 | 9203 | int skipped_disks = 0; |
276d77db AK |
9204 | |
9205 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
9206 | if (err < 1) | |
9207 | return 1; | |
9208 | ||
9209 | /* recover data only during assemblation */ | |
9210 | if (strncmp(buffer, "inactive", 8) != 0) | |
9211 | return 0; | |
9212 | /* no data to recover */ | |
9213 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
9214 | return 0; | |
9215 | if (curr_migr_unit >= num_migr_units) | |
9216 | return 1; | |
9217 | ||
9218 | /* find device during reshape */ | |
9219 | for (id = super->devlist; id; id = id->next) | |
9220 | if (is_gen_migration(id->dev)) | |
9221 | break; | |
9222 | if (id == NULL) | |
9223 | return 1; | |
9224 | ||
238c0a71 | 9225 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
9226 | new_disks = map_dest->num_members; |
9227 | ||
9228 | read_offset = (unsigned long long) | |
9229 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
9230 | ||
9231 | write_offset = ((unsigned long long) | |
9232 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 9233 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
9234 | |
9235 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9236 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
9237 | goto abort; | |
503975b9 | 9238 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 9239 | |
9a717282 | 9240 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 9241 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
9242 | goto abort; |
9243 | } | |
276d77db AK |
9244 | |
9245 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
9246 | if (targets[i] < 0) { |
9247 | skipped_disks++; | |
9248 | continue; | |
9249 | } | |
276d77db | 9250 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
9251 | pr_err("Cannot seek to block: %s\n", |
9252 | strerror(errno)); | |
137debce AK |
9253 | skipped_disks++; |
9254 | continue; | |
276d77db | 9255 | } |
9ec11d1a | 9256 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9257 | pr_err("Cannot read copy area block: %s\n", |
9258 | strerror(errno)); | |
137debce AK |
9259 | skipped_disks++; |
9260 | continue; | |
276d77db AK |
9261 | } |
9262 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
9263 | pr_err("Cannot seek to block: %s\n", |
9264 | strerror(errno)); | |
137debce AK |
9265 | skipped_disks++; |
9266 | continue; | |
276d77db | 9267 | } |
9ec11d1a | 9268 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9269 | pr_err("Cannot restore block: %s\n", |
9270 | strerror(errno)); | |
137debce AK |
9271 | skipped_disks++; |
9272 | continue; | |
276d77db AK |
9273 | } |
9274 | } | |
9275 | ||
137debce AK |
9276 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
9277 | new_disks, | |
9278 | super, | |
9279 | id->dev)) { | |
e7b84f9d N |
9280 | pr_err("Cannot restore data from backup." |
9281 | " Too many failed disks\n"); | |
6c3560c0 AK |
9282 | goto abort; |
9283 | } | |
9284 | ||
befb629b AK |
9285 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
9286 | /* ignore error == 2, this can mean end of reshape here | |
9287 | */ | |
9288 | dprintf("imsm: Cannot write checkpoint to " | |
9289 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
9290 | } else | |
276d77db | 9291 | retval = 0; |
276d77db AK |
9292 | |
9293 | abort: | |
9294 | if (targets) { | |
9295 | for (i = 0; i < new_disks; i++) | |
9296 | if (targets[i]) | |
9297 | close(targets[i]); | |
9298 | free(targets); | |
9299 | } | |
9300 | free(buf); | |
9301 | return retval; | |
9302 | } | |
9303 | ||
2cda7640 ML |
9304 | static char disk_by_path[] = "/dev/disk/by-path/"; |
9305 | ||
9306 | static const char *imsm_get_disk_controller_domain(const char *path) | |
9307 | { | |
2cda7640 | 9308 | char disk_path[PATH_MAX]; |
96234762 LM |
9309 | char *drv=NULL; |
9310 | struct stat st; | |
2cda7640 | 9311 | |
96234762 LM |
9312 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
9313 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
9314 | if (stat(disk_path, &st) == 0) { | |
9315 | struct sys_dev* hba; | |
9316 | char *path=NULL; | |
9317 | ||
9318 | path = devt_to_devpath(st.st_rdev); | |
9319 | if (path == NULL) | |
9320 | return "unknown"; | |
9321 | hba = find_disk_attached_hba(-1, path); | |
9322 | if (hba && hba->type == SYS_DEV_SAS) | |
9323 | drv = "isci"; | |
9324 | else if (hba && hba->type == SYS_DEV_SATA) | |
9325 | drv = "ahci"; | |
9326 | else | |
9327 | drv = "unknown"; | |
9328 | dprintf("path: %s hba: %s attached: %s\n", | |
9329 | path, (hba) ? hba->path : "NULL", drv); | |
9330 | free(path); | |
2cda7640 | 9331 | } |
96234762 | 9332 | return drv; |
2cda7640 ML |
9333 | } |
9334 | ||
78b10e66 N |
9335 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
9336 | { | |
9337 | char subdev_name[20]; | |
9338 | struct mdstat_ent *mdstat; | |
9339 | ||
9340 | sprintf(subdev_name, "%d", subdev); | |
9341 | mdstat = mdstat_by_subdev(subdev_name, container); | |
9342 | if (!mdstat) | |
9343 | return -1; | |
9344 | ||
9345 | *minor = mdstat->devnum; | |
9346 | free_mdstat(mdstat); | |
9347 | return 0; | |
9348 | } | |
9349 | ||
9350 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
9351 | struct geo_params *geo, | |
fbf3d202 AK |
9352 | int *old_raid_disks, |
9353 | int direction) | |
78b10e66 | 9354 | { |
694575e7 KW |
9355 | /* currently we only support increasing the number of devices |
9356 | * for a container. This increases the number of device for each | |
9357 | * member array. They must all be RAID0 or RAID5. | |
9358 | */ | |
78b10e66 N |
9359 | int ret_val = 0; |
9360 | struct mdinfo *info, *member; | |
9361 | int devices_that_can_grow = 0; | |
9362 | ||
9363 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
9364 | "st->devnum = (%i)\n", | |
9365 | st->devnum); | |
9366 | ||
d04f65f4 | 9367 | if (geo->size > 0 || |
78b10e66 N |
9368 | geo->level != UnSet || |
9369 | geo->layout != UnSet || | |
9370 | geo->chunksize != 0 || | |
9371 | geo->raid_disks == UnSet) { | |
9372 | dprintf("imsm: Container operation is allowed for " | |
9373 | "raid disks number change only.\n"); | |
9374 | return ret_val; | |
9375 | } | |
9376 | ||
fbf3d202 AK |
9377 | if (direction == ROLLBACK_METADATA_CHANGES) { |
9378 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9379 | "container operation.\n"); | |
