]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
78 | MPB_ATTRIB_2TB | \ |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
418f9b36 N |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE) |
85 | ||
86 | /* Define attributes that are unused but not harmful */ | |
87 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 88 | |
8e59f3d8 | 89 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 90 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 91 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 92 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
93 | |
94 | /* Disk configuration info. */ | |
95 | #define IMSM_MAX_DEVICES 255 | |
96 | struct imsm_disk { | |
97 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 98 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 99 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
100 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
101 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
102 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 103 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb | 104 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
105 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
106 | #define IMSM_DISK_FILLERS 3 | |
107 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
108 | }; |
109 | ||
3b451610 AK |
110 | /* map selector for map managment |
111 | */ | |
238c0a71 AK |
112 | #define MAP_0 0 |
113 | #define MAP_1 1 | |
114 | #define MAP_X -1 | |
3b451610 | 115 | |
cdddbdbc DW |
116 | /* RAID map configuration infos. */ |
117 | struct imsm_map { | |
5551b113 CA |
118 | __u32 pba_of_lba0_lo; /* start address of partition */ |
119 | __u32 blocks_per_member_lo;/* blocks per member */ | |
120 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
121 | __u16 blocks_per_strip; |
122 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
123 | #define IMSM_T_STATE_NORMAL 0 | |
124 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
125 | #define IMSM_T_STATE_DEGRADED 2 |
126 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
127 | __u8 raid_level; |
128 | #define IMSM_T_RAID0 0 | |
129 | #define IMSM_T_RAID1 1 | |
130 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
131 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
132 | __u8 num_domains; /* number of parity domains */ |
133 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 134 | __u8 ddf; |
5551b113 CA |
135 | __u32 pba_of_lba0_hi; |
136 | __u32 blocks_per_member_hi; | |
137 | __u32 num_data_stripes_hi; | |
138 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 139 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 140 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
141 | * top byte contains some flags |
142 | */ | |
cdddbdbc DW |
143 | } __attribute__ ((packed)); |
144 | ||
145 | struct imsm_vol { | |
f8f603f1 | 146 | __u32 curr_migr_unit; |
fe7ed8cb | 147 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 148 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
149 | #define MIGR_INIT 0 |
150 | #define MIGR_REBUILD 1 | |
151 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
152 | #define MIGR_GEN_MIGR 3 | |
153 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 154 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
155 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
156 | __u8 dirty; | |
fe7ed8cb DW |
157 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
158 | __u16 verify_errors; /* number of mismatches */ | |
159 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
160 | __u32 filler[4]; | |
cdddbdbc DW |
161 | struct imsm_map map[1]; |
162 | /* here comes another one if migr_state */ | |
163 | } __attribute__ ((packed)); | |
164 | ||
165 | struct imsm_dev { | |
fe7ed8cb | 166 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
167 | __u32 size_low; |
168 | __u32 size_high; | |
fe7ed8cb DW |
169 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
170 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
171 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
172 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
173 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
174 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
175 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
176 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
177 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
178 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
179 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
180 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
181 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
182 | __u32 status; /* Persistent RaidDev status */ |
183 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
184 | __u8 migr_priority; |
185 | __u8 num_sub_vols; | |
186 | __u8 tid; | |
187 | __u8 cng_master_disk; | |
188 | __u16 cache_policy; | |
189 | __u8 cng_state; | |
190 | __u8 cng_sub_state; | |
191 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
192 | __u32 filler[IMSM_DEV_FILLERS]; |
193 | struct imsm_vol vol; | |
194 | } __attribute__ ((packed)); | |
195 | ||
196 | struct imsm_super { | |
197 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
198 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
199 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
200 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
201 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
202 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
203 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
204 | __u8 num_disks; /* 0x38 Number of configured disks */ |
205 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
206 | __u8 error_log_pos; /* 0x3A */ |
207 | __u8 fill[1]; /* 0x3B */ | |
208 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
209 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
210 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
211 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
212 | #define IMSM_FILLERS 35 | |
213 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
214 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
215 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 216 | /* here comes BBM logs */ |
cdddbdbc DW |
217 | } __attribute__ ((packed)); |
218 | ||
604b746f JD |
219 | #define BBM_LOG_MAX_ENTRIES 254 |
220 | ||
221 | struct bbm_log_entry { | |
222 | __u64 defective_block_start; | |
223 | #define UNREADABLE 0xFFFFFFFF | |
224 | __u32 spare_block_offset; | |
225 | __u16 remapped_marked_count; | |
226 | __u16 disk_ordinal; | |
227 | } __attribute__ ((__packed__)); | |
228 | ||
229 | struct bbm_log { | |
230 | __u32 signature; /* 0xABADB10C */ | |
231 | __u32 entry_count; | |
232 | __u32 reserved_spare_block_count; /* 0 */ | |
233 | __u32 reserved; /* 0xFFFF */ | |
234 | __u64 first_spare_lba; | |
235 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
236 | } __attribute__ ((__packed__)); | |
237 | ||
238 | ||
cdddbdbc DW |
239 | #ifndef MDASSEMBLE |
240 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
241 | #endif | |
242 | ||
8e59f3d8 AK |
243 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
244 | ||
245 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
246 | ||
17a4eaf9 AK |
247 | #define MIGR_REC_BUF_SIZE 512 /* size of migr_record i/o buffer */ |
248 | #define MIGR_REC_POSITION 512 /* migr_record position offset on disk, | |
249 | * MIGR_REC_BUF_SIZE <= MIGR_REC_POSITION | |
250 | */ | |
251 | ||
252 | ||
8e59f3d8 AK |
253 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
254 | * be recovered using srcMap */ | |
255 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
256 | * already been migrated and must | |
257 | * be recovered from checkpoint area */ | |
258 | struct migr_record { | |
259 | __u32 rec_status; /* Status used to determine how to restart | |
260 | * migration in case it aborts | |
261 | * in some fashion */ | |
262 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
263 | __u32 family_num; /* Family number of MPB | |
264 | * containing the RaidDev | |
265 | * that is migrating */ | |
266 | __u32 ascending_migr; /* True if migrating in increasing | |
267 | * order of lbas */ | |
268 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
269 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
270 | * member disk | |
271 | * advances per unit-of-operation */ | |
272 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
273 | __u32 dest_1st_member_lba; /* First member lba on first | |
274 | * stripe of destination */ | |
275 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
276 | __u32 post_migr_vol_cap; /* Size of volume after | |
277 | * migration completes */ | |
278 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
279 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
280 | * migration ckpt record was read from | |
281 | * (for recovered migrations) */ | |
282 | } __attribute__ ((__packed__)); | |
283 | ||
ec50f7b6 LM |
284 | struct md_list { |
285 | /* usage marker: | |
286 | * 1: load metadata | |
287 | * 2: metadata does not match | |
288 | * 4: already checked | |
289 | */ | |
290 | int used; | |
291 | char *devname; | |
292 | int found; | |
293 | int container; | |
294 | dev_t st_rdev; | |
295 | struct md_list *next; | |
296 | }; | |
297 | ||
e7b84f9d | 298 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 299 | |
1484e727 DW |
300 | static __u8 migr_type(struct imsm_dev *dev) |
301 | { | |
302 | if (dev->vol.migr_type == MIGR_VERIFY && | |
303 | dev->status & DEV_VERIFY_AND_FIX) | |
304 | return MIGR_REPAIR; | |
305 | else | |
306 | return dev->vol.migr_type; | |
307 | } | |
308 | ||
309 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
310 | { | |
311 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
312 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
313 | */ | |
314 | if (migr_type == MIGR_REPAIR) { | |
315 | dev->vol.migr_type = MIGR_VERIFY; | |
316 | dev->status |= DEV_VERIFY_AND_FIX; | |
317 | } else { | |
318 | dev->vol.migr_type = migr_type; | |
319 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
320 | } | |
321 | } | |
322 | ||
87eb16df | 323 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 324 | { |
654a3817 | 325 | return ROUND_UP(bytes, 512) / 512; |
87eb16df | 326 | } |
cdddbdbc | 327 | |
87eb16df DW |
328 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
329 | { | |
330 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
331 | } |
332 | ||
ba2de7ba DW |
333 | struct intel_dev { |
334 | struct imsm_dev *dev; | |
335 | struct intel_dev *next; | |
f21e18ca | 336 | unsigned index; |
ba2de7ba DW |
337 | }; |
338 | ||
88654014 LM |
339 | struct intel_hba { |
340 | enum sys_dev_type type; | |
341 | char *path; | |
342 | char *pci_id; | |
343 | struct intel_hba *next; | |
344 | }; | |
345 | ||
1a64be56 LM |
346 | enum action { |
347 | DISK_REMOVE = 1, | |
348 | DISK_ADD | |
349 | }; | |
cdddbdbc DW |
350 | /* internal representation of IMSM metadata */ |
351 | struct intel_super { | |
352 | union { | |
949c47a0 DW |
353 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
354 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 355 | }; |
8e59f3d8 AK |
356 | union { |
357 | void *migr_rec_buf; /* buffer for I/O operations */ | |
358 | struct migr_record *migr_rec; /* migration record */ | |
359 | }; | |
51d83f5d AK |
360 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
361 | array, it indicates that mdmon is allowed to clean migration | |
362 | record */ | |
949c47a0 | 363 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
364 | void *next_buf; /* for realloc'ing buf from the manager */ |
365 | size_t next_len; | |
c2c087e6 | 366 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 367 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 368 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 369 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 370 | struct intel_dev *devlist; |
cdddbdbc DW |
371 | struct dl { |
372 | struct dl *next; | |
373 | int index; | |
374 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
375 | int major, minor; | |
376 | char *devname; | |
b9f594fe | 377 | struct imsm_disk disk; |
cdddbdbc | 378 | int fd; |
0dcecb2e DW |
379 | int extent_cnt; |
380 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 381 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 382 | enum action action; |
ca0748fa | 383 | } *disks, *current_disk; |
1a64be56 LM |
384 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
385 | active */ | |
47ee5a45 | 386 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 387 | struct bbm_log *bbm_log; |
88654014 | 388 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 389 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
390 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
391 | }; | |
392 | ||
393 | struct intel_disk { | |
394 | struct imsm_disk disk; | |
395 | #define IMSM_UNKNOWN_OWNER (-1) | |
396 | int owner; | |
397 | struct intel_disk *next; | |
cdddbdbc DW |
398 | }; |
399 | ||
c2c087e6 DW |
400 | struct extent { |
401 | unsigned long long start, size; | |
402 | }; | |
403 | ||
694575e7 KW |
404 | /* definitions of reshape process types */ |
405 | enum imsm_reshape_type { | |
406 | CH_TAKEOVER, | |
b5347799 | 407 | CH_MIGRATION, |
7abc9871 | 408 | CH_ARRAY_SIZE, |
694575e7 KW |
409 | }; |
410 | ||
88758e9d DW |
411 | /* definition of messages passed to imsm_process_update */ |
412 | enum imsm_update_type { | |
413 | update_activate_spare, | |
8273f55e | 414 | update_create_array, |
33414a01 | 415 | update_kill_array, |
aa534678 | 416 | update_rename_array, |
1a64be56 | 417 | update_add_remove_disk, |
78b10e66 | 418 | update_reshape_container_disks, |
48c5303a | 419 | update_reshape_migration, |
2d40f3a1 AK |
420 | update_takeover, |
421 | update_general_migration_checkpoint, | |
f3871fdc | 422 | update_size_change, |
88758e9d DW |
423 | }; |
424 | ||
425 | struct imsm_update_activate_spare { | |
426 | enum imsm_update_type type; | |
d23fe947 | 427 | struct dl *dl; |
88758e9d DW |
428 | int slot; |
429 | int array; | |
430 | struct imsm_update_activate_spare *next; | |
431 | }; | |
432 | ||
78b10e66 N |
433 | struct geo_params { |
434 | int dev_id; | |
435 | char *dev_name; | |
d04f65f4 | 436 | unsigned long long size; |
78b10e66 N |
437 | int level; |
438 | int layout; | |
439 | int chunksize; | |
440 | int raid_disks; | |
441 | }; | |
442 | ||
bb025c2f KW |
443 | enum takeover_direction { |
444 | R10_TO_R0, | |
445 | R0_TO_R10 | |
446 | }; | |
447 | struct imsm_update_takeover { | |
448 | enum imsm_update_type type; | |
449 | int subarray; | |
450 | enum takeover_direction direction; | |
451 | }; | |
78b10e66 N |
452 | |
453 | struct imsm_update_reshape { | |
454 | enum imsm_update_type type; | |
455 | int old_raid_disks; | |
456 | int new_raid_disks; | |
48c5303a PC |
457 | |
458 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
459 | }; | |
460 | ||
461 | struct imsm_update_reshape_migration { | |
462 | enum imsm_update_type type; | |
463 | int old_raid_disks; | |
464 | int new_raid_disks; | |
465 | /* fields for array migration changes | |
466 | */ | |
467 | int subdev; | |
468 | int new_level; | |
469 | int new_layout; | |
4bba0439 | 470 | int new_chunksize; |
48c5303a | 471 | |
d195167d | 472 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
473 | }; |
474 | ||
f3871fdc AK |
475 | struct imsm_update_size_change { |
476 | enum imsm_update_type type; | |
477 | int subdev; | |
478 | long long new_size; | |
479 | }; | |
480 | ||
2d40f3a1 AK |
481 | struct imsm_update_general_migration_checkpoint { |
482 | enum imsm_update_type type; | |
483 | __u32 curr_migr_unit; | |
484 | }; | |
485 | ||
54c2c1ea DW |
486 | struct disk_info { |
487 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
488 | }; | |
489 | ||
8273f55e DW |
490 | struct imsm_update_create_array { |
491 | enum imsm_update_type type; | |
8273f55e | 492 | int dev_idx; |
6a3e913e | 493 | struct imsm_dev dev; |
8273f55e DW |
494 | }; |
495 | ||
33414a01 DW |
496 | struct imsm_update_kill_array { |
497 | enum imsm_update_type type; | |
498 | int dev_idx; | |
499 | }; | |
500 | ||
aa534678 DW |
501 | struct imsm_update_rename_array { |
502 | enum imsm_update_type type; | |
503 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
504 | int dev_idx; | |
505 | }; | |
506 | ||
1a64be56 | 507 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
508 | enum imsm_update_type type; |
509 | }; | |
510 | ||
88654014 LM |
511 | |
512 | static const char *_sys_dev_type[] = { | |
513 | [SYS_DEV_UNKNOWN] = "Unknown", | |
514 | [SYS_DEV_SAS] = "SAS", | |
515 | [SYS_DEV_SATA] = "SATA" | |
516 | }; | |
517 | ||
518 | const char *get_sys_dev_type(enum sys_dev_type type) | |
519 | { | |
520 | if (type >= SYS_DEV_MAX) | |
521 | type = SYS_DEV_UNKNOWN; | |
522 | ||
523 | return _sys_dev_type[type]; | |
524 | } | |
525 | ||
526 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
527 | { | |
503975b9 N |
528 | struct intel_hba *result = xmalloc(sizeof(*result)); |
529 | ||
530 | result->type = device->type; | |
531 | result->path = xstrdup(device->path); | |
532 | result->next = NULL; | |
533 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
534 | result->pci_id++; | |
535 | ||
88654014 LM |
536 | return result; |
537 | } | |
538 | ||
539 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
540 | { | |
541 | struct intel_hba *result=NULL; | |
542 | for (result = hba; result; result = result->next) { | |
543 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
544 | break; | |
545 | } | |
546 | return result; | |
547 | } | |
548 | ||
b4cf4cba | 549 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
550 | { |
551 | struct intel_hba *hba; | |
552 | ||
553 | /* check if disk attached to Intel HBA */ | |
554 | hba = find_intel_hba(super->hba, device); | |
555 | if (hba != NULL) | |
556 | return 1; | |
557 | /* Check if HBA is already attached to super */ | |
558 | if (super->hba == NULL) { | |
559 | super->hba = alloc_intel_hba(device); | |
560 | return 1; | |
561 | } | |
562 | ||
563 | hba = super->hba; | |
564 | /* Intel metadata allows for all disks attached to the same type HBA. | |
565 | * Do not sypport odf HBA types mixing | |
566 | */ | |
567 | if (device->type != hba->type) | |
568 | return 2; | |
569 | ||
570 | while (hba->next) | |
571 | hba = hba->next; | |
572 | ||
573 | hba->next = alloc_intel_hba(device); | |
574 | return 1; | |
575 | } | |
576 | ||
577 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
578 | { | |
579 | struct sys_dev *list, *elem, *prev; | |
580 | char *disk_path; | |
581 | ||
582 | if ((list = find_intel_devices()) == NULL) | |
583 | return 0; | |
584 | ||
585 | if (fd < 0) | |
586 | disk_path = (char *) devname; | |
587 | else | |
588 | disk_path = diskfd_to_devpath(fd); | |
589 | ||
590 | if (!disk_path) { | |
591 | free_sys_dev(&list); | |
592 | return 0; | |
593 | } | |
594 | ||
595 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
596 | if (path_attached_to_hba(disk_path, elem->path)) { | |
597 | if (prev == NULL) | |
598 | list = list->next; | |
599 | else | |
600 | prev->next = elem->next; | |
601 | elem->next = NULL; | |
602 | if (disk_path != devname) | |
603 | free(disk_path); | |
604 | free_sys_dev(&list); | |
605 | return elem; | |
606 | } | |
607 | } | |
608 | if (disk_path != devname) | |
609 | free(disk_path); | |
610 | free_sys_dev(&list); | |
611 | ||
612 | return NULL; | |
613 | } | |
614 | ||
615 | ||
d424212e N |
616 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
617 | char *devname); | |
f2f5c343 | 618 | |
cdddbdbc DW |
619 | static struct supertype *match_metadata_desc_imsm(char *arg) |
620 | { | |
621 | struct supertype *st; | |
622 | ||
623 | if (strcmp(arg, "imsm") != 0 && | |
624 | strcmp(arg, "default") != 0 | |
625 | ) | |
626 | return NULL; | |
627 | ||
503975b9 | 628 | st = xcalloc(1, sizeof(*st)); |
d1d599ea | 629 | st->container_dev = NoMdDev; |
cdddbdbc DW |
630 | st->ss = &super_imsm; |
631 | st->max_devs = IMSM_MAX_DEVICES; | |
632 | st->minor_version = 0; | |
633 | st->sb = NULL; | |
634 | return st; | |
635 | } | |
636 | ||
0e600426 | 637 | #ifndef MDASSEMBLE |
cdddbdbc DW |
638 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
639 | { | |
640 | return &mpb->sig[MPB_SIG_LEN]; | |
641 | } | |
9e2d750d | 642 | #endif |
cdddbdbc | 643 | |
949c47a0 DW |
644 | /* retrieve a disk directly from the anchor when the anchor is known to be |
645 | * up-to-date, currently only at load time | |
646 | */ | |
647 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 648 | { |
949c47a0 | 649 | if (index >= mpb->num_disks) |
cdddbdbc DW |
650 | return NULL; |
651 | return &mpb->disk[index]; | |
652 | } | |
653 | ||
95d07a2c LM |
654 | /* retrieve the disk description based on a index of the disk |
655 | * in the sub-array | |
656 | */ | |
657 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 658 | { |
b9f594fe DW |
659 | struct dl *d; |
660 | ||
661 | for (d = super->disks; d; d = d->next) | |
662 | if (d->index == index) | |
95d07a2c LM |
663 | return d; |
664 | ||
665 | return NULL; | |
666 | } | |
667 | /* retrieve a disk from the parsed metadata */ | |
668 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
669 | { | |
670 | struct dl *dl; | |
671 | ||
672 | dl = get_imsm_dl_disk(super, index); | |
673 | if (dl) | |
674 | return &dl->disk; | |
675 | ||
b9f594fe | 676 | return NULL; |
949c47a0 DW |
677 | } |
678 | ||
679 | /* generate a checksum directly from the anchor when the anchor is known to be | |
680 | * up-to-date, currently only at load or write_super after coalescing | |
681 | */ | |
682 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
683 | { |
684 | __u32 end = mpb->mpb_size / sizeof(end); | |
685 | __u32 *p = (__u32 *) mpb; | |
686 | __u32 sum = 0; | |
687 | ||
97f734fd N |
688 | while (end--) { |
689 | sum += __le32_to_cpu(*p); | |
690 | p++; | |
691 | } | |
cdddbdbc DW |
692 | |
693 | return sum - __le32_to_cpu(mpb->check_sum); | |
694 | } | |
695 | ||
a965f303 DW |
696 | static size_t sizeof_imsm_map(struct imsm_map *map) |
697 | { | |
698 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
699 | } | |
700 | ||
701 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 702 | { |
5e7b0330 AK |
703 | /* A device can have 2 maps if it is in the middle of a migration. |
704 | * If second_map is: | |
238c0a71 AK |
705 | * MAP_0 - we return the first map |
706 | * MAP_1 - we return the second map if it exists, else NULL | |
707 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 708 | */ |
a965f303 | 709 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 710 | struct imsm_map *map2 = NULL; |
a965f303 | 711 | |
9535fc47 AK |
712 | if (dev->vol.migr_state) |
713 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 714 | |
9535fc47 | 715 | switch (second_map) { |
3b451610 | 716 | case MAP_0: |
9535fc47 | 717 | break; |
3b451610 | 718 | case MAP_1: |
9535fc47 AK |
719 | map = map2; |
720 | break; | |
238c0a71 | 721 | case MAP_X: |
9535fc47 AK |
722 | if (map2) |
723 | map = map2; | |
724 | break; | |
9535fc47 AK |
725 | default: |
726 | map = NULL; | |
727 | } | |
728 | return map; | |
5e7b0330 | 729 | |
a965f303 | 730 | } |
cdddbdbc | 731 | |
3393c6af DW |
732 | /* return the size of the device. |
733 | * migr_state increases the returned size if map[0] were to be duplicated | |
734 | */ | |
735 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
736 | { |
737 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 738 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
739 | |
740 | /* migrating means an additional map */ | |
a965f303 | 741 | if (dev->vol.migr_state) |
238c0a71 | 742 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 743 | else if (migr_state) |
238c0a71 | 744 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
745 | |
746 | return size; | |
747 | } | |
748 | ||
54c2c1ea DW |
749 | #ifndef MDASSEMBLE |
750 | /* retrieve disk serial number list from a metadata update */ | |
751 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
752 | { | |
753 | void *u = update; | |
754 | struct disk_info *inf; | |
755 | ||
756 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
757 | sizeof_imsm_dev(&update->dev, 0); | |
758 | ||
759 | return inf; | |
760 | } | |
761 | #endif | |
762 | ||
949c47a0 | 763 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
764 | { |
765 | int offset; | |
766 | int i; | |
767 | void *_mpb = mpb; | |
768 | ||
949c47a0 | 769 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
770 | return NULL; |
771 | ||
772 | /* devices start after all disks */ | |
773 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
774 | ||
775 | for (i = 0; i <= index; i++) | |
776 | if (i == index) | |
777 | return _mpb + offset; | |
778 | else | |
3393c6af | 779 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
780 | |
781 | return NULL; | |
782 | } | |
783 | ||
949c47a0 DW |
784 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
785 | { | |
ba2de7ba DW |
786 | struct intel_dev *dv; |
787 | ||
949c47a0 DW |
788 | if (index >= super->anchor->num_raid_devs) |
789 | return NULL; | |
ba2de7ba DW |
790 | for (dv = super->devlist; dv; dv = dv->next) |
791 | if (dv->index == index) | |
792 | return dv->dev; | |
793 | return NULL; | |
949c47a0 DW |
794 | } |
795 | ||
98130f40 AK |
796 | /* |
797 | * for second_map: | |
238c0a71 AK |
798 | * == MAP_0 get first map |
799 | * == MAP_1 get second map | |
800 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
801 | */ |
802 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
803 | int slot, | |
804 | int second_map) | |
7eef0453 DW |
805 | { |
806 | struct imsm_map *map; | |
807 | ||
5e7b0330 | 808 | map = get_imsm_map(dev, second_map); |
7eef0453 | 809 | |
ff077194 DW |
810 | /* top byte identifies disk under rebuild */ |
811 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
812 | } | |
813 | ||
814 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 815 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 816 | { |
98130f40 | 817 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
818 | |
819 | return ord_to_idx(ord); | |
7eef0453 DW |
820 | } |
821 | ||
be73972f DW |
822 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
823 | { | |
824 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
825 | } | |
826 | ||
f21e18ca | 827 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
828 | { |
829 | int slot; | |
830 | __u32 ord; | |
831 | ||
832 | for (slot = 0; slot < map->num_members; slot++) { | |
833 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
834 | if (ord_to_idx(ord) == idx) | |
835 | return slot; | |
836 | } | |
837 | ||
838 | return -1; | |
839 | } | |
840 | ||
cdddbdbc DW |
841 | static int get_imsm_raid_level(struct imsm_map *map) |
842 | { | |
843 | if (map->raid_level == 1) { | |
844 | if (map->num_members == 2) | |
845 | return 1; | |
846 | else | |
847 | return 10; | |
848 | } | |
849 | ||
850 | return map->raid_level; | |
851 | } | |
852 | ||
c2c087e6 DW |
853 | static int cmp_extent(const void *av, const void *bv) |
854 | { | |
855 | const struct extent *a = av; | |
856 | const struct extent *b = bv; | |
857 | if (a->start < b->start) | |
858 | return -1; | |
859 | if (a->start > b->start) | |
860 | return 1; | |
861 | return 0; | |
862 | } | |
863 | ||
0dcecb2e | 864 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 865 | { |
c2c087e6 | 866 | int memberships = 0; |
620b1713 | 867 | int i; |
c2c087e6 | 868 | |
949c47a0 DW |
869 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
870 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 871 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 872 | |
620b1713 DW |
873 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
874 | memberships++; | |
c2c087e6 | 875 | } |
0dcecb2e DW |
876 | |
877 | return memberships; | |
878 | } | |
879 | ||
b81221b7 CA |
880 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
881 | ||
5551b113 CA |
882 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
883 | { | |
884 | if (lo == 0 || hi == 0) | |
885 | return 1; | |
886 | *lo = __le32_to_cpu((unsigned)n); | |
887 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
888 | return 0; | |
889 | } | |
890 | ||
891 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
892 | { | |
893 | return (unsigned long long)__le32_to_cpu(lo) | | |
894 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
895 | } | |
896 | ||
897 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
898 | { | |
899 | if (disk == NULL) | |
900 | return 0; | |
901 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
902 | } | |
903 | ||
904 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
905 | { | |
906 | if (map == NULL) | |
907 | return 0; | |
908 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
909 | } | |
910 | ||
911 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
912 | { | |
913 | if (map == NULL) | |
914 | return 0; | |
915 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
916 | } | |
917 | ||
ae2416e7 | 918 | #ifndef MDASSEMBLE |
5551b113 CA |
919 | static unsigned long long num_data_stripes(struct imsm_map *map) |
920 | { | |
921 | if (map == NULL) | |
922 | return 0; | |
923 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
924 | } | |
925 | ||
926 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
927 | { | |
928 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
929 | } | |
ae2416e7 | 930 | #endif |
5551b113 CA |
931 | |
932 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
933 | { | |
934 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
935 | } | |
936 | ||
937 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
938 | { | |
939 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
940 | } | |
941 | ||
942 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
943 | { | |
944 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
945 | } | |
946 | ||
0dcecb2e DW |
947 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
948 | { | |
949 | /* find a list of used extents on the given physical device */ | |
950 | struct extent *rv, *e; | |
620b1713 | 951 | int i; |
0dcecb2e | 952 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
953 | __u32 reservation; |
954 | ||
955 | /* trim the reserved area for spares, so they can join any array | |
956 | * regardless of whether the OROM has assigned sectors from the | |
957 | * IMSM_RESERVED_SECTORS region | |
958 | */ | |
959 | if (dl->index == -1) | |
b81221b7 | 960 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
961 | else |
962 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 963 | |
503975b9 | 964 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
965 | e = rv; |
966 | ||
949c47a0 DW |
967 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
968 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 969 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 970 | |
620b1713 | 971 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
972 | e->start = pba_of_lba0(map); |
973 | e->size = blocks_per_member(map); | |
620b1713 | 974 | e++; |
c2c087e6 DW |
975 | } |
976 | } | |
977 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
978 | ||
14e8215b DW |
979 | /* determine the start of the metadata |
980 | * when no raid devices are defined use the default | |
981 | * ...otherwise allow the metadata to truncate the value | |
982 | * as is the case with older versions of imsm | |
983 | */ | |
984 | if (memberships) { | |
985 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 986 | unsigned long long remainder; |
14e8215b | 987 | |
5551b113 | 988 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
989 | /* round down to 1k block to satisfy precision of the kernel |
990 | * 'size' interface | |
991 | */ | |
992 | remainder &= ~1UL; | |
993 | /* make sure remainder is still sane */ | |
f21e18ca | 994 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 995 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
996 | if (reservation > remainder) |
997 | reservation = remainder; | |
998 | } | |
5551b113 | 999 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1000 | e->size = 0; |
1001 | return rv; | |
1002 | } | |
1003 | ||
14e8215b DW |
1004 | /* try to determine how much space is reserved for metadata from |
1005 | * the last get_extents() entry, otherwise fallback to the | |
1006 | * default | |
1007 | */ | |
1008 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1009 | { | |
1010 | struct extent *e; | |
1011 | int i; | |
1012 | __u32 rv; | |
1013 | ||
1014 | /* for spares just return a minimal reservation which will grow | |
1015 | * once the spare is picked up by an array | |
1016 | */ | |
1017 | if (dl->index == -1) | |
1018 | return MPB_SECTOR_CNT; | |
1019 | ||
1020 | e = get_extents(super, dl); | |
1021 | if (!e) | |
1022 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1023 | ||
1024 | /* scroll to last entry */ | |
1025 | for (i = 0; e[i].size; i++) | |
1026 | continue; | |
1027 | ||
5551b113 | 1028 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1029 | |
1030 | free(e); | |
1031 | ||
1032 | return rv; | |
1033 | } | |
1034 | ||
25ed7e59 DW |
1035 | static int is_spare(struct imsm_disk *disk) |
1036 | { | |
1037 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1038 | } | |
1039 | ||
1040 | static int is_configured(struct imsm_disk *disk) | |
1041 | { | |
1042 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1043 | } | |
1044 | ||
1045 | static int is_failed(struct imsm_disk *disk) | |
1046 | { | |
1047 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1048 | } | |
1049 | ||
b81221b7 CA |
1050 | /* try to determine how much space is reserved for metadata from |
1051 | * the last get_extents() entry on the smallest active disk, | |
1052 | * otherwise fallback to the default | |
1053 | */ | |
1054 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1055 | { | |
1056 | struct extent *e; | |
1057 | int i; | |
5551b113 CA |
1058 | unsigned long long min_active; |
1059 | __u32 remainder; | |
b81221b7 CA |
1060 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1061 | struct dl *dl, *dl_min = NULL; | |
1062 | ||
1063 | if (!super) | |
1064 | return rv; | |
1065 | ||
1066 | min_active = 0; | |
1067 | for (dl = super->disks; dl; dl = dl->next) { | |
1068 | if (dl->index < 0) | |
1069 | continue; | |
5551b113 CA |
1070 | unsigned long long blocks = total_blocks(&dl->disk); |
1071 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1072 | dl_min = dl; |
5551b113 | 1073 | min_active = blocks; |
b81221b7 CA |
1074 | } |
1075 | } | |
1076 | if (!dl_min) | |
1077 | return rv; | |
1078 | ||
1079 | /* find last lba used by subarrays on the smallest active disk */ | |
1080 | e = get_extents(super, dl_min); | |
1081 | if (!e) | |
1082 | return rv; | |
1083 | for (i = 0; e[i].size; i++) | |
1084 | continue; | |
1085 | ||
1086 | remainder = min_active - e[i].start; | |
1087 | free(e); | |
1088 | ||
1089 | /* to give priority to recovery we should not require full | |
1090 | IMSM_RESERVED_SECTORS from the spare */ | |
1091 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1092 | ||
1093 | /* if real reservation is smaller use that value */ | |
1094 | return (remainder < rv) ? remainder : rv; | |
1095 | } | |
1096 | ||
80e7f8c3 AC |
1097 | /* Return minimum size of a spare that can be used in this array*/ |
1098 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1099 | { | |
1100 | struct intel_super *super = st->sb; | |
1101 | struct dl *dl; | |
1102 | struct extent *e; | |
1103 | int i; | |
1104 | unsigned long long rv = 0; | |
1105 | ||
1106 | if (!super) | |
1107 | return rv; | |
1108 | /* find first active disk in array */ | |
1109 | dl = super->disks; | |
1110 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1111 | dl = dl->next; | |
1112 | if (!dl) | |
1113 | return rv; | |
1114 | /* find last lba used by subarrays */ | |
1115 | e = get_extents(super, dl); | |
1116 | if (!e) | |
1117 | return rv; | |
1118 | for (i = 0; e[i].size; i++) | |
1119 | continue; | |
1120 | if (i > 0) | |
1121 | rv = e[i-1].start + e[i-1].size; | |
1122 | free(e); | |
b81221b7 | 1123 | |
80e7f8c3 | 1124 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1125 | rv = rv + imsm_min_reserved_sectors(super); |
1126 | ||
80e7f8c3 AC |
1127 | return rv * 512; |
1128 | } | |
1129 | ||
d1e02575 AK |
1130 | static int is_gen_migration(struct imsm_dev *dev); |
1131 | ||
1799c9e8 | 1132 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1133 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1134 | struct imsm_dev *dev); | |
1e5c6983 | 1135 | |
c47b0ff6 AK |
1136 | static void print_imsm_dev(struct intel_super *super, |
1137 | struct imsm_dev *dev, | |
1138 | char *uuid, | |
1139 | int disk_idx) | |
cdddbdbc DW |
1140 | { |
1141 | __u64 sz; | |
0d80bb2f | 1142 | int slot, i; |
238c0a71 AK |
1143 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1144 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1145 | __u32 ord; |
cdddbdbc DW |
1146 | |
1147 | printf("\n"); | |
1e7bc0ed | 1148 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1149 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1150 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1151 | if (map2) | |
1152 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1153 | printf("\n"); | |
1154 | printf(" Members : %d", map->num_members); | |
1155 | if (map2) | |
1156 | printf(" <-- %d", map2->num_members); | |
1157 | printf("\n"); | |
0d80bb2f DW |
1158 | printf(" Slots : ["); |
1159 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1160 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1161 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1162 | } | |
dd8bcb3b AK |
1163 | printf("]"); |
1164 | if (map2) { | |
1165 | printf(" <-- ["); | |
1166 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1167 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1168 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1169 | } | |
1170 | printf("]"); | |
1171 | } | |
1172 | printf("\n"); | |
7095bccb AK |
1173 | printf(" Failed disk : "); |
1174 | if (map->failed_disk_num == 0xff) | |
1175 | printf("none"); | |
1176 | else | |
1177 | printf("%i", map->failed_disk_num); | |
1178 | printf("\n"); | |
620b1713 DW |
1179 | slot = get_imsm_disk_slot(map, disk_idx); |
1180 | if (slot >= 0) { | |
238c0a71 | 1181 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1182 | printf(" This Slot : %d%s\n", slot, |
1183 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1184 | } else | |
cdddbdbc DW |
1185 | printf(" This Slot : ?\n"); |
1186 | sz = __le32_to_cpu(dev->size_high); | |
1187 | sz <<= 32; | |
1188 | sz += __le32_to_cpu(dev->size_low); | |
1189 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1190 | human_size(sz * 512)); | |
5551b113 | 1191 | sz = blocks_per_member(map); |
cdddbdbc DW |
1192 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, |
1193 | human_size(sz * 512)); | |
5551b113 CA |
1194 | printf(" Sector Offset : %llu\n", |
1195 | pba_of_lba0(map)); | |
1196 | printf(" Num Stripes : %llu\n", | |
1197 | num_data_stripes(map)); | |
dd8bcb3b | 1198 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1199 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1200 | if (map2) |
1201 | printf(" <-- %u KiB", | |
1202 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1203 | printf("\n"); | |
cdddbdbc | 1204 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1205 | printf(" Migrate State : "); |
1484e727 DW |
1206 | if (dev->vol.migr_state) { |
1207 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1208 | printf("initialize\n"); |
1484e727 | 1209 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1210 | printf("rebuild\n"); |
1484e727 | 1211 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1212 | printf("check\n"); |
1484e727 | 1213 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1214 | printf("general migration\n"); |
1484e727 | 1215 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1216 | printf("state change\n"); |
1484e727 | 1217 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1218 | printf("repair\n"); |
1484e727 | 1219 | else |
8655a7b1 DW |
1220 | printf("<unknown:%d>\n", migr_type(dev)); |
1221 | } else | |
1222 | printf("idle\n"); | |
3393c6af DW |
1223 | printf(" Map State : %s", map_state_str[map->map_state]); |
1224 | if (dev->vol.migr_state) { | |
238c0a71 | 1225 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1226 | |
b10b37b8 | 1227 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1228 | printf("\n Checkpoint : %u ", |
1229 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
3136abe5 | 1230 | if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0))) |
464d40e8 LD |
1231 | printf("(N/A)"); |
1232 | else | |
1233 | printf("(%llu)", (unsigned long long) | |
1234 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1235 | } |
1236 | printf("\n"); | |
cdddbdbc | 1237 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1238 | } |
1239 | ||
0ec1f4e8 | 1240 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1241 | { |
1f24f035 | 1242 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1243 | __u64 sz; |
1244 | ||
0ec1f4e8 | 1245 | if (index < -1 || !disk) |
e9d82038 DW |
1246 | return; |
1247 | ||
cdddbdbc | 1248 | printf("\n"); |
1f24f035 | 1249 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1250 | if (index >= 0) |
1251 | printf(" Disk%02d Serial : %s\n", index, str); | |
1252 | else | |
1253 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1254 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1255 | is_configured(disk) ? " active" : "", | |
1256 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1257 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1258 | sz = total_blocks(disk) - reserved; |
cdddbdbc DW |
1259 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1260 | human_size(sz * 512)); | |
1261 | } | |
1262 | ||
520e69e2 AK |
1263 | void examine_migr_rec_imsm(struct intel_super *super) |
1264 | { | |
1265 | struct migr_record *migr_rec = super->migr_rec; | |
1266 | struct imsm_super *mpb = super->anchor; | |
1267 | int i; | |
1268 | ||
1269 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1270 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1271 | struct imsm_map *map; |
b4ab44d8 | 1272 | int slot = -1; |
3136abe5 | 1273 | |
520e69e2 AK |
1274 | if (is_gen_migration(dev) == 0) |
1275 | continue; | |
1276 | ||
1277 | printf("\nMigration Record Information:"); | |
3136abe5 | 1278 | |
44bfe6df AK |
1279 | /* first map under migration */ |
1280 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1281 | if (map) |
1282 | slot = get_imsm_disk_slot(map, super->disks->index); | |
1283 | if ((map == NULL) || (slot > 1) || (slot < 0)) { | |
520e69e2 AK |
1284 | printf(" Empty\n "); |
1285 | printf("Examine one of first two disks in array\n"); | |
1286 | break; | |
1287 | } | |
1288 | printf("\n Status : "); | |
1289 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1290 | printf("Normal\n"); | |
1291 | else | |
1292 | printf("Contains Data\n"); | |
1293 | printf(" Current Unit : %u\n", | |
1294 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1295 | printf(" Family : %u\n", | |
1296 | __le32_to_cpu(migr_rec->family_num)); | |
1297 | printf(" Ascending : %u\n", | |
1298 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1299 | printf(" Blocks Per Unit : %u\n", | |
1300 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1301 | printf(" Dest. Depth Per Unit : %u\n", | |
1302 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1303 | printf(" Checkpoint Area pba : %u\n", | |
1304 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1305 | printf(" First member lba : %u\n", | |
1306 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1307 | printf(" Total Number of Units : %u\n", | |
1308 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1309 | printf(" Size of volume : %u\n", | |
1310 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1311 | printf(" Expansion space for LBA64 : %u\n", | |
1312 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1313 | printf(" Record was read from : %u\n", | |
1314 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1315 | ||
1316 | break; | |
1317 | } | |
1318 | } | |
9e2d750d | 1319 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1320 | /******************************************************************************* |
1321 | * function: imsm_check_attributes | |
1322 | * Description: Function checks if features represented by attributes flags | |
1323 | * are supported by mdadm. | |
1324 | * Parameters: | |
1325 | * attributes - Attributes read from metadata | |
1326 | * Returns: | |
1327 | * 0 - passed attributes contains unsupported features flags | |
1328 | * 1 - all features are supported | |
1329 | ******************************************************************************/ | |
1330 | static int imsm_check_attributes(__u32 attributes) | |
1331 | { | |
1332 | int ret_val = 1; | |
418f9b36 N |
1333 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1334 | ||
1335 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1336 | |
1337 | not_supported &= attributes; | |
1338 | if (not_supported) { | |
e7b84f9d | 1339 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1340 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1341 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1342 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1343 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1344 | } | |
1345 | if (not_supported & MPB_ATTRIB_2TB) { | |
1346 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1347 | not_supported ^= MPB_ATTRIB_2TB; | |
1348 | } | |
1349 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1350 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1351 | not_supported ^= MPB_ATTRIB_RAID0; | |
1352 | } | |
1353 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1354 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1355 | not_supported ^= MPB_ATTRIB_RAID1; | |
1356 | } | |
1357 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1358 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1359 | not_supported ^= MPB_ATTRIB_RAID10; | |
1360 | } | |
1361 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1362 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1363 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1364 | } | |
1365 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1366 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1367 | not_supported ^= MPB_ATTRIB_RAID5; | |
1368 | } | |
1369 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1370 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1371 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1372 | } | |
1373 | if (not_supported & MPB_ATTRIB_BBM) { | |
1374 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1375 | not_supported ^= MPB_ATTRIB_BBM; | |
1376 | } | |
1377 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1378 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1379 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1380 | } | |
1381 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1382 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1383 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1384 | } | |
1385 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1386 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1387 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1388 | } | |
1389 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1390 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1391 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1392 | } | |
1393 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1394 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1395 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1396 | } | |
1397 | ||
1398 | if (not_supported) | |
1399 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1400 | ||
1401 | ret_val = 0; | |
1402 | } | |
1403 | ||
1404 | return ret_val; | |
1405 | } | |
1406 | ||
9e2d750d | 1407 | #ifndef MDASSEMBLE |
a5d85af7 | 1408 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1409 | |
cdddbdbc DW |
1410 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1411 | { | |
1412 | struct intel_super *super = st->sb; | |
949c47a0 | 1413 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1414 | char str[MAX_SIGNATURE_LENGTH]; |
1415 | int i; | |
27fd6274 DW |
1416 | struct mdinfo info; |
1417 | char nbuf[64]; | |
cdddbdbc | 1418 | __u32 sum; |
14e8215b | 1419 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1420 | struct dl *dl; |
27fd6274 | 1421 | |
cdddbdbc DW |
1422 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1423 | printf(" Magic : %s\n", str); | |
1424 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1425 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1426 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1427 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1428 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1429 | printf(" Attributes : "); |
1430 | if (imsm_check_attributes(mpb->attributes)) | |
1431 | printf("All supported\n"); | |
1432 | else | |
1433 | printf("not supported\n"); | |
a5d85af7 | 1434 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1435 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1436 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1437 | sum = __le32_to_cpu(mpb->check_sum); |
1438 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1439 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1440 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1441 | printf(" Disks : %d\n", mpb->num_disks); |
1442 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1443 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1444 | if (super->bbm_log) { |
1445 | struct bbm_log *log = super->bbm_log; | |
1446 | ||
1447 | printf("\n"); | |
1448 | printf("Bad Block Management Log:\n"); | |
1449 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1450 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1451 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1452 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1453 | printf(" First Spare : %llx\n", |
1454 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1455 | } |
44470971 DW |
1456 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1457 | struct mdinfo info; | |
1458 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1459 | ||
1460 | super->current_vol = i; | |
a5d85af7 | 1461 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1462 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1463 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1464 | } |
cdddbdbc DW |
1465 | for (i = 0; i < mpb->num_disks; i++) { |
1466 | if (i == super->disks->index) | |
1467 | continue; | |
0ec1f4e8 | 1468 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1469 | } |
94827db3 | 1470 | |
0ec1f4e8 DW |
1471 | for (dl = super->disks; dl; dl = dl->next) |
1472 | if (dl->index == -1) | |
1473 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1474 | |
1475 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1476 | } |
1477 | ||
061f2c6a | 1478 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1479 | { |
27fd6274 | 1480 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1481 | struct mdinfo info; |
1482 | char nbuf[64]; | |
1e7bc0ed | 1483 | struct intel_super *super = st->sb; |
1e7bc0ed | 1484 | |
0d5a423f DW |
1485 | if (!super->anchor->num_raid_devs) { |
1486 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1487 | return; |
0d5a423f | 1488 | } |
ff54de6e | 1489 | |
a5d85af7 | 1490 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1491 | fname_from_uuid(st, &info, nbuf, ':'); |
1492 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1493 | } | |
1494 | ||
1495 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1496 | { | |
1497 | /* We just write a generic IMSM ARRAY entry */ | |
1498 | struct mdinfo info; | |
1499 | char nbuf[64]; | |
1500 | char nbuf1[64]; | |
1501 | struct intel_super *super = st->sb; | |
1502 | int i; | |
1503 | ||
1504 | if (!super->anchor->num_raid_devs) | |
1505 | return; | |
1506 | ||
a5d85af7 | 1507 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1508 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1509 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1510 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1511 | ||
1512 | super->current_vol = i; | |
a5d85af7 | 1513 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1514 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1515 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1516 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1517 | } |
cdddbdbc DW |
1518 | } |
1519 | ||
9d84c8ea DW |
1520 | static void export_examine_super_imsm(struct supertype *st) |
1521 | { | |
1522 | struct intel_super *super = st->sb; | |
1523 | struct imsm_super *mpb = super->anchor; | |
1524 | struct mdinfo info; | |
1525 | char nbuf[64]; | |
1526 | ||
a5d85af7 | 1527 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1528 | fname_from_uuid(st, &info, nbuf, ':'); |
1529 | printf("MD_METADATA=imsm\n"); | |
1530 | printf("MD_LEVEL=container\n"); | |
1531 | printf("MD_UUID=%s\n", nbuf+5); | |
1532 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1533 | } | |
1534 | ||
cdddbdbc DW |
1535 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1536 | { | |
3ebe00a1 DW |
1537 | struct mdinfo info; |
1538 | char nbuf[64]; | |
1539 | ||
a5d85af7 | 1540 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1541 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1542 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1543 | } |
1544 | ||
1545 | static void brief_detail_super_imsm(struct supertype *st) | |
1546 | { | |
ff54de6e N |
1547 | struct mdinfo info; |
1548 | char nbuf[64]; | |
a5d85af7 | 1549 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1550 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1551 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1552 | } |
d665cc31 DW |
1553 | |
1554 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1555 | static void fd2devname(int fd, char *name); | |
1556 | ||
120dc887 | 1557 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1558 | { |
120dc887 LM |
1559 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1560 | * controller, as well as non-connected ports | |
d665cc31 DW |
1561 | */ |
1562 | int hba_len = strlen(hba_path) + 1; | |
1563 | struct dirent *ent; | |
1564 | DIR *dir; | |
1565 | char *path = NULL; | |
1566 | int err = 0; | |
1567 | unsigned long port_mask = (1 << port_count) - 1; | |
1568 | ||
f21e18ca | 1569 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 1570 | if (verbose > 0) |
e7b84f9d | 1571 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
1572 | return 2; |
1573 | } | |
1574 | ||
1575 | /* scroll through /sys/dev/block looking for devices attached to | |
1576 | * this hba | |
1577 | */ | |
1578 | dir = opendir("/sys/dev/block"); | |
1579 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1580 | int fd; | |
1581 | char model[64]; | |
1582 | char vendor[64]; | |
1583 | char buf[1024]; | |
1584 | int major, minor; | |
1585 | char *device; | |
1586 | char *c; | |
1587 | int port; | |
1588 | int type; | |
1589 | ||
1590 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1591 | continue; | |
1592 | path = devt_to_devpath(makedev(major, minor)); | |
1593 | if (!path) | |
1594 | continue; | |
1595 | if (!path_attached_to_hba(path, hba_path)) { | |
1596 | free(path); | |
1597 | path = NULL; | |
1598 | continue; | |
1599 | } | |
1600 | ||
1601 | /* retrieve the scsi device type */ | |
1602 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 1603 | if (verbose > 0) |
e7b84f9d | 1604 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
1605 | err = 2; |
1606 | break; | |
1607 | } | |
1608 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1609 | if (load_sys(device, buf) != 0) { | |
ba728be7 | 1610 | if (verbose > 0) |
e7b84f9d | 1611 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
1612 | path); |
1613 | err = 2; | |
1614 | free(device); | |
1615 | break; | |
1616 | } | |
1617 | type = strtoul(buf, NULL, 10); | |
1618 | ||
1619 | /* if it's not a disk print the vendor and model */ | |
1620 | if (!(type == 0 || type == 7 || type == 14)) { | |
1621 | vendor[0] = '\0'; | |
1622 | model[0] = '\0'; | |
1623 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1624 | if (load_sys(device, buf) == 0) { | |
1625 | strncpy(vendor, buf, sizeof(vendor)); | |
1626 | vendor[sizeof(vendor) - 1] = '\0'; | |
1627 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1628 | while (isspace(*c) || *c == '\0') | |
1629 | *c-- = '\0'; | |
1630 | ||
1631 | } | |
1632 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1633 | if (load_sys(device, buf) == 0) { | |
1634 | strncpy(model, buf, sizeof(model)); | |
1635 | model[sizeof(model) - 1] = '\0'; | |
1636 | c = (char *) &model[sizeof(model) - 1]; | |
1637 | while (isspace(*c) || *c == '\0') | |
1638 | *c-- = '\0'; | |
1639 | } | |
1640 | ||
1641 | if (vendor[0] && model[0]) | |
1642 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1643 | else | |
1644 | switch (type) { /* numbers from hald/linux/device.c */ | |
1645 | case 1: sprintf(buf, "tape"); break; | |
1646 | case 2: sprintf(buf, "printer"); break; | |
1647 | case 3: sprintf(buf, "processor"); break; | |
1648 | case 4: | |
1649 | case 5: sprintf(buf, "cdrom"); break; | |
1650 | case 6: sprintf(buf, "scanner"); break; | |
1651 | case 8: sprintf(buf, "media_changer"); break; | |
1652 | case 9: sprintf(buf, "comm"); break; | |
1653 | case 12: sprintf(buf, "raid"); break; | |
1654 | default: sprintf(buf, "unknown"); | |
1655 | } | |
1656 | } else | |
1657 | buf[0] = '\0'; | |
1658 | free(device); | |
1659 | ||
1660 | /* chop device path to 'host%d' and calculate the port number */ | |
1661 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 1662 | if (!c) { |
ba728be7 | 1663 | if (verbose > 0) |
e7b84f9d | 1664 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
1665 | err = 2; |
1666 | break; | |
1667 | } | |
d665cc31 DW |
1668 | *c = '\0'; |
1669 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1670 | port -= host_base; | |
1671 | else { | |
ba728be7 | 1672 | if (verbose > 0) { |
d665cc31 | 1673 | *c = '/'; /* repair the full string */ |
e7b84f9d | 1674 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
1675 | path); |
1676 | } | |
1677 | err = 2; | |
1678 | break; | |
1679 | } | |
1680 | ||
1681 | /* mark this port as used */ | |
1682 | port_mask &= ~(1 << port); | |
1683 | ||
1684 | /* print out the device information */ | |
1685 | if (buf[0]) { | |
1686 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1687 | continue; | |
1688 | } | |
1689 | ||
1690 | fd = dev_open(ent->d_name, O_RDONLY); | |
1691 | if (fd < 0) | |
1692 | printf(" Port%d : - disk info unavailable -\n", port); | |
1693 | else { | |
1694 | fd2devname(fd, buf); | |
1695 | printf(" Port%d : %s", port, buf); | |
1696 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1697 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1698 | else |
664d5325 | 1699 | printf(" ()\n"); |
4dab422a | 1700 | close(fd); |
d665cc31 | 1701 | } |
d665cc31 DW |
1702 | free(path); |
1703 | path = NULL; | |
1704 | } | |
1705 | if (path) | |
1706 | free(path); | |
1707 | if (dir) | |
1708 | closedir(dir); | |
1709 | if (err == 0) { | |
1710 | int i; | |
1711 | ||
1712 | for (i = 0; i < port_count; i++) | |
1713 | if (port_mask & (1 << i)) | |
1714 | printf(" Port%d : - no device attached -\n", i); | |
1715 | } | |
1716 | ||
1717 | return err; | |
1718 | } | |
1719 | ||
120dc887 LM |
1720 | static void print_found_intel_controllers(struct sys_dev *elem) |
1721 | { | |
1722 | for (; elem; elem = elem->next) { | |
e7b84f9d | 1723 | pr_err("found Intel(R) "); |
120dc887 LM |
1724 | if (elem->type == SYS_DEV_SATA) |
1725 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1726 | else if (elem->type == SYS_DEV_SAS) |
1727 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1728 | fprintf(stderr, "RAID controller"); |
1729 | if (elem->pci_id) | |
1730 | fprintf(stderr, " at %s", elem->pci_id); | |
1731 | fprintf(stderr, ".\n"); | |
1732 | } | |
1733 | fflush(stderr); | |
1734 | } | |
1735 | ||
120dc887 LM |
1736 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1737 | { | |
1738 | struct dirent *ent; | |
1739 | DIR *dir; | |
1740 | int host_base = -1; | |
1741 | ||
1742 | *port_count = 0; | |
1743 | if ((dir = opendir(hba_path)) == NULL) | |
1744 | return -1; | |
1745 | ||
1746 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1747 | int host; | |
1748 | ||
1749 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1750 | continue; | |
1751 | if (*port_count == 0) | |
1752 | host_base = host; | |
1753 | else if (host < host_base) | |
1754 | host_base = host; | |
1755 | ||
1756 | if (host + 1 > *port_count + host_base) | |
1757 | *port_count = host + 1 - host_base; | |
1758 | } | |
1759 | closedir(dir); | |
1760 | return host_base; | |
1761 | } | |
1762 | ||
a891a3c2 LM |
1763 | static void print_imsm_capability(const struct imsm_orom *orom) |
1764 | { | |
1765 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1766 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1767 | orom->hotfix_ver, orom->build); | |
1768 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1769 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1770 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1771 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1772 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1773 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1774 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1775 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1776 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1777 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1778 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1779 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1780 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1781 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1782 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1783 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1784 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1785 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1786 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1787 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1788 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1789 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1790 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
1791 | printf(" 2TB volumes :%s supported\n", |
1792 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
1793 | printf(" 2TB disks :%s supported\n", | |
1794 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 LM |
1795 | printf(" Max Disks : %d\n", orom->tds); |
1796 | printf(" Max Volumes : %d per array, %d per controller\n", | |
1797 | orom->vpa, orom->vphba); | |
a891a3c2 LM |
1798 | return; |
1799 | } | |
1800 | ||
e50cf220 MN |
1801 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
1802 | { | |
1803 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
1804 | printf("IMSM_VERSION=%d.%d.%d.%d\n",orom->major_ver, orom->minor_ver, | |
1805 | orom->hotfix_ver, orom->build); | |
1806 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", | |
1807 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
1808 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
1809 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
1810 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
1811 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
1812 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1813 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
1814 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
1815 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
1816 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
1817 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
1818 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
1819 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
1820 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
1821 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
1822 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
1823 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
1824 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
1825 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
1826 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
1827 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
1828 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
1829 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
1830 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
1831 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
1832 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
1833 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
1834 | } | |
1835 | ||
9eafa1de | 1836 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
1837 | { |
1838 | /* There are two components to imsm platform support, the ahci SATA | |
1839 | * controller and the option-rom. To find the SATA controller we | |
1840 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1841 | * controller with the Intel vendor id is present. This approach | |
1842 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1843 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1844 | * detect platform raid capabilities. The option-rom resides in a | |
1845 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1846 | * platform capabilities. If raid support is disabled in the BIOS the | |
1847 | * option-rom capability structure will not be available. | |
1848 | */ | |
1849 | const struct imsm_orom *orom; | |
1850 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1851 | int host_base = 0; |
1852 | int port_count = 0; | |
9eafa1de | 1853 | int result=1; |
d665cc31 | 1854 | |
5615172f | 1855 | if (enumerate_only) { |
a891a3c2 | 1856 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1857 | return 0; |
a891a3c2 LM |
1858 | list = find_intel_devices(); |
1859 | if (!list) | |
1860 | return 2; | |
1861 | for (hba = list; hba; hba = hba->next) { | |
1862 | orom = find_imsm_capability(hba->type); | |
1863 | if (!orom) { | |
1864 | result = 2; | |
1865 | break; | |
1866 | } | |
9eafa1de MN |
1867 | else |
1868 | result = 0; | |
a891a3c2 LM |
1869 | } |
1870 | free_sys_dev(&list); | |
1871 | return result; | |
5615172f DW |
1872 | } |
1873 | ||
155cbb4c LM |
1874 | list = find_intel_devices(); |
1875 | if (!list) { | |
ba728be7 | 1876 | if (verbose > 0) |
e7b84f9d | 1877 | pr_err("no active Intel(R) RAID " |
155cbb4c | 1878 | "controller found.\n"); |
d665cc31 DW |
1879 | free_sys_dev(&list); |
1880 | return 2; | |
ba728be7 | 1881 | } else if (verbose > 0) |
155cbb4c | 1882 | print_found_intel_controllers(list); |
d665cc31 | 1883 | |
a891a3c2 | 1884 | for (hba = list; hba; hba = hba->next) { |
9eafa1de MN |
1885 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
1886 | continue; | |
a891a3c2 LM |
1887 | orom = find_imsm_capability(hba->type); |
1888 | if (!orom) | |
e7b84f9d | 1889 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
a891a3c2 | 1890 | hba->path, get_sys_dev_type(hba->type)); |
9eafa1de MN |
1891 | else { |
1892 | result = 0; | |
a891a3c2 | 1893 | print_imsm_capability(orom); |
9eafa1de MN |
1894 | printf(" I/O Controller : %s (%s)\n", |
1895 | hba->path, get_sys_dev_type(hba->type)); | |
1896 | if (hba->type == SYS_DEV_SATA) { | |
1897 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1898 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1899 | if (verbose > 0) | |
1900 | pr_err("failed to enumerate " | |
1901 | "ports on SATA controller at %s.\n", hba->pci_id); | |
1902 | result |= 2; | |
1903 | } | |
120dc887 LM |
1904 | } |
1905 | } | |
d665cc31 | 1906 | } |
155cbb4c | 1907 | |
9eafa1de MN |