9380 | return ret_val; | |
9381 | } | |
9382 | ||
78b10e66 N |
9383 | info = container_content_imsm(st, NULL); |
9384 | for (member = info; member; member = member->next) { | |
9385 | int result; | |
9386 | int minor; | |
9387 | ||
9388 | dprintf("imsm: checking device_num: %i\n", | |
9389 | member->container_member); | |
9390 | ||
d7d205bd | 9391 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
9392 | /* we work on container for Online Capacity Expansion |
9393 | * only so raid_disks has to grow | |
9394 | */ | |
9395 | dprintf("imsm: for container operation raid disks " | |
9396 | "increase is required\n"); | |
9397 | break; | |
9398 | } | |
9399 | ||
9400 | if ((info->array.level != 0) && | |
9401 | (info->array.level != 5)) { | |
9402 | /* we cannot use this container with other raid level | |
9403 | */ | |
690aae1a | 9404 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
9405 | " raid level (%i) detected\n", |
9406 | info->array.level); | |
9407 | break; | |
9408 | } else { | |
9409 | /* check for platform support | |
9410 | * for this raid level configuration | |
9411 | */ | |
9412 | struct intel_super *super = st->sb; | |
9413 | if (!is_raid_level_supported(super->orom, | |
9414 | member->array.level, | |
9415 | geo->raid_disks)) { | |
690aae1a | 9416 | dprintf("platform does not support raid%d with" |
78b10e66 N |
9417 | " %d disk%s\n", |
9418 | info->array.level, | |
9419 | geo->raid_disks, | |
9420 | geo->raid_disks > 1 ? "s" : ""); | |
9421 | break; | |
9422 | } | |
2a4a08e7 AK |
9423 | /* check if component size is aligned to chunk size |
9424 | */ | |
9425 | if (info->component_size % | |
9426 | (info->array.chunk_size/512)) { | |
9427 | dprintf("Component size is not aligned to " | |
9428 | "chunk size\n"); | |
9429 | break; | |
9430 | } | |
78b10e66 N |
9431 | } |
9432 | ||
9433 | if (*old_raid_disks && | |
9434 | info->array.raid_disks != *old_raid_disks) | |
9435 | break; | |
9436 | *old_raid_disks = info->array.raid_disks; | |
9437 | ||
9438 | /* All raid5 and raid0 volumes in container | |
9439 | * have to be ready for Online Capacity Expansion | |
9440 | * so they need to be assembled. We have already | |
9441 | * checked that no recovery etc is happening. | |
9442 | */ | |
9443 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
9444 | st->container_dev, | |
9445 | &minor); | |
9446 | if (result < 0) { | |
9447 | dprintf("imsm: cannot find array\n"); | |
9448 | break; | |
9449 | } | |
9450 | devices_that_can_grow++; | |
9451 | } | |
9452 | sysfs_free(info); | |
9453 | if (!member && devices_that_can_grow) | |
9454 | ret_val = 1; | |
9455 | ||
9456 | if (ret_val) | |
9457 | dprintf("\tContainer operation allowed\n"); | |
9458 | else | |
9459 | dprintf("\tError: %i\n", ret_val); | |
9460 | ||
9461 | return ret_val; | |
9462 | } | |
9463 | ||
9464 | /* Function: get_spares_for_grow | |
9465 | * Description: Allocates memory and creates list of spare devices | |
9466 | * avaliable in container. Checks if spare drive size is acceptable. | |
9467 | * Parameters: Pointer to the supertype structure | |
9468 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
9469 | * NULL if fail | |
9470 | */ | |
9471 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
9472 | { | |
78b10e66 | 9473 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 9474 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
9475 | } |
9476 | ||
9477 | /****************************************************************************** | |
9478 | * function: imsm_create_metadata_update_for_reshape | |
9479 | * Function creates update for whole IMSM container. | |
9480 | * | |
9481 | ******************************************************************************/ | |
9482 | static int imsm_create_metadata_update_for_reshape( | |
9483 | struct supertype *st, | |
9484 | struct geo_params *geo, | |
9485 | int old_raid_disks, | |
9486 | struct imsm_update_reshape **updatep) | |
9487 | { | |
9488 | struct intel_super *super = st->sb; | |
9489 | struct imsm_super *mpb = super->anchor; | |
9490 | int update_memory_size = 0; | |
9491 | struct imsm_update_reshape *u = NULL; | |
9492 | struct mdinfo *spares = NULL; | |
9493 | int i; | |
9494 | int delta_disks = 0; | |
bbd24d86 | 9495 | struct mdinfo *dev; |
78b10e66 N |
9496 | |
9497 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
9498 | geo->raid_disks); | |
9499 | ||
9500 | delta_disks = geo->raid_disks - old_raid_disks; | |
9501 | ||
9502 | /* size of all update data without anchor */ | |
9503 | update_memory_size = sizeof(struct imsm_update_reshape); | |
9504 | ||
9505 | /* now add space for spare disks that we need to add. */ | |
9506 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
9507 | ||
503975b9 | 9508 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
9509 | u->type = update_reshape_container_disks; |
9510 | u->old_raid_disks = old_raid_disks; | |
9511 | u->new_raid_disks = geo->raid_disks; | |
9512 | ||
9513 | /* now get spare disks list | |
9514 | */ | |
9515 | spares = get_spares_for_grow(st); | |
9516 | ||
9517 | if (spares == NULL | |
9518 | || delta_disks > spares->array.spare_disks) { | |
e7b84f9d | 9519 | pr_err("imsm: ERROR: Cannot get spare devices " |
e14e5960 | 9520 | "for %s.\n", geo->dev_name); |
e4c72d1d | 9521 | i = -1; |
78b10e66 N |
9522 | goto abort; |
9523 | } | |
9524 | ||
9525 | /* we have got spares | |
9526 | * update disk list in imsm_disk list table in anchor | |
9527 | */ | |
9528 | dprintf("imsm: %i spares are available.\n\n", | |
9529 | spares->array.spare_disks); | |
9530 | ||
bbd24d86 | 9531 | dev = spares->devs; |
78b10e66 | 9532 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
9533 | struct dl *dl; |
9534 | ||
bbd24d86 AK |
9535 | if (dev == NULL) |
9536 | break; | |
78b10e66 N |
9537 | u->new_disks[i] = makedev(dev->disk.major, |
9538 | dev->disk.minor); | |
9539 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
9540 | dl->index = mpb->num_disks; |
9541 | mpb->num_disks++; | |
bbd24d86 | 9542 | dev = dev->next; |
78b10e66 | 9543 | } |
78b10e66 N |
9544 | |
9545 | abort: | |
9546 | /* free spares | |
9547 | */ | |
9548 | sysfs_free(spares); | |
9549 | ||
d677e0b8 | 9550 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 9551 | if (i == delta_disks) { |
d677e0b8 | 9552 | dprintf(" OK\n"); |
78b10e66 N |
9553 | *updatep = u; |
9554 | return update_memory_size; | |
9555 | } | |
9556 | free(u); | |
d677e0b8 | 9557 | dprintf(" Error\n"); |
78b10e66 N |
9558 | |
9559 | return 0; | |
9560 | } | |
9561 | ||
f3871fdc AK |
9562 | |
9563 | /****************************************************************************** | |
9564 | * function: imsm_create_metadata_update_for_size_change() | |
9565 | * Creates update for IMSM array for array size change. | |
9566 | * | |
9567 | ******************************************************************************/ | |
9568 | static int imsm_create_metadata_update_for_size_change( | |
9569 | struct supertype *st, | |
9570 | struct geo_params *geo, | |
9571 | struct imsm_update_size_change **updatep) | |
9572 | { | |
9573 | struct intel_super *super = st->sb; | |
9574 | int update_memory_size = 0; | |
9575 | struct imsm_update_size_change *u = NULL; | |
9576 | ||
9577 | dprintf("imsm_create_metadata_update_for_size_change(enter)" | |
9578 | " New size = %llu\n", geo->size); | |
9579 | ||
9580 | /* size of all update data without anchor */ | |
9581 | update_memory_size = sizeof(struct imsm_update_size_change); | |
9582 | ||
503975b9 | 9583 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
9584 | u->type = update_size_change; |
9585 | u->subdev = super->current_vol; | |
9586 | u->new_size = geo->size; | |
9587 | ||
9588 | dprintf("imsm: reshape update preparation : OK\n"); | |
9589 | *updatep = u; | |
9590 | ||
9591 | return update_memory_size; | |
9592 | } | |
9593 | ||
48c5303a PC |
9594 | /****************************************************************************** |
9595 | * function: imsm_create_metadata_update_for_migration() | |
9596 | * Creates update for IMSM array. | |
9597 | * | |
9598 | ******************************************************************************/ | |
9599 | static int imsm_create_metadata_update_for_migration( | |
9600 | struct supertype *st, | |
9601 | struct geo_params *geo, | |
9602 | struct imsm_update_reshape_migration **updatep) | |
9603 | { | |
9604 | struct intel_super *super = st->sb; | |
9605 | int update_memory_size = 0; | |
9606 | struct imsm_update_reshape_migration *u = NULL; | |
9607 | struct imsm_dev *dev; | |
9608 | int previous_level = -1; | |
9609 | ||
9610 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
9611 | " New Level = %i\n", geo->level); | |
9612 | ||
9613 | /* size of all update data without anchor */ | |
9614 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
9615 | ||
503975b9 | 9616 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
9617 | u->type = update_reshape_migration; |
9618 | u->subdev = super->current_vol; | |
9619 | u->new_level = geo->level; | |
9620 | u->new_layout = geo->layout; | |
9621 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
9622 | u->new_disks[0] = -1; | |
4bba0439 | 9623 | u->new_chunksize = -1; |
48c5303a PC |
9624 | |
9625 | dev = get_imsm_dev(super, u->subdev); | |
9626 | if (dev) { | |
9627 | struct imsm_map *map; | |
9628 | ||
238c0a71 | 9629 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
9630 | if (map) { |
9631 | int current_chunk_size = | |
9632 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
9633 | ||
9634 | if (geo->chunksize != current_chunk_size) { | |
9635 | u->new_chunksize = geo->chunksize / 1024; | |
9636 | dprintf("imsm: " | |
9637 | "chunk size change from %i to %i\n", | |
9638 | current_chunk_size, u->new_chunksize); | |
9639 | } | |
48c5303a | 9640 | previous_level = map->raid_level; |
4bba0439 | 9641 | } |
48c5303a PC |
9642 | } |
9643 | if ((geo->level == 5) && (previous_level == 0)) { | |
9644 | struct mdinfo *spares = NULL; | |
9645 | ||
9646 | u->new_raid_disks++; | |
9647 | spares = get_spares_for_grow(st); | |
9648 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
9649 | free(u); | |
9650 | sysfs_free(spares); | |
9651 | update_memory_size = 0; | |
9652 | dprintf("error: cannot get spare device " | |
9653 | "for requested migration"); | |
9654 | return 0; | |
9655 | } | |
9656 | sysfs_free(spares); | |
9657 | } | |
9658 | dprintf("imsm: reshape update preparation : OK\n"); | |
9659 | *updatep = u; | |
9660 | ||
9661 | return update_memory_size; | |
9662 | } | |
9663 | ||
8dd70bce AK |
9664 | static void imsm_update_metadata_locally(struct supertype *st, |
9665 | void *buf, int len) | |
9666 | { | |
9667 | struct metadata_update mu; | |
9668 | ||
9669 | mu.buf = buf; | |
9670 | mu.len = len; | |
9671 | mu.space = NULL; | |
9672 | mu.space_list = NULL; | |
9673 | mu.next = NULL; | |
9674 | imsm_prepare_update(st, &mu); | |
9675 | imsm_process_update(st, &mu); | |
9676 | ||
9677 | while (mu.space_list) { | |
9678 | void **space = mu.space_list; | |
9679 | mu.space_list = *space; | |
9680 | free(space); | |
9681 | } | |
9682 | } | |
78b10e66 | 9683 | |
471bceb6 | 9684 | /*************************************************************************** |
694575e7 | 9685 | * Function: imsm_analyze_change |
471bceb6 KW |
9686 | * Description: Function analyze change for single volume |
9687 | * and validate if transition is supported | |
fbf3d202 AK |
9688 | * Parameters: Geometry parameters, supertype structure, |
9689 | * metadata change direction (apply/rollback) | |
694575e7 | 9690 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
9691 | ****************************************************************************/ |
9692 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
9693 | struct geo_params *geo, |
9694 | int direction) | |
694575e7 | 9695 | { |
471bceb6 KW |
9696 | struct mdinfo info; |
9697 | int change = -1; | |
9698 | int check_devs = 0; | |
c21e737b | 9699 | int chunk; |
67a2db32 AK |
9700 | /* number of added/removed disks in operation result */ |
9701 | int devNumChange = 0; | |
9702 | /* imsm compatible layout value for array geometry verification */ | |
9703 | int imsm_layout = -1; | |
7abc9871 AK |
9704 | int data_disks; |
9705 | struct imsm_dev *dev; | |
9706 | struct intel_super *super; | |
d04f65f4 | 9707 | unsigned long long current_size; |
65d38cca | 9708 | unsigned long long free_size; |
d04f65f4 | 9709 | unsigned long long max_size; |
65d38cca | 9710 | int rv; |
471bceb6 KW |
9711 | |
9712 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