1908 | if (controller_path && result == 1) |
1909 | pr_err("no active Intel(R) RAID " | |
1910 | "controller found under %s\n",controller_path); | |
1911 | ||
120dc887 LM |
1912 | free_sys_dev(&list); |
1913 | return result; | |
d665cc31 | 1914 | } |
e50cf220 | 1915 | |
9eafa1de | 1916 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 MN |
1917 | { |
1918 | const struct imsm_orom *orom; | |
1919 | struct sys_dev *list, *hba; | |
1920 | int result=1; | |
1921 | ||
1922 | list = find_intel_devices(); | |
1923 | if (!list) { | |
1924 | if (verbose > 0) | |
1925 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
1926 | result = 2; | |
1927 | free_sys_dev(&list); | |
1928 | return result; | |
1929 | } | |
1930 | ||
1931 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
1932 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
1933 | continue; | |
e50cf220 MN |
1934 | orom = find_imsm_capability(hba->type); |
1935 | if (!orom) { | |
1936 | if (verbose > 0) | |
1937 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",hba->path); | |
1938 | } | |
1939 | else { | |
1940 | print_imsm_capability_export(orom); | |
1941 | result = 0; | |
1942 | } | |
1943 | } | |
1944 | ||
e50cf220 MN |
1945 | return result; |
1946 | } | |
1947 | ||
cdddbdbc DW |
1948 | #endif |
1949 | ||
1950 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1951 | { | |
5115ca67 DW |
1952 | /* the imsm metadata format does not specify any host |
1953 | * identification information. We return -1 since we can never | |
1954 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1955 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1956 | * exclude member disks that do not belong, and we rely on |
1957 | * mdadm.conf to specify the arrays that should be assembled. | |
1958 | * Auto-assembly may still pick up "foreign" arrays. | |
1959 | */ | |
cdddbdbc | 1960 | |
9362c1c8 | 1961 | return -1; |
cdddbdbc DW |
1962 | } |
1963 | ||
1964 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1965 | { | |
51006d85 N |
1966 | /* The uuid returned here is used for: |
1967 | * uuid to put into bitmap file (Create, Grow) | |
1968 | * uuid for backup header when saving critical section (Grow) | |
1969 | * comparing uuids when re-adding a device into an array | |
1970 | * In these cases the uuid required is that of the data-array, | |
1971 | * not the device-set. | |
1972 | * uuid to recognise same set when adding a missing device back | |
1973 | * to an array. This is a uuid for the device-set. | |
1974 | * | |
1975 | * For each of these we can make do with a truncated | |
1976 | * or hashed uuid rather than the original, as long as | |
1977 | * everyone agrees. | |
1978 | * In each case the uuid required is that of the data-array, | |
1979 | * not the device-set. | |
43dad3d6 | 1980 | */ |
51006d85 N |
1981 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1982 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1983 | * - the orig_family_num of the container |
51006d85 N |
1984 | * - the index number of the volume |
1985 | * - the 'serial' number of the volume. | |
1986 | * Hopefully these are all constant. | |
1987 | */ | |
1988 | struct intel_super *super = st->sb; | |
43dad3d6 | 1989 | |
51006d85 N |
1990 | char buf[20]; |
1991 | struct sha1_ctx ctx; | |
1992 | struct imsm_dev *dev = NULL; | |
148acb7b | 1993 | __u32 family_num; |
51006d85 | 1994 | |
148acb7b DW |
1995 | /* some mdadm versions failed to set ->orig_family_num, in which |
1996 | * case fall back to ->family_num. orig_family_num will be | |
1997 | * fixed up with the first metadata update. | |
1998 | */ | |
1999 | family_num = super->anchor->orig_family_num; | |
2000 | if (family_num == 0) | |
2001 | family_num = super->anchor->family_num; | |
51006d85 | 2002 | sha1_init_ctx(&ctx); |
92bd8f8d | 2003 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2004 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2005 | if (super->current_vol >= 0) |
2006 | dev = get_imsm_dev(super, super->current_vol); | |
2007 | if (dev) { | |
2008 | __u32 vol = super->current_vol; | |
2009 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2010 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2011 | } | |
2012 | sha1_finish_ctx(&ctx, buf); | |
2013 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2014 | } |
2015 | ||
0d481d37 | 2016 | #if 0 |
4f5bc454 DW |
2017 | static void |
2018 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2019 | { |
cdddbdbc DW |
2020 | __u8 *v = get_imsm_version(mpb); |
2021 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2022 | char major[] = { 0, 0, 0 }; | |
2023 | char minor[] = { 0 ,0, 0 }; | |
2024 | char patch[] = { 0, 0, 0 }; | |
2025 | char *ver_parse[] = { major, minor, patch }; | |
2026 | int i, j; | |
2027 | ||
2028 | i = j = 0; | |
2029 | while (*v != '\0' && v < end) { | |
2030 | if (*v != '.' && j < 2) | |
2031 | ver_parse[i][j++] = *v; | |
2032 | else { | |
2033 | i++; | |
2034 | j = 0; | |
2035 | } | |
2036 | v++; | |
2037 | } | |
2038 | ||
4f5bc454 DW |
2039 | *m = strtol(minor, NULL, 0); |
2040 | *p = strtol(patch, NULL, 0); | |
2041 | } | |
0d481d37 | 2042 | #endif |
4f5bc454 | 2043 | |
1e5c6983 DW |
2044 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2045 | { | |
2046 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2047 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2048 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2049 | ||
2050 | switch (get_imsm_raid_level(map)) { | |
2051 | case 5: | |
2052 | case 10: | |
2053 | return chunk; | |
2054 | default: | |
2055 | return 128*1024 >> 9; | |
2056 | } | |
2057 | } | |
2058 | ||
2059 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2060 | { | |
2061 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2062 | * this is different than migr_strip_size_resync(), but it's good | |
2063 | * to be compatible | |
2064 | */ | |
238c0a71 | 2065 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2066 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2067 | ||
2068 | switch (get_imsm_raid_level(map)) { | |
2069 | case 1: | |
2070 | case 10: | |
2071 | if (map->num_members % map->num_domains == 0) | |
2072 | return 128*1024 >> 9; | |
2073 | else | |
2074 | return chunk; | |
2075 | case 5: | |
2076 | return max((__u32) 64*1024 >> 9, chunk); | |
2077 | default: | |
2078 | return 128*1024 >> 9; | |
2079 | } | |
2080 | } | |
2081 | ||
2082 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2083 | { | |
238c0a71 AK |
2084 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2085 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2086 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2087 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2088 | ||
2089 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2090 | } | |
2091 | ||
2092 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2093 | { | |
238c0a71 | 2094 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2095 | int level = get_imsm_raid_level(lo); |
2096 | ||
2097 | if (level == 1 || level == 10) { | |
238c0a71 | 2098 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2099 | |
2100 | return hi->num_domains; | |
2101 | } else | |
2102 | return num_stripes_per_unit_resync(dev); | |
2103 | } | |
2104 | ||
98130f40 | 2105 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2106 | { |
2107 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2108 | * counter-intuitively have the same number of data disks | |
2109 | */ | |
98130f40 | 2110 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2111 | |
2112 | switch (get_imsm_raid_level(map)) { | |
2113 | case 0: | |
36fd8ccc AK |
2114 | return map->num_members; |
2115 | break; | |
1e5c6983 DW |
2116 | case 1: |
2117 | case 10: | |
36fd8ccc | 2118 | return map->num_members/2; |
1e5c6983 DW |
2119 | case 5: |
2120 | return map->num_members - 1; | |
2121 | default: | |
2122 | dprintf("%s: unsupported raid level\n", __func__); | |
2123 | return 0; | |
2124 | } | |
2125 | } | |
2126 | ||
2127 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2128 | { | |
238c0a71 | 2129 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2130 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2131 | ||
2132 | switch(get_imsm_raid_level(map)) { | |
2133 | case 1: | |
2134 | case 10: | |
2135 | return chunk * map->num_domains; | |
2136 | case 5: | |
2137 | return chunk * map->num_members; | |
2138 | default: | |
2139 | return chunk; | |
2140 | } | |
2141 | } | |
2142 | ||
2143 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2144 | { | |
238c0a71 | 2145 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2146 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2147 | __u32 strip = block / chunk; | |
2148 | ||
2149 | switch (get_imsm_raid_level(map)) { | |
2150 | case 1: | |
2151 | case 10: { | |
2152 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2153 | __u32 vol_stripe = vol_strip / map->num_members; | |
2154 | ||
2155 | return vol_stripe * chunk + block % chunk; | |
2156 | } case 5: { | |
2157 | __u32 stripe = strip / (map->num_members - 1); | |
2158 | ||
2159 | return stripe * chunk + block % chunk; | |
2160 | } | |
2161 | default: | |
2162 | return 0; | |
2163 | } | |
2164 | } | |
2165 | ||
c47b0ff6 AK |
2166 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2167 | struct imsm_dev *dev) | |
1e5c6983 DW |
2168 | { |
2169 | /* calculate the conversion factor between per member 'blocks' | |
2170 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2171 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2172 | */ | |
2173 | if (!dev->vol.migr_state) | |
2174 | return 0; | |
2175 | ||
2176 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2177 | case MIGR_GEN_MIGR: { |
2178 | struct migr_record *migr_rec = super->migr_rec; | |
2179 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2180 | } | |
1e5c6983 DW |
2181 | case MIGR_VERIFY: |
2182 | case MIGR_REPAIR: | |
2183 | case MIGR_INIT: { | |
238c0a71 | 2184 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2185 | __u32 stripes_per_unit; |
2186 | __u32 blocks_per_unit; | |
2187 | __u32 parity_depth; | |
2188 | __u32 migr_chunk; | |
2189 | __u32 block_map; | |
2190 | __u32 block_rel; | |
2191 | __u32 segment; | |
2192 | __u32 stripe; | |
2193 | __u8 disks; | |
2194 | ||
2195 | /* yes, this is really the translation of migr_units to | |
2196 | * per-member blocks in the 'resync' case | |
2197 | */ | |
2198 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2199 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2200 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2201 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2202 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2203 | segment = blocks_per_unit / stripe; |
2204 | block_rel = blocks_per_unit - segment * stripe; | |
2205 | parity_depth = parity_segment_depth(dev); | |
2206 | block_map = map_migr_block(dev, block_rel); | |
2207 | return block_map + parity_depth * segment; | |
2208 | } | |
2209 | case MIGR_REBUILD: { | |
2210 | __u32 stripes_per_unit; | |
2211 | __u32 migr_chunk; | |
2212 | ||
2213 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2214 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2215 | return migr_chunk * stripes_per_unit; | |
2216 | } | |
1e5c6983 DW |
2217 | case MIGR_STATE_CHANGE: |
2218 | default: | |
2219 | return 0; | |
2220 | } | |
2221 | } | |
2222 | ||
c2c087e6 DW |
2223 | static int imsm_level_to_layout(int level) |
2224 | { | |
2225 | switch (level) { | |
2226 | case 0: | |
2227 | case 1: | |
2228 | return 0; | |
2229 | case 5: | |
2230 | case 6: | |
a380c027 | 2231 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2232 | case 10: |
c92a2527 | 2233 | return 0x102; |
c2c087e6 | 2234 | } |
a18a888e | 2235 | return UnSet; |
c2c087e6 DW |
2236 | } |
2237 | ||
8e59f3d8 AK |
2238 | /******************************************************************************* |
2239 | * Function: read_imsm_migr_rec | |
2240 | * Description: Function reads imsm migration record from last sector of disk | |
2241 | * Parameters: | |
2242 | * fd : disk descriptor | |
2243 | * super : metadata info | |
2244 | * Returns: | |
2245 | * 0 : success, | |
2246 | * -1 : fail | |
2247 | ******************************************************************************/ | |
2248 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2249 | { | |
2250 | int ret_val = -1; | |
2251 | unsigned long long dsize; | |
2252 | ||
2253 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2254 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2255 | pr_err("Cannot seek to anchor block: %s\n", |
2256 | strerror(errno)); | |
8e59f3d8 AK |
2257 | goto out; |
2258 | } | |
17a4eaf9 AK |
2259 | if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2260 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2261 | pr_err("Cannot read migr record block: %s\n", |
2262 | strerror(errno)); | |
8e59f3d8 AK |
2263 | goto out; |
2264 | } | |
2265 | ret_val = 0; | |
2266 | ||
2267 | out: | |
2268 | return ret_val; | |
2269 | } | |
2270 | ||
3136abe5 AK |
2271 | static struct imsm_dev *imsm_get_device_during_migration( |
2272 | struct intel_super *super) | |
2273 | { | |
2274 | ||
2275 | struct intel_dev *dv; | |
2276 | ||
2277 | for (dv = super->devlist; dv; dv = dv->next) { | |
2278 | if (is_gen_migration(dv->dev)) | |
2279 | return dv->dev; | |
2280 | } | |
2281 | return NULL; | |
2282 | } | |
2283 | ||
8e59f3d8 AK |
2284 | /******************************************************************************* |
2285 | * Function: load_imsm_migr_rec | |
2286 | * Description: Function reads imsm migration record (it is stored at the last | |
2287 | * sector of disk) | |
2288 | * Parameters: | |
2289 | * super : imsm internal array info | |
2290 | * info : general array info | |
2291 | * Returns: | |
2292 | * 0 : success | |
2293 | * -1 : fail | |
4c965cc9 | 2294 | * -2 : no migration in progress |
8e59f3d8 AK |
2295 | ******************************************************************************/ |
2296 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2297 | { | |
2298 | struct mdinfo *sd; | |
2299 | struct dl *dl = NULL; | |
2300 | char nm[30]; | |
2301 | int retval = -1; | |
2302 | int fd = -1; | |
3136abe5 AK |
2303 | struct imsm_dev *dev; |
2304 | struct imsm_map *map = NULL; | |
b4ab44d8 | 2305 | int slot = -1; |
3136abe5 AK |
2306 | |
2307 | /* find map under migration */ | |
2308 | dev = imsm_get_device_during_migration(super); | |
2309 | /* nothing to load,no migration in progress? | |
2310 | */ | |
2311 | if (dev == NULL) | |
4c965cc9 | 2312 | return -2; |
44bfe6df | 2313 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 AK |
2314 | |
2315 | if (info) { | |
2316 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3136abe5 AK |
2317 | /* skip spare and failed disks |
2318 | */ | |
2319 | if (sd->disk.raid_disk < 0) | |
2320 | continue; | |
8e59f3d8 | 2321 | /* read only from one of the first two slots */ |
3136abe5 AK |
2322 | if (map) |
2323 | slot = get_imsm_disk_slot(map, | |
2324 | sd->disk.raid_disk); | |
2325 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 | 2326 | continue; |
3136abe5 | 2327 | |
8e59f3d8 AK |
2328 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2329 | fd = dev_open(nm, O_RDONLY); | |
2330 | if (fd >= 0) | |
2331 | break; | |
2332 | } | |
2333 | } | |
2334 | if (fd < 0) { | |
2335 | for (dl = super->disks; dl; dl = dl->next) { | |
3136abe5 AK |
2336 | /* skip spare and failed disks |
2337 | */ | |
2338 | if (dl->index < 0) | |
2339 | continue; | |
8e59f3d8 | 2340 | /* read only from one of the first two slots */ |
3136abe5 AK |
2341 | if (map) |
2342 | slot = get_imsm_disk_slot(map, dl->index); | |
2343 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 AK |
2344 | continue; |
2345 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2346 | fd = dev_open(nm, O_RDONLY); | |
2347 | if (fd >= 0) | |
2348 | break; | |
2349 | } | |
2350 | } | |
2351 | if (fd < 0) | |
2352 | goto out; | |
2353 | retval = read_imsm_migr_rec(fd, super); | |
2354 | ||
2355 | out: | |
2356 | if (fd >= 0) | |
2357 | close(fd); | |
2358 | return retval; | |
2359 | } | |
2360 | ||
9e2d750d | 2361 | #ifndef MDASSEMBLE |
c17608ea AK |
2362 | /******************************************************************************* |
2363 | * function: imsm_create_metadata_checkpoint_update | |
2364 | * Description: It creates update for checkpoint change. | |
2365 | * Parameters: | |
2366 | * super : imsm internal array info | |
2367 | * u : pointer to prepared update | |
2368 | * Returns: | |
2369 | * Uptate length. | |
2370 | * If length is equal to 0, input pointer u contains no update | |
2371 | ******************************************************************************/ | |
2372 | static int imsm_create_metadata_checkpoint_update( | |
2373 | struct intel_super *super, | |
2374 | struct imsm_update_general_migration_checkpoint **u) | |
2375 | { | |
2376 | ||
2377 | int update_memory_size = 0; | |
2378 | ||
2379 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2380 | ||
2381 | if (u == NULL) | |
2382 | return 0; | |
2383 | *u = NULL; | |
2384 | ||
2385 | /* size of all update data without anchor */ | |
2386 | update_memory_size = | |
2387 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2388 | ||
503975b9 | 2389 | *u = xcalloc(1, update_memory_size); |
c17608ea AK |
2390 | if (*u == NULL) { |
2391 | dprintf("error: cannot get memory for " | |
2392 | "imsm_create_metadata_checkpoint_update update\n"); | |
2393 | return 0; | |
2394 | } | |
2395 | (*u)->type = update_general_migration_checkpoint; | |
2396 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2397 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2398 | (*u)->curr_migr_unit); | |
2399 | ||
2400 | return update_memory_size; | |
2401 | } | |
2402 | ||
2403 | ||
2404 | static void imsm_update_metadata_locally(struct supertype *st, | |
2405 | void *buf, int len); | |
2406 | ||
687629c2 AK |
2407 | /******************************************************************************* |
2408 | * Function: write_imsm_migr_rec | |
2409 | * Description: Function writes imsm migration record | |
2410 | * (at the last sector of disk) | |
2411 | * Parameters: | |
2412 | * super : imsm internal array info | |
2413 | * Returns: | |
2414 | * 0 : success | |
2415 | * -1 : if fail | |
2416 | ******************************************************************************/ | |
2417 | static int write_imsm_migr_rec(struct supertype *st) | |
2418 | { | |
2419 | struct intel_super *super = st->sb; | |
2420 | unsigned long long dsize; | |
2421 | char nm[30]; | |
2422 | int fd = -1; | |
2423 | int retval = -1; | |
2424 | struct dl *sd; | |
c17608ea AK |
2425 | int len; |
2426 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 AK |
2427 | struct imsm_dev *dev; |
2428 | struct imsm_map *map = NULL; | |
2429 | ||
2430 | /* find map under migration */ | |
2431 | dev = imsm_get_device_during_migration(super); | |
2432 | /* if no migration, write buffer anyway to clear migr_record | |
2433 | * on disk based on first available device | |
2434 | */ | |
2435 | if (dev == NULL) | |
2436 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2437 | super->current_vol); | |
2438 | ||
44bfe6df | 2439 | map = get_imsm_map(dev, MAP_0); |
687629c2 AK |
2440 | |
2441 | for (sd = super->disks ; sd ; sd = sd->next) { | |
b4ab44d8 | 2442 | int slot = -1; |
3136abe5 AK |
2443 | |
2444 | /* skip failed and spare devices */ | |
2445 | if (sd->index < 0) | |
2446 | continue; | |
687629c2 | 2447 | /* write to 2 first slots only */ |
3136abe5 AK |
2448 | if (map) |
2449 | slot = get_imsm_disk_slot(map, sd->index); | |
2450 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
687629c2 | 2451 | continue; |
3136abe5 | 2452 | |
687629c2 AK |
2453 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
2454 | fd = dev_open(nm, O_RDWR); | |
2455 | if (fd < 0) | |
2456 | continue; | |
2457 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2458 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2459 | pr_err("Cannot seek to anchor block: %s\n", |
2460 | strerror(errno)); | |
687629c2 AK |
2461 | goto out; |
2462 | } | |
17a4eaf9 AK |
2463 | if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2464 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2465 | pr_err("Cannot write migr record block: %s\n", |
2466 | strerror(errno)); | |
687629c2 AK |
2467 | goto out; |
2468 | } | |
2469 | close(fd); | |
2470 | fd = -1; | |
2471 | } | |
c17608ea AK |
2472 | /* update checkpoint information in metadata */ |
2473 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2474 | ||
2475 | if (len <= 0) { | |
2476 | dprintf("imsm: Cannot prepare update\n"); | |
2477 | goto out; | |
2478 | } | |
2479 | /* update metadata locally */ | |
2480 | imsm_update_metadata_locally(st, u, len); | |
2481 | /* and possibly remotely */ | |
2482 | if (st->update_tail) { | |
2483 | append_metadata_update(st, u, len); | |
2484 | /* during reshape we do all work inside metadata handler | |
2485 | * manage_reshape(), so metadata update has to be triggered | |
2486 | * insida it | |
2487 | */ | |
2488 | flush_metadata_updates(st); | |
2489 | st->update_tail = &st->updates; | |
2490 | } else | |
2491 | free(u); | |
687629c2 AK |
2492 | |
2493 | retval = 0; | |
2494 | out: | |
2495 | if (fd >= 0) | |
2496 | close(fd); | |
2497 | return retval; | |
2498 | } | |
9e2d750d | 2499 | #endif /* MDASSEMBLE */ |
687629c2 | 2500 | |
e2962bfc AK |
2501 | /* spare/missing disks activations are not allowe when |
2502 | * array/container performs reshape operation, because | |
2503 | * all arrays in container works on the same disks set | |
2504 | */ | |
2505 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2506 | { | |
2507 | int rv = 0; | |
2508 | struct intel_dev *i_dev; | |
2509 | struct imsm_dev *dev; | |
2510 | ||
2511 | /* check whole container | |
2512 | */ | |
2513 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2514 | dev = i_dev->dev; | |
3ad25638 | 2515 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2516 | /* No repair during any migration in container |
2517 | */ | |
2518 | rv = 1; | |
2519 | break; | |
2520 | } | |
2521 | } | |
2522 | return rv; | |
2523 | } | |
c41e00b2 AK |
2524 | static unsigned long long imsm_component_size_aligment_check(int level, |
2525 | int chunk_size, | |
2526 | unsigned long long component_size) | |
2527 | { | |
2528 | unsigned int component_size_alligment; | |
2529 | ||
2530 | /* check component size aligment | |
2531 | */ | |
2532 | component_size_alligment = component_size % (chunk_size/512); | |
2533 | ||
2534 | dprintf("imsm_component_size_aligment_check(Level: %i, " | |
2535 | "chunk_size = %i, component_size = %llu), " | |
2536 | "component_size_alligment = %u\n", | |
2537 | level, chunk_size, component_size, | |
2538 | component_size_alligment); | |
2539 | ||
2540 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
2541 | dprintf("imsm: reported component size alligned from %llu ", | |
2542 | component_size); | |
2543 | component_size -= component_size_alligment; | |
2544 | dprintf("to %llu (%i).\n", | |
2545 | component_size, component_size_alligment); | |
2546 | } | |
2547 | ||
2548 | return component_size; | |
2549 | } | |
e2962bfc | 2550 | |
a5d85af7 | 2551 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2552 | { |
2553 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2554 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2555 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2556 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2557 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2558 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2559 | struct dl *dl; |
e207da2f | 2560 | char *devname; |
a5d85af7 | 2561 | int map_disks = info->array.raid_disks; |
bf5a934a | 2562 | |
95eeceeb | 2563 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2564 | if (prev_map) |
2565 | map_to_analyse = prev_map; | |
2566 | ||
ca0748fa | 2567 | dl = super->current_disk; |
9894ec0d | 2568 | |
bf5a934a | 2569 | info->container_member = super->current_vol; |
cd0430a1 | 2570 | info->array.raid_disks = map->num_members; |
b335e593 | 2571 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2572 | info->array.layout = imsm_level_to_layout(info->array.level); |
2573 | info->array.md_minor = -1; | |
2574 | info->array.ctime = 0; | |
2575 | info->array.utime = 0; | |
b335e593 AK |
2576 | info->array.chunk_size = |
2577 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2578 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2579 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2580 | info->custom_array_size <<= 32; | |
2581 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2582 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2583 | ||
3f510843 | 2584 | if (is_gen_migration(dev)) { |
3f83228a | 2585 | info->reshape_active = 1; |
b335e593 AK |
2586 | info->new_level = get_imsm_raid_level(map); |
2587 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2588 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2589 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2590 | if (info->delta_disks) { |
2591 | /* this needs to be applied to every array | |
2592 | * in the container. | |
2593 | */ | |
81219e70 | 2594 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2595 | } |
3f83228a N |
2596 | /* We shape information that we give to md might have to be |
2597 | * modify to cope with md's requirement for reshaping arrays. | |
2598 | * For example, when reshaping a RAID0, md requires it to be | |
2599 | * presented as a degraded RAID4. | |
2600 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2601 | * the array as already being RAID5, but the 'before' layout | |
2602 | * is a RAID4-like layout. | |
2603 | */ | |
2604 | switch (info->array.level) { | |
2605 | case 0: | |
2606 | switch(info->new_level) { | |
2607 | case 0: | |
2608 | /* conversion is happening as RAID4 */ | |
2609 | info->array.level = 4; | |
2610 | info->array.raid_disks += 1; | |
2611 | break; | |
2612 | case 5: | |
2613 | /* conversion is happening as RAID5 */ | |
2614 | info->array.level = 5; | |
2615 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2616 | info->delta_disks -= 1; |
2617 | break; | |
2618 | default: | |
2619 | /* FIXME error message */ | |
2620 | info->array.level = UnSet; | |
2621 | break; | |
2622 | } | |
2623 | break; | |
2624 | } | |
b335e593 AK |
2625 | } else { |
2626 | info->new_level = UnSet; | |
2627 | info->new_layout = UnSet; | |
2628 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2629 | info->delta_disks = 0; |
b335e593 | 2630 | } |
ca0748fa | 2631 | |
efb30e7f DW |
2632 | if (dl) { |
2633 | info->disk.major = dl->major; | |
2634 | info->disk.minor = dl->minor; | |
ca0748fa | 2635 | info->disk.number = dl->index; |
656b6b5a N |
2636 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2637 | dl->index); | |
efb30e7f | 2638 | } |
bf5a934a | 2639 | |
5551b113 CA |
2640 | info->data_offset = pba_of_lba0(map_to_analyse); |
2641 | info->component_size = blocks_per_member(map_to_analyse); | |
139dae11 | 2642 | |
c41e00b2 AK |
2643 | info->component_size = imsm_component_size_aligment_check( |
2644 | info->array.level, | |
2645 | info->array.chunk_size, | |
2646 | info->component_size); | |
139dae11 | 2647 | |
301406c9 | 2648 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2649 | info->recovery_start = MaxSector; |
bf5a934a | 2650 | |
d2e6d5d6 | 2651 | info->reshape_progress = 0; |
b6796ce1 | 2652 | info->resync_start = MaxSector; |
b9172665 AK |
2653 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2654 | dev->vol.dirty) && | |
2655 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2656 | info->resync_start = 0; |
b6796ce1 AK |
2657 | } |
2658 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2659 | switch (migr_type(dev)) { |
2660 | case MIGR_REPAIR: | |
2661 | case MIGR_INIT: { | |
c47b0ff6 AK |
2662 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2663 | dev); | |
1e5c6983 DW |
2664 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2665 | ||
2666 | info->resync_start = blocks_per_unit * units; | |
2667 | break; | |
2668 | } | |
d2e6d5d6 | 2669 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2670 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2671 | dev); | |
2672 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2673 | unsigned long long array_blocks; |
2674 | int used_disks; | |
d2e6d5d6 | 2675 | |
befb629b AK |
2676 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2677 | (units < | |
2678 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2679 | (super->migr_rec->rec_status == | |
2680 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2681 | units++; | |
2682 | ||
d2e6d5d6 | 2683 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2684 | |
d2e6d5d6 AK |
2685 | dprintf("IMSM: General Migration checkpoint : %llu " |
2686 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2687 | (unsigned long long)units, |
2688 | (unsigned long long)blocks_per_unit, | |
2689 | info->reshape_progress); | |
75156c46 | 2690 | |
238c0a71 | 2691 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 2692 | if (used_disks > 0) { |
5551b113 | 2693 | array_blocks = blocks_per_member(map) * |
75156c46 AK |
2694 | used_disks; |
2695 | /* round array size down to closest MB | |
2696 | */ | |
2697 | info->custom_array_size = (array_blocks | |
2698 | >> SECT_PER_MB_SHIFT) | |
2699 | << SECT_PER_MB_SHIFT; | |
2700 | } | |
d2e6d5d6 | 2701 | } |
1e5c6983 DW |
2702 | case MIGR_VERIFY: |
2703 | /* we could emulate the checkpointing of | |
2704 | * 'sync_action=check' migrations, but for now | |
2705 | * we just immediately complete them | |
2706 | */ | |
2707 | case MIGR_REBUILD: | |
2708 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2709 | case MIGR_STATE_CHANGE: |
2710 | /* FIXME handle other migrations */ | |
2711 | default: | |
2712 | /* we are not dirty, so... */ | |
2713 | info->resync_start = MaxSector; | |
2714 | } | |
b6796ce1 | 2715 | } |
301406c9 DW |
2716 | |
2717 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2718 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2719 | |
f35f2525 N |
2720 | info->array.major_version = -1; |
2721 | info->array.minor_version = -2; | |
e207da2f AW |
2722 | devname = devnum2devname(st->container_dev); |
2723 | *info->text_version = '\0'; | |
2724 | if (devname) | |
2725 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2726 | free(devname); | |
a67dd8cc | 2727 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2728 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2729 | |
2730 | if (dmap) { | |
2731 | int i, j; | |
2732 | for (i=0; i<map_disks; i++) { | |
2733 | dmap[i] = 0; | |
2734 | if (i < info->array.raid_disks) { | |
2735 | struct imsm_disk *dsk; | |
238c0a71 | 2736 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2737 | dsk = get_imsm_disk(super, j); |
2738 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2739 | dmap[i] = 1; | |
2740 | } | |
2741 | } | |
2742 | } | |
81ac8b4d | 2743 | } |
bf5a934a | 2744 | |
3b451610 AK |
2745 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2746 | int failed, int look_in_map); | |
2747 | ||
2748 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2749 | int look_in_map); | |
2750 | ||
b4ab44d8 AK |
2751 | |
2752 | #ifndef MDASSEMBLE | |
3b451610 AK |
2753 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
2754 | { | |
2755 | if (is_gen_migration(dev)) { | |
2756 | int failed; | |
2757 | __u8 map_state; | |
2758 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2759 | ||
2760 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2761 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2762 | if (map2->map_state != map_state) { |
2763 | map2->map_state = map_state; | |
2764 | super->updates_pending++; | |
2765 | } | |
2766 | } | |
2767 | } | |
b4ab44d8 | 2768 | #endif |
97b4d0e9 DW |
2769 | |
2770 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2771 | { | |
2772 | struct dl *d; | |
2773 | ||
2774 | for (d = super->missing; d; d = d->next) | |
2775 | if (d->index == index) | |
2776 | return &d->disk; | |
2777 | return NULL; | |
2778 | } | |
2779 | ||
a5d85af7 | 2780 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2781 | { |
2782 | struct intel_super *super = st->sb; | |
4f5bc454 | 2783 | struct imsm_disk *disk; |
a5d85af7 | 2784 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2785 | int max_enough = -1; |
2786 | int i; | |
2787 | struct imsm_super *mpb; | |
4f5bc454 | 2788 | |
bf5a934a | 2789 | if (super->current_vol >= 0) { |
a5d85af7 | 2790 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2791 | return; |
2792 | } | |
95eeceeb | 2793 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2794 | |
2795 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2796 | * spares | |
2797 | */ | |
2798 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2799 | info->array.level = LEVEL_CONTAINER; |
2800 | info->array.layout = 0; | |
2801 | info->array.md_minor = -1; | |
c2c087e6 | 2802 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2803 | info->array.utime = 0; |
2804 | info->array.chunk_size = 0; | |
2805 | ||
2806 | info->disk.major = 0; | |
2807 | info->disk.minor = 0; | |
cdddbdbc | 2808 | info->disk.raid_disk = -1; |
c2c087e6 | 2809 | info->reshape_active = 0; |
f35f2525 N |
2810 | info->array.major_version = -1; |
2811 | info->array.minor_version = -2; | |
c2c087e6 | 2812 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2813 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2814 | info->disk.number = -1; |
2815 | info->disk.state = 0; | |
c5afc314 | 2816 | info->name[0] = 0; |
921d9e16 | 2817 | info->recovery_start = MaxSector; |
3ad25638 | 2818 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2819 | |
97b4d0e9 | 2820 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2821 | mpb = super->anchor; |
97b4d0e9 | 2822 | |
ab3cb6b3 N |
2823 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2824 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2825 | int failed, enough, j, missing = 0; | |
2826 | struct imsm_map *map; | |
2827 | __u8 state; | |
97b4d0e9 | 2828 | |
3b451610 AK |
2829 | failed = imsm_count_failed(super, dev, MAP_0); |
2830 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2831 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2832 | |
2833 | /* any newly missing disks? | |
2834 | * (catches single-degraded vs double-degraded) | |
2835 | */ | |
2836 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2837 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2838 | __u32 idx = ord_to_idx(ord); |
2839 | ||
2840 | if (!(ord & IMSM_ORD_REBUILD) && | |
2841 | get_imsm_missing(super, idx)) { | |
2842 | missing = 1; | |
2843 | break; | |
2844 | } | |
97b4d0e9 | 2845 | } |
ab3cb6b3 N |
2846 | |
2847 | if (state == IMSM_T_STATE_FAILED) | |
2848 | enough = -1; | |
2849 | else if (state == IMSM_T_STATE_DEGRADED && | |
2850 | (state != map->map_state || missing)) | |
2851 | enough = 0; | |
2852 | else /* we're normal, or already degraded */ | |
2853 | enough = 1; | |
d2bde6d3 AK |
2854 | if (is_gen_migration(dev) && missing) { |
2855 | /* during general migration we need all disks | |
2856 | * that process is running on. | |
2857 | * No new missing disk is allowed. | |
2858 | */ | |
2859 | max_enough = -1; | |
2860 | enough = -1; | |
2861 | /* no more checks necessary | |
2862 | */ | |
2863 | break; | |
2864 | } | |
ab3cb6b3 N |
2865 | /* in the missing/failed disk case check to see |
2866 | * if at least one array is runnable | |
2867 | */ | |
2868 | max_enough = max(max_enough, enough); | |
2869 | } | |
2870 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2871 | info->container_enough = max_enough; | |
97b4d0e9 | 2872 | |
4a04ec6c | 2873 | if (super->disks) { |
14e8215b DW |
2874 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2875 | ||
b9f594fe | 2876 | disk = &super->disks->disk; |
5551b113 | 2877 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 2878 | info->component_size = reserved; |
25ed7e59 | 2879 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2880 | /* we don't change info->disk.raid_disk here because |
2881 | * this state will be finalized in mdmon after we have | |
2882 | * found the 'most fresh' version of the metadata | |
2883 | */ | |
25ed7e59 DW |
2884 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2885 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2886 | } |
a575e2a7 DW |
2887 | |
2888 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2889 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2890 | */ | |
2891 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2892 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2893 | else |
2894 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2895 | |
2896 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2897 | if (map) { | |
2898 | int i; | |
2899 | for (i = 0; i < map_disks; i++) | |
2900 | map[i] = 1; | |
2901 | } | |
2902 | ||
cdddbdbc DW |
2903 | } |
2904 | ||
5c4cd5da AC |
2905 | /* allocates memory and fills disk in mdinfo structure |
2906 | * for each disk in array */ | |
2907 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2908 | { | |
2909 | struct mdinfo *mddev = NULL; | |
2910 | struct intel_super *super = st->sb; | |
2911 | struct imsm_disk *disk; | |
2912 | int count = 0; | |
2913 | struct dl *dl; | |
2914 | if (!super || !super->disks) | |
2915 | return NULL; | |
2916 | dl = super->disks; | |
503975b9 | 2917 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
2918 | while (dl) { |
2919 | struct mdinfo *tmp; | |
2920 | disk = &dl->disk; | |
503975b9 | 2921 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
2922 | if (mddev->devs) |
2923 | tmp->next = mddev->devs; | |
2924 | mddev->devs = tmp; | |
2925 | tmp->disk.number = count++; | |
2926 | tmp->disk.major = dl->major; | |
2927 | tmp->disk.minor = dl->minor; | |
2928 | tmp->disk.state = is_configured(disk) ? | |
2929 | (1 << MD_DISK_ACTIVE) : 0; | |
2930 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2931 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2932 | tmp->disk.raid_disk = -1; | |
2933 | dl = dl->next; | |
2934 | } | |
2935 | return mddev; | |
2936 | } | |
2937 | ||
cdddbdbc DW |
2938 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2939 | char *update, char *devname, int verbose, | |
2940 | int uuid_set, char *homehost) | |
2941 | { | |
f352c545 DW |
2942 | /* For 'assemble' and 'force' we need to return non-zero if any |
2943 | * change was made. For others, the return value is ignored. | |
2944 | * Update options are: | |
2945 | * force-one : This device looks a bit old but needs to be included, | |
2946 | * update age info appropriately. | |
2947 | * assemble: clear any 'faulty' flag to allow this device to | |
2948 | * be assembled. | |
2949 | * force-array: Array is degraded but being forced, mark it clean | |
2950 | * if that will be needed to assemble it. | |
2951 | * | |
2952 | * newdev: not used ???? | |
2953 | * grow: Array has gained a new device - this is currently for | |
2954 | * linear only | |
2955 | * resync: mark as dirty so a resync will happen. | |
2956 | * name: update the name - preserving the homehost | |
6e46bf34 | 2957 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2958 | * |
2959 | * Following are not relevant for this imsm: | |
2960 | * sparc2.2 : update from old dodgey metadata | |
2961 | * super-minor: change the preferred_minor number | |
2962 | * summaries: update redundant counters. | |
f352c545 DW |
2963 | * homehost: update the recorded homehost |
2964 | * _reshape_progress: record new reshape_progress position. | |
2965 | */ | |
6e46bf34 DW |
2966 | int rv = 1; |
2967 | struct intel_super *super = st->sb; | |
2968 | struct imsm_super *mpb; | |
f352c545 | 2969 | |
6e46bf34 DW |
2970 | /* we can only update container info */ |
2971 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2972 | return 1; | |
2973 | ||
2974 | mpb = super->anchor; | |
2975 | ||
81a5b4f5 N |
2976 | if (strcmp(update, "uuid") == 0) { |
2977 | /* We take this to mean that the family_num should be updated. | |
2978 | * However that is much smaller than the uuid so we cannot really | |
2979 | * allow an explicit uuid to be given. And it is hard to reliably | |
2980 | * know if one was. | |
2981 | * So if !uuid_set we know the current uuid is random and just used | |
2982 | * the first 'int' and copy it to the other 3 positions. | |
2983 | * Otherwise we require the 4 'int's to be the same as would be the | |
2984 | * case if we are using a random uuid. So an explicit uuid will be | |
2985 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 2986 | */ |
81a5b4f5 N |
2987 | if (!uuid_set) { |
2988 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 2989 | rv = 0; |
81a5b4f5 N |
2990 | } else { |
2991 | if (info->uuid[0] != info->uuid[1] || | |
2992 | info->uuid[1] != info->uuid[2] || | |
2993 | info->uuid[2] != info->uuid[3]) | |
2994 | rv = -1; | |
2995 | else | |
2996 | rv = 0; | |
6e46bf34 | 2997 | } |
81a5b4f5 N |
2998 | if (rv == 0) |
2999 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3000 | } else if (strcmp(update, "assemble") == 0) |
3001 | rv = 0; | |
3002 | else | |
1e2b2765 | 3003 | rv = -1; |
f352c545 | 3004 | |
6e46bf34 DW |
3005 | /* successful update? recompute checksum */ |
3006 | if (rv == 0) | |
3007 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3008 | |
3009 | return rv; | |
cdddbdbc DW |
3010 | } |
3011 | ||
c2c087e6 | 3012 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3013 | { |
c2c087e6 | 3014 | size_t size; |
cdddbdbc | 3015 | |
c2c087e6 DW |
3016 | size = sizeof(struct imsm_super); |
3017 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3018 | size += 2 * sizeof(struct imsm_dev); | |
3019 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3020 | size += (4 - 2) * sizeof(struct imsm_map); | |
3021 | /* 4 possible disk_ord_tbl's */ | |
3022 | size += 4 * (disks - 1) * sizeof(__u32); | |
3023 | ||
3024 | return size; | |
3025 | } | |
3026 | ||
3027 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
3028 | { | |
3029 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3030 | return 0; | |
3031 | ||
3032 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3033 | } |
3034 | ||
ba2de7ba DW |
3035 | static void free_devlist(struct intel_super *super) |
3036 | { | |
3037 | struct intel_dev *dv; | |
3038 | ||
3039 | while (super->devlist) { | |
3040 | dv = super->devlist->next; | |
3041 | free(super->devlist->dev); | |
3042 | free(super->devlist); | |
3043 | super->devlist = dv; | |
3044 | } | |
3045 | } | |
3046 | ||
3047 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3048 | { | |
3049 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3050 | } | |
3051 | ||
cdddbdbc DW |
3052 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3053 | { | |
3054 | /* | |
3055 | * return: | |
3056 | * 0 same, or first was empty, and second was copied | |
3057 | * 1 second had wrong number | |
3058 | * 2 wrong uuid | |
3059 | * 3 wrong other info | |
3060 | */ | |
3061 | struct intel_super *first = st->sb; | |
3062 | struct intel_super *sec = tst->sb; | |
3063 | ||
3064 | if (!first) { | |
3065 | st->sb = tst->sb; | |
3066 | tst->sb = NULL; | |
3067 | return 0; | |
3068 | } | |
8603ea6f LM |
3069 | /* in platform dependent environment test if the disks |
3070 | * use the same Intel hba | |
3071 | */ | |
3072 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
3073 | if (!first->hba || !sec->hba || |
3074 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
3075 | fprintf(stderr, |
3076 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
3077 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
3078 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
3079 | return 3; |
3080 | } | |
3081 | } | |
cdddbdbc | 3082 | |
d23fe947 DW |
3083 | /* if an anchor does not have num_raid_devs set then it is a free |
3084 | * floating spare | |
3085 | */ | |
3086 | if (first->anchor->num_raid_devs > 0 && | |
3087 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3088 | /* Determine if these disks might ever have been |
3089 | * related. Further disambiguation can only take place | |
3090 | * in load_super_imsm_all | |
3091 | */ | |
3092 | __u32 first_family = first->anchor->orig_family_num; | |
3093 | __u32 sec_family = sec->anchor->orig_family_num; | |
3094 | ||
f796af5d DW |
3095 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3096 | MAX_SIGNATURE_LENGTH) != 0) | |
3097 | return 3; | |
3098 | ||
a2b97981 DW |
3099 | if (first_family == 0) |
3100 | first_family = first->anchor->family_num; | |
3101 | if (sec_family == 0) | |
3102 | sec_family = sec->anchor->family_num; | |
3103 | ||
3104 | if (first_family != sec_family) | |
d23fe947 | 3105 | return 3; |
f796af5d | 3106 | |
d23fe947 | 3107 | } |
cdddbdbc | 3108 | |
f796af5d | 3109 | |
3e372e5a DW |
3110 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3111 | * family number | |
3112 | */ | |
3113 | if (first->anchor->num_raid_devs == 0 && | |
3114 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3115 | int i; |
ba2de7ba DW |
3116 | struct intel_dev *dv; |
3117 | struct imsm_dev *dev; | |
78d30f94 DW |
3118 | |
3119 | /* we need to copy raid device info from sec if an allocation | |
3120 | * fails here we don't associate the spare | |
3121 | */ | |
3122 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3123 | dv = xmalloc(sizeof(*dv)); |
3124 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3125 | dv->dev = dev; |
3126 | dv->index = i; | |
3127 | dv->next = first->devlist; | |
3128 | first->devlist = dv; | |
78d30f94 | 3129 | } |
709743c5 | 3130 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3131 | /* allocation failure */ |
3132 | free_devlist(first); | |
3133 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
3134 | return 3; | |
78d30f94 | 3135 | } |
3e372e5a | 3136 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3137 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3138 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3139 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3140 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3141 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3142 | } |
3143 | ||
cdddbdbc DW |
3144 | return 0; |
3145 | } | |
3146 | ||
0030e8d6 DW |
3147 | static void fd2devname(int fd, char *name) |
3148 | { | |
3149 | struct stat st; | |
3150 | char path[256]; | |
33a6535d | 3151 | char dname[PATH_MAX]; |
0030e8d6 DW |
3152 | char *nm; |
3153 | int rv; | |
3154 | ||
3155 | name[0] = '\0'; | |
3156 | if (fstat(fd, &st) != 0) | |
3157 | return; | |
3158 | sprintf(path, "/sys/dev/block/%d:%d", | |
3159 | major(st.st_rdev), minor(st.st_rdev)); | |
3160 | ||
9cf014ec | 3161 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3162 | if (rv <= 0) |
3163 | return; | |
9587c373 | 3164 | |
0030e8d6 DW |
3165 | dname[rv] = '\0'; |
3166 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3167 | if (nm) { |
3168 | nm++; | |
3169 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3170 | } | |
0030e8d6 DW |
3171 | } |
3172 | ||
cdddbdbc DW |
3173 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3174 | ||
3175 | static int imsm_read_serial(int fd, char *devname, | |
3176 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3177 | { | |
3178 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
3179 | int rv; |
3180 | int rsp_len; | |
1f24f035 | 3181 | int len; |
316e2bf4 DW |
3182 | char *dest; |
3183 | char *src; | |
3184 | char *rsp_buf; | |
3185 | int i; | |
cdddbdbc DW |
3186 | |
3187 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 3188 | |
f9ba0ff1 DW |
3189 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
3190 | ||
40ebbb9c | 3191 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3192 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3193 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3194 | return 0; |
3195 | } | |
3196 | ||
cdddbdbc DW |
3197 | if (rv != 0) { |
3198 | if (devname) | |
e7b84f9d N |
3199 | pr_err("Failed to retrieve serial for %s\n", |
3200 | devname); | |
cdddbdbc DW |
3201 | return rv; |
3202 | } | |
3203 | ||
3204 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
3205 | if (!rsp_len) { |
3206 | if (devname) | |
e7b84f9d N |
3207 | pr_err("Failed to retrieve serial for %s\n", |
3208 | devname); | |
03cd4cc8 DW |
3209 | return 2; |
3210 | } | |
1f24f035 | 3211 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 3212 | |
316e2bf4 DW |
3213 | /* trim all whitespace and non-printable characters and convert |
3214 | * ':' to ';' | |
3215 | */ | |
3216 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
3217 | src = &rsp_buf[i]; | |
3218 | if (*src > 0x20) { | |
3219 | /* ':' is reserved for use in placeholder serial | |
3220 | * numbers for missing disks | |
3221 | */ | |
3222 | if (*src == ':') | |
3223 | *dest++ = ';'; | |
3224 | else | |
3225 | *dest++ = *src; | |
3226 | } | |
3227 | } | |
3228 | len = dest - rsp_buf; | |
3229 | dest = rsp_buf; | |
3230 | ||
3231 | /* truncate leading characters */ | |
3232 | if (len > MAX_RAID_SERIAL_LEN) { | |
3233 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3234 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3235 | } |
5c3db629 | 3236 | |
5c3db629 | 3237 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3238 | memcpy(serial, dest, len); |
cdddbdbc DW |
3239 | |
3240 | return 0; | |
3241 | } | |
3242 | ||
1f24f035 DW |
3243 | static int serialcmp(__u8 *s1, __u8 *s2) |
3244 | { | |
3245 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3246 | } | |
3247 | ||
3248 | static void serialcpy(__u8 *dest, __u8 *src) | |
3249 | { | |
3250 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3251 | } | |
3252 | ||
54c2c1ea DW |
3253 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3254 | { | |
3255 | struct dl *dl; | |
3256 | ||
3257 | for (dl = super->disks; dl; dl = dl->next) | |
3258 | if (serialcmp(dl->serial, serial) == 0) | |
3259 | break; | |
3260 | ||
3261 | return dl; | |
3262 | } | |
3263 | ||
a2b97981 DW |
3264 | static struct imsm_disk * |
3265 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3266 | { | |
3267 | int i; | |
3268 | ||
3269 | for (i = 0; i < mpb->num_disks; i++) { | |
3270 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3271 | ||
3272 | if (serialcmp(disk->serial, serial) == 0) { | |
3273 | if (idx) | |
3274 | *idx = i; | |
3275 | return disk; | |
3276 | } | |
3277 | } | |
3278 | ||
3279 | return NULL; | |
3280 | } | |
3281 | ||
cdddbdbc DW |
3282 | static int |
3283 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3284 | { | |
a2b97981 | 3285 | struct imsm_disk *disk; |
cdddbdbc DW |
3286 | struct dl *dl; |
3287 | struct stat stb; | |
cdddbdbc | 3288 | int rv; |
a2b97981 | 3289 | char name[40]; |
d23fe947 DW |
3290 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3291 | ||
3292 | rv = imsm_read_serial(fd, devname, serial); | |
3293 | ||
3294 | if (rv != 0) | |
3295 | return 2; | |
3296 | ||
503975b9 | 3297 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3298 | |
a2b97981 DW |
3299 | fstat(fd, &stb); |
3300 | dl->major = major(stb.st_rdev); | |
3301 | dl->minor = minor(stb.st_rdev); | |
3302 | dl->next = super->disks; | |
3303 | dl->fd = keep_fd ? fd : -1; | |
3304 | assert(super->disks == NULL); | |
3305 | super->disks = dl; | |
3306 | serialcpy(dl->serial, serial); | |
3307 | dl->index = -2; | |
3308 | dl->e = NULL; | |
3309 | fd2devname(fd, name); | |
3310 | if (devname) | |
503975b9 | 3311 | dl->devname = xstrdup(devname); |
a2b97981 | 3312 | else |
503975b9 | 3313 | dl->devname = xstrdup(name); |
cdddbdbc | 3314 | |
d23fe947 | 3315 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3316 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3317 | if (disk) { | |
3318 | dl->disk = *disk; | |
3319 | /* only set index on disks that are a member of a | |
3320 | * populated contianer, i.e. one with raid_devs | |
3321 | */ | |
3322 | if (is_failed(&dl->disk)) | |
3f6efecc | 3323 | dl->index = -2; |
a2b97981 DW |
3324 | else if (is_spare(&dl->disk)) |
3325 | dl->index = -1; | |
3f6efecc DW |
3326 | } |
3327 | ||
949c47a0 DW |
3328 | return 0; |
3329 | } | |
3330 | ||
0e600426 | 3331 | #ifndef MDASSEMBLE |
0c046afd DW |
3332 | /* When migrating map0 contains the 'destination' state while map1 |
3333 | * contains the current state. When not migrating map0 contains the | |
3334 | * current state. This routine assumes that map[0].map_state is set to | |
3335 | * the current array state before being called. | |
3336 | * | |
3337 | * Migration is indicated by one of the following states | |
3338 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3339 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3340 | * map1state=unitialized) |
1484e727 | 3341 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3342 | * map1state=normal) |
e3bba0e0 | 3343 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3344 | * map1state=degraded) |
8e59f3d8 AK |
3345 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3346 | * map1state=normal) | |
0c046afd | 3347 | */ |
8e59f3d8 AK |
3348 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3349 | __u8 to_state, int migr_type) | |
3393c6af | 3350 | { |
0c046afd | 3351 | struct imsm_map *dest; |
238c0a71 | 3352 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3353 | |
0c046afd | 3354 | dev->vol.migr_state = 1; |
1484e727 | 3355 | set_migr_type(dev, migr_type); |
f8f603f1 | 3356 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3357 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3358 | |
0556e1a2 | 3359 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3360 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3361 | if ((migr_type == MIGR_REBUILD) || |
3362 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3363 | __u32 ord; |
3364 | int i; | |
3365 | ||
3366 | for (i = 0; i < src->num_members; i++) { | |
3367 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3368 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3369 | } | |
3370 | } | |
3371 | ||
8e59f3d8 AK |
3372 | if (migr_type == MIGR_GEN_MIGR) |
3373 | /* Clear migration record */ | |
3374 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3375 | ||
0c046afd | 3376 | src->map_state = to_state; |
949c47a0 | 3377 | } |
f8f603f1 | 3378 | |
809da78e AK |
3379 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3380 | __u8 map_state) | |
f8f603f1 | 3381 | { |
238c0a71 AK |
3382 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3383 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3384 | MAP_0 : MAP_1); | |
28bce06f | 3385 | int i, j; |
0556e1a2 DW |
3386 | |
3387 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3388 | * completed in the last migration. | |
3389 | * | |
28bce06f | 3390 | * FIXME add support for raid-level-migration |
0556e1a2 | 3391 | */ |
809da78e AK |
3392 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3393 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3394 | /* when final map state is other than expected | |
3395 | * merge maps (not for migration) | |
3396 | */ | |
3397 | int failed; | |
3398 | ||
3399 | for (i = 0; i < prev->num_members; i++) | |
3400 | for (j = 0; j < map->num_members; j++) | |
3401 | /* during online capacity expansion | |
3402 | * disks position can be changed | |
3403 | * if takeover is used | |
3404 | */ | |
3405 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3406 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3407 | map->disk_ord_tbl[j] |= | |
3408 | prev->disk_ord_tbl[i]; | |
3409 | break; | |
3410 | } | |
3411 | failed = imsm_count_failed(super, dev, MAP_0); | |
3412 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3413 | } | |
f8f603f1 DW |
3414 | |
3415 | dev->vol.migr_state = 0; | |
ea672ee1 | 3416 | set_migr_type(dev, 0); |
f8f603f1 DW |
3417 | dev->vol.curr_migr_unit = 0; |
3418 | map->map_state = map_state; | |
3419 | } | |
0e600426 | 3420 | #endif |
949c47a0 DW |
3421 | |
3422 | static int parse_raid_devices(struct intel_super *super) | |
3423 | { | |
3424 | int i; | |
3425 | struct imsm_dev *dev_new; | |
4d7b1503 | 3426 | size_t len, len_migr; |
401d313b | 3427 | size_t max_len = 0; |
4d7b1503 DW |
3428 | size_t space_needed = 0; |
3429 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3430 | |
3431 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3432 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3433 | struct intel_dev *dv; |
949c47a0 | 3434 | |
4d7b1503 DW |
3435 | len = sizeof_imsm_dev(dev_iter, 0); |
3436 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3437 | if (len_migr > len) | |
3438 | space_needed += len_migr - len; | |
ca9de185 | 3439 | |
503975b9 | 3440 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
3441 | if (max_len < len_migr) |
3442 | max_len = len_migr; | |
3443 | if (max_len > len_migr) | |
3444 | space_needed += max_len - len_migr; | |
503975b9 | 3445 | dev_new = xmalloc(max_len); |
949c47a0 | 3446 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3447 | dv->dev = dev_new; |
3448 | dv->index = i; | |
3449 | dv->next = super->devlist; | |
3450 | super->devlist = dv; | |
949c47a0 | 3451 | } |
cdddbdbc | 3452 | |
4d7b1503 DW |
3453 | /* ensure that super->buf is large enough when all raid devices |
3454 | * are migrating | |
3455 | */ | |
3456 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3457 | void *buf; | |
3458 | ||
3459 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3460 | if (posix_memalign(&buf, 512, len) != 0) | |
3461 | return 1; | |
3462 | ||
1f45a8ad DW |
3463 | memcpy(buf, super->buf, super->len); |
3464 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3465 | free(super->buf); |
3466 | super->buf = buf; | |
3467 | super->len = len; | |
3468 | } | |
ca9de185 | 3469 | |
cdddbdbc DW |
3470 | return 0; |
3471 | } | |
3472 | ||
604b746f JD |
3473 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3474 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3475 | { | |
3476 | void *ptr = NULL; | |
3477 | ||
3478 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3479 | ptr = mpb; | |
3480 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3481 | } | |
3482 | ||
3483 | return ptr; | |
3484 | } | |
3485 | ||
e2f41b2c AK |
3486 | /******************************************************************************* |
3487 | * Function: check_mpb_migr_compatibility | |
3488 | * Description: Function checks for unsupported migration features: | |
3489 | * - migration optimization area (pba_of_lba0) | |
3490 | * - descending reshape (ascending_migr) | |
3491 | * Parameters: | |
3492 | * super : imsm metadata information | |
3493 | * Returns: | |
3494 | * 0 : migration is compatible | |
3495 | * -1 : migration is not compatible | |
3496 | ******************************************************************************/ | |
3497 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3498 | { | |
3499 | struct imsm_map *map0, *map1; | |
3500 | struct migr_record *migr_rec = super->migr_rec; | |
3501 | int i; | |
3502 | ||
3503 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3504 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3505 | ||
3506 | if (dev_iter && | |
3507 | dev_iter->vol.migr_state == 1 && | |
3508 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3509 | /* This device is migrating */ | |
238c0a71 AK |
3510 | map0 = get_imsm_map(dev_iter, MAP_0); |
3511 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 3512 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
3513 | /* migration optimization area was used */ |
3514 | return -1; | |
3515 | if (migr_rec->ascending_migr == 0 | |
3516 | && migr_rec->dest_depth_per_unit > 0) | |
3517 | /* descending reshape not supported yet */ | |
3518 | return -1; | |
3519 | } | |
3520 | } | |
3521 | return 0; | |
3522 | } | |
3523 | ||
d23fe947 | 3524 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3525 | |
cdddbdbc | 3526 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3527 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3528 | */ |
3529 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3530 | { | |
3531 | unsigned long long dsize; | |
cdddbdbc DW |
3532 | unsigned long long sectors; |
3533 | struct stat; | |
6416d527 | 3534 | struct imsm_super *anchor; |
cdddbdbc DW |
3535 | __u32 check_sum; |
3536 | ||
cdddbdbc | 3537 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3538 | if (dsize < 1024) { |
3539 | if (devname) | |
e7b84f9d N |
3540 | pr_err("%s: device to small for imsm\n", |
3541 | devname); | |
64436f06 N |
3542 | return 1; |
3543 | } | |
cdddbdbc DW |
3544 | |
3545 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3546 | if (devname) | |
e7b84f9d N |
3547 | pr_err("Cannot seek to anchor block on %s: %s\n", |
3548 | devname, strerror(errno)); | |
cdddbdbc DW |
3549 | return 1; |
3550 | } | |
3551 | ||
949c47a0 | 3552 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e | 3553 | if (devname) |
e7b84f9d N |
3554 | pr_err("Failed to allocate imsm anchor buffer" |
3555 | " on %s\n", devname); | |
ad97895e DW |
3556 | return 1; |
3557 | } | |
949c47a0 | 3558 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc | 3559 | if (devname) |
e7b84f9d N |
3560 | pr_err("Cannot read anchor block on %s: %s\n", |
3561 | devname, strerror(errno)); | |
6416d527 | 3562 | free(anchor); |
cdddbdbc DW |
3563 | return 1; |
3564 | } | |
3565 | ||
6416d527 | 3566 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 3567 | if (devname) |
e7b84f9d | 3568 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 3569 | free(anchor); |
cdddbdbc DW |
3570 | return 2; |
3571 | } | |
3572 | ||
d23fe947 | 3573 | __free_imsm(super, 0); |
f2f5c343 LM |
3574 | /* reload capability and hba */ |
3575 | ||
3576 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3577 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3578 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3579 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc | 3580 | if (devname) |
e7b84f9d N |
3581 | pr_err("unable to allocate %zu byte mpb buffer\n", |
3582 | super->len); | |
6416d527 | 3583 | free(anchor); |
cdddbdbc DW |
3584 | return 2; |
3585 | } | |
949c47a0 | 3586 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3587 | |
6416d527 NB |
3588 | sectors = mpb_sectors(anchor) - 1; |
3589 | free(anchor); | |
8e59f3d8 | 3590 | |
17a4eaf9 | 3591 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
e7b84f9d | 3592 | pr_err("%s could not allocate migr_rec buffer\n", __func__); |
8e59f3d8 AK |
3593 | free(super->buf); |
3594 | return 2; | |
3595 | } | |
51d83f5d | 3596 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 3597 | |
949c47a0 | 3598 | if (!sectors) { |
ecf45690 DW |
3599 | check_sum = __gen_imsm_checksum(super->anchor); |
3600 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3601 | if (devname) | |
e7b84f9d N |
3602 | pr_err("IMSM checksum %x != %x on %s\n", |
3603 | check_sum, | |
3604 | __le32_to_cpu(super->anchor->check_sum), | |
3605 | devname); | |
ecf45690 DW |
3606 | return 2; |
3607 | } | |
3608 | ||
a2b97981 | 3609 | return 0; |
949c47a0 | 3610 | } |
cdddbdbc DW |
3611 | |
3612 | /* read the extended mpb */ | |
3613 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3614 | if (devname) | |
e7b84f9d N |
3615 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
3616 | devname, strerror(errno)); | |
cdddbdbc DW |
3617 | return 1; |
3618 | } | |
3619 | ||
f21e18ca | 3620 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc | 3621 | if (devname) |
e7b84f9d N |
3622 | pr_err("Cannot read extended mpb on %s: %s\n", |
3623 | devname, strerror(errno)); | |
cdddbdbc DW |
3624 | return 2; |
3625 | } | |
3626 | ||
949c47a0 DW |
3627 | check_sum = __gen_imsm_checksum(super->anchor); |
3628 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 3629 | if (devname) |
e7b84f9d N |
3630 | pr_err("IMSM checksum %x != %x on %s\n", |
3631 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
3632 | devname); | |
db575f3b | 3633 | return 3; |
cdddbdbc DW |
3634 | } |
3635 | ||
604b746f JD |
3636 | /* FIXME the BBM log is disk specific so we cannot use this global |
3637 | * buffer for all disks. Ok for now since we only look at the global | |
3638 | * bbm_log_size parameter to gate assembly | |
3639 | */ | |
3640 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3641 | ||
a2b97981 DW |
3642 | return 0; |
3643 | } | |
3644 | ||
8e59f3d8 AK |
3645 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3646 | ||
97f81ee2 CA |
3647 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
3648 | static void clear_hi(struct intel_super *super) | |
3649 | { | |
3650 | struct imsm_super *mpb = super->anchor; | |
3651 | int i, n; | |
3652 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
3653 | return; | |
3654 | for (i = 0; i < mpb->num_disks; ++i) { | |
3655 | struct imsm_disk *disk = &mpb->disk[i]; | |
3656 | disk->total_blocks_hi = 0; | |
3657 | } | |
3658 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
3659 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3660 | if (!dev) | |
3661 | return; | |
3662 | for (n = 0; n < 2; ++n) { | |
3663 | struct imsm_map *map = get_imsm_map(dev, n); | |
3664 | if (!map) | |
3665 | continue; | |
3666 | map->pba_of_lba0_hi = 0; | |
3667 | map->blocks_per_member_hi = 0; | |
3668 | map->num_data_stripes_hi = 0; | |
3669 | } | |
3670 | } | |
3671 | } | |
3672 | ||
a2b97981 DW |
3673 | static int |
3674 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3675 | { | |
3676 | int err; | |
3677 | ||
3678 | err = load_imsm_mpb(fd, super, devname); | |
3679 | if (err) | |
3680 | return err; | |
3681 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3682 | if (err) | |
3683 | return err; | |
3684 | err = parse_raid_devices(super); | |
97f81ee2 | 3685 | clear_hi(super); |
a2b97981 | 3686 | return err; |
cdddbdbc DW |
3687 | } |
3688 | ||
ae6aad82 DW |
3689 | static void __free_imsm_disk(struct dl *d) |
3690 | { | |
3691 | if (d->fd >= 0) | |
3692 | close(d->fd); | |
3693 | if (d->devname) | |
3694 | free(d->devname); | |
0dcecb2e DW |
3695 | if (d->e) |
3696 | free(d->e); | |
ae6aad82 DW |
3697 | free(d); |
3698 | ||
3699 | } | |
1a64be56 | 3700 | |
cdddbdbc DW |
3701 | static void free_imsm_disks(struct intel_super *super) |
3702 | { | |
47ee5a45 | 3703 | struct dl *d; |
cdddbdbc | 3704 | |
47ee5a45 DW |
3705 | while (super->disks) { |
3706 | d = super->disks; | |
cdddbdbc | 3707 | super->disks = d->next; |
ae6aad82 | 3708 | __free_imsm_disk(d); |
cdddbdbc | 3709 | } |
cb82edca AK |
3710 | while (super->disk_mgmt_list) { |
3711 | d = super->disk_mgmt_list; | |
3712 | super->disk_mgmt_list = d->next; | |
3713 | __free_imsm_disk(d); | |
3714 | } | |
47ee5a45 DW |
3715 | while (super->missing) { |
3716 | d = super->missing; | |
3717 | super->missing = d->next; | |
3718 | __free_imsm_disk(d); | |
3719 | } | |
3720 | ||
cdddbdbc DW |
3721 | } |
3722 | ||
9ca2c81c | 3723 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3724 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3725 | { |
88654014 LM |
3726 | struct intel_hba *elem, *next; |
3727 | ||
9ca2c81c | 3728 | if (super->buf) { |
949c47a0 | 3729 | free(super->buf); |
9ca2c81c DW |
3730 | super->buf = NULL; |
3731 | } | |
f2f5c343 LM |
3732 | /* unlink capability description */ |
3733 | super->orom = NULL; | |
8e59f3d8 AK |
3734 | if (super->migr_rec_buf) { |
3735 | free(super->migr_rec_buf); | |
3736 | super->migr_rec_buf = NULL; | |
3737 | } | |
d23fe947 DW |
3738 | if (free_disks) |
3739 | free_imsm_disks(super); | |
ba2de7ba | 3740 | free_devlist(super); |
88654014 LM |
3741 | elem = super->hba; |
3742 | while (elem) { | |
3743 | if (elem->path) | |
3744 | free((void *)elem->path); | |
3745 | next = elem->next; | |
3746 | free(elem); | |
3747 | elem = next; | |
88c32bb1 | 3748 | } |
88654014 | 3749 | super->hba = NULL; |
cdddbdbc DW |
3750 | } |
3751 | ||
9ca2c81c DW |
3752 | static void free_imsm(struct intel_super *super) |
3753 | { | |
d23fe947 | 3754 | __free_imsm(super, 1); |
9ca2c81c DW |
3755 | free(super); |
3756 | } | |
cdddbdbc DW |
3757 | |