9713 | if ((geo->level != info.array.level) && |
9714 | (geo->level >= 0) && | |
9715 | (geo->level != UnSet)) { | |
9716 | switch (info.array.level) { | |
9717 | case 0: | |
9718 | if (geo->level == 5) { | |
b5347799 | 9719 | change = CH_MIGRATION; |
e13ce846 | 9720 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
e7b84f9d N |
9721 | pr_err("Error. Requested Layout " |
9722 | "not supported (left-asymmetric layout " | |
9723 | "is supported only)!\n"); | |
e13ce846 AK |
9724 | change = -1; |
9725 | goto analyse_change_exit; | |
9726 | } | |
67a2db32 | 9727 | imsm_layout = geo->layout; |
471bceb6 | 9728 | check_devs = 1; |
e91a3bad LM |
9729 | devNumChange = 1; /* parity disk added */ |
9730 | } else if (geo->level == 10) { | |
471bceb6 KW |
9731 | change = CH_TAKEOVER; |
9732 | check_devs = 1; | |
e91a3bad | 9733 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 9734 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 9735 | } |
dfe77a9e KW |
9736 | break; |
9737 | case 1: | |
471bceb6 KW |
9738 | case 10: |
9739 | if (geo->level == 0) { | |
9740 | change = CH_TAKEOVER; | |
9741 | check_devs = 1; | |
e91a3bad | 9742 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 9743 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
9744 | } |
9745 | break; | |
9746 | } | |
9747 | if (change == -1) { | |
e7b84f9d N |
9748 | pr_err("Error. Level Migration from %d to %d " |
9749 | "not supported!\n", | |
9750 | info.array.level, geo->level); | |
471bceb6 KW |
9751 | goto analyse_change_exit; |
9752 | } | |
9753 | } else | |
9754 | geo->level = info.array.level; | |
9755 | ||
9756 | if ((geo->layout != info.array.layout) | |
9757 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 9758 | change = CH_MIGRATION; |
471bceb6 KW |
9759 | if ((info.array.layout == 0) |
9760 | && (info.array.level == 5) | |
9761 | && (geo->layout == 5)) { | |
9762 | /* reshape 5 -> 4 */ | |
9763 | } else if ((info.array.layout == 5) | |
9764 | && (info.array.level == 5) | |
9765 | && (geo->layout == 0)) { | |
9766 | /* reshape 4 -> 5 */ | |
9767 | geo->layout = 0; | |
9768 | geo->level = 5; | |
9769 | } else { | |
e7b84f9d N |
9770 | pr_err("Error. Layout Migration from %d to %d " |
9771 | "not supported!\n", | |
9772 | info.array.layout, geo->layout); | |
471bceb6 KW |
9773 | change = -1; |
9774 | goto analyse_change_exit; | |
9775 | } | |
67a2db32 | 9776 | } else { |
471bceb6 | 9777 | geo->layout = info.array.layout; |
67a2db32 AK |
9778 | if (imsm_layout == -1) |
9779 | imsm_layout = info.array.layout; | |
9780 | } | |
471bceb6 KW |
9781 | |
9782 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
9783 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 9784 | change = CH_MIGRATION; |
471bceb6 KW |
9785 | else |
9786 | geo->chunksize = info.array.chunk_size; | |
9787 | ||
c21e737b | 9788 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
9789 | |
9790 | super = st->sb; | |
9791 | dev = get_imsm_dev(super, super->current_vol); | |
9792 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 9793 | /* compute current size per disk member |
7abc9871 | 9794 | */ |
c41e00b2 AK |
9795 | current_size = info.custom_array_size / data_disks; |
9796 | ||
d5d2c614 | 9797 | if ((geo->size > 0) && (geo->size != MAX_SIZE)) { |
c41e00b2 AK |
9798 | /* align component size |
9799 | */ | |
9800 | geo->size = imsm_component_size_aligment_check( | |
9801 | get_imsm_raid_level(dev->vol.map), | |
9802 | chunk * 1024, | |
9803 | geo->size * 2); | |
65d0b4ce LD |
9804 | if (geo->size == 0) { |
9805 | pr_err("Error. Size expansion is " \ | |
9806 | "supported only (current size is %llu, " \ | |
9807 | "requested size /rounded/ is 0).\n", | |
9808 | current_size); | |
9809 | goto analyse_change_exit; | |
9810 | } | |
c41e00b2 | 9811 | } |
7abc9871 | 9812 | |
d04f65f4 | 9813 | if ((current_size != geo->size) && (geo->size > 0)) { |
7abc9871 | 9814 | if (change != -1) { |
e7b84f9d | 9815 | pr_err("Error. Size change should be the only " |
7abc9871 AK |
9816 | "one at a time.\n"); |
9817 | change = -1; | |
9818 | goto analyse_change_exit; | |
9819 | } | |
9820 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
e7b84f9d N |
9821 | pr_err("Error. The last volume in container " |
9822 | "can be expanded only (%i/%i).\n", | |
9823 | super->current_vol, st->devnum); | |
7abc9871 AK |
9824 | goto analyse_change_exit; |
9825 | } | |
65d38cca LD |
9826 | /* check the maximum available size |
9827 | */ | |
9828 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
9829 | 0, chunk, &free_size); | |
9830 | if (rv == 0) | |
9831 | /* Cannot find maximum available space | |
9832 | */ | |
9833 | max_size = 0; | |
9834 | else { | |
9835 | max_size = free_size + current_size; | |
9836 | /* align component size | |
9837 | */ | |
9838 | max_size = imsm_component_size_aligment_check( | |
9839 | get_imsm_raid_level(dev->vol.map), | |
9840 | chunk * 1024, | |
9841 | max_size); | |
9842 | } | |
d04f65f4 | 9843 | if (geo->size == MAX_SIZE) { |
b130333f AK |
9844 | /* requested size change to the maximum available size |
9845 | */ | |
65d38cca | 9846 | if (max_size == 0) { |
e7b84f9d | 9847 | pr_err("Error. Cannot find " |
b130333f AK |
9848 | "maximum available space.\n"); |
9849 | change = -1; | |
9850 | goto analyse_change_exit; | |
65d38cca LD |
9851 | } else |
9852 | geo->size = max_size; | |
c41e00b2 | 9853 | } |
b130333f | 9854 | |
fbf3d202 AK |
9855 | if ((direction == ROLLBACK_METADATA_CHANGES)) { |
9856 | /* accept size for rollback only | |
9857 | */ | |
9858 | } else { | |
9859 | /* round size due to metadata compatibility | |
9860 | */ | |
9861 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
9862 | << SECT_PER_MB_SHIFT; | |
9863 | dprintf("Prepare update for size change to %llu\n", | |
9864 | geo->size ); | |
9865 | if (current_size >= geo->size) { | |
e7b84f9d N |
9866 | pr_err("Error. Size expansion is " |
9867 | "supported only (current size is %llu, " | |
9868 | "requested size /rounded/ is %llu).\n", | |
9869 | current_size, geo->size); | |
fbf3d202 AK |
9870 | goto analyse_change_exit; |
9871 | } | |
65d38cca | 9872 | if (max_size && geo->size > max_size) { |
e7b84f9d N |
9873 | pr_err("Error. Requested size is larger " |
9874 | "than maximum available size (maximum " | |
9875 | "available size is %llu, " | |
9876 | "requested size /rounded/ is %llu).\n", | |
9877 | max_size, geo->size); | |
65d38cca LD |
9878 | goto analyse_change_exit; |
9879 | } | |
7abc9871 AK |
9880 | } |
9881 | geo->size *= data_disks; | |
9882 | geo->raid_disks = dev->vol.map->num_members; | |
9883 | change = CH_ARRAY_SIZE; | |
9884 | } | |
471bceb6 KW |
9885 | if (!validate_geometry_imsm(st, |
9886 | geo->level, | |
67a2db32 | 9887 | imsm_layout, |
e91a3bad | 9888 | geo->raid_disks + devNumChange, |
c21e737b | 9889 | &chunk, |
af4348dd | 9890 | geo->size, INVALID_SECTORS, |
471bceb6 KW |
9891 | 0, 0, 1)) |
9892 | change = -1; | |
9893 | ||
9894 | if (check_devs) { | |
9895 | struct intel_super *super = st->sb; | |
9896 | struct imsm_super *mpb = super->anchor; | |
9897 | ||
9898 | if (mpb->num_raid_devs > 1) { | |
e7b84f9d N |
9899 | pr_err("Error. Cannot perform operation on %s" |
9900 | "- for this operation it MUST be single " | |
9901 | "array in container\n", | |
9902 | geo->dev_name); | |
471bceb6 KW |
9903 | change = -1; |
9904 | } | |
9905 | } | |
9906 | ||
9907 | analyse_change_exit: | |
fbf3d202 AK |
9908 | if ((direction == ROLLBACK_METADATA_CHANGES) && |
9909 | ((change == CH_MIGRATION) || (change == CH_TAKEOVER))) { | |
9910 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9911 | "migration and takeover operations.\n"); | |
9912 | change = -1; | |
9913 | } | |
471bceb6 | 9914 | return change; |
694575e7 KW |
9915 | } |
9916 | ||
bb025c2f KW |
9917 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
9918 | { | |
9919 | struct intel_super *super = st->sb; | |
9920 | struct imsm_update_takeover *u; | |
9921 | ||
503975b9 | 9922 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
9923 | |
9924 | u->type = update_takeover; | |
9925 | u->subarray = super->current_vol; | |
9926 | ||
9927 | /* 10->0 transition */ | |
9928 | if (geo->level == 0) | |
9929 | u->direction = R10_TO_R0; | |
9930 | ||
0529c688 KW |
9931 | /* 0->10 transition */ |
9932 | if (geo->level == 10) | |
9933 | u->direction = R0_TO_R10; | |
9934 | ||
bb025c2f KW |
9935 | /* update metadata locally */ |
9936 | imsm_update_metadata_locally(st, u, | |
9937 | sizeof(struct imsm_update_takeover)); | |
9938 | /* and possibly remotely */ | |
9939 | if (st->update_tail) | |
9940 | append_metadata_update(st, u, | |
9941 | sizeof(struct imsm_update_takeover)); | |
9942 | else | |
9943 | free(u); | |
9944 | ||
9945 | return 0; | |
9946 | } | |
9947 | ||
d04f65f4 N |
9948 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
9949 | int level, | |
78b10e66 | 9950 | int layout, int chunksize, int raid_disks, |
41784c88 | 9951 | int delta_disks, char *backup, char *dev, |
016e00f5 | 9952 | int direction, int verbose) |
78b10e66 | 9953 | { |
78b10e66 N |
9954 | int ret_val = 1; |
9955 | struct geo_params geo; | |
9956 | ||
9957 | dprintf("imsm: reshape_super called.\n"); | |
9958 | ||
71204a50 | 9959 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
9960 | |
9961 | geo.dev_name = dev; | |
694575e7 | 9962 | geo.dev_id = st->devnum; |
78b10e66 N |
9963 | geo.size = size; |
9964 | geo.level = level; | |
9965 | geo.layout = layout; | |
9966 | geo.chunksize = chunksize; | |
9967 | geo.raid_disks = raid_disks; | |
41784c88 AK |
9968 | if (delta_disks != UnSet) |
9969 | geo.raid_disks += delta_disks; | |
78b10e66 N |
9970 | |
9971 | dprintf("\tfor level : %i\n", geo.level); | |
9972 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
9973 | ||
9974 | if (experimental() == 0) | |
9975 | return ret_val; | |
9976 | ||
78b10e66 | 9977 | if (st->container_dev == st->devnum) { |
694575e7 KW |
9978 | /* On container level we can only increase number of devices. */ |
9979 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 9980 | int old_raid_disks = 0; |
6dc0be30 | 9981 | |
78b10e66 | 9982 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 9983 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
9984 | struct imsm_update_reshape *u = NULL; |
9985 | int len; | |
9986 | ||
9987 | len = imsm_create_metadata_update_for_reshape( | |
9988 | st, &geo, old_raid_disks, &u); | |
9989 | ||
ed08d51c AK |
9990 | if (len <= 0) { |
9991 | dprintf("imsm: Cannot prepare update\n"); | |
9992 | goto exit_imsm_reshape_super; | |
9993 | } | |
9994 | ||
8dd70bce AK |
9995 | ret_val = 0; |
9996 | /* update metadata locally */ | |
9997 | imsm_update_metadata_locally(st, u, len); | |
9998 | /* and possibly remotely */ | |
9999 | if (st->update_tail) | |
10000 | append_metadata_update(st, u, len); | |
10001 | else | |
ed08d51c | 10002 | free(u); |
8dd70bce | 10003 | |
694575e7 | 10004 | } else { |
e7b84f9d | 10005 | pr_err("(imsm) Operation " |
e7ff7e40 | 10006 | "is not allowed on this container\n"); |
694575e7 KW |
10007 | } |
10008 | } else { | |
10009 | /* On volume level we support following operations | |
471bceb6 KW |
10010 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
10011 | * - chunk size migration | |
10012 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
10013 | */ | |
10014 | struct intel_super *super = st->sb; | |
10015 | struct intel_dev *dev = super->devlist; | |
10016 | int change, devnum; | |
694575e7 | 10017 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
10018 | /* find requested device */ |
10019 | while (dev) { | |
19986c72 MB |
10020 | if (imsm_find_array_minor_by_subdev( |
10021 | dev->index, st->container_dev, &devnum) == 0 | |
10022 | && devnum == geo.dev_id) | |
471bceb6 KW |
10023 | break; |
10024 | dev = dev->next; | |
10025 | } | |
10026 | if (dev == NULL) { | |
e7b84f9d | 10027 | pr_err("Cannot find %s (%i) subarray\n", |
471bceb6 KW |
10028 | geo.dev_name, geo.dev_id); |
10029 | goto exit_imsm_reshape_super; | |
10030 | } | |
10031 | super->current_vol = dev->index; | |
fbf3d202 | 10032 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 10033 | switch (change) { |
471bceb6 | 10034 | case CH_TAKEOVER: |
bb025c2f | 10035 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 10036 | break; |
48c5303a PC |
10037 | case CH_MIGRATION: { |
10038 | struct imsm_update_reshape_migration *u = NULL; | |
10039 | int len = | |
10040 | imsm_create_metadata_update_for_migration( | |
10041 | st, &geo, &u); | |
10042 | if (len < 1) { | |
10043 | dprintf("imsm: " | |