3758 | static void free_super_imsm(struct supertype *st) | |
3759 | { | |
3760 | struct intel_super *super = st->sb; | |
3761 | ||
3762 | if (!super) | |
3763 | return; | |
3764 | ||
3765 | free_imsm(super); | |
3766 | st->sb = NULL; | |
3767 | } | |
3768 | ||
49133e57 | 3769 | static struct intel_super *alloc_super(void) |
c2c087e6 | 3770 | { |
503975b9 | 3771 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 3772 | |
503975b9 N |
3773 | super->current_vol = -1; |
3774 | super->create_offset = ~((unsigned long long) 0); | |
c2c087e6 DW |
3775 | return super; |
3776 | } | |
3777 | ||
f0f5a016 LM |
3778 | /* |
3779 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3780 | */ | |
d424212e | 3781 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3782 | { |
3783 | struct sys_dev *hba_name; | |
3784 | int rv = 0; | |
3785 | ||
3786 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3787 | super->orom = NULL; |
f0f5a016 LM |
3788 | super->hba = NULL; |
3789 | return 0; | |
3790 | } | |
3791 | hba_name = find_disk_attached_hba(fd, NULL); | |
3792 | if (!hba_name) { | |
d424212e | 3793 | if (devname) |
e7b84f9d N |
3794 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
3795 | devname); | |
f0f5a016 LM |
3796 | return 1; |
3797 | } | |
3798 | rv = attach_hba_to_super(super, hba_name); | |
3799 | if (rv == 2) { | |
d424212e N |
3800 | if (devname) { |
3801 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3802 | |
e7b84f9d | 3803 | pr_err("%s is attached to Intel(R) %s RAID " |
f0f5a016 LM |
3804 | "controller (%s),\n" |
3805 | " but the container is assigned to Intel(R) " | |
3806 | "%s RAID controller (", | |
d424212e | 3807 | devname, |
f0f5a016 LM |
3808 | hba_name->path, |
3809 | hba_name->pci_id ? : "Err!", | |
3810 | get_sys_dev_type(hba_name->type)); | |
3811 | ||
f0f5a016 LM |
3812 | while (hba) { |
3813 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3814 | if (hba->next) | |
3815 | fprintf(stderr, ", "); | |
3816 | hba = hba->next; | |
3817 | } | |
3818 | ||
3819 | fprintf(stderr, ").\n" | |
3820 | " Mixing devices attached to different controllers " | |
3821 | "is not allowed.\n"); | |
3822 | } | |
3823 | free_sys_dev(&hba_name); | |
3824 | return 2; | |
3825 | } | |
f2f5c343 | 3826 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3827 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3828 | if (!super->orom) |
3829 | return 3; | |
f0f5a016 LM |
3830 | return 0; |
3831 | } | |
3832 | ||
47ee5a45 DW |
3833 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3834 | * disks that have disappeared from the system. This routine relies on | |
3835 | * the mpb being uptodate, which it is at load time. | |
3836 | */ | |
3837 | static int find_missing(struct intel_super *super) | |
3838 | { | |
3839 | int i; | |
3840 | struct imsm_super *mpb = super->anchor; | |
3841 | struct dl *dl; | |
3842 | struct imsm_disk *disk; | |
47ee5a45 DW |
3843 | |
3844 | for (i = 0; i < mpb->num_disks; i++) { | |
3845 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3846 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3847 | if (dl) |
3848 | continue; | |
47ee5a45 | 3849 | |
503975b9 | 3850 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
3851 | dl->major = 0; |
3852 | dl->minor = 0; | |
3853 | dl->fd = -1; | |
503975b9 | 3854 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
3855 | dl->index = i; |
3856 | serialcpy(dl->serial, disk->serial); | |
3857 | dl->disk = *disk; | |
689c9bf3 | 3858 | dl->e = NULL; |
47ee5a45 DW |
3859 | dl->next = super->missing; |
3860 | super->missing = dl; | |
3861 | } | |
3862 | ||
3863 | return 0; | |
3864 | } | |
3865 | ||
3960e579 | 3866 | #ifndef MDASSEMBLE |
a2b97981 DW |
3867 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3868 | { | |
3869 | struct intel_disk *idisk = disk_list; | |
3870 | ||
3871 | while (idisk) { | |
3872 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3873 | break; | |
3874 | idisk = idisk->next; | |
3875 | } | |
3876 | ||
3877 | return idisk; | |
3878 | } | |
3879 | ||
3880 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3881 | struct intel_super *super, | |
3882 | struct intel_disk **disk_list) | |
3883 | { | |
3884 | struct imsm_disk *d = &super->disks->disk; | |
3885 | struct imsm_super *mpb = super->anchor; | |
3886 | int i, j; | |
3887 | ||
3888 | for (i = 0; i < tbl_size; i++) { | |
3889 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3890 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3891 | ||
3892 | if (tbl_mpb->family_num == mpb->family_num) { | |
3893 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3894 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3895 | __func__, super->disks->major, | |
3896 | super->disks->minor, | |
3897 | table[i]->disks->major, | |
3898 | table[i]->disks->minor); | |
3899 | break; | |
3900 | } | |
3901 | ||
3902 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3903 | is_configured(d) == is_configured(tbl_d)) && | |
3904 | tbl_mpb->generation_num < mpb->generation_num) { | |
3905 | /* current version of the mpb is a | |
3906 | * better candidate than the one in | |
3907 | * super_table, but copy over "cross | |
3908 | * generational" status | |
3909 | */ | |
3910 | struct intel_disk *idisk; | |
3911 | ||
3912 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3913 | __func__, super->disks->major, | |
3914 | super->disks->minor, | |
3915 | table[i]->disks->major, | |
3916 | table[i]->disks->minor); | |
3917 | ||
3918 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3919 | if (idisk && is_failed(&idisk->disk)) | |
3920 | tbl_d->status |= FAILED_DISK; | |
3921 | break; | |
3922 | } else { | |
3923 | struct intel_disk *idisk; | |
3924 | struct imsm_disk *disk; | |
3925 | ||
3926 | /* tbl_mpb is more up to date, but copy | |
3927 | * over cross generational status before | |
3928 | * returning | |
3929 | */ | |
3930 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3931 | if (disk && is_failed(disk)) | |
3932 | d->status |= FAILED_DISK; | |
3933 | ||
3934 | idisk = disk_list_get(d->serial, *disk_list); | |
3935 | if (idisk) { | |
3936 | idisk->owner = i; | |
3937 | if (disk && is_configured(disk)) | |
3938 | idisk->disk.status |= CONFIGURED_DISK; | |
3939 | } | |
3940 | ||
3941 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3942 | __func__, super->disks->major, | |
3943 | super->disks->minor, | |
3944 | table[i]->disks->major, | |
3945 | table[i]->disks->minor); | |
3946 | ||
3947 | return tbl_size; | |
3948 | } | |
3949 | } | |
3950 | } | |
3951 | ||
3952 | if (i >= tbl_size) | |
3953 | table[tbl_size++] = super; | |
3954 | else | |
3955 | table[i] = super; | |
3956 | ||
3957 | /* update/extend the merged list of imsm_disk records */ | |
3958 | for (j = 0; j < mpb->num_disks; j++) { | |
3959 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3960 | struct intel_disk *idisk; | |
3961 | ||
3962 | idisk = disk_list_get(disk->serial, *disk_list); | |
3963 | if (idisk) { | |
3964 | idisk->disk.status |= disk->status; | |
3965 | if (is_configured(&idisk->disk) || | |
3966 | is_failed(&idisk->disk)) | |
3967 | idisk->disk.status &= ~(SPARE_DISK); | |
3968 | } else { | |
503975b9 | 3969 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
3970 | idisk->owner = IMSM_UNKNOWN_OWNER; |
3971 | idisk->disk = *disk; | |
3972 | idisk->next = *disk_list; | |
3973 | *disk_list = idisk; | |
3974 | } | |
3975 | ||
3976 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3977 | idisk->owner = i; | |
3978 | } | |
3979 | ||
3980 | return tbl_size; | |
3981 | } | |
3982 | ||
3983 | static struct intel_super * | |
3984 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3985 | const int owner) | |
3986 | { | |
3987 | struct imsm_super *mpb = super->anchor; | |
3988 | int ok_count = 0; | |
3989 | int i; | |
3990 | ||
3991 | for (i = 0; i < mpb->num_disks; i++) { | |
3992 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3993 | struct intel_disk *idisk; | |
3994 | ||
3995 | idisk = disk_list_get(disk->serial, disk_list); | |
3996 | if (idisk) { | |
3997 | if (idisk->owner == owner || | |
3998 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3999 | ok_count++; | |
4000 | else | |
4001 | dprintf("%s: '%.16s' owner %d != %d\n", | |
4002 | __func__, disk->serial, idisk->owner, | |
4003 | owner); | |
4004 | } else { | |
4005 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
4006 | __func__, __le32_to_cpu(mpb->family_num), i, | |
4007 | disk->serial); | |
4008 | break; | |
4009 | } | |
4010 | } | |
4011 | ||
4012 | if (ok_count == mpb->num_disks) | |
4013 | return super; | |
4014 | return NULL; | |
4015 | } | |
4016 | ||
4017 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4018 | { | |
4019 | struct intel_super *s; | |
4020 | ||
4021 | for (s = super_list; s; s = s->next) { | |
4022 | if (family_num != s->anchor->family_num) | |
4023 | continue; | |
4024 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
4025 | __le32_to_cpu(family_num), s->disks->devname); | |
4026 | } | |
4027 | } | |
4028 | ||
4029 | static struct intel_super * | |
4030 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4031 | { | |
4032 | struct intel_super *super_table[len]; | |
4033 | struct intel_disk *disk_list = NULL; | |
4034 | struct intel_super *champion, *spare; | |
4035 | struct intel_super *s, **del; | |
4036 | int tbl_size = 0; | |
4037 | int conflict; | |
4038 | int i; | |
4039 | ||
4040 | memset(super_table, 0, sizeof(super_table)); | |
4041 | for (s = *super_list; s; s = s->next) | |
4042 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4043 | ||
4044 | for (i = 0; i < tbl_size; i++) { | |
4045 | struct imsm_disk *d; | |
4046 | struct intel_disk *idisk; | |
4047 | struct imsm_super *mpb = super_table[i]->anchor; | |
4048 | ||
4049 | s = super_table[i]; | |
4050 | d = &s->disks->disk; | |
4051 | ||
4052 | /* 'd' must appear in merged disk list for its | |
4053 | * configuration to be valid | |
4054 | */ | |
4055 | idisk = disk_list_get(d->serial, disk_list); | |
4056 | if (idisk && idisk->owner == i) | |
4057 | s = validate_members(s, disk_list, i); | |
4058 | else | |
4059 | s = NULL; | |
4060 | ||
4061 | if (!s) | |
4062 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
4063 | __func__, mpb->family_num, | |
4064 | super_table[i]->disks->major, | |
4065 | super_table[i]->disks->minor); | |
4066 | super_table[i] = s; | |
4067 | } | |
4068 | ||
4069 | /* This is where the mdadm implementation differs from the Windows | |
4070 | * driver which has no strict concept of a container. We can only | |
4071 | * assemble one family from a container, so when returning a prodigal | |
4072 | * array member to this system the code will not be able to disambiguate | |
4073 | * the container contents that should be assembled ("foreign" versus | |
4074 | * "local"). It requires user intervention to set the orig_family_num | |
4075 | * to a new value to establish a new container. The Windows driver in | |
4076 | * this situation fixes up the volume name in place and manages the | |
4077 | * foreign array as an independent entity. | |
4078 | */ | |
4079 | s = NULL; | |
4080 | spare = NULL; | |
4081 | conflict = 0; | |
4082 | for (i = 0; i < tbl_size; i++) { | |
4083 | struct intel_super *tbl_ent = super_table[i]; | |
4084 | int is_spare = 0; | |
4085 | ||
4086 | if (!tbl_ent) | |
4087 | continue; | |
4088 | ||
4089 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4090 | spare = tbl_ent; | |
4091 | is_spare = 1; | |
4092 | } | |
4093 | ||
4094 | if (s && !is_spare) { | |
4095 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4096 | conflict++; | |
4097 | } else if (!s && !is_spare) | |
4098 | s = tbl_ent; | |
4099 | } | |
4100 | ||
4101 | if (!s) | |
4102 | s = spare; | |
4103 | if (!s) { | |
4104 | champion = NULL; | |
4105 | goto out; | |
4106 | } | |
4107 | champion = s; | |
4108 | ||
4109 | if (conflict) | |
4110 | fprintf(stderr, "Chose family %#x on '%s', " | |
4111 | "assemble conflicts to new container with '--update=uuid'\n", | |
4112 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
4113 | ||
4114 | /* collect all dl's onto 'champion', and update them to | |
4115 | * champion's version of the status | |
4116 | */ | |
4117 | for (s = *super_list; s; s = s->next) { | |
4118 | struct imsm_super *mpb = champion->anchor; | |
4119 | struct dl *dl = s->disks; | |
4120 | ||
4121 | if (s == champion) | |
4122 | continue; | |
4123 | ||
5d7b407a CA |
4124 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4125 | ||
a2b97981 DW |
4126 | for (i = 0; i < mpb->num_disks; i++) { |
4127 | struct imsm_disk *disk; | |
4128 | ||
4129 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4130 | if (disk) { | |
4131 | dl->disk = *disk; | |
4132 | /* only set index on disks that are a member of | |
4133 | * a populated contianer, i.e. one with | |
4134 | * raid_devs | |
4135 | */ | |
4136 | if (is_failed(&dl->disk)) | |
4137 | dl->index = -2; | |
4138 | else if (is_spare(&dl->disk)) | |
4139 | dl->index = -1; | |
4140 | break; | |
4141 | } | |
4142 | } | |
4143 | ||
4144 | if (i >= mpb->num_disks) { | |
4145 | struct intel_disk *idisk; | |
4146 | ||
4147 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4148 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4149 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4150 | dl->index = -1; | |
4151 | else { | |
4152 | dl->index = -2; | |
4153 | continue; | |
4154 | } | |
4155 | } | |
4156 | ||
4157 | dl->next = champion->disks; | |
4158 | champion->disks = dl; | |
4159 | s->disks = NULL; | |
4160 | } | |
4161 | ||
4162 | /* delete 'champion' from super_list */ | |
4163 | for (del = super_list; *del; ) { | |
4164 | if (*del == champion) { | |
4165 | *del = (*del)->next; | |
4166 | break; | |
4167 | } else | |
4168 | del = &(*del)->next; | |
4169 | } | |
4170 | champion->next = NULL; | |
4171 | ||
4172 | out: | |
4173 | while (disk_list) { | |
4174 | struct intel_disk *idisk = disk_list; | |
4175 | ||
4176 | disk_list = disk_list->next; | |
4177 | free(idisk); | |
4178 | } | |
4179 | ||
4180 | return champion; | |
4181 | } | |
4182 | ||
9587c373 LM |
4183 | |
4184 | static int | |
4185 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
9587c373 LM |
4186 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, |
4187 | int major, int minor, int keep_fd); | |
ec50f7b6 LM |
4188 | static int |
4189 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4190 | int *max, int keep_fd); | |
4191 | ||
9587c373 | 4192 | |
cdddbdbc | 4193 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4194 | char *devname, struct md_list *devlist, |
4195 | int keep_fd) | |
cdddbdbc | 4196 | { |
a2b97981 DW |
4197 | struct intel_super *super_list = NULL; |
4198 | struct intel_super *super = NULL; | |
a2b97981 | 4199 | int err = 0; |
9587c373 | 4200 | int i = 0; |
dab4a513 | 4201 | |
9587c373 LM |
4202 | if (fd >= 0) |
4203 | /* 'fd' is an opened container */ | |
4204 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4205 | else | |
ec50f7b6 LM |
4206 | /* get super block from devlist devices */ |
4207 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4208 | if (err) |
1602d52c | 4209 | goto error; |
a2b97981 DW |
4210 | /* all mpbs enter, maybe one leaves */ |
4211 | super = imsm_thunderdome(&super_list, i); | |
4212 | if (!super) { | |
4213 | err = 1; | |
4214 | goto error; | |
cdddbdbc DW |
4215 | } |
4216 | ||
47ee5a45 DW |
4217 | if (find_missing(super) != 0) { |
4218 | free_imsm(super); | |
a2b97981 DW |
4219 | err = 2; |
4220 | goto error; | |
47ee5a45 | 4221 | } |
8e59f3d8 AK |
4222 | |
4223 | /* load migration record */ | |
4224 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4225 | if (err == -1) { |
4226 | /* migration is in progress, | |
4227 | * but migr_rec cannot be loaded, | |
4228 | */ | |
8e59f3d8 AK |
4229 | err = 4; |
4230 | goto error; | |
4231 | } | |
e2f41b2c AK |
4232 | |
4233 | /* Check migration compatibility */ | |
4c965cc9 | 4234 | if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) { |
e7b84f9d | 4235 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4236 | if (devname) |
4237 | fprintf(stderr, " on %s\n", devname); | |
4238 | else | |
4239 | fprintf(stderr, " (IMSM).\n"); | |
4240 | ||
4241 | err = 5; | |
4242 | goto error; | |
4243 | } | |
4244 | ||
a2b97981 DW |
4245 | err = 0; |
4246 | ||
4247 | error: | |
4248 | while (super_list) { | |
4249 | struct intel_super *s = super_list; | |
4250 | ||
4251 | super_list = super_list->next; | |
4252 | free_imsm(s); | |
4253 | } | |
9587c373 | 4254 | |
a2b97981 DW |
4255 | |
4256 | if (err) | |
4257 | return err; | |
f7e7067b | 4258 | |
cdddbdbc | 4259 | *sbp = super; |
9587c373 LM |
4260 | if (fd >= 0) |
4261 | st->container_dev = fd2devnum(fd); | |
4262 | else | |
4263 | st->container_dev = NoMdDev; | |
a2b97981 | 4264 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4265 | st->ss = &super_imsm; |
cdddbdbc DW |
4266 | st->minor_version = 0; |
4267 | st->max_devs = IMSM_MAX_DEVICES; | |
4268 | } | |
cdddbdbc DW |
4269 | return 0; |
4270 | } | |
2b959fbf | 4271 | |
9587c373 | 4272 | |
ec50f7b6 LM |
4273 | static int |
4274 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4275 | int *max, int keep_fd) | |
4276 | { | |
4277 | struct md_list *tmpdev; | |
4278 | int err = 0; | |
4279 | int i = 0; | |
9587c373 | 4280 | |
ec50f7b6 LM |
4281 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4282 | if (tmpdev->used != 1) | |
4283 | continue; | |
4284 | if (tmpdev->container == 1) { | |
ca9de185 | 4285 | int lmax = 0; |
ec50f7b6 LM |
4286 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4287 | if (fd < 0) { | |
e7b84f9d | 4288 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4289 | tmpdev->devname, strerror(errno)); |
4290 | err = 8; | |
4291 | goto error; | |
4292 | } | |
4293 | err = get_sra_super_block(fd, super_list, | |
4294 | tmpdev->devname, &lmax, | |
4295 | keep_fd); | |
4296 | i += lmax; | |
4297 | close(fd); | |
4298 | if (err) { | |
4299 | err = 7; | |
4300 | goto error; | |
4301 | } | |
4302 | } else { | |
4303 | int major = major(tmpdev->st_rdev); | |
4304 | int minor = minor(tmpdev->st_rdev); | |
4305 | err = get_super_block(super_list, | |
4306 | -1, | |
4307 | tmpdev->devname, | |
4308 | major, minor, | |
4309 | keep_fd); | |
4310 | i++; | |
4311 | if (err) { | |
4312 | err = 6; | |
4313 | goto error; | |
4314 | } | |
4315 | } | |
4316 | } | |
4317 | error: | |
4318 | *max = i; | |
4319 | return err; | |
4320 | } | |
9587c373 LM |
4321 | |
4322 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, | |
4323 | int major, int minor, int keep_fd) | |
4324 | { | |
4325 | struct intel_super*s = NULL; | |
4326 | char nm[32]; | |
4327 | int dfd = -1; | |
4328 | int rv; | |
4329 | int err = 0; | |
4330 | int retry; | |
4331 | ||
4332 | s = alloc_super(); | |
4333 | if (!s) { | |
4334 | err = 1; | |
4335 | goto error; | |
4336 | } | |
4337 | ||
4338 | sprintf(nm, "%d:%d", major, minor); | |
4339 | dfd = dev_open(nm, O_RDWR); | |
4340 | if (dfd < 0) { | |
4341 | err = 2; | |
4342 | goto error; | |
4343 | } | |
4344 | ||
4345 | rv = find_intel_hba_capability(dfd, s, devname); | |
4346 | /* no orom/efi or non-intel hba of the disk */ | |
4347 | if (rv != 0) { | |
4348 | err = 4; | |
4349 | goto error; | |
4350 | } | |
4351 | ||
4352 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4353 | ||
4354 | /* retry the load if we might have raced against mdmon */ | |
4355 | if (err == 3 && (devnum != -1) && mdmon_running(devnum)) | |
4356 | for (retry = 0; retry < 3; retry++) { | |
4357 | usleep(3000); | |
4358 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4359 | if (err != 3) | |
4360 | break; | |
4361 | } | |
4362 | error: | |
4363 | if (!err) { | |
4364 | s->next = *super_list; | |
4365 | *super_list = s; | |
4366 | } else { | |
4367 | if (s) | |
4368 | free(s); | |
36614e95 | 4369 | if (dfd >= 0) |
9587c373 LM |
4370 | close(dfd); |
4371 | } | |
4372 | if ((dfd >= 0) && (!keep_fd)) | |
4373 | close(dfd); | |
4374 | return err; | |
4375 | ||
4376 | } | |
4377 | ||
4378 | static int | |
4379 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4380 | { | |
4381 | struct mdinfo *sra; | |
4382 | int devnum; | |
4383 | struct mdinfo *sd; | |
4384 | int err = 0; | |
4385 | int i = 0; | |
4386 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
4387 | if (!sra) | |
4388 | return 1; | |
4389 | ||
4390 | if (sra->array.major_version != -1 || | |
4391 | sra->array.minor_version != -2 || | |
4392 | strcmp(sra->text_version, "imsm") != 0) { | |
4393 | err = 1; | |
4394 | goto error; | |
4395 | } | |
4396 | /* load all mpbs */ | |
4397 | devnum = fd2devnum(fd); | |
4398 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
4399 | if (get_super_block(super_list, devnum, devname, | |
4400 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { | |
4401 | err = 7; | |
4402 | goto error; | |
4403 | } | |
4404 | } | |
4405 | error: | |
4406 | sysfs_free(sra); | |
4407 | *max = i; | |
4408 | return err; | |
4409 | } | |
4410 | ||
2b959fbf N |
4411 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4412 | { | |
ec50f7b6 | 4413 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4414 | } |
cdddbdbc DW |
4415 | #endif |
4416 | ||
4417 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4418 | { | |
4419 | struct intel_super *super; | |
4420 | int rv; | |
4421 | ||
691c6ee1 N |
4422 | if (test_partition(fd)) |
4423 | /* IMSM not allowed on partitions */ | |
4424 | return 1; | |
4425 | ||
37424f13 DW |
4426 | free_super_imsm(st); |
4427 | ||
49133e57 | 4428 | super = alloc_super(); |
ea2bc72b LM |
4429 | /* Load hba and capabilities if they exist. |
4430 | * But do not preclude loading metadata in case capabilities or hba are | |
4431 | * non-compliant and ignore_hw_compat is set. | |
4432 | */ | |
d424212e | 4433 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4434 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4435 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 | 4436 | if (devname) |
e7b84f9d | 4437 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
4438 | free_imsm(super); |
4439 | return 2; | |
4440 | } | |
a2b97981 | 4441 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
4442 | |
4443 | if (rv) { | |
4444 | if (devname) | |
e7b84f9d N |
4445 | pr_err("Failed to load all information " |
4446 | "sections on %s\n", devname); | |
cdddbdbc DW |
4447 | free_imsm(super); |
4448 | return rv; | |
4449 | } | |
4450 | ||
4451 | st->sb = super; | |
4452 | if (st->ss == NULL) { | |
4453 | st->ss = &super_imsm; | |
4454 | st->minor_version = 0; | |
4455 | st->max_devs = IMSM_MAX_DEVICES; | |
4456 | } | |
8e59f3d8 AK |
4457 | |
4458 | /* load migration record */ | |
2e062e82 AK |
4459 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4460 | /* Check for unsupported migration features */ | |
4461 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 4462 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
4463 | if (devname) |
4464 | fprintf(stderr, " on %s\n", devname); | |
4465 | else | |
4466 | fprintf(stderr, " (IMSM).\n"); | |
4467 | return 3; | |
4468 | } | |
e2f41b2c AK |
4469 | } |
4470 | ||
cdddbdbc DW |
4471 | return 0; |
4472 | } | |
4473 | ||
ef6ffade DW |
4474 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4475 | { | |
4476 | if (info->level == 1) | |
4477 | return 128; | |
4478 | return info->chunk_size >> 9; | |
4479 | } | |
4480 | ||
5551b113 CA |
4481 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
4482 | unsigned long long size) | |
fcfd9599 | 4483 | { |
4025c288 | 4484 | if (info->level == 1) |
5551b113 | 4485 | return size * 2; |
4025c288 | 4486 | else |
5551b113 | 4487 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
4488 | } |
4489 | ||
4d1313e9 DW |
4490 | static void imsm_update_version_info(struct intel_super *super) |
4491 | { | |
4492 | /* update the version and attributes */ | |
4493 | struct imsm_super *mpb = super->anchor; | |
4494 | char *version; | |
4495 | struct imsm_dev *dev; | |
4496 | struct imsm_map *map; | |
4497 | int i; | |
4498 | ||
4499 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4500 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4501 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4502 | if (__le32_to_cpu(dev->size_high) > 0) |
4503 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4504 | ||
4505 | /* FIXME detect when an array spans a port multiplier */ | |
4506 | #if 0 | |
4507 | mpb->attributes |= MPB_ATTRIB_PM; | |
4508 | #endif | |
4509 | ||
4510 | if (mpb->num_raid_devs > 1 || | |
4511 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4512 | version = MPB_VERSION_ATTRIBS; | |
4513 | switch (get_imsm_raid_level(map)) { | |
4514 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4515 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4516 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4517 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4518 | } | |
4519 | } else { | |
4520 | if (map->num_members >= 5) | |
4521 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4522 | else if (dev->status == DEV_CLONE_N_GO) | |
4523 | version = MPB_VERSION_CNG; | |
4524 | else if (get_imsm_raid_level(map) == 5) | |
4525 | version = MPB_VERSION_RAID5; | |
4526 | else if (map->num_members >= 3) | |
4527 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4528 | else if (get_imsm_raid_level(map) == 1) | |
4529 | version = MPB_VERSION_RAID1; | |
4530 | else | |
4531 | version = MPB_VERSION_RAID0; | |
4532 | } | |
4533 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4534 | } | |
4535 | } | |
4536 | ||
aa534678 DW |
4537 | static int check_name(struct intel_super *super, char *name, int quiet) |
4538 | { | |
4539 | struct imsm_super *mpb = super->anchor; | |
4540 | char *reason = NULL; | |
4541 | int i; | |
4542 | ||
4543 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4544 | reason = "must be 16 characters or less"; | |
4545 | ||
4546 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4547 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4548 | ||
4549 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4550 | reason = "already exists"; | |
4551 | break; | |
4552 | } | |
4553 | } | |
4554 | ||
4555 | if (reason && !quiet) | |
e7b84f9d | 4556 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
4557 | |
4558 | return !reason; | |
4559 | } | |
4560 | ||
8b353278 DW |
4561 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4562 | unsigned long long size, char *name, | |
4563 | char *homehost, int *uuid) | |
cdddbdbc | 4564 | { |
c2c087e6 DW |
4565 | /* We are creating a volume inside a pre-existing container. |
4566 | * so st->sb is already set. | |
4567 | */ | |
4568 | struct intel_super *super = st->sb; | |
949c47a0 | 4569 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4570 | struct intel_dev *dv; |
c2c087e6 DW |
4571 | struct imsm_dev *dev; |
4572 | struct imsm_vol *vol; | |
4573 | struct imsm_map *map; | |
4574 | int idx = mpb->num_raid_devs; | |
4575 | int i; | |
4576 | unsigned long long array_blocks; | |
2c092cad | 4577 | size_t size_old, size_new; |
5551b113 | 4578 | unsigned long long num_data_stripes; |
cdddbdbc | 4579 | |
88c32bb1 | 4580 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
e7b84f9d | 4581 | pr_err("This imsm-container already has the " |
88c32bb1 | 4582 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4583 | return 0; |
4584 | } | |
4585 | ||
2c092cad DW |
4586 | /* ensure the mpb is large enough for the new data */ |
4587 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4588 | size_new = disks_to_mpb_size(info->nr_disks); | |
4589 | if (size_new > size_old) { | |
4590 | void *mpb_new; | |
4591 | size_t size_round = ROUND_UP(size_new, 512); | |
4592 | ||
4593 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
e7b84f9d | 4594 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
4595 | return 0; |
4596 | } | |
17a4eaf9 AK |
4597 | if (posix_memalign(&super->migr_rec_buf, 512, |
4598 | MIGR_REC_BUF_SIZE) != 0) { | |
e7b84f9d N |
4599 | pr_err("%s could not allocate migr_rec buffer\n", |
4600 | __func__); | |
8e59f3d8 AK |
4601 | free(super->buf); |
4602 | free(super); | |
ea944c8f | 4603 | free(mpb_new); |
8e59f3d8 AK |
4604 | return 0; |
4605 | } | |
2c092cad DW |
4606 | memcpy(mpb_new, mpb, size_old); |
4607 | free(mpb); | |
4608 | mpb = mpb_new; | |
949c47a0 | 4609 | super->anchor = mpb_new; |
2c092cad DW |
4610 | mpb->mpb_size = __cpu_to_le32(size_new); |
4611 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4612 | } | |
bf5a934a | 4613 | super->current_vol = idx; |
3960e579 DW |
4614 | |
4615 | /* handle 'failed_disks' by either: | |
4616 | * a) create dummy disk entries in the table if this the first | |
4617 | * volume in the array. We add them here as this is the only | |
4618 | * opportunity to add them. add_to_super_imsm_volume() | |
4619 | * handles the non-failed disks and continues incrementing | |
4620 | * mpb->num_disks. | |
4621 | * b) validate that 'failed_disks' matches the current number | |
4622 | * of missing disks if the container is populated | |
d23fe947 | 4623 | */ |
3960e579 | 4624 | if (super->current_vol == 0) { |
d23fe947 | 4625 | mpb->num_disks = 0; |
3960e579 DW |
4626 | for (i = 0; i < info->failed_disks; i++) { |
4627 | struct imsm_disk *disk; | |
4628 | ||
4629 | mpb->num_disks++; | |
4630 | disk = __get_imsm_disk(mpb, i); | |
4631 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4632 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4633 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4634 | "missing:%d", i); | |
4635 | } | |
4636 | find_missing(super); | |
4637 | } else { | |
4638 | int missing = 0; | |
4639 | struct dl *d; | |
4640 | ||
4641 | for (d = super->missing; d; d = d->next) | |
4642 | missing++; | |
4643 | if (info->failed_disks > missing) { | |
e7b84f9d | 4644 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
4645 | return 0; |
4646 | } | |
4647 | } | |
5a038140 | 4648 | |
aa534678 DW |
4649 | if (!check_name(super, name, 0)) |
4650 | return 0; | |
503975b9 N |
4651 | dv = xmalloc(sizeof(*dv)); |
4652 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 4653 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 4654 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 4655 | info->layout, info->chunk_size, |
5551b113 | 4656 | size * 2); |
979d38be DW |
4657 | /* round array size down to closest MB */ |
4658 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4659 | ||
c2c087e6 DW |
4660 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4661 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4662 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4663 | vol = &dev->vol; |
4664 | vol->migr_state = 0; | |
1484e727 | 4665 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4666 | vol->dirty = !info->state; |
f8f603f1 | 4667 | vol->curr_migr_unit = 0; |
238c0a71 | 4668 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
4669 | set_pba_of_lba0(map, super->create_offset); |
4670 | set_blocks_per_member(map, info_to_blocks_per_member(info, size)); | |
ef6ffade | 4671 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4672 | map->failed_disk_num = ~0; |
bf4442ab AK |
4673 | if (info->level > 0) |
4674 | map->map_state = IMSM_T_STATE_UNINITIALIZED; | |
4675 | else | |
4676 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4677 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4678 | map->ddf = 1; |
ef6ffade DW |
4679 | |
4680 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4681 | free(dev); |
4682 | free(dv); | |
e7b84f9d | 4683 | pr_err("imsm does not support more than 2 disks" |
ef6ffade DW |
4684 | "in a raid1 volume\n"); |
4685 | return 0; | |
4686 | } | |
81062a36 DW |
4687 | |
4688 | map->raid_level = info->level; | |
4d1313e9 | 4689 | if (info->level == 10) { |
c2c087e6 | 4690 | map->raid_level = 1; |
4d1313e9 | 4691 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4692 | } else if (info->level == 1) |
4693 | map->num_domains = info->raid_disks; | |
4694 | else | |
ff596308 | 4695 | map->num_domains = 1; |
81062a36 | 4696 | |
5551b113 CA |
4697 | /* info->size is only int so use the 'size' parameter instead */ |
4698 | num_data_stripes = (size * 2) / info_to_blocks_per_strip(info); | |
4699 | num_data_stripes /= map->num_domains; | |
4700 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 4701 | |
c2c087e6 DW |
4702 | map->num_members = info->raid_disks; |
4703 | for (i = 0; i < map->num_members; i++) { | |
4704 | /* initialized in add_to_super */ | |
4eb26970 | 4705 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4706 | } |
949c47a0 | 4707 | mpb->num_raid_devs++; |
ba2de7ba DW |
4708 | |
4709 | dv->dev = dev; | |
4710 | dv->index = super->current_vol; | |
4711 | dv->next = super->devlist; | |
4712 | super->devlist = dv; | |
c2c087e6 | 4713 | |
4d1313e9 DW |
4714 | imsm_update_version_info(super); |
4715 | ||
c2c087e6 | 4716 | return 1; |
cdddbdbc DW |
4717 | } |
4718 | ||
bf5a934a DW |
4719 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4720 | unsigned long long size, char *name, | |
4721 | char *homehost, int *uuid) | |
4722 | { | |
4723 | /* This is primarily called by Create when creating a new array. | |
4724 | * We will then get add_to_super called for each component, and then | |
4725 | * write_init_super called to write it out to each device. | |
4726 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4727 | * array. | |
4728 | * To create on a pre-existing array a different method will be called. | |
4729 | * This one is just for fresh drives. | |
4730 | */ | |
4731 | struct intel_super *super; | |
4732 | struct imsm_super *mpb; | |
4733 | size_t mpb_size; | |
4d1313e9 | 4734 | char *version; |
bf5a934a | 4735 | |
bf5a934a | 4736 | if (st->sb) |
e683ca88 DW |
4737 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
4738 | ||
4739 | if (info) | |
4740 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4741 | else | |
4742 | mpb_size = 512; | |
bf5a934a | 4743 | |
49133e57 | 4744 | super = alloc_super(); |
e683ca88 | 4745 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4746 | free(super); |