10044 | "Cannot prepare update\n"); | |
10045 | break; | |
10046 | } | |
471bceb6 | 10047 | ret_val = 0; |
48c5303a PC |
10048 | /* update metadata locally */ |
10049 | imsm_update_metadata_locally(st, u, len); | |
10050 | /* and possibly remotely */ | |
10051 | if (st->update_tail) | |
10052 | append_metadata_update(st, u, len); | |
10053 | else | |
10054 | free(u); | |
10055 | } | |
10056 | break; | |
7abc9871 | 10057 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
10058 | struct imsm_update_size_change *u = NULL; |
10059 | int len = | |
10060 | imsm_create_metadata_update_for_size_change( | |
10061 | st, &geo, &u); | |
10062 | if (len < 1) { | |
10063 | dprintf("imsm: " | |
10064 | "Cannot prepare update\n"); | |
10065 | break; | |
10066 | } | |
10067 | ret_val = 0; | |
10068 | /* update metadata locally */ | |
10069 | imsm_update_metadata_locally(st, u, len); | |
10070 | /* and possibly remotely */ | |
10071 | if (st->update_tail) | |
10072 | append_metadata_update(st, u, len); | |
10073 | else | |
10074 | free(u); | |
7abc9871 AK |
10075 | } |
10076 | break; | |
471bceb6 KW |
10077 | default: |
10078 | ret_val = 1; | |
694575e7 | 10079 | } |
694575e7 | 10080 | } |
78b10e66 | 10081 | |
ed08d51c | 10082 | exit_imsm_reshape_super: |
78b10e66 N |
10083 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
10084 | return ret_val; | |
10085 | } | |
2cda7640 | 10086 | |
eee67a47 AK |
10087 | /******************************************************************************* |
10088 | * Function: wait_for_reshape_imsm | |
10089 | * Description: Function writes new sync_max value and waits until | |
10090 | * reshape process reach new position | |
10091 | * Parameters: | |
10092 | * sra : general array info | |
eee67a47 AK |
10093 | * ndata : number of disks in new array's layout |
10094 | * Returns: | |
10095 | * 0 : success, | |
10096 | * 1 : there is no reshape in progress, | |
10097 | * -1 : fail | |
10098 | ******************************************************************************/ | |
ae9f01f8 | 10099 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
10100 | { |
10101 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
10102 | unsigned long long completed; | |
ae9f01f8 AK |
10103 | /* to_complete : new sync_max position */ |
10104 | unsigned long long to_complete = sra->reshape_progress; | |
10105 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 10106 | |
ae9f01f8 AK |
10107 | if (fd < 0) { |
10108 | dprintf("imsm: wait_for_reshape_imsm() " | |
10109 | "cannot open reshape_position\n"); | |
eee67a47 | 10110 | return 1; |
ae9f01f8 | 10111 | } |
eee67a47 | 10112 | |
ae9f01f8 AK |
10113 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
10114 | dprintf("imsm: wait_for_reshape_imsm() " | |
10115 | "cannot read reshape_position (no reshape in progres)\n"); | |
10116 | close(fd); | |
10117 | return 0; | |
10118 | } | |
eee67a47 | 10119 | |
ae9f01f8 AK |
10120 | if (completed > to_complete) { |
10121 | dprintf("imsm: wait_for_reshape_imsm() " | |
10122 | "wrong next position to set %llu (%llu)\n", | |
10123 | to_complete, completed); | |
10124 | close(fd); | |
10125 | return -1; | |
10126 | } | |
10127 | dprintf("Position set: %llu\n", position_to_set); | |
10128 | if (sysfs_set_num(sra, NULL, "sync_max", | |
10129 | position_to_set) != 0) { | |
10130 | dprintf("imsm: wait_for_reshape_imsm() " | |
10131 | "cannot set reshape position to %llu\n", | |
10132 | position_to_set); | |
10133 | close(fd); | |
10134 | return -1; | |
eee67a47 AK |
10135 | } |
10136 | ||
eee67a47 AK |
10137 | do { |
10138 | char action[20]; | |
10139 | fd_set rfds; | |
10140 | FD_ZERO(&rfds); | |
10141 | FD_SET(fd, &rfds); | |
a47e44fb AK |
10142 | select(fd+1, &rfds, NULL, NULL, NULL); |
10143 | if (sysfs_get_str(sra, NULL, "sync_action", | |
10144 | action, 20) > 0 && | |
10145 | strncmp(action, "reshape", 7) != 0) | |
10146 | break; | |
eee67a47 | 10147 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
10148 | dprintf("imsm: wait_for_reshape_imsm() " |
10149 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
10150 | close(fd); |
10151 | return 1; | |
10152 | } | |
eee67a47 AK |
10153 | } while (completed < to_complete); |
10154 | close(fd); | |
10155 | return 0; | |
10156 | ||
10157 | } | |
10158 | ||
b915c95f AK |
10159 | /******************************************************************************* |
10160 | * Function: check_degradation_change | |
10161 | * Description: Check that array hasn't become failed. | |
10162 | * Parameters: | |
10163 | * info : for sysfs access | |
10164 | * sources : source disks descriptors | |
10165 | * degraded: previous degradation level | |
10166 | * Returns: | |
10167 | * degradation level | |
10168 | ******************************************************************************/ | |
10169 | int check_degradation_change(struct mdinfo *info, | |
10170 | int *sources, | |
10171 | int degraded) | |
10172 | { | |
10173 | unsigned long long new_degraded; | |
e1993023 LD |
10174 | int rv; |
10175 | ||
10176 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
10177 | if ((rv == -1) || (new_degraded != (unsigned long long)degraded)) { | |
b915c95f AK |
10178 | /* check each device to ensure it is still working */ |
10179 | struct mdinfo *sd; | |
10180 | new_degraded = 0; | |
10181 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10182 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
10183 | continue; | |
10184 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
10185 | char sbuf[20]; | |
10186 | if (sysfs_get_str(info, | |
10187 | sd, "state", sbuf, 20) < 0 || | |
10188 | strstr(sbuf, "faulty") || | |
10189 | strstr(sbuf, "in_sync") == NULL) { | |
10190 | /* this device is dead */ | |
10191 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
10192 | if (sd->disk.raid_disk >= 0 && | |
10193 | sources[sd->disk.raid_disk] >= 0) { | |
10194 | close(sources[ | |
10195 | sd->disk.raid_disk]); | |
10196 | sources[sd->disk.raid_disk] = | |
10197 | -1; | |
10198 | } | |
10199 | new_degraded++; | |
10200 | } | |
10201 | } | |
10202 | } | |
10203 | } | |
10204 | ||
10205 | return new_degraded; | |
10206 | } | |
10207 | ||
10f22854 AK |
10208 | /******************************************************************************* |
10209 | * Function: imsm_manage_reshape | |
10210 | * Description: Function finds array under reshape and it manages reshape | |
10211 | * process. It creates stripes backups (if required) and sets | |
10212 | * checheckpoits. | |
10213 | * Parameters: | |
10214 | * afd : Backup handle (nattive) - not used | |
10215 | * sra : general array info | |
10216 | * reshape : reshape parameters - not used | |
10217 | * st : supertype structure | |
10218 | * blocks : size of critical section [blocks] | |
10219 | * fds : table of source device descriptor | |
10220 | * offsets : start of array (offest per devices) | |
10221 | * dests : not used | |
10222 | * destfd : table of destination device descriptor | |
10223 | * destoffsets : table of destination offsets (per device) | |
10224 | * Returns: | |
10225 | * 1 : success, reshape is done | |
10226 | * 0 : fail | |
10227 | ******************************************************************************/ | |
999b4972 N |
10228 | static int imsm_manage_reshape( |
10229 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 10230 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
10231 | int *fds, unsigned long long *offsets, |
10232 | int dests, int *destfd, unsigned long long *destoffsets) | |
10233 | { | |
10f22854 AK |
10234 | int ret_val = 0; |
10235 | struct intel_super *super = st->sb; | |
10236 | struct intel_dev *dv = NULL; | |
10237 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 10238 | struct imsm_map *map_src; |
10f22854 AK |
10239 | int migr_vol_qan = 0; |
10240 | int ndata, odata; /* [bytes] */ | |
10241 | int chunk; /* [bytes] */ | |
10242 | struct migr_record *migr_rec; | |
10243 | char *buf = NULL; | |
10244 | unsigned int buf_size; /* [bytes] */ | |
10245 | unsigned long long max_position; /* array size [bytes] */ | |
10246 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
10247 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
10248 | unsigned long long start_src; /* [bytes] */ |
10249 | unsigned long long start; /* [bytes] */ | |
10250 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 10251 | int degraded = 0; |
ab724b98 | 10252 | int source_layout = 0; |
10f22854 | 10253 | |
1ab242d8 | 10254 | if (!fds || !offsets || !sra) |
10f22854 AK |
10255 | goto abort; |
10256 | ||
10257 | /* Find volume during the reshape */ | |
10258 | for (dv = super->devlist; dv; dv = dv->next) { | |
10259 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
10260 | && dv->dev->vol.migr_state == 1) { | |
10261 | dev = dv->dev; | |
10262 | migr_vol_qan++; | |
10263 | } | |
10264 | } | |
10265 | /* Only one volume can migrate at the same time */ | |
10266 | if (migr_vol_qan != 1) { | |
e7b84f9d | 10267 | pr_err(": %s", migr_vol_qan ? |
10f22854 AK |
10268 | "Number of migrating volumes greater than 1\n" : |
10269 | "There is no volume during migrationg\n"); | |
10270 | goto abort; | |
10271 | } | |
10272 | ||
238c0a71 | 10273 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
10274 | if (map_src == NULL) |
10275 | goto abort; | |
10f22854 | 10276 | |
238c0a71 AK |
10277 | ndata = imsm_num_data_members(dev, MAP_0); |
10278 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 10279 | |
7b1ab482 | 10280 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
10281 | old_data_stripe_length = odata * chunk; |
10282 | ||
10283 | migr_rec = super->migr_rec; | |
10284 | ||
10f22854 AK |
10285 | /* initialize migration record for start condition */ |
10286 | if (sra->reshape_progress == 0) | |
10287 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
10288 | else { |
10289 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
10290 | dprintf("imsm: cannot restart migration when data " | |
10291 | "are present in copy area.\n"); | |
10292 | goto abort; | |
10293 | } | |
6a75c8ca AK |
10294 | /* Save checkpoint to update migration record for current |
10295 | * reshape position (in md). It can be farther than current | |
10296 | * reshape position in metadata. | |
10297 | */ | |
10298 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
10299 | /* ignore error == 2, this can mean end of reshape here | |
10300 | */ | |
10301 | dprintf("imsm: Cannot write checkpoint to " | |
10302 | "migration record (UNIT_SRC_NORMAL, " | |
10303 | "initial save)\n"); | |
10304 | goto abort; | |
10305 | } | |
b2c59438 | 10306 | } |
10f22854 AK |
10307 | |
10308 | /* size for data */ | |
10309 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
10310 | /* extend buffer size for parity disk */ | |
10311 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
10312 | /* add space for stripe aligment */ | |
10313 | buf_size += old_data_stripe_length; | |
10314 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
10315 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
10316 | goto abort; | |
10317 | } | |
10318 | ||
3ef4403c | 10319 | max_position = sra->component_size * ndata; |
68eb8bc6 | 10320 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
10321 | |
10322 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
10323 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
10324 | /* current reshape position [blocks] */ | |
10325 | unsigned long long current_position = | |
10326 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
10327 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
10328 | unsigned long long border; | |
10329 | ||
b915c95f AK |
10330 | /* Check that array hasn't become failed. |
10331 | */ | |
10332 | degraded = check_degradation_change(sra, fds, degraded); | |
10333 | if (degraded > 1) { | |
10334 | dprintf("imsm: Abort reshape due to degradation" | |
10335 | " level (%i)\n", degraded); | |
10336 | goto abort; | |
10337 | } | |
10338 | ||
10f22854 AK |
10339 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
10340 | ||
10341 | if ((current_position + next_step) > max_position) | |
10342 | next_step = max_position - current_position; | |
10343 | ||
92144abf | 10344 | start = current_position * 512; |
10f22854 AK |
10345 | |
10346 | /* allign reading start to old geometry */ | |
10347 | start_buf_shift = start % old_data_stripe_length; | |
10348 | start_src = start - start_buf_shift; | |
10349 | ||
10350 | border = (start_src / odata) - (start / ndata); | |
10351 | border /= 512; | |
10352 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