e683ca88 DW |
4747 | super = NULL; |
4748 | } | |
4749 | if (!super) { | |
e7b84f9d | 4750 | pr_err("%s could not allocate superblock\n", __func__); |
bf5a934a DW |
4751 | return 0; |
4752 | } | |
17a4eaf9 | 4753 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
e7b84f9d | 4754 | pr_err("%s could not allocate migr_rec buffer\n", __func__); |
8e59f3d8 AK |
4755 | free(super->buf); |
4756 | free(super); | |
4757 | return 0; | |
4758 | } | |
e683ca88 | 4759 | memset(super->buf, 0, mpb_size); |
ef649044 | 4760 | mpb = super->buf; |
e683ca88 DW |
4761 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4762 | st->sb = super; | |
4763 | ||
4764 | if (info == NULL) { | |
4765 | /* zeroing superblock */ | |
4766 | return 0; | |
4767 | } | |
bf5a934a | 4768 | |
4d1313e9 DW |
4769 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4770 | ||
4771 | version = (char *) mpb->sig; | |
4772 | strcpy(version, MPB_SIGNATURE); | |
4773 | version += strlen(MPB_SIGNATURE); | |
4774 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4775 | |
bf5a934a DW |
4776 | return 1; |
4777 | } | |
4778 | ||
0e600426 | 4779 | #ifndef MDASSEMBLE |
f20c3968 | 4780 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4781 | int fd, char *devname) |
4782 | { | |
4783 | struct intel_super *super = st->sb; | |
d23fe947 | 4784 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4785 | struct imsm_disk *_disk; |
bf5a934a DW |
4786 | struct imsm_dev *dev; |
4787 | struct imsm_map *map; | |
3960e579 | 4788 | struct dl *dl, *df; |
4eb26970 | 4789 | int slot; |
bf5a934a | 4790 | |
949c47a0 | 4791 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4792 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4793 | |
208933a7 | 4794 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 4795 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
4796 | devname); |
4797 | return 1; | |
4798 | } | |
4799 | ||
efb30e7f DW |
4800 | if (fd == -1) { |
4801 | /* we're doing autolayout so grab the pre-marked (in | |
4802 | * validate_geometry) raid_disk | |
4803 | */ | |
4804 | for (dl = super->disks; dl; dl = dl->next) | |
4805 | if (dl->raiddisk == dk->raid_disk) | |
4806 | break; | |
4807 | } else { | |
4808 | for (dl = super->disks; dl ; dl = dl->next) | |
4809 | if (dl->major == dk->major && | |
4810 | dl->minor == dk->minor) | |
4811 | break; | |
4812 | } | |
d23fe947 | 4813 | |
208933a7 | 4814 | if (!dl) { |
e7b84f9d | 4815 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 4816 | return 1; |
208933a7 | 4817 | } |
bf5a934a | 4818 | |
d23fe947 DW |
4819 | /* add a pristine spare to the metadata */ |
4820 | if (dl->index < 0) { | |
4821 | dl->index = super->anchor->num_disks; | |
4822 | super->anchor->num_disks++; | |
4823 | } | |
4eb26970 DW |
4824 | /* Check the device has not already been added */ |
4825 | slot = get_imsm_disk_slot(map, dl->index); | |
4826 | if (slot >= 0 && | |
238c0a71 | 4827 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 4828 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
4829 | devname); |
4830 | return 1; | |
4831 | } | |
656b6b5a | 4832 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4833 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4834 | |
3960e579 DW |
4835 | /* update size of 'missing' disks to be at least as large as the |
4836 | * largest acitve member (we only have dummy missing disks when | |
4837 | * creating the first volume) | |
4838 | */ | |
4839 | if (super->current_vol == 0) { | |
4840 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
4841 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
4842 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
4843 | _disk = __get_imsm_disk(mpb, df->index); |
4844 | *_disk = df->disk; | |
4845 | } | |
4846 | } | |
4847 | ||
4848 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
4849 | for (df = super->missing; df; df = df->next) | |
4850 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 4851 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
4852 | |
4853 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
4854 | continue; | |
4855 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 4856 | if (is_gen_migration(dev)) { |
238c0a71 AK |
4857 | struct imsm_map *map2 = get_imsm_map(dev, |
4858 | MAP_1); | |
0a108d63 AK |
4859 | int slot2 = get_imsm_disk_slot(map2, df->index); |
4860 | if ((slot2 < map2->num_members) && | |
4861 | (slot2 >= 0)) { | |
1ace8403 | 4862 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
4863 | slot2, |
4864 | MAP_1); | |
1ace8403 AK |
4865 | if ((unsigned)df->index == |
4866 | ord_to_idx(ord2)) | |
4867 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 4868 | slot2, |
1ace8403 AK |
4869 | df->index | |
4870 | IMSM_ORD_REBUILD); | |
4871 | } | |
4872 | } | |
3960e579 DW |
4873 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
4874 | break; | |
4875 | } | |
4876 | ||
d23fe947 DW |
4877 | /* if we are creating the first raid device update the family number */ |
4878 | if (super->current_vol == 0) { | |
4879 | __u32 sum; | |
4880 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 4881 | |
3960e579 | 4882 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 4883 | if (!_dev || !_disk) { |
e7b84f9d | 4884 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
4885 | return 1; |
4886 | } | |
d23fe947 DW |
4887 | *_dev = *dev; |
4888 | *_disk = dl->disk; | |
148acb7b DW |
4889 | sum = random32(); |
4890 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4891 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4892 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4893 | } |
ca0748fa | 4894 | super->current_disk = dl; |
f20c3968 | 4895 | return 0; |
bf5a934a DW |
4896 | } |
4897 | ||
a8619d23 AK |
4898 | /* mark_spare() |
4899 | * Function marks disk as spare and restores disk serial | |
4900 | * in case it was previously marked as failed by takeover operation | |
4901 | * reruns: | |
4902 | * -1 : critical error | |
4903 | * 0 : disk is marked as spare but serial is not set | |
4904 | * 1 : success | |
4905 | */ | |
4906 | int mark_spare(struct dl *disk) | |
4907 | { | |
4908 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
4909 | int ret_val = -1; | |
4910 | ||
4911 | if (!disk) | |
4912 | return ret_val; | |
4913 | ||
4914 | ret_val = 0; | |
4915 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
4916 | /* Restore disk serial number, because takeover marks disk | |
4917 | * as failed and adds to serial ':0' before it becomes | |
4918 | * a spare disk. | |
4919 | */ | |
4920 | serialcpy(disk->serial, serial); | |
4921 | serialcpy(disk->disk.serial, serial); | |
4922 | ret_val = 1; | |
4923 | } | |
4924 | disk->disk.status = SPARE_DISK; | |
4925 | disk->index = -1; | |
4926 | ||
4927 | return ret_val; | |
4928 | } | |
88654014 | 4929 | |
f20c3968 | 4930 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4931 | int fd, char *devname) |
cdddbdbc | 4932 | { |
c2c087e6 | 4933 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4934 | struct dl *dd; |
4935 | unsigned long long size; | |
f2f27e63 | 4936 | __u32 id; |
c2c087e6 DW |
4937 | int rv; |
4938 | struct stat stb; | |
4939 | ||
88654014 LM |
4940 | /* If we are on an RAID enabled platform check that the disk is |
4941 | * attached to the raid controller. | |
4942 | * We do not need to test disks attachment for container based additions, | |
4943 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4944 | */ |
d424212e | 4945 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4946 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4947 | if (rv != 0) { |
4948 | dprintf("capability: %p fd: %d ret: %d\n", | |
4949 | super->orom, fd, rv); | |
4950 | return 1; | |
88c32bb1 DW |
4951 | } |
4952 | ||
f20c3968 DW |
4953 | if (super->current_vol >= 0) |
4954 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4955 | |
c2c087e6 | 4956 | fstat(fd, &stb); |
503975b9 | 4957 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
4958 | dd->major = major(stb.st_rdev); |
4959 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 4960 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 4961 | dd->fd = fd; |
689c9bf3 | 4962 | dd->e = NULL; |
1a64be56 | 4963 | dd->action = DISK_ADD; |
c2c087e6 | 4964 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4965 | if (rv) { |
e7b84f9d | 4966 | pr_err("failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4967 | free(dd); |
0030e8d6 | 4968 | abort(); |
c2c087e6 DW |
4969 | } |
4970 | ||
c2c087e6 DW |
4971 | get_dev_size(fd, NULL, &size); |
4972 | size /= 512; | |
1f24f035 | 4973 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
4974 | set_total_blocks(&dd->disk, size); |
4975 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
4976 | struct imsm_super *mpb = super->anchor; | |
4977 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
4978 | } | |
a8619d23 | 4979 | mark_spare(dd); |
c2c087e6 | 4980 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4981 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4982 | else |
b9f594fe | 4983 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4984 | |
4985 | if (st->update_tail) { | |
1a64be56 LM |
4986 | dd->next = super->disk_mgmt_list; |
4987 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4988 | } else { |
4989 | dd->next = super->disks; | |
4990 | super->disks = dd; | |
ceaf0ee1 | 4991 | super->updates_pending++; |
43dad3d6 | 4992 | } |
f20c3968 DW |
4993 | |
4994 | return 0; | |
cdddbdbc DW |
4995 | } |
4996 | ||
1a64be56 LM |
4997 | |
4998 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4999 | { | |
5000 | struct intel_super *super = st->sb; | |
5001 | struct dl *dd; | |
5002 | ||
5003 | /* remove from super works only in mdmon - for communication | |
5004 | * manager - monitor. Check if communication memory buffer | |
5005 | * is prepared. | |
5006 | */ | |
5007 | if (!st->update_tail) { | |
e7b84f9d N |
5008 | pr_err("%s shall be used in mdmon context only" |
5009 | "(line %d).\n", __func__, __LINE__); | |
1a64be56 LM |
5010 | return 1; |
5011 | } | |
503975b9 | 5012 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5013 | dd->major = dk->major; |
5014 | dd->minor = dk->minor; | |
1a64be56 | 5015 | dd->fd = -1; |
a8619d23 | 5016 | mark_spare(dd); |
1a64be56 LM |
5017 | dd->action = DISK_REMOVE; |
5018 | ||
5019 | dd->next = super->disk_mgmt_list; | |
5020 | super->disk_mgmt_list = dd; | |
5021 | ||
5022 | ||
5023 | return 0; | |
5024 | } | |
5025 | ||
f796af5d DW |
5026 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5027 | ||
5028 | static union { | |
5029 | char buf[512]; | |
5030 | struct imsm_super anchor; | |
5031 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 5032 | |
d23fe947 DW |
5033 | /* spare records have their own family number and do not have any defined raid |
5034 | * devices | |
5035 | */ | |
5036 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5037 | { | |
d23fe947 | 5038 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5039 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5040 | __u32 sum; |
5041 | struct dl *d; | |
5042 | ||
f796af5d DW |
5043 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
5044 | spare->generation_num = __cpu_to_le32(1UL), | |
5045 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
5046 | spare->num_disks = 1, | |
5047 | spare->num_raid_devs = 0, | |
5048 | spare->cache_size = mpb->cache_size, | |
5049 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
5050 | ||
5051 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5052 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5053 | |
5054 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5055 | if (d->index != -1) |
d23fe947 DW |
5056 | continue; |
5057 | ||
f796af5d | 5058 | spare->disk[0] = d->disk; |
027c374f CA |
5059 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5060 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5061 | ||
f796af5d DW |
5062 | sum = __gen_imsm_checksum(spare); |
5063 | spare->family_num = __cpu_to_le32(sum); | |
5064 | spare->orig_family_num = 0; | |
5065 | sum = __gen_imsm_checksum(spare); | |
5066 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5067 | |
f796af5d | 5068 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
5069 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
5070 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 5071 | return 1; |
d23fe947 DW |
5072 | } |
5073 | if (doclose) { | |
5074 | close(d->fd); | |
5075 | d->fd = -1; | |
5076 | } | |
5077 | } | |
5078 | ||
e74255d9 | 5079 | return 0; |
d23fe947 DW |
5080 | } |
5081 | ||
36988a3d | 5082 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5083 | { |
36988a3d | 5084 | struct intel_super *super = st->sb; |
949c47a0 | 5085 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5086 | struct dl *d; |
5087 | __u32 generation; | |
5088 | __u32 sum; | |
d23fe947 | 5089 | int spares = 0; |
949c47a0 | 5090 | int i; |
a48ac0a8 | 5091 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5092 | int num_disks = 0; |
146c6260 | 5093 | int clear_migration_record = 1; |
cdddbdbc | 5094 | |
c2c087e6 DW |
5095 | /* 'generation' is incremented everytime the metadata is written */ |
5096 | generation = __le32_to_cpu(mpb->generation_num); | |
5097 | generation++; | |
5098 | mpb->generation_num = __cpu_to_le32(generation); | |
5099 | ||
148acb7b DW |
5100 | /* fix up cases where previous mdadm releases failed to set |
5101 | * orig_family_num | |
5102 | */ | |
5103 | if (mpb->orig_family_num == 0) | |
5104 | mpb->orig_family_num = mpb->family_num; | |
5105 | ||
d23fe947 | 5106 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5107 | if (d->index == -1) |
d23fe947 | 5108 | spares++; |
36988a3d | 5109 | else { |
d23fe947 | 5110 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5111 | num_disks++; |
5112 | } | |
d23fe947 | 5113 | } |
36988a3d | 5114 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5115 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5116 | num_disks++; |
5117 | } | |
5118 | mpb->num_disks = num_disks; | |
5119 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5120 | |
949c47a0 DW |
5121 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5122 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5123 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5124 | if (dev && dev2) { | |
5125 | imsm_copy_dev(dev, dev2); | |
5126 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5127 | } | |
146c6260 AK |
5128 | if (is_gen_migration(dev2)) |
5129 | clear_migration_record = 0; | |
949c47a0 | 5130 | } |
a48ac0a8 DW |
5131 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
5132 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 5133 | |
c2c087e6 | 5134 | /* recalculate checksum */ |
949c47a0 | 5135 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5136 | mpb->check_sum = __cpu_to_le32(sum); |
5137 | ||
51d83f5d AK |
5138 | if (super->clean_migration_record_by_mdmon) { |
5139 | clear_migration_record = 1; | |
5140 | super->clean_migration_record_by_mdmon = 0; | |
5141 | } | |
146c6260 | 5142 | if (clear_migration_record) |
17a4eaf9 | 5143 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); |
146c6260 | 5144 | |
d23fe947 | 5145 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5146 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5147 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5148 | continue; |
30602f53 | 5149 | |
146c6260 AK |
5150 | if (clear_migration_record) { |
5151 | unsigned long long dsize; | |
5152 | ||
5153 | get_dev_size(d->fd, NULL, &dsize); | |
5154 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
17a4eaf9 AK |
5155 | if (write(d->fd, super->migr_rec_buf, |
5156 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
9e2d750d | 5157 | perror("Write migr_rec failed"); |
146c6260 AK |
5158 | } |
5159 | } | |
51d83f5d AK |
5160 | |
5161 | if (store_imsm_mpb(d->fd, mpb)) | |
5162 | fprintf(stderr, | |
5163 | "%s: failed for device %d:%d (fd: %d)%s\n", | |
5164 | __func__, d->major, d->minor, | |
5165 | d->fd, strerror(errno)); | |
5166 | ||
c2c087e6 DW |
5167 | if (doclose) { |
5168 | close(d->fd); | |
5169 | d->fd = -1; | |
5170 | } | |
5171 | } | |
5172 | ||
d23fe947 DW |
5173 | if (spares) |
5174 | return write_super_imsm_spares(super, doclose); | |
5175 | ||
e74255d9 | 5176 | return 0; |
c2c087e6 DW |
5177 | } |
5178 | ||
0e600426 | 5179 | |
9b1fb677 | 5180 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5181 | { |
5182 | size_t len; | |
5183 | struct imsm_update_create_array *u; | |
5184 | struct intel_super *super = st->sb; | |
9b1fb677 | 5185 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5186 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5187 | struct disk_info *inf; |
5188 | struct imsm_disk *disk; | |
5189 | int i; | |
43dad3d6 | 5190 | |
54c2c1ea DW |
5191 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5192 | sizeof(*inf) * map->num_members; | |
503975b9 | 5193 | u = xmalloc(len); |
43dad3d6 | 5194 | u->type = update_create_array; |
9b1fb677 | 5195 | u->dev_idx = dev_idx; |
43dad3d6 | 5196 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5197 | inf = get_disk_info(u); |
5198 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5199 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5200 | |
54c2c1ea DW |
5201 | disk = get_imsm_disk(super, idx); |
5202 | serialcpy(inf[i].serial, disk->serial); | |
5203 | } | |
43dad3d6 DW |
5204 | append_metadata_update(st, u, len); |
5205 | ||
5206 | return 0; | |
5207 | } | |
5208 | ||
1a64be56 | 5209 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5210 | { |
5211 | struct intel_super *super = st->sb; | |
5212 | size_t len; | |
1a64be56 | 5213 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5214 | |
1a64be56 | 5215 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5216 | return 0; |
5217 | ||
5218 | len = sizeof(*u); | |
503975b9 | 5219 | u = xmalloc(len); |
1a64be56 | 5220 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5221 | append_metadata_update(st, u, len); |
5222 | ||
5223 | return 0; | |
5224 | } | |
5225 | ||
c2c087e6 DW |
5226 | static int write_init_super_imsm(struct supertype *st) |
5227 | { | |
9b1fb677 DW |
5228 | struct intel_super *super = st->sb; |
5229 | int current_vol = super->current_vol; | |
5230 | ||
5231 | /* we are done with current_vol reset it to point st at the container */ | |
5232 | super->current_vol = -1; | |
5233 | ||
8273f55e | 5234 | if (st->update_tail) { |
43dad3d6 DW |
5235 | /* queue the recently created array / added disk |
5236 | * as a metadata update */ | |
43dad3d6 | 5237 | int rv; |
8273f55e | 5238 | |
43dad3d6 | 5239 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5240 | if (current_vol < 0) { |
1a64be56 LM |
5241 | /* in the mgmt (add/remove) disk case we are running |
5242 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5243 | */ |
1a64be56 | 5244 | return mgmt_disk(st); |
43dad3d6 | 5245 | } else |
9b1fb677 | 5246 | rv = create_array(st, current_vol); |
8273f55e | 5247 | |
43dad3d6 | 5248 | return rv; |
d682f344 N |
5249 | } else { |
5250 | struct dl *d; | |
5251 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 5252 | Kill(d->devname, NULL, 0, -1, 1); |
36988a3d | 5253 | return write_super_imsm(st, 1); |
d682f344 | 5254 | } |
cdddbdbc | 5255 | } |
0e600426 | 5256 | #endif |
cdddbdbc | 5257 | |
e683ca88 | 5258 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5259 | { |
e683ca88 DW |
5260 | struct intel_super *super = st->sb; |
5261 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5262 | |
e683ca88 | 5263 | if (!mpb) |
ad97895e DW |
5264 | return 1; |
5265 | ||
1799c9e8 | 5266 | #ifndef MDASSEMBLE |
e683ca88 | 5267 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5268 | #else |
5269 | return 1; | |
5270 | #endif | |
cdddbdbc DW |
5271 | } |
5272 | ||
0e600426 N |
5273 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
5274 | { | |
5275 | return __le32_to_cpu(mpb->bbm_log_size); | |
5276 | } | |
5277 | ||
5278 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
5279 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5280 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 5281 | unsigned long long size, char *dev, |
2c514b71 NB |
5282 | unsigned long long *freesize, |
5283 | int verbose) | |
cdddbdbc | 5284 | { |
c2c087e6 DW |
5285 | int fd; |
5286 | unsigned long long ldsize; | |
f2f5c343 LM |
5287 | struct intel_super *super=NULL; |
5288 | int rv = 0; | |
cdddbdbc | 5289 | |
c2c087e6 DW |
5290 | if (level != LEVEL_CONTAINER) |
5291 | return 0; | |
5292 | if (!dev) | |
5293 | return 1; | |
5294 | ||
5295 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5296 | if (fd < 0) { | |
ba728be7 | 5297 | if (verbose > 0) |
e7b84f9d | 5298 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 5299 | dev, strerror(errno)); |
c2c087e6 DW |
5300 | return 0; |
5301 | } | |
5302 | if (!get_dev_size(fd, dev, &ldsize)) { | |
5303 | close(fd); | |
5304 | return 0; | |
5305 | } | |
f2f5c343 LM |
5306 | |
5307 | /* capabilities retrieve could be possible | |
5308 | * note that there is no fd for the disks in array. | |
5309 | */ | |
5310 | super = alloc_super(); | |
ba728be7 | 5311 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
5312 | if (rv != 0) { |
5313 | #if DEBUG | |
5314 | char str[256]; | |
5315 | fd2devname(fd, str); | |
5316 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
5317 | fd, str, super->orom, rv, raiddisks); | |
5318 | #endif | |
5319 | /* no orom/efi or non-intel hba of the disk */ | |
5320 | close(fd); | |
5321 | free_imsm(super); | |
5322 | return 0; | |
5323 | } | |
c2c087e6 | 5324 | close(fd); |
9126b9a8 CA |
5325 | if (super->orom) { |
5326 | if (raiddisks > super->orom->tds) { | |
5327 | if (verbose) | |
e7b84f9d | 5328 | pr_err("%d exceeds maximum number of" |
9126b9a8 CA |
5329 | " platform supported disks: %d\n", |
5330 | raiddisks, super->orom->tds); | |
5331 | free_imsm(super); | |
5332 | return 0; | |
5333 | } | |
5334 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
5335 | (ldsize >> 9) >> 32 > 0) { | |
5336 | if (verbose) | |
e7b84f9d | 5337 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
5338 | free_imsm(super); |
5339 | return 0; | |
5340 | } | |
f2f5c343 | 5341 | } |
c2c087e6 DW |
5342 | |
5343 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 5344 | free_imsm(super); |
c2c087e6 DW |
5345 | |
5346 | return 1; | |
cdddbdbc DW |
5347 | } |
5348 | ||
0dcecb2e DW |
5349 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5350 | { | |
5351 | const unsigned long long base_start = e[*idx].start; | |
5352 | unsigned long long end = base_start + e[*idx].size; | |
5353 | int i; | |
5354 | ||
5355 | if (base_start == end) | |
5356 | return 0; | |
5357 | ||
5358 | *idx = *idx + 1; | |
5359 | for (i = *idx; i < num_extents; i++) { | |
5360 | /* extend overlapping extents */ | |
5361 | if (e[i].start >= base_start && | |
5362 | e[i].start <= end) { | |
5363 | if (e[i].size == 0) | |
5364 | return 0; | |
5365 | if (e[i].start + e[i].size > end) | |
5366 | end = e[i].start + e[i].size; | |
5367 | } else if (e[i].start > end) { | |
5368 | *idx = i; | |
5369 | break; | |
5370 | } | |
5371 | } | |
5372 | ||
5373 | return end - base_start; | |
5374 | } | |
5375 | ||
5376 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5377 | { | |
5378 | /* build a composite disk with all known extents and generate a new | |
5379 | * 'maxsize' given the "all disks in an array must share a common start | |
5380 | * offset" constraint | |
5381 | */ | |
503975b9 | 5382 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
5383 | struct dl *dl; |
5384 | int i, j; | |
5385 | int start_extent; | |
5386 | unsigned long long pos; | |
b9d77223 | 5387 | unsigned long long start = 0; |
0dcecb2e DW |
5388 | unsigned long long maxsize; |
5389 | unsigned long reserve; | |
5390 | ||
0dcecb2e DW |
5391 | /* coalesce and sort all extents. also, check to see if we need to |
5392 | * reserve space between member arrays | |
5393 | */ | |
5394 | j = 0; | |
5395 | for (dl = super->disks; dl; dl = dl->next) { | |
5396 | if (!dl->e) | |
5397 | continue; | |
5398 | for (i = 0; i < dl->extent_cnt; i++) | |
5399 | e[j++] = dl->e[i]; | |
5400 | } | |
5401 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5402 | ||
5403 | /* merge extents */ | |
5404 | i = 0; | |
5405 | j = 0; | |
5406 | while (i < sum_extents) { | |
5407 | e[j].start = e[i].start; | |
5408 | e[j].size = find_size(e, &i, sum_extents); | |
5409 | j++; | |
5410 | if (e[j-1].size == 0) | |
5411 | break; | |
5412 | } | |
5413 | ||
5414 | pos = 0; | |
5415 | maxsize = 0; | |
5416 | start_extent = 0; | |
5417 | i = 0; | |
5418 | do { | |
5419 | unsigned long long esize; | |
5420 | ||
5421 | esize = e[i].start - pos; | |
5422 | if (esize >= maxsize) { | |
5423 | maxsize = esize; | |
5424 | start = pos; | |
5425 | start_extent = i; | |
5426 | } | |
5427 | pos = e[i].start + e[i].size; | |
5428 | i++; | |
5429 | } while (e[i-1].size); | |
5430 | free(e); | |
5431 | ||
a7dd165b DW |
5432 | if (maxsize == 0) |
5433 | return 0; | |
5434 | ||
5435 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5436 | * container | |
5437 | */ | |
0dcecb2e DW |
5438 | if (start_extent > 0) |
5439 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5440 | else | |
5441 | reserve = 0; | |
5442 | ||
5443 | if (maxsize < reserve) | |
a7dd165b | 5444 | return 0; |
0dcecb2e | 5445 | |
5551b113 | 5446 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 5447 | if (start + reserve > super->create_offset) |
a7dd165b | 5448 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5449 | super->create_offset = start + reserve; |
5450 | ||
5451 | return maxsize - reserve; | |
5452 | } | |
5453 | ||
88c32bb1 DW |
5454 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5455 | { | |
5456 | if (level < 0 || level == 6 || level == 4) | |
5457 | return 0; | |
5458 | ||
5459 | /* if we have an orom prevent invalid raid levels */ | |
5460 | if (orom) | |
5461 | switch (level) { | |
5462 | case 0: return imsm_orom_has_raid0(orom); | |
5463 | case 1: | |
5464 | if (raiddisks > 2) | |
5465 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5466 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5467 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5468 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5469 | } |
5470 | else | |
5471 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5472 | ||
5473 | return 0; | |
5474 | } | |
5475 | ||
ca9de185 LM |
5476 | |
5477 | static int | |
5478 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
5479 | int dpa, int verbose) | |
5480 | { | |
5481 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
5482 | struct mdstat_ent *memb = NULL; | |
5483 | int count = 0; | |
5484 | int num = 0; | |
5485 | struct md_list *dv = NULL; | |
5486 | int found; | |
5487 | ||
5488 | for (memb = mdstat ; memb ; memb = memb->next) { | |
5489 | if (memb->metadata_version && | |
5490 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
5491 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
5492 | !is_subarray(memb->metadata_version+9) && | |
5493 | memb->members) { | |
5494 | struct dev_member *dev = memb->members; | |
5495 | int fd = -1; | |
5496 | while(dev && (fd < 0)) { | |
503975b9 N |
5497 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
5498 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
5499 | if (num > 0) | |
5500 | fd = open(path, O_RDONLY, 0); | |
5501 | if ((num <= 0) || (fd < 0)) { | |
5502 | pr_vrb(": Cannot open %s: %s\n", | |
5503 | dev->name, strerror(errno)); | |
ca9de185 | 5504 | } |
503975b9 | 5505 | free(path); |
ca9de185 LM |
5506 | dev = dev->next; |
5507 | } | |
5508 | found = 0; | |
5509 | if ((fd >= 0) && disk_attached_to_hba(fd, hba)) { | |
5510 | struct mdstat_ent *vol; | |
5511 | for (vol = mdstat ; vol ; vol = vol->next) { | |
5512 | if ((vol->active > 0) && | |
5513 | vol->metadata_version && | |
5514 | is_container_member(vol, memb->dev)) { | |
5515 | found++; | |
5516 | count++; | |
5517 | } | |
5518 | } | |
5519 | if (*devlist && (found < dpa)) { | |
503975b9 N |
5520 | dv = xcalloc(1, sizeof(*dv)); |
5521 | dv->devname = xmalloc(strlen(memb->dev) + strlen("/dev/") + 1); | |
5522 | sprintf(dv->devname, "%s%s", "/dev/", memb->dev); | |
5523 | dv->found = found; | |
5524 | dv->used = 0; | |
5525 | dv->next = *devlist; | |
5526 | *devlist = dv; | |
ca9de185 LM |
5527 | } |
5528 | } | |
5529 | if (fd >= 0) | |
5530 | close(fd); | |
5531 | } | |
5532 | } | |
5533 | free_mdstat(mdstat); | |
5534 | return count; | |
5535 | } | |
5536 | ||
5537 | #ifdef DEBUG_LOOP | |
5538 | static struct md_list* | |
5539 | get_loop_devices(void) | |
5540 | { | |
5541 | int i; | |
5542 | struct md_list *devlist = NULL; | |
5543 | struct md_list *dv = NULL; | |
5544 | ||
5545 | for(i = 0; i < 12; i++) { | |
503975b9 N |
5546 | dv = xcalloc(1, sizeof(*dv)); |
5547 | dv->devname = xmalloc(40); | |
ca9de185 LM |
5548 | sprintf(dv->devname, "/dev/loop%d", i); |
5549 | dv->next = devlist; | |
5550 | devlist = dv; | |
5551 | } | |
5552 | return devlist; | |
5553 | } | |
5554 | #endif | |
5555 | ||
5556 | static struct md_list* | |
5557 | get_devices(const char *hba_path) | |
5558 | { | |
5559 | struct md_list *devlist = NULL; | |
5560 | struct md_list *dv = NULL; | |
5561 | struct dirent *ent; | |
5562 | DIR *dir; | |
5563 | int err = 0; | |
5564 | ||
5565 | #if DEBUG_LOOP | |
5566 | devlist = get_loop_devices(); | |
5567 | return devlist; | |
5568 | #endif | |
5569 | /* scroll through /sys/dev/block looking for devices attached to | |
5570 | * this hba | |
5571 | */ | |
5572 | dir = opendir("/sys/dev/block"); | |
5573 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
5574 | int fd; | |
5575 | char buf[1024]; | |
5576 | int major, minor; | |
5577 | char *path = NULL; | |
5578 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
5579 | continue; | |
5580 | path = devt_to_devpath(makedev(major, minor)); | |
5581 | if (!path) | |
5582 | continue; | |
5583 | if (!path_attached_to_hba(path, hba_path)) { | |
5584 | free(path); | |
5585 | path = NULL; | |
5586 | continue; | |
5587 | } | |
5588 | free(path); | |
5589 | path = NULL; | |
5590 | fd = dev_open(ent->d_name, O_RDONLY); | |
5591 | if (fd >= 0) { | |
5592 | fd2devname(fd, buf); | |
5593 | close(fd); | |
5594 | } else { | |
e7b84f9d | 5595 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
5596 | ent->d_name); |
5597 | continue; | |
5598 | } | |
5599 | ||
5600 | ||
503975b9 N |
5601 | dv = xcalloc(1, sizeof(*dv)); |
5602 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
5603 | dv->next = devlist; |
5604 | devlist = dv; | |
5605 | } | |
5606 | if (err) { | |
5607 | while(devlist) { | |
5608 | dv = devlist; | |
5609 | devlist = devlist->next; | |
5610 | free(dv->devname); | |
5611 | free(dv); | |
5612 | } | |
5613 | } | |
562aa102 | 5614 | closedir(dir); |
ca9de185 LM |
5615 | return devlist; |
5616 | } | |
5617 | ||
5618 | static int | |
5619 | count_volumes_list(struct md_list *devlist, char *homehost, | |
5620 | int verbose, int *found) | |
5621 | { | |
5622 | struct md_list *tmpdev; | |
5623 | int count = 0; | |
5624 | struct supertype *st = NULL; | |
5625 | ||
5626 | /* first walk the list of devices to find a consistent set | |
5627 | * that match the criterea, if that is possible. | |
5628 | * We flag the ones we like with 'used'. | |
5629 | */ | |
5630 | *found = 0; | |
5631 | st = match_metadata_desc_imsm("imsm"); | |
5632 | if (st == NULL) { | |
5633 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5634 | return 0; | |
5635 | } | |
5636 | ||
5637 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5638 | char *devname = tmpdev->devname; | |
5639 | struct stat stb; | |
5640 | struct supertype *tst; | |
5641 | int dfd; | |
5642 | if (tmpdev->used > 1) | |
5643 | continue; | |
5644 | tst = dup_super(st); | |
5645 | if (tst == NULL) { | |
5646 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5647 | goto err_1; | |
5648 | } | |
5649 | tmpdev->container = 0; | |
5650 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
5651 | if (dfd < 0) { | |
5652 | dprintf(": cannot open device %s: %s\n", | |
5653 | devname, strerror(errno)); | |
5654 | tmpdev->used = 2; | |
5655 | } else if (fstat(dfd, &stb)< 0) { | |
5656 | /* Impossible! */ | |
5657 | dprintf(": fstat failed for %s: %s\n", | |
5658 | devname, strerror(errno)); | |
5659 | tmpdev->used = 2; | |
5660 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
5661 | dprintf(": %s is not a block device.\n", | |
5662 | devname); | |
5663 | tmpdev->used = 2; | |
5664 | } else if (must_be_container(dfd)) { | |
5665 | struct supertype *cst; | |
5666 | cst = super_by_fd(dfd, NULL); | |
5667 | if (cst == NULL) { | |
5668 | dprintf(": cannot recognize container type %s\n", | |
5669 | devname); | |
5670 | tmpdev->used = 2; | |
5671 | } else if (tst->ss != st->ss) { | |
5672 | dprintf(": non-imsm container - ignore it: %s\n", | |
5673 | devname); | |
5674 | tmpdev->used = 2; | |
5675 | } else if (!tst->ss->load_container || | |
5676 | tst->ss->load_container(tst, dfd, NULL)) | |
5677 | tmpdev->used = 2; | |
5678 | else { | |
5679 | tmpdev->container = 1; | |
5680 | } | |
5681 | if (cst) | |
5682 | cst->ss->free_super(cst); | |
5683 | } else { | |
5684 | tmpdev->st_rdev = stb.st_rdev; | |
5685 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
5686 | dprintf(": no RAID superblock on %s\n", | |
5687 | devname); | |
5688 | tmpdev->used = 2; | |
5689 | } else if (tst->ss->compare_super == NULL) { | |
5690 | dprintf(": Cannot assemble %s metadata on %s\n", | |
5691 | tst->ss->name, devname); | |
5692 | tmpdev->used = 2; | |
5693 | } | |
5694 | } | |
5695 | if (dfd >= 0) | |
5696 | close(dfd); | |
5697 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
5698 | /* Ignore unrecognised devices during auto-assembly */ | |
5699 | goto loop; | |
5700 | } | |
5701 | else { | |
5702 | struct mdinfo info; | |
5703 | tst->ss->getinfo_super(tst, &info, NULL); | |
5704 | ||
5705 | if (st->minor_version == -1) | |
5706 | st->minor_version = tst->minor_version; | |
5707 | ||