10353 | /* save critical stripes to buf | |
10354 | * start - start address of current unit | |
10355 | * to backup [bytes] | |
10356 | * start_src - start address of current unit | |
10357 | * to backup alligned to source array | |
10358 | * [bytes] | |
10359 | */ | |
10360 | unsigned long long next_step_filler = 0; | |
10361 | unsigned long long copy_length = next_step * 512; | |
10362 | ||
10363 | /* allign copy area length to stripe in old geometry */ | |
10364 | next_step_filler = ((copy_length + start_buf_shift) | |
10365 | % old_data_stripe_length); | |
10366 | if (next_step_filler) | |
10367 | next_step_filler = (old_data_stripe_length | |
10368 | - next_step_filler); | |
10369 | dprintf("save_stripes() parameters: start = %llu," | |
10370 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
10371 | "\tstart_in_buf_shift = %llu," | |
10372 | "\tnext_step_filler = %llu\n", | |
10373 | start, start_src, copy_length, | |
10374 | start_buf_shift, next_step_filler); | |
10375 | ||
10376 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
10377 | chunk, map_src->raid_level, |
10378 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
10379 | copy_length + |
10380 | next_step_filler + start_buf_shift, | |
10381 | buf)) { | |
10382 | dprintf("imsm: Cannot save stripes" | |
10383 | " to buffer\n"); | |
10384 | goto abort; | |
10385 | } | |
10386 | /* Convert data to destination format and store it | |
10387 | * in backup general migration area | |
10388 | */ | |
10389 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 10390 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
10391 | dprintf("imsm: Cannot save stripes to " |
10392 | "target devices\n"); | |
10393 | goto abort; | |
10394 | } | |
10395 | if (save_checkpoint_imsm(st, sra, | |
10396 | UNIT_SRC_IN_CP_AREA)) { | |
10397 | dprintf("imsm: Cannot write checkpoint to " | |
10398 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
10399 | goto abort; | |
10400 | } | |
8016a6d4 AK |
10401 | } else { |
10402 | /* set next step to use whole border area */ | |
10403 | border /= next_step; | |
10404 | if (border > 1) | |
10405 | next_step *= border; | |
10f22854 AK |
10406 | } |
10407 | /* When data backed up, checkpoint stored, | |
10408 | * kick the kernel to reshape unit of data | |
10409 | */ | |
10410 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
10411 | /* limit next step to array max position */ |
10412 | if (next_step > max_position) | |
10413 | next_step = max_position; | |
10f22854 AK |
10414 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
10415 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 10416 | sra->reshape_progress = next_step; |
10f22854 AK |
10417 | |
10418 | /* wait until reshape finish */ | |
ae9f01f8 | 10419 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
10420 | dprintf("wait_for_reshape_imsm returned error!\n"); |
10421 | goto abort; | |
10422 | } | |
10f22854 | 10423 | |
0228d92c AK |
10424 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
10425 | /* ignore error == 2, this can mean end of reshape here | |
10426 | */ | |
10f22854 AK |
10427 | dprintf("imsm: Cannot write checkpoint to " |
10428 | "migration record (UNIT_SRC_NORMAL)\n"); | |
10429 | goto abort; | |
10430 | } | |
10431 | ||
10432 | } | |
10433 | ||
10434 | /* return '1' if done */ | |
10435 | ret_val = 1; | |
10436 | abort: | |
10437 | free(buf); | |
10438 | abort_reshape(sra); | |
10439 | ||
10440 | return ret_val; | |
999b4972 | 10441 | } |
71204a50 | 10442 | #endif /* MDASSEMBLE */ |
999b4972 | 10443 | |
cdddbdbc DW |
10444 | struct superswitch super_imsm = { |
10445 | #ifndef MDASSEMBLE | |
10446 | .examine_super = examine_super_imsm, | |
10447 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 10448 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 10449 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
10450 | .detail_super = detail_super_imsm, |
10451 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 10452 | .write_init_super = write_init_super_imsm, |
0e600426 N |
10453 | .validate_geometry = validate_geometry_imsm, |
10454 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 10455 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 10456 | .detail_platform = detail_platform_imsm, |
e50cf220 | 10457 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 10458 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 10459 | .update_subarray = update_subarray_imsm, |
2b959fbf | 10460 | .load_container = load_container_imsm, |
71204a50 N |
10461 | .default_geometry = default_geometry_imsm, |
10462 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
10463 | .reshape_super = imsm_reshape_super, | |
10464 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 10465 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
10466 | #endif |
10467 | .match_home = match_home_imsm, | |
10468 | .uuid_from_super= uuid_from_super_imsm, | |
10469 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 10470 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
10471 | .update_super = update_super_imsm, |
10472 | ||
10473 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 10474 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
10475 | |
10476 | .compare_super = compare_super_imsm, | |
10477 | ||
10478 | .load_super = load_super_imsm, | |
bf5a934a | 10479 | .init_super = init_super_imsm, |
e683ca88 | 10480 | .store_super = store_super_imsm, |
cdddbdbc DW |
10481 | .free_super = free_super_imsm, |
10482 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 10483 | .container_content = container_content_imsm, |
cdddbdbc | 10484 | |
276d77db | 10485 | |
cdddbdbc | 10486 | .external = 1, |
4cce4069 | 10487 | .name = "imsm", |
845dea95 | 10488 | |
0e600426 | 10489 | #ifndef MDASSEMBLE |
845dea95 NB |
10490 | /* for mdmon */ |
10491 | .open_new = imsm_open_new, | |
ed9d66aa | 10492 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
10493 | .set_disk = imsm_set_disk, |
10494 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 10495 | .activate_spare = imsm_activate_spare, |
e8319a19 | 10496 | .process_update = imsm_process_update, |
8273f55e | 10497 | .prepare_update = imsm_prepare_update, |
0e600426 | 10498 | #endif /* MDASSEMBLE */ |
cdddbdbc | 10499 | }; |