5708 | if (memcmp(info.uuid, uuid_zero, | |
5709 | sizeof(int[4])) == 0) { | |
5710 | /* this is a floating spare. It cannot define | |
5711 | * an array unless there are no more arrays of | |
5712 | * this type to be found. It can be included | |
5713 | * in an array of this type though. | |
5714 | */ | |
5715 | tmpdev->used = 3; | |
5716 | goto loop; | |
5717 | } | |
5718 | ||
5719 | if (st->ss != tst->ss || | |
5720 | st->minor_version != tst->minor_version || | |
5721 | st->ss->compare_super(st, tst) != 0) { | |
5722 | /* Some mismatch. If exactly one array matches this host, | |
5723 | * we can resolve on that one. | |
5724 | * Or, if we are auto assembling, we just ignore the second | |
5725 | * for now. | |
5726 | */ | |
5727 | dprintf(": superblock on %s doesn't match others - assembly aborted\n", | |
5728 | devname); | |
5729 | goto loop; | |
5730 | } | |
5731 | tmpdev->used = 1; | |
5732 | *found = 1; | |
5733 | dprintf("found: devname: %s\n", devname); | |
5734 | } | |
5735 | loop: | |
5736 | if (tst) | |
5737 | tst->ss->free_super(tst); | |
5738 | } | |
5739 | if (*found != 0) { | |
5740 | int err; | |
5741 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
5742 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
5743 | for (iter = head; iter; iter = iter->next) { | |
5744 | dprintf("content->text_version: %s vol\n", | |
5745 | iter->text_version); | |
5746 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
5747 | /* do not assemble arrays with unsupported | |
5748 | configurations */ | |
5749 | dprintf(": Cannot activate member %s.\n", | |
5750 | iter->text_version); | |
5751 | } else | |
5752 | count++; | |
5753 | } | |
5754 | sysfs_free(head); | |
5755 | ||
5756 | } else { | |
5757 | dprintf(" no valid super block on device list: err: %d %p\n", | |
5758 | err, st->sb); | |
5759 | } | |
5760 | } else { | |
5761 | dprintf(" no more devices to examin\n"); | |
5762 | } | |
5763 | ||
5764 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5765 | if ((tmpdev->used == 1) && (tmpdev->found)) { | |
5766 | if (count) { | |
5767 | if (count < tmpdev->found) | |
5768 | count = 0; | |
5769 | else | |
5770 | count -= tmpdev->found; | |
5771 | } | |
5772 | } | |
5773 | if (tmpdev->used == 1) | |
5774 | tmpdev->used = 4; | |
5775 | } | |
5776 | err_1: | |
5777 | if (st) | |
5778 | st->ss->free_super(st); | |
5779 | return count; | |
5780 | } | |
5781 | ||
5782 | ||
5783 | static int | |
5784 | count_volumes(char *hba, int dpa, int verbose) | |
5785 | { | |
5786 | struct md_list *devlist = NULL; | |
5787 | int count = 0; | |
5788 | int found = 0;; | |
5789 | ||
5790 | devlist = get_devices(hba); | |
5791 | /* if no intel devices return zero volumes */ | |
5792 | if (devlist == NULL) | |
5793 | return 0; | |
5794 | ||
5795 | count = active_arrays_by_format("imsm", hba, &devlist, dpa, verbose); | |
5796 | dprintf(" path: %s active arrays: %d\n", hba, count); | |
5797 | if (devlist == NULL) | |
5798 | return 0; | |
5799 | do { | |
5800 | found = 0; | |
5801 | count += count_volumes_list(devlist, | |
5802 | NULL, | |
5803 | verbose, | |
5804 | &found); | |
5805 | dprintf("found %d count: %d\n", found, count); | |
5806 | } while (found); | |
5807 | ||
5808 | dprintf("path: %s total number of volumes: %d\n", hba, count); | |
5809 | ||
5810 | while(devlist) { | |
5811 | struct md_list *dv = devlist; | |
5812 | devlist = devlist->next; | |
5813 | free(dv->devname); | |
5814 | free(dv); | |
5815 | } | |
5816 | return count; | |
5817 | } | |
5818 | ||
cd9d1ac7 DW |
5819 | static int imsm_default_chunk(const struct imsm_orom *orom) |
5820 | { | |
5821 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
5822 | * 128 if no platform detected | |
5823 | */ | |
5824 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
5825 | ||
5826 | return min(512, (1 << fs)); | |
5827 | } | |
73408129 | 5828 | |
6592ce37 DW |
5829 | static int |
5830 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 5831 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 5832 | { |
660260d0 DW |
5833 | /* check/set platform and metadata limits/defaults */ |
5834 | if (super->orom && raiddisks > super->orom->dpa) { | |
5835 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
5836 | super->orom->dpa); | |
73408129 LM |
5837 | return 0; |
5838 | } | |
5839 | ||
5840 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
660260d0 | 5841 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
5842 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
5843 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
5844 | return 0; | |
5845 | } | |
cd9d1ac7 DW |
5846 | |
5847 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
5848 | *chunk = imsm_default_chunk(super->orom); | |
5849 | ||
5850 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
5851 | pr_vrb(": platform does not support a chunk size of: " | |
5852 | "%d\n", *chunk); | |
5853 | return 0; | |
6592ce37 | 5854 | } |
cd9d1ac7 | 5855 | |
6592ce37 DW |
5856 | if (layout != imsm_level_to_layout(level)) { |
5857 | if (level == 5) | |
5858 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
5859 | else if (level == 10) | |
5860 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
5861 | else | |
5862 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
5863 | layout, level); | |
5864 | return 0; | |
5865 | } | |
2cc699af CA |
5866 | |
5867 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && chunk && | |
5868 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { | |
5869 | pr_vrb(": platform does not support a volume size over 2TB\n"); | |
5870 | return 0; | |
5871 | } | |
6592ce37 DW |
5872 | return 1; |
5873 | } | |
5874 | ||
c2c087e6 DW |
5875 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
5876 | * FIX ME add ahci details | |
5877 | */ | |
8b353278 | 5878 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 5879 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 5880 | unsigned long long size, char *dev, |
2c514b71 NB |
5881 | unsigned long long *freesize, |
5882 | int verbose) | |
cdddbdbc | 5883 | { |
c2c087e6 DW |
5884 | struct stat stb; |
5885 | struct intel_super *super = st->sb; | |
b2916f25 | 5886 | struct imsm_super *mpb; |
c2c087e6 DW |
5887 | struct dl *dl; |
5888 | unsigned long long pos = 0; | |
5889 | unsigned long long maxsize; | |
5890 | struct extent *e; | |
5891 | int i; | |
cdddbdbc | 5892 | |
88c32bb1 DW |
5893 | /* We must have the container info already read in. */ |
5894 | if (!super) | |
c2c087e6 DW |
5895 | return 0; |
5896 | ||
b2916f25 JS |
5897 | mpb = super->anchor; |
5898 | ||
2cc699af | 5899 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
e7b84f9d | 5900 | pr_err("RAID gemetry validation failed. " |
d54559f0 | 5901 | "Cannot proceed with the action(s).\n"); |
c2c087e6 | 5902 | return 0; |
d54559f0 | 5903 | } |
c2c087e6 DW |
5904 | if (!dev) { |
5905 | /* General test: make sure there is space for | |
2da8544a DW |
5906 | * 'raiddisks' device extents of size 'size' at a given |
5907 | * offset | |
c2c087e6 | 5908 | */ |
e46273eb | 5909 | unsigned long long minsize = size; |
b7528a20 | 5910 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5911 | int dcnt = 0; |
5912 | if (minsize == 0) | |
5913 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5914 | for (dl = super->disks; dl ; dl = dl->next) { | |
5915 | int found = 0; | |
5916 | ||
bf5a934a | 5917 | pos = 0; |
c2c087e6 DW |
5918 | i = 0; |
5919 | e = get_extents(super, dl); | |
5920 | if (!e) continue; | |
5921 | do { | |
5922 | unsigned long long esize; | |
5923 | esize = e[i].start - pos; | |
5924 | if (esize >= minsize) | |
5925 | found = 1; | |
b7528a20 | 5926 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5927 | start_offset = pos; |
5928 | break; | |
5929 | } else if (found && pos != start_offset) { | |
5930 | found = 0; | |
5931 | break; | |
5932 | } | |
c2c087e6 DW |
5933 | pos = e[i].start + e[i].size; |
5934 | i++; | |
5935 | } while (e[i-1].size); | |
5936 | if (found) | |
5937 | dcnt++; | |
5938 | free(e); | |
5939 | } | |
5940 | if (dcnt < raiddisks) { | |
2c514b71 | 5941 | if (verbose) |
e7b84f9d | 5942 | pr_err("imsm: Not enough " |
2c514b71 NB |
5943 | "devices with space for this array " |
5944 | "(%d < %d)\n", | |
5945 | dcnt, raiddisks); | |
c2c087e6 DW |
5946 | return 0; |
5947 | } | |
5948 | return 1; | |
5949 | } | |
0dcecb2e | 5950 | |
c2c087e6 DW |
5951 | /* This device must be a member of the set */ |
5952 | if (stat(dev, &stb) < 0) | |
5953 | return 0; | |
5954 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
5955 | return 0; | |
5956 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
5957 | if (dl->major == (int)major(stb.st_rdev) && |
5958 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
5959 | break; |
5960 | } | |
5961 | if (!dl) { | |
2c514b71 | 5962 | if (verbose) |
e7b84f9d | 5963 | pr_err("%s is not in the " |
2c514b71 | 5964 | "same imsm set\n", dev); |
c2c087e6 | 5965 | return 0; |
a20d2ba5 DW |
5966 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
5967 | /* If a volume is present then the current creation attempt | |
5968 | * cannot incorporate new spares because the orom may not | |
5969 | * understand this configuration (all member disks must be | |
5970 | * members of each array in the container). | |
5971 | */ | |
e7b84f9d | 5972 | pr_err("%s is a spare and a volume" |
a20d2ba5 | 5973 | " is already defined for this container\n", dev); |
e7b84f9d | 5974 | pr_err("The option-rom requires all member" |
a20d2ba5 DW |
5975 | " disks to be a member of all volumes\n"); |
5976 | return 0; | |
5fe62b94 WD |
5977 | } else if (super->orom && mpb->num_raid_devs > 0 && |
5978 | mpb->num_disks != raiddisks) { | |
e7b84f9d | 5979 | pr_err("The option-rom requires all member" |
5fe62b94 WD |
5980 | " disks to be a member of all volumes\n"); |
5981 | return 0; | |
c2c087e6 | 5982 | } |
0dcecb2e DW |
5983 | |
5984 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
5985 | e = get_extents(super, dl); |
5986 | maxsize = 0; | |
5987 | i = 0; | |
0dcecb2e DW |
5988 | if (e) { |
5989 | do { | |
5990 | unsigned long long esize; | |
5991 | ||
5992 | esize = e[i].start - pos; | |
5993 | if (esize >= maxsize) | |
5994 | maxsize = esize; | |
5995 | pos = e[i].start + e[i].size; | |
5996 | i++; | |
5997 | } while (e[i-1].size); | |
5998 | dl->e = e; | |
5999 | dl->extent_cnt = i; | |
6000 | } else { | |
6001 | if (verbose) | |
e7b84f9d | 6002 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
6003 | dev); |
6004 | return 0; | |
6005 | } | |
6006 | if (maxsize < size) { | |
6007 | if (verbose) | |
e7b84f9d | 6008 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
6009 | dev, maxsize, size); |
6010 | return 0; | |
6011 | } | |
6012 | ||
6013 | /* count total number of extents for merge */ | |
6014 | i = 0; | |
6015 | for (dl = super->disks; dl; dl = dl->next) | |
6016 | if (dl->e) | |
6017 | i += dl->extent_cnt; | |
6018 | ||
6019 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
6020 | |
6021 | if (!check_env("IMSM_NO_PLATFORM") && | |
6022 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
e7b84f9d | 6023 | pr_err("attempting to create a second " |
3baa56ab LO |
6024 | "volume with size less then remaining space. " |
6025 | "Aborting...\n"); | |
6026 | return 0; | |
6027 | } | |
6028 | ||
a7dd165b | 6029 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
6030 | if (verbose) { |
6031 | if (maxsize == 0) | |
e7b84f9d | 6032 | pr_err("no free space" |
b3071342 LD |
6033 | " left on device. Aborting...\n"); |
6034 | else | |
e7b84f9d | 6035 | pr_err("not enough space" |
b3071342 LD |
6036 | " to create volume of given size" |
6037 | " (%llu < %llu). Aborting...\n", | |
6038 | maxsize, size); | |
6039 | } | |
0dcecb2e | 6040 | return 0; |
0dcecb2e DW |
6041 | } |
6042 | ||
c2c087e6 DW |
6043 | *freesize = maxsize; |
6044 | ||
ca9de185 LM |
6045 | if (super->orom) { |
6046 | int count = count_volumes(super->hba->path, | |
6047 | super->orom->dpa, verbose); | |
6048 | if (super->orom->vphba <= count) { | |
40110b90 | 6049 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6050 | super->orom->vphba); |
6051 | return 0; | |
6052 | } | |
6053 | } | |
c2c087e6 | 6054 | return 1; |
cdddbdbc DW |
6055 | } |
6056 | ||
13bcac90 | 6057 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
6058 | unsigned long long size, int chunk, |
6059 | unsigned long long *freesize) | |
6060 | { | |
6061 | struct intel_super *super = st->sb; | |
6062 | struct imsm_super *mpb = super->anchor; | |
6063 | struct dl *dl; | |
6064 | int i; | |
6065 | int extent_cnt; | |
6066 | struct extent *e; | |
6067 | unsigned long long maxsize; | |
6068 | unsigned long long minsize; | |
6069 | int cnt; | |
6070 | int used; | |
6071 | ||
6072 | /* find the largest common start free region of the possible disks */ | |
6073 | used = 0; | |
6074 | extent_cnt = 0; | |
6075 | cnt = 0; | |
6076 | for (dl = super->disks; dl; dl = dl->next) { | |
6077 | dl->raiddisk = -1; | |
6078 | ||
6079 | if (dl->index >= 0) | |
6080 | used++; | |
6081 | ||
6082 | /* don't activate new spares if we are orom constrained | |
6083 | * and there is already a volume active in the container | |
6084 | */ | |
6085 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
6086 | continue; | |
6087 | ||
6088 | e = get_extents(super, dl); | |
6089 | if (!e) | |
6090 | continue; | |
6091 | for (i = 1; e[i-1].size; i++) | |
6092 | ; | |
6093 | dl->e = e; | |
6094 | dl->extent_cnt = i; | |
6095 | extent_cnt += i; | |
6096 | cnt++; | |
6097 | } | |
6098 | ||
6099 | maxsize = merge_extents(super, extent_cnt); | |
6100 | minsize = size; | |
6101 | if (size == 0) | |
612e59d8 CA |
6102 | /* chunk is in K */ |
6103 | minsize = chunk * 2; | |
efb30e7f DW |
6104 | |
6105 | if (cnt < raiddisks || | |
6106 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6107 | maxsize < minsize || |
6108 | maxsize == 0) { | |
e7b84f9d | 6109 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
6110 | return 0; /* No enough free spaces large enough */ |
6111 | } | |
6112 | ||
6113 | if (size == 0) { | |
6114 | size = maxsize; | |
6115 | if (chunk) { | |
612e59d8 CA |
6116 | size /= 2 * chunk; |
6117 | size *= 2 * chunk; | |
efb30e7f | 6118 | } |
f878b242 LM |
6119 | maxsize = size; |
6120 | } | |
6121 | if (!check_env("IMSM_NO_PLATFORM") && | |
6122 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
e7b84f9d | 6123 | pr_err("attempting to create a second " |
f878b242 LM |
6124 | "volume with size less then remaining space. " |
6125 | "Aborting...\n"); | |
6126 | return 0; | |
efb30e7f | 6127 | } |
efb30e7f DW |
6128 | cnt = 0; |
6129 | for (dl = super->disks; dl; dl = dl->next) | |
6130 | if (dl->e) | |
6131 | dl->raiddisk = cnt++; | |
6132 | ||
6133 | *freesize = size; | |
6134 | ||
13bcac90 AK |
6135 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
6136 | ||
efb30e7f DW |
6137 | return 1; |
6138 | } | |
6139 | ||
13bcac90 AK |
6140 | static int reserve_space(struct supertype *st, int raiddisks, |
6141 | unsigned long long size, int chunk, | |
6142 | unsigned long long *freesize) | |
6143 | { | |
6144 | struct intel_super *super = st->sb; | |
6145 | struct dl *dl; | |
6146 | int cnt; | |
6147 | int rv = 0; | |
6148 | ||
6149 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
6150 | if (rv) { | |
6151 | cnt = 0; | |
6152 | for (dl = super->disks; dl; dl = dl->next) | |
6153 | if (dl->e) | |
6154 | dl->raiddisk = cnt++; | |
6155 | rv = 1; | |
6156 | } | |
6157 | ||
6158 | return rv; | |
6159 | } | |
6160 | ||
bf5a934a | 6161 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6162 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
6163 | char *dev, unsigned long long *freesize, |
6164 | int verbose) | |
6165 | { | |
6166 | int fd, cfd; | |
6167 | struct mdinfo *sra; | |
20cbe8d2 | 6168 | int is_member = 0; |
bf5a934a | 6169 | |
d54559f0 LM |
6170 | /* load capability |
6171 | * if given unused devices create a container | |
bf5a934a DW |
6172 | * if given given devices in a container create a member volume |
6173 | */ | |
6174 | if (level == LEVEL_CONTAINER) { | |
6175 | /* Must be a fresh device to add to a container */ | |
6176 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
6177 | raiddisks, |
6178 | chunk?*chunk:0, size, | |
bf5a934a DW |
6179 | dev, freesize, |
6180 | verbose); | |
6181 | } | |
9587c373 | 6182 | |
8592f29d | 6183 | if (!dev) { |
e91a3bad | 6184 | if (st->sb) { |
ca9de185 | 6185 | struct intel_super *super = st->sb; |
e91a3bad | 6186 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 6187 | raiddisks, chunk, size, |
e91a3bad LM |
6188 | verbose)) |
6189 | return 0; | |
efb30e7f DW |
6190 | /* we are being asked to automatically layout a |
6191 | * new volume based on the current contents of | |
6192 | * the container. If the the parameters can be | |
6193 | * satisfied reserve_space will record the disks, | |
6194 | * start offset, and size of the volume to be | |
6195 | * created. add_to_super and getinfo_super | |
6196 | * detect when autolayout is in progress. | |
6197 | */ | |
ca9de185 LM |
6198 | /* assuming that freesize is always given when array is |
6199 | created */ | |
6200 | if (super->orom && freesize) { | |
6201 | int count; | |
6202 | count = count_volumes(super->hba->path, | |
6203 | super->orom->dpa, verbose); | |
6204 | if (super->orom->vphba <= count) { | |
6205 | pr_vrb(": platform does not support more" | |
40110b90 | 6206 | " than %d raid volumes.\n", |
ca9de185 LM |
6207 | super->orom->vphba); |
6208 | return 0; | |
6209 | } | |
6210 | } | |
e91a3bad LM |
6211 | if (freesize) |
6212 | return reserve_space(st, raiddisks, size, | |
6213 | chunk?*chunk:0, freesize); | |
8592f29d N |
6214 | } |
6215 | return 1; | |
6216 | } | |
bf5a934a DW |
6217 | if (st->sb) { |
6218 | /* creating in a given container */ | |
6219 | return validate_geometry_imsm_volume(st, level, layout, | |
6220 | raiddisks, chunk, size, | |
6221 | dev, freesize, verbose); | |
6222 | } | |
6223 | ||
bf5a934a DW |
6224 | /* This device needs to be a device in an 'imsm' container */ |
6225 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6226 | if (fd >= 0) { | |
6227 | if (verbose) | |
e7b84f9d N |
6228 | pr_err("Cannot create this array on device %s\n", |
6229 | dev); | |
bf5a934a DW |
6230 | close(fd); |
6231 | return 0; | |
6232 | } | |
6233 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6234 | if (verbose) | |
e7b84f9d | 6235 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
6236 | dev, strerror(errno)); |
6237 | return 0; | |
6238 | } | |
6239 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6240 | cfd = open_container(fd); | |
20cbe8d2 | 6241 | close(fd); |
bf5a934a | 6242 | if (cfd < 0) { |
bf5a934a | 6243 | if (verbose) |
e7b84f9d | 6244 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
6245 | dev); |
6246 | return 0; | |
6247 | } | |
6248 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 6249 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
6250 | strcmp(sra->text_version, "imsm") == 0) |
6251 | is_member = 1; | |
6252 | sysfs_free(sra); | |
6253 | if (is_member) { | |
bf5a934a DW |
6254 | /* This is a member of a imsm container. Load the container |
6255 | * and try to create a volume | |
6256 | */ | |
6257 | struct intel_super *super; | |
6258 | ||
ec50f7b6 | 6259 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a DW |
6260 | st->sb = super; |
6261 | st->container_dev = fd2devnum(cfd); | |
6262 | close(cfd); | |
6263 | return validate_geometry_imsm_volume(st, level, layout, | |
6264 | raiddisks, chunk, | |
6265 | size, dev, | |
ecbd9e81 N |
6266 | freesize, 1) |
6267 | ? 1 : -1; | |
bf5a934a | 6268 | } |
20cbe8d2 | 6269 | } |
bf5a934a | 6270 | |
20cbe8d2 | 6271 | if (verbose) |
e7b84f9d | 6272 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
6273 | |
6274 | close(cfd); | |
6275 | return 0; | |
bf5a934a | 6276 | } |
0bd16cf2 | 6277 | |
30f58b22 | 6278 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
6279 | { |
6280 | struct intel_super *super = st->sb; | |
6281 | ||
30f58b22 DW |
6282 | if (level && *level == UnSet) |
6283 | *level = LEVEL_CONTAINER; | |
6284 | ||
6285 | if (level && layout && *layout == UnSet) | |
6286 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 6287 | |
cd9d1ac7 DW |
6288 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
6289 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
6290 | } |
6291 | ||
33414a01 DW |
6292 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
6293 | ||
6294 | static int kill_subarray_imsm(struct supertype *st) | |
6295 | { | |
6296 | /* remove the subarray currently referenced by ->current_vol */ | |
6297 | __u8 i; | |
6298 | struct intel_dev **dp; | |
6299 | struct intel_super *super = st->sb; | |
6300 | __u8 current_vol = super->current_vol; | |
6301 | struct imsm_super *mpb = super->anchor; | |
6302 | ||
6303 | if (super->current_vol < 0) | |
6304 | return 2; | |
6305 | super->current_vol = -1; /* invalidate subarray cursor */ | |
6306 | ||
6307 | /* block deletions that would change the uuid of active subarrays | |
6308 | * | |
6309 | * FIXME when immutable ids are available, but note that we'll | |
6310 | * also need to fixup the invalidated/active subarray indexes in | |
6311 | * mdstat | |
6312 | */ | |
6313 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6314 | char subarray[4]; | |
6315 | ||
6316 | if (i < current_vol) | |
6317 | continue; | |
6318 | sprintf(subarray, "%u", i); | |
6319 | if (is_subarray_active(subarray, st->devname)) { | |
e7b84f9d N |
6320 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
6321 | current_vol, i); | |
33414a01 DW |
6322 | |
6323 | return 2; | |
6324 | } | |
6325 | } | |
6326 | ||
6327 | if (st->update_tail) { | |
503975b9 | 6328 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 6329 | |
33414a01 DW |
6330 | u->type = update_kill_array; |
6331 | u->dev_idx = current_vol; | |
6332 | append_metadata_update(st, u, sizeof(*u)); | |
6333 | ||
6334 | return 0; | |
6335 | } | |
6336 | ||
6337 | for (dp = &super->devlist; *dp;) | |
6338 | if ((*dp)->index == current_vol) { | |
6339 | *dp = (*dp)->next; | |
6340 | } else { | |
6341 | handle_missing(super, (*dp)->dev); | |
6342 | if ((*dp)->index > current_vol) | |
6343 | (*dp)->index--; | |
6344 | dp = &(*dp)->next; | |
6345 | } | |
6346 | ||
6347 | /* no more raid devices, all active components are now spares, | |
6348 | * but of course failed are still failed | |
6349 | */ | |
6350 | if (--mpb->num_raid_devs == 0) { | |
6351 | struct dl *d; | |
6352 | ||
6353 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
6354 | if (d->index > -2) |
6355 | mark_spare(d); | |
33414a01 DW |
6356 | } |
6357 | ||
6358 | super->updates_pending++; | |
6359 | ||
6360 | return 0; | |
6361 | } | |
aa534678 | 6362 | |
a951a4f7 | 6363 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 6364 | char *update, struct mddev_ident *ident) |
aa534678 DW |
6365 | { |
6366 | /* update the subarray currently referenced by ->current_vol */ | |
6367 | struct intel_super *super = st->sb; | |
6368 | struct imsm_super *mpb = super->anchor; | |
6369 | ||
aa534678 DW |
6370 | if (strcmp(update, "name") == 0) { |
6371 | char *name = ident->name; | |
a951a4f7 N |
6372 | char *ep; |
6373 | int vol; | |
aa534678 | 6374 | |
a951a4f7 | 6375 | if (is_subarray_active(subarray, st->devname)) { |
e7b84f9d | 6376 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
6377 | return 2; |
6378 | } | |
6379 | ||
6380 | if (!check_name(super, name, 0)) | |
6381 | return 2; | |
6382 | ||
a951a4f7 N |
6383 | vol = strtoul(subarray, &ep, 10); |
6384 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
6385 | return 2; | |
6386 | ||
aa534678 | 6387 | if (st->update_tail) { |
503975b9 | 6388 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 6389 | |
aa534678 | 6390 | u->type = update_rename_array; |
a951a4f7 | 6391 | u->dev_idx = vol; |
aa534678 DW |
6392 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
6393 | append_metadata_update(st, u, sizeof(*u)); | |
6394 | } else { | |
6395 | struct imsm_dev *dev; | |
6396 | int i; | |
6397 | ||
a951a4f7 | 6398 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
6399 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
6400 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6401 | dev = get_imsm_dev(super, i); | |
6402 | handle_missing(super, dev); | |
6403 | } | |
6404 | super->updates_pending++; | |
6405 | } | |
6406 | } else | |
6407 | return 2; | |
6408 | ||
6409 | return 0; | |
6410 | } | |
d1e02575 | 6411 | #endif /* MDASSEMBLE */ |
bf5a934a | 6412 | |
28bce06f AK |
6413 | static int is_gen_migration(struct imsm_dev *dev) |
6414 | { | |
7534230b AK |
6415 | if (dev == NULL) |
6416 | return 0; | |
6417 | ||
28bce06f AK |
6418 | if (!dev->vol.migr_state) |
6419 | return 0; | |
6420 | ||
6421 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
6422 | return 1; | |
6423 | ||
6424 | return 0; | |
6425 | } | |
6426 | ||
1e5c6983 DW |
6427 | static int is_rebuilding(struct imsm_dev *dev) |
6428 | { | |
6429 | struct imsm_map *migr_map; | |
6430 | ||
6431 | if (!dev->vol.migr_state) | |
6432 | return 0; | |
6433 | ||
6434 | if (migr_type(dev) != MIGR_REBUILD) | |
6435 | return 0; | |
6436 | ||
238c0a71 | 6437 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
6438 | |
6439 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
6440 | return 1; | |
6441 | else | |
6442 | return 0; | |
6443 | } | |
6444 | ||
b4ab44d8 | 6445 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
6446 | static int is_initializing(struct imsm_dev *dev) |
6447 | { | |
6448 | struct imsm_map *migr_map; | |
6449 | ||
6450 | if (!dev->vol.migr_state) | |
6451 | return 0; | |
6452 | ||
6453 | if (migr_type(dev) != MIGR_INIT) | |
6454 | return 0; | |
6455 | ||
238c0a71 | 6456 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
6457 | |
6458 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6459 | return 1; | |
6460 | ||
6461 | return 0; | |
6ce1fbf1 | 6462 | } |
b4ab44d8 | 6463 | #endif |
6ce1fbf1 | 6464 | |
c47b0ff6 AK |
6465 | static void update_recovery_start(struct intel_super *super, |
6466 | struct imsm_dev *dev, | |
6467 | struct mdinfo *array) | |
1e5c6983 DW |
6468 | { |
6469 | struct mdinfo *rebuild = NULL; | |
6470 | struct mdinfo *d; | |
6471 | __u32 units; | |
6472 | ||
6473 | if (!is_rebuilding(dev)) | |
6474 | return; | |
6475 | ||
6476 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
6477 | for (d = array->devs; d; d = d->next) | |
6478 | if (d->recovery_start == 0) { | |
6479 | if (rebuild) | |
6480 | return; | |
6481 | rebuild = d; | |
6482 | } | |
6483 | ||
4363fd80 DW |
6484 | if (!rebuild) { |
6485 | /* (?) none of the disks are marked with | |
6486 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
6487 | * disk_ord_tbl was not correctly updated | |
6488 | */ | |
6489 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
6490 | return; | |
6491 | } | |
6492 | ||
1e5c6983 | 6493 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 6494 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
6495 | } |
6496 | ||
9e2d750d | 6497 | #ifndef MDASSEMBLE |
276d77db | 6498 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 6499 | #endif |
1e5c6983 | 6500 | |
00bbdbda | 6501 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 6502 | { |
4f5bc454 DW |
6503 | /* Given a container loaded by load_super_imsm_all, |
6504 | * extract information about all the arrays into | |
6505 | * an mdinfo tree. | |
00bbdbda | 6506 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
6507 | * |
6508 | * For each imsm_dev create an mdinfo, fill it in, | |
6509 | * then look for matching devices in super->disks | |
6510 | * and create appropriate device mdinfo. | |
6511 | */ | |
6512 | struct intel_super *super = st->sb; | |
949c47a0 | 6513 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 6514 | struct mdinfo *rest = NULL; |
00bbdbda | 6515 | unsigned int i; |
81219e70 | 6516 | int sb_errors = 0; |
abef11a3 AK |
6517 | struct dl *d; |
6518 | int spare_disks = 0; | |
cdddbdbc | 6519 | |
19482bcc AK |
6520 | /* do not assemble arrays when not all attributes are supported */ |
6521 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 6522 | sb_errors = 1; |
e7b84f9d | 6523 | pr_err("Unsupported attributes in IMSM metadata." |
81219e70 | 6524 | "Arrays activation is blocked.\n"); |
19482bcc AK |
6525 | } |
6526 | ||
a06d022d | 6527 | /* check for bad blocks */ |
81219e70 | 6528 | if (imsm_bbm_log_size(super->anchor)) { |
e7b84f9d N |
6529 | pr_err("BBM log found in IMSM metadata." |
6530 | "Arrays activation is blocked.\n"); | |
81219e70 LM |
6531 | sb_errors = 1; |
6532 | } | |
6533 | ||
604b746f | 6534 | |
abef11a3 AK |
6535 | /* count spare devices, not used in maps |
6536 | */ | |
6537 | for (d = super->disks; d; d = d->next) | |
6538 | if (d->index == -1) | |
6539 | spare_disks++; | |
6540 | ||
4f5bc454 | 6541 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
6542 | struct imsm_dev *dev; |
6543 | struct imsm_map *map; | |
86e3692b | 6544 | struct imsm_map *map2; |
4f5bc454 | 6545 | struct mdinfo *this; |
a6482415 N |
6546 | int slot; |
6547 | #ifndef MDASSEMBLE | |
6548 | int chunk; | |
6549 | #endif | |
00bbdbda N |
6550 | char *ep; |
6551 | ||
6552 | if (subarray && | |
6553 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
6554 | continue; | |
6555 | ||
6556 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
6557 | map = get_imsm_map(dev, MAP_0); |
6558 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 6559 | |
1ce0101c DW |
6560 | /* do not publish arrays that are in the middle of an |
6561 | * unsupported migration | |
6562 | */ | |
6563 | if (dev->vol.migr_state && | |
28bce06f | 6564 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
e7b84f9d | 6565 | pr_err("cannot assemble volume '%.16s':" |
1ce0101c DW |
6566 | " unsupported migration in progress\n", |
6567 | dev->volume); | |
6568 | continue; | |
6569 | } | |
2db86302 LM |
6570 | /* do not publish arrays that are not support by controller's |
6571 | * OROM/EFI | |
6572 | */ | |
1ce0101c | 6573 | |
503975b9 | 6574 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 6575 | |
301406c9 | 6576 | super->current_vol = i; |
a5d85af7 | 6577 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 6578 | this->next = rest; |
81219e70 | 6579 | #ifndef MDASSEMBLE |
a6482415 | 6580 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
6581 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
6582 | if (!validate_geometry_imsm_orom(super, | |
6583 | get_imsm_raid_level(map), /* RAID level */ | |
6584 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
6585 | map->num_members, /* raid disks */ | |
2cc699af | 6586 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 6587 | 1 /* verbose */)) { |
e7b84f9d | 6588 | pr_err("IMSM RAID geometry validation" |
446894ea | 6589 | " failed. Array %s activation is blocked.\n", |
81219e70 LM |
6590 | dev->volume); |
6591 | this->array.state |= | |
6592 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6593 | (1<<MD_SB_BLOCK_VOLUME); | |
6594 | } | |
6595 | #endif | |
6596 | ||
6597 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
6598 | if (sb_errors) | |
6599 | this->array.state |= | |
6600 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6601 | (1<<MD_SB_BLOCK_VOLUME); | |
6602 | ||
4f5bc454 | 6603 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 6604 | unsigned long long recovery_start; |
4f5bc454 DW |
6605 | struct mdinfo *info_d; |
6606 | struct dl *d; | |
6607 | int idx; | |
9a1608e5 | 6608 | int skip; |
7eef0453 | 6609 | __u32 ord; |
4f5bc454 | 6610 | |
9a1608e5 | 6611 | skip = 0; |
238c0a71 AK |
6612 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
6613 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
6614 | for (d = super->disks; d ; d = d->next) |
6615 | if (d->index == idx) | |
0fbd635c | 6616 | break; |
4f5bc454 | 6617 | |
1e5c6983 | 6618 | recovery_start = MaxSector; |
4f5bc454 | 6619 | if (d == NULL) |
9a1608e5 | 6620 | skip = 1; |
25ed7e59 | 6621 | if (d && is_failed(&d->disk)) |
9a1608e5 | 6622 | skip = 1; |
7eef0453 | 6623 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 6624 | recovery_start = 0; |
9a1608e5 DW |
6625 | |
6626 | /* | |
6627 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
6628 | * reset resync start to avoid a dirty-degraded |
6629 | * situation when performing the intial sync | |
9a1608e5 DW |
6630 | * |
6631 | * FIXME handle dirty degraded | |
6632 | */ | |
1e5c6983 | 6633 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 6634 | this->resync_start = MaxSector; |
9a1608e5 DW |
6635 | if (skip) |
6636 | continue; | |
4f5bc454 | 6637 | |
503975b9 | 6638 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
6639 | info_d->next = this->devs; |
6640 | this->devs = info_d; | |
6641 | ||
4f5bc454 DW |
6642 | info_d->disk.number = d->index; |
6643 | info_d->disk.major = d->major; | |
6644 | info_d->disk.minor = d->minor; | |
6645 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 6646 | info_d->recovery_start = recovery_start; |
86e3692b AK |
6647 | if (map2) { |
6648 | if (slot < map2->num_members) | |
6649 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6650 | else |
6651 | this->array.spare_disks++; | |
86e3692b AK |
6652 | } else { |
6653 | if (slot < map->num_members) | |
6654 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6655 | else |
6656 | this->array.spare_disks++; | |
86e3692b | 6657 | } |
1e5c6983 DW |
6658 | if (info_d->recovery_start == MaxSector) |
6659 | this->array.working_disks++; | |
4f5bc454 DW |
6660 | |
6661 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 CA |
6662 | info_d->data_offset = pba_of_lba0(map); |
6663 | info_d->component_size = blocks_per_member(map); | |
4f5bc454 | 6664 | } |
1e5c6983 | 6665 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 6666 | update_recovery_start(super, dev, this); |
abef11a3 | 6667 | this->array.spare_disks += spare_disks; |
276d77db | 6668 | |
9e2d750d | 6669 | #ifndef MDASSEMBLE |
276d77db AK |
6670 | /* check for reshape */ |
6671 | if (this->reshape_active == 1) | |
6672 | recover_backup_imsm(st, this); | |
9e2d750d | 6673 | #endif |
9a1608e5 | 6674 | rest = this; |
4f5bc454 DW |
6675 | } |
6676 | ||
6677 | return rest; | |
cdddbdbc DW |
6678 | } |
6679 | ||
845dea95 | 6680 | |
3b451610 AK |
6681 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
6682 | int failed, int look_in_map) | |
c2a1e7da | 6683 | { |
3b451610 AK |
6684 | struct imsm_map *map; |
6685 | ||
6686 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
6687 | |
6688 | if (!failed) | |
3393c6af DW |
6689 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
6690 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
6691 | |
6692 | switch (get_imsm_raid_level(map)) { | |
6693 | case 0: | |
6694 | return IMSM_T_STATE_FAILED; | |
6695 | break; | |
6696 | case 1: | |
6697 | if (failed < map->num_members) | |
6698 | return IMSM_T_STATE_DEGRADED; | |
6699 | else | |
6700 | return IMSM_T_STATE_FAILED; | |
6701 | break; | |
6702 | case 10: | |
6703 | { | |
6704 | /** | |
c92a2527 DW |
6705 | * check to see if any mirrors have failed, otherwise we |
6706 | * are degraded. Even numbered slots are mirrored on | |
6707 | * slot+1 | |
c2a1e7da | 6708 | */ |
c2a1e7da | 6709 | int i; |
d9b420a5 N |
6710 | /* gcc -Os complains that this is unused */ |
6711 | int insync = insync; | |
c2a1e7da DW |
6712 | |
6713 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6714 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6715 | int idx = ord_to_idx(ord); |
6716 | struct imsm_disk *disk; | |
c2a1e7da | 6717 | |
c92a2527 DW |
6718 | /* reset the potential in-sync count on even-numbered |
6719 | * slots. num_copies is always 2 for imsm raid10 | |
6720 | */ | |
6721 | if ((i & 1) == 0) | |
6722 | insync = 2; | |
c2a1e7da | 6723 | |
c92a2527 | 6724 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6725 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6726 | insync--; |
c2a1e7da | 6727 | |
c92a2527 DW |
6728 | /* no in-sync disks left in this mirror the |
6729 | * array has failed | |
6730 | */ | |
6731 | if (insync == 0) | |
6732 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6733 | } |
6734 | ||
6735 | return IMSM_T_STATE_DEGRADED; | |
6736 | } | |
6737 | case 5: | |
6738 | if (failed < 2) | |
6739 | return IMSM_T_STATE_DEGRADED; | |
6740 | else | |
6741 | return IMSM_T_STATE_FAILED; | |
6742 | break; | |
6743 | default: | |
6744 | break; | |
6745 | } | |
6746 | ||
6747 | return map->map_state; | |
6748 | } | |
6749 | ||
3b451610 AK |
6750 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6751 | int look_in_map) | |
c2a1e7da DW |
6752 | { |
6753 | int i; | |
6754 | int failed = 0; | |
6755 | struct imsm_disk *disk; | |
d5985138 AK |
6756 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6757 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6758 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6759 | __u32 ord; |
6760 | int idx; | |
d5985138 | 6761 | int idx_1; |
c2a1e7da | 6762 | |
0556e1a2 DW |
6763 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6764 | * disks that are being rebuilt. New failures are recorded to | |
6765 | * map[0]. So we look through all the disks we started with and | |
6766 | * see if any failures are still present, or if any new ones | |
6767 | * have arrived | |
0556e1a2 | 6768 | */ |
d5985138 AK |
6769 | map_for_loop = map; |
6770 | if (prev && (map->num_members < prev->num_members)) | |
6771 | map_for_loop = prev; | |
68fe4598 LD |
6772 | |
6773 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6774 | idx_1 = -255; |
238c0a71 AK |
6775 | /* when MAP_X is passed both maps failures are counted |
6776 | */ | |
d5985138 | 6777 | if (prev && |
238c0a71 AK |
6778 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6779 | (i < prev->num_members)) { | |
d5985138 AK |
6780 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6781 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6782 | |
d5985138 AK |
6783 | disk = get_imsm_disk(super, idx_1); |
6784 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6785 | failed++; | |
6786 | } | |
238c0a71 AK |
6787 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6788 | (i < map->num_members)) { | |
d5985138 AK |
6789 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6790 | idx = ord_to_idx(ord); | |
6791 | ||
6792 | if (idx != idx_1) { | |
6793 | disk = get_imsm_disk(super, idx); | |
6794 | if (!disk || is_failed(disk) || | |
6795 | ord & IMSM_ORD_REBUILD) | |
6796 | failed++; | |
6797 | } | |
6798 | } | |
c2a1e7da DW |
6799 | } |
6800 | ||
6801 | return failed; | |
845dea95 NB |
6802 | } |
6803 | ||
97b4d0e9 DW |
6804 | #ifndef MDASSEMBLE |
6805 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
6806 | char *inst) | |
6807 | { | |
6808 | struct intel_super *super = c->sb; | |
6809 | struct imsm_super *mpb = super->anchor; | |
9587c373 | 6810 | |
97b4d0e9 DW |
6811 | if (atoi(inst) >= mpb->num_raid_devs) { |
6812 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
6813 | __func__, atoi(inst)); | |
6814 | return -ENODEV; | |
6815 | } | |
6816 | ||
6817 | dprintf("imsm: open_new %s\n", inst); | |
6818 | a->info.container_member = atoi(inst); | |
6819 | return 0; | |
6820 | } | |
6821 | ||
0c046afd DW |
6822 | static int is_resyncing(struct imsm_dev *dev) |
6823 | { | |
6824 | struct imsm_map *migr_map; | |
6825 | ||
6826 | if (!dev->vol.migr_state) | |
6827 | return 0; | |
6828 | ||
1484e727 DW |
6829 | if (migr_type(dev) == MIGR_INIT || |
6830 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
6831 | return 1; |
6832 | ||
4c9bc37b AK |
6833 | if (migr_type(dev) == MIGR_GEN_MIGR) |
6834 | return 0; | |
6835 | ||
238c0a71 | 6836 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 6837 | |
4c9bc37b AK |
6838 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
6839 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
6840 | return 1; |
6841 | else | |
6842 | return 0; | |
6843 | } | |
6844 | ||
0556e1a2 DW |
6845 | /* return true if we recorded new information */ |
6846 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 6847 | { |
0556e1a2 DW |
6848 | __u32 ord; |
6849 | int slot; | |
6850 | struct imsm_map *map; | |
86c54047 DW |
6851 | char buf[MAX_RAID_SERIAL_LEN+3]; |
6852 | unsigned int len, shift = 0; | |
0556e1a2 DW |
6853 | |
6854 | /* new failures are always set in map[0] */ | |
238c0a71 | 6855 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
6856 | |
6857 | slot = get_imsm_disk_slot(map, idx); | |
6858 | if (slot < 0) | |
6859 | return 0; | |
6860 | ||
6861 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 6862 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
6863 | return 0; |
6864 | ||
7d0c5e24 LD |
6865 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
6866 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
6867 | strcat(buf, ":0"); | |
86c54047 DW |
6868 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
6869 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
6870 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
6871 | ||
f2f27e63 | 6872 | disk->status |= FAILED_DISK; |
0556e1a2 | 6873 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
6874 | /* mark failures in second map if second map exists and this disk |
6875 | * in this slot. | |
6876 | * This is valid for migration, initialization and rebuild | |
6877 | */ | |
6878 | if (dev->vol.migr_state) { | |
238c0a71 | 6879 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
6880 | int slot2 = get_imsm_disk_slot(map2, idx); |
6881 | ||
6882 | if ((slot2 < map2->num_members) && | |
6883 | (slot2 >= 0)) | |
6884 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
6885 | idx | IMSM_ORD_REBUILD); |
6886 | } | |
f21e18ca | 6887 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
6888 | map->failed_disk_num = slot; |
6889 | return 1; | |
6890 | } | |
6891 | ||
6892 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
6893 | { | |
6894 | mark_failure(dev, disk, idx); | |
6895 | ||
6896 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
6897 | return; | |
6898 | ||
47ee5a45 DW |
6899 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
6900 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
6901 | } | |
6902 | ||
33414a01 DW |
6903 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
6904 | { | |
33414a01 | 6905 | struct dl *dl; |
33414a01 DW |
6906 | |
6907 | if (!super->missing) | |
6908 | return; | |
33414a01 DW |
6909 | |
6910 | dprintf("imsm: mark missing\n"); | |
3d59f0c0 AK |
6911 | /* end process for initialization and rebuild only |
6912 | */ | |
6913 | if (is_gen_migration(dev) == 0) { | |
6914 | __u8 map_state; | |
6915 | int failed; | |
6916 | ||
6917 | failed = imsm_count_failed(super, dev, MAP_0); | |
6918 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
6919 | ||
6920 | end_migration(dev, super, map_state); | |
6921 | } | |
33414a01 DW |
6922 | for (dl = super->missing; dl; dl = dl->next) |
6923 | mark_missing(dev, &dl->disk, dl->index); | |
6924 | super->updates_pending++; | |
6925 | } | |
6926 | ||
f3871fdc AK |
6927 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
6928 | long long new_size) | |
70bdf0dc | 6929 | { |
238c0a71 | 6930 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
6931 | unsigned long long array_blocks; |
6932 | struct imsm_map *map; | |
6933 | ||
6934 | if (used_disks == 0) { | |
6935 | /* when problems occures | |
6936 | * return current array_blocks value | |
6937 | */ | |
6938 | array_blocks = __le32_to_cpu(dev->size_high); | |
6939 | array_blocks = array_blocks << 32; | |
6940 | array_blocks += __le32_to_cpu(dev->size_low); | |
6941 | ||
6942 | return array_blocks; | |
6943 | } | |
6944 | ||
6945 | /* set array size in metadata | |
6946 | */ | |
f3871fdc AK |
6947 | if (new_size <= 0) { |
6948 | /* OLCE size change is caused by added disks | |
6949 | */ | |
6950 | map = get_imsm_map(dev, MAP_0); | |
6951 | array_blocks = blocks_per_member(map) * used_disks; | |
6952 | } else { | |
6953 | /* Online Volume Size Change | |
6954 | * Using available free space | |
6955 | */ | |
6956 | array_blocks = new_size; | |
6957 | } | |
70bdf0dc AK |
6958 | |
6959 | /* round array size down to closest MB | |
6960 | */ | |
6961 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
6962 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
6963 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
6964 | ||
6965 | return array_blocks; | |
6966 | } | |
6967 | ||
28bce06f AK |
6968 | static void imsm_set_disk(struct active_array *a, int n, int state); |
6969 | ||
0e2d1a4e AK |
6970 | static void imsm_progress_container_reshape(struct intel_super *super) |
6971 | { | |
6972 | /* if no device has a migr_state, but some device has a | |
6973 | * different number of members than the previous device, start | |
6974 | * changing the number of devices in this device to match | |
6975 | * previous. | |
6976 | */ | |
6977 | struct imsm_super *mpb = super->anchor; | |
6978 | int prev_disks = -1; | |
6979 | int i; | |
1dfaa380 | 6980 | int copy_map_size; |
0e2d1a4e AK |
6981 | |
6982 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6983 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 6984 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
6985 | struct imsm_map *map2; |
6986 | int prev_num_members; | |
0e2d1a4e AK |
6987 | |
6988 | if (dev->vol.migr_state) | |
6989 | return; | |
6990 | ||
6991 | if (prev_disks == -1) | |
6992 | prev_disks = map->num_members; | |
6993 | if (prev_disks == map->num_members) | |
6994 | continue; | |
6995 | ||
6996 | /* OK, this array needs to enter reshape mode. | |
6997 | * i.e it needs a migr_state | |
6998 | */ | |
6999 | ||
1dfaa380 | 7000 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
7001 | prev_num_members = map->num_members; |
7002 | map->num_members = prev_disks; | |
7003 | dev->vol.migr_state = 1; | |
7004 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7005 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
7006 | for (i = prev_num_members; |
7007 | i < map->num_members; i++) | |
7008 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 7009 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 7010 | /* Copy the current map */ |
1dfaa380 | 7011 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
7012 | map2->num_members = prev_num_members; |
7013 | ||
f3871fdc | 7014 | imsm_set_array_size(dev, -1); |
51d83f5d | 7015 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
7016 | super->updates_pending++; |
7017 | } | |
7018 | } | |
7019 | ||
aad6f216 | 7020 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
7021 | * states are handled in imsm_set_disk() with one exception, when a |
7022 | * resync is stopped due to a new failure this routine will set the | |
7023 | * 'degraded' state for the array. | |
7024 | */ | |
01f157d7 | 7025 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
7026 | { |
7027 | int inst = a->info.container_member; | |
7028 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7029 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7030 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
7031 | int failed = imsm_count_failed(super, dev, MAP_0); |
7032 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 7033 | __u32 blocks_per_unit; |
a862209d | 7034 | |
1af97990 AK |
7035 | if (dev->vol.migr_state && |
7036 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
7037 | /* array state change is blocked due to reshape action | |
aad6f216 N |
7038 | * We might need to |
7039 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
7040 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
7041 | * - update curr_migr_unit | |
1af97990 | 7042 | */ |
aad6f216 N |
7043 | if (a->curr_action == reshape) { |
7044 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 7045 | goto mark_checkpoint; |
aad6f216 N |
7046 | } else { |
7047 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
7048 | /* for some reason we aborted the reshape. | |
b66e591b AK |
7049 | * |
7050 | * disable automatic metadata rollback | |
7051 | * user action is required to recover process | |
aad6f216 | 7052 | */ |
b66e591b | 7053 | if (0) { |
238c0a71 AK |
7054 | struct imsm_map *map2 = |
7055 | get_imsm_map(dev, MAP_1); | |
7056 | dev->vol.migr_state = 0; | |
7057 | set_migr_type(dev, 0); | |
7058 | dev->vol.curr_migr_unit = 0; | |
7059 | memcpy(map, map2, | |
7060 | sizeof_imsm_map(map2)); | |
7061 | super->updates_pending++; | |
b66e591b | 7062 | } |
aad6f216 N |
7063 | } |
7064 | if (a->last_checkpoint >= a->info.component_size) { | |
7065 | unsigned long long array_blocks; | |
7066 | int used_disks; | |
e154ced3 | 7067 | struct mdinfo *mdi; |
aad6f216 | 7068 | |
238c0a71 | 7069 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
7070 | if (used_disks > 0) { |
7071 | array_blocks = | |
5551b113 | 7072 | blocks_per_member(map) * |
d55adef9 AK |
7073 | used_disks; |
7074 | /* round array size down to closest MB | |
7075 | */ | |
7076 | array_blocks = (array_blocks | |
7077 | >> SECT_PER_MB_SHIFT) | |
7078 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
7079 | a->info.custom_array_size = array_blocks; |
7080 | /* encourage manager to update array | |
7081 | * size | |
7082 | */ | |
e154ced3 | 7083 | |
d55adef9 | 7084 | a->check_reshape = 1; |
633b5610 | 7085 | } |
e154ced3 AK |
7086 | /* finalize online capacity expansion/reshape */ |
7087 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
7088 | imsm_set_disk(a, | |
7089 | mdi->disk.raid_disk, | |
7090 | mdi->curr_state); | |
7091 | ||
0e2d1a4e | 7092 | imsm_progress_container_reshape(super); |
e154ced3 | 7093 | } |
aad6f216 | 7094 | } |
1af97990 AK |
7095 | } |
7096 | ||
47ee5a45 | 7097 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
7098 | if (consistent == 2) |
7099 | handle_missing(super, dev); | |
1e5c6983 | 7100 | |
0c046afd | 7101 | if (consistent == 2 && |
b7941fd6 | 7102 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7103 | map_state != IMSM_T_STATE_NORMAL || |
7104 | dev->vol.migr_state)) | |
01f157d7 | 7105 | consistent = 0; |
272906ef | 7106 | |
b7941fd6 | 7107 | if (is_resync_complete(&a->info)) { |
0c046afd | 7108 | /* complete intialization / resync, |
0556e1a2 DW |
7109 | * recovery and interrupted recovery is completed in |
7110 | * ->set_disk | |
0c046afd DW |
7111 | */ |
7112 | if (is_resyncing(dev)) { | |
7113 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7114 | end_migration(dev, super, map_state); |
115c3803 | 7115 | super->updates_pending++; |
484240d8 | 7116 | a->last_checkpoint = 0; |
115c3803 | 7117 | } |
b9172665 AK |
7118 | } else if ((!is_resyncing(dev) && !failed) && |
7119 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7120 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7121 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7122 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7123 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7124 | else |
8e59f3d8 | 7125 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7126 | super->updates_pending++; |
115c3803 | 7127 | } |
a862209d | 7128 | |
633b5610 | 7129 | mark_checkpoint: |
5b83bacf AK |
7130 | /* skip checkpointing for general migration, |
7131 | * it is controlled in mdadm | |
7132 | */ | |
7133 | if (is_gen_migration(dev)) | |
7134 | goto skip_mark_checkpoint; | |
7135 | ||
1e5c6983 | 7136 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7137 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7138 | if (blocks_per_unit) { |
1e5c6983 DW |
7139 | __u32 units32; |
7140 | __u64 units; | |
7141 | ||
4f0a7acc | 7142 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7143 | units32 = units; |
7144 | ||
7145 | /* check that we did not overflow 32-bits, and that | |
7146 | * curr_migr_unit needs updating | |
7147 | */ | |
7148 | if (units32 == units && | |
bfd80a56 | 7149 | units32 != 0 && |
1e5c6983 DW |
7150 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7151 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7152 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7153 | super->updates_pending++; | |
7154 | } | |
7155 | } | |
f8f603f1 | 7156 | |
5b83bacf | 7157 | skip_mark_checkpoint: |
3393c6af | 7158 | /* mark dirty / clean */ |
0c046afd | 7159 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7160 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7161 | if (consistent) |
7162 | dev->vol.dirty = 0; | |
7163 | else | |
7164 | dev->vol.dirty = 1; | |
a862209d DW |
7165 | super->updates_pending++; |
7166 | } | |
28bce06f | 7167 | |
01f157d7 | 7168 | return consistent; |
a862209d DW |
7169 | } |
7170 | ||
8d45d196 | 7171 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7172 | { |
8d45d196 DW |
7173 | int inst = a->info.container_member; |
7174 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7175 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7176 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7177 | struct imsm_disk *disk; |
7ce05701 LD |
7178 | struct mdinfo *mdi; |
7179 | int recovery_not_finished = 0; | |
0c046afd | 7180 | int failed; |
b10b37b8 | 7181 | __u32 ord; |
0c046afd | 7182 | __u8 map_state; |
8d45d196 DW |
7183 | |
7184 | if (n > map->num_members) | |
7185 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
7186 | n, map->num_members - 1); | |
7187 | ||
7188 | if (n < 0) | |
7189 | return; | |
7190 | ||
4e6e574a | 7191 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 7192 | |
238c0a71 | 7193 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 7194 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7195 | |
5802a811 | 7196 | /* check for new failures */ |
0556e1a2 DW |
7197 | if (state & DS_FAULTY) { |
7198 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
7199 | super->updates_pending++; | |
8d45d196 | 7200 | } |
47ee5a45 | 7201 | |
19859edc | 7202 | /* check if in_sync */ |
0556e1a2 | 7203 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7204 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7205 | |
7206 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7207 | super->updates_pending++; |
7208 | } | |
8d45d196 | 7209 | |
3b451610 AK |
7210 | failed = imsm_count_failed(super, dev, MAP_0); |
7211 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7212 | |
0c046afd | 7213 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7214 | dprintf("imsm: Detected transition to state "); |
7215 | switch (map_state) { | |
7216 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
7217 | dprintf("normal: "); | |
7218 | if (is_rebuilding(dev)) { | |
7219 | dprintf("while rebuilding"); | |
7ce05701 LD |
7220 | /* check if recovery is really finished */ |
7221 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
7222 | if (mdi->recovery_start != MaxSector) { | |
7223 | recovery_not_finished = 1; | |
7224 | break; | |
7225 | } | |
7226 | if (recovery_not_finished) { | |
7227 | dprintf("\nimsm: Rebuild has not finished yet, " | |
7228 | "state not changed"); | |
7229 | if (a->last_checkpoint < mdi->recovery_start) { | |
7230 | a->last_checkpoint = mdi->recovery_start; | |
7231 | super->updates_pending++; | |
7232 | } | |
7233 | break; | |
7234 | } | |
94002678 | 7235 | end_migration(dev, super, map_state); |
238c0a71 | 7236 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
7237 | map->failed_disk_num = ~0; |
7238 | super->updates_pending++; | |
7239 | a->last_checkpoint = 0; | |
7240 | break; | |
7241 | } | |
7242 | if (is_gen_migration(dev)) { | |
7243 | dprintf("while general migration"); | |
bf2f0071 | 7244 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7245 | end_migration(dev, super, map_state); |
94002678 AK |
7246 | else |
7247 | map->map_state = map_state; | |
238c0a71 | 7248 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 7249 | map->failed_disk_num = ~0; |
94002678 | 7250 | super->updates_pending++; |
bf2f0071 | 7251 | break; |
94002678 AK |
7252 | } |
7253 | break; | |
7254 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
7255 | dprintf("degraded: "); | |
7256 | if ((map->map_state != map_state) && | |
7257 | !dev->vol.migr_state) { | |
7258 | dprintf("mark degraded"); | |
7259 | map->map_state = map_state; | |
7260 | super->updates_pending++; | |
7261 | a->last_checkpoint = 0; | |
7262 | break; | |
7263 | } | |
7264 | if (is_rebuilding(dev)) { | |
7265 | dprintf("while rebuilding."); | |
7266 | if (map->map_state != map_state) { | |
7267 | dprintf(" Map state change"); | |
7268 | end_migration(dev, super, map_state); | |
7269 | super->updates_pending++; | |
7270 | } | |
7271 | break; | |
7272 | } | |
7273 | if (is_gen_migration(dev)) { | |
7274 | dprintf("while general migration"); | |
bf2f0071 | 7275 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7276 | end_migration(dev, super, map_state); |
94002678 AK |
7277 | else { |
7278 | map->map_state = map_state; | |
3b451610 | 7279 | manage_second_map(super, dev); |
94002678 AK |
7280 | } |
7281 | super->updates_pending++; | |
bf2f0071 | 7282 | break; |
28bce06f | 7283 | } |
6ce1fbf1 AK |
7284 | if (is_initializing(dev)) { |
7285 | dprintf("while initialization."); | |
7286 | map->map_state = map_state; | |
7287 | super->updates_pending++; | |
7288 | break; | |
7289 | } | |
94002678 AK |
7290 | break; |
7291 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
7292 | dprintf("failed: "); | |
7293 | if (is_gen_migration(dev)) { | |
7294 | dprintf("while general migration"); | |
7295 | map->map_state = map_state; | |
7296 | super->updates_pending++; | |
7297 | break; | |
7298 | } | |
7299 | if (map->map_state != map_state) { | |
7300 | dprintf("mark failed"); | |
7301 | end_migration(dev, super, map_state); | |
7302 | super->updates_pending++; | |
7303 | a->last_checkpoint = 0; | |
7304 | break; | |
7305 | } | |
7306 | break; | |
7307 | default: | |
7308 | dprintf("state %i\n", map_state); | |
5802a811 | 7309 | } |
94002678 AK |
7310 | dprintf("\n"); |
7311 | ||
845dea95 NB |
7312 | } |
7313 | ||
f796af5d | 7314 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 7315 | { |
f796af5d | 7316 | void *buf = mpb; |
c2a1e7da DW |
7317 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
7318 | unsigned long long dsize; | |
7319 | unsigned long long sectors; | |
7320 | ||
7321 | get_dev_size(fd, NULL, &dsize); | |
7322 | ||
272f648f DW |
7323 | if (mpb_size > 512) { |
7324 | /* -1 to account for anchor */ | |
7325 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 7326 | |
272f648f DW |
7327 | /* write the extended mpb to the sectors preceeding the anchor */ |
7328 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
7329 | return 1; | |
c2a1e7da | 7330 | |
f21e18ca N |
7331 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
7332 | != 512 * sectors) | |
272f648f DW |
7333 | return 1; |
7334 | } | |
c2a1e7da | 7335 | |
272f648f DW |
7336 | /* first block is stored on second to last sector of the disk */ |
7337 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
7338 | return 1; |
7339 | ||
f796af5d | 7340 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
7341 | return 1; |
7342 | ||
c2a1e7da DW |
7343 | return 0; |
7344 | } | |
7345 | ||
2e735d19 | 7346 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 7347 | { |
2e735d19 | 7348 | struct intel_super *super = container->sb; |
c2a1e7da | 7349 | |
1a64be56 | 7350 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
7351 | if (!super->updates_pending) |
7352 | return; | |
7353 | ||
36988a3d | 7354 | write_super_imsm(container, 0); |
c2a1e7da DW |
7355 | |
7356 | super->updates_pending = 0; | |
845dea95 NB |
7357 | } |
7358 | ||
272906ef DW |
7359 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
7360 | { | |
7361 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7362 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
7363 | struct dl *dl; |
7364 | ||
7365 | for (dl = super->disks; dl; dl = dl->next) | |
7366 | if (dl->index == i) | |
7367 | break; | |
7368 | ||
25ed7e59 | 7369 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
7370 | dl = NULL; |
7371 | ||
7372 | if (dl) | |
7373 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
7374 | ||
7375 | return dl; | |
7376 | } | |
7377 | ||
a20d2ba5 | 7378 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
7379 | struct active_array *a, int activate_new, |
7380 | struct mdinfo *additional_test_list) | |
272906ef DW |
7381 | { |
7382 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7383 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
7384 | struct imsm_super *mpb = super->anchor; |
7385 | struct imsm_map *map; | |
272906ef DW |
7386 | unsigned long long pos; |
7387 | struct mdinfo *d; | |
7388 | struct extent *ex; | |
a20d2ba5 | 7389 | int i, j; |
272906ef | 7390 | int found; |
569cc43f DW |
7391 | __u32 array_start = 0; |
7392 | __u32 array_end = 0; | |
272906ef | 7393 | struct dl *dl; |
6c932028 | 7394 | struct mdinfo *test_list; |
272906ef DW |
7395 | |
7396 | for (dl = super->disks; dl; dl = dl->next) { | |
7397 | /* If in this array, skip */ | |
7398 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
7399 | if (d->state_fd >= 0 && |
7400 | d->disk.major == dl->major && | |
272906ef | 7401 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
7402 | dprintf("%x:%x already in array\n", |
7403 | dl->major, dl->minor); | |
272906ef DW |
7404 | break; |
7405 | } | |
7406 | if (d) | |
7407 | continue; | |
6c932028 AK |
7408 | test_list = additional_test_list; |
7409 | while (test_list) { | |
7410 | if (test_list->disk.major == dl->major && | |
7411 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
7412 | dprintf("%x:%x already in additional test list\n", |
7413 | dl->major, dl->minor); | |
7414 | break; | |
7415 | } | |
6c932028 | 7416 | test_list = test_list->next; |
8ba77d32 | 7417 | } |
6c932028 | 7418 | if (test_list) |
8ba77d32 | 7419 | continue; |
272906ef | 7420 | |
e553d2a4 | 7421 | /* skip in use or failed drives */ |
25ed7e59 | 7422 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
7423 | dl->index == -2) { |
7424 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 7425 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
7426 | continue; |
7427 | } | |
7428 | ||
a20d2ba5 DW |
7429 | /* skip pure spares when we are looking for partially |
7430 | * assimilated drives | |
7431 | */ | |
7432 | if (dl->index == -1 && !activate_new) | |
7433 | continue; | |
7434 | ||
272906ef | 7435 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 7436 | * It needs to be able to cover all member volumes |
272906ef DW |
7437 | */ |
7438 | ex = get_extents(super, dl); | |
7439 | if (!ex) { | |
7440 | dprintf("cannot get extents\n"); | |
7441 | continue; | |
7442 | } | |
a20d2ba5 DW |
7443 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7444 | dev = get_imsm_dev(super, i); | |
238c0a71 | 7445 | map = get_imsm_map(dev, MAP_0); |
272906ef | 7446 | |
a20d2ba5 DW |
7447 | /* check if this disk is already a member of |
7448 | * this array | |
272906ef | 7449 | */ |
620b1713 | 7450 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
7451 | continue; |
7452 | ||
7453 | found = 0; | |
7454 | j = 0; | |
7455 | pos = 0; | |
5551b113 | 7456 | array_start = pba_of_lba0(map); |
329c8278 | 7457 | array_end = array_start + |
5551b113 | 7458 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
7459 | |
7460 | do { | |
7461 | /* check that we can start at pba_of_lba0 with | |
7462 | * blocks_per_member of space | |
7463 | */ | |
329c8278 | 7464 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
7465 | found = 1; |
7466 | break; | |
7467 | } | |
7468 | pos = ex[j].start + ex[j].size; | |
7469 | j++; | |
7470 | } while (ex[j-1].size); | |
7471 | ||
7472 | if (!found) | |
272906ef | 7473 | break; |
a20d2ba5 | 7474 | } |
272906ef DW |
7475 | |
7476 | free(ex); | |
a20d2ba5 | 7477 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
7478 | dprintf("%x:%x does not have %u to %u available\n", |
7479 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
7480 | /* No room */ |
7481 | continue; | |
a20d2ba5 DW |
7482 | } |
7483 | return dl; | |
272906ef DW |
7484 | } |
7485 | ||
7486 | return dl; | |
7487 | } | |
7488 | ||
95d07a2c LM |
7489 | |
7490 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
7491 | { | |
7492 | struct imsm_dev *dev2; | |
7493 | struct imsm_map *map; | |
7494 | struct dl *idisk; | |
7495 | int slot; | |
7496 | int idx; | |
7497 | __u8 state; | |
7498 | ||
7499 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
7500 | if (dev2) { | |
238c0a71 | 7501 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 7502 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 7503 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
7504 | if (!map) |
7505 | return 1; | |
7506 | for (slot = 0; slot < map->num_members; slot++) { | |
7507 | /* | |
7508 | * Check if failed disks are deleted from intel | |
7509 | * disk list or are marked to be deleted | |
7510 | */ | |
238c0a71 | 7511 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
7512 | idisk = get_imsm_dl_disk(cont->sb, idx); |
7513 | /* | |
7514 | * Do not rebuild the array if failed disks | |
7515 | * from failed sub-array are not removed from | |
7516 | * container. | |
7517 | */ | |
7518 | if (idisk && | |
7519 | is_failed(&idisk->disk) && | |
7520 | (idisk->action != DISK_REMOVE)) | |
7521 | return 0; | |
7522 | } | |
7523 | } | |
7524 | } | |
7525 | return 1; | |
7526 | } | |
7527 | ||
88758e9d DW |
7528 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
7529 | struct metadata_update **updates) | |
7530 | { | |
7531 | /** | |
d23fe947 DW |
7532 | * Find a device with unused free space and use it to replace a |
7533 | * failed/vacant region in an array. We replace failed regions one a | |
7534 | * array at a time. The result is that a new spare disk will be added | |
7535 | * to the first failed array and after the monitor has finished | |
7536 | * propagating failures the remainder will be consumed. | |
88758e9d | 7537 | * |
d23fe947 DW |
7538 | * FIXME add a capability for mdmon to request spares from another |
7539 | * container. | |
88758e9d DW |
7540 | */ |
7541 | ||
7542 | struct intel_super *super = a->container->sb; | |
88758e9d | 7543 | int inst = a->info.container_member; |
949c47a0 | 7544 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7545 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
7546 | int failed = a->info.array.raid_disks; |
7547 | struct mdinfo *rv = NULL; | |
7548 | struct mdinfo *d; | |
7549 | struct mdinfo *di; | |
7550 | struct metadata_update *mu; | |
7551 | struct dl *dl; | |
7552 | struct imsm_update_activate_spare *u; | |
7553 | int num_spares = 0; | |
7554 | int i; | |
95d07a2c | 7555 | int allowed; |
88758e9d DW |
7556 | |
7557 | for (d = a->info.devs ; d ; d = d->next) { | |
7558 | if ((d->curr_state & DS_FAULTY) && | |
7559 | d->state_fd >= 0) | |
7560 | /* wait for Removal to happen */ | |
7561 | return NULL; | |
7562 | if (d->state_fd >= 0) | |
7563 | failed--; | |
7564 | } | |
7565 | ||
7566 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
7567 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 7568 | |
e2962bfc AK |
7569 | if (imsm_reshape_blocks_arrays_changes(super)) |
7570 | return NULL; | |
1af97990 | 7571 | |
fc8ca064 AK |
7572 | /* Cannot activate another spare if rebuild is in progress already |
7573 | */ | |
7574 | if (is_rebuilding(dev)) { | |
7575 | dprintf("imsm: No spare activation allowed. " | |
7576 | "Rebuild in progress already.\n"); | |
7577 | return NULL; | |
7578 | } | |
7579 | ||
89c67882 AK |
7580 | if (a->info.array.level == 4) |
7581 | /* No repair for takeovered array | |
7582 | * imsm doesn't support raid4 | |
7583 | */ | |
7584 | return NULL; | |
7585 | ||
3b451610 AK |
7586 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
7587 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
7588 | return NULL; |
7589 | ||
95d07a2c LM |
7590 | /* |
7591 | * If there are any failed disks check state of the other volume. | |
7592 | * Block rebuild if the another one is failed until failed disks | |
7593 | * are removed from container. | |
7594 | */ | |
7595 | if (failed) { | |
c4acd1e5 | 7596 | dprintf("found failed disks in %.*s, check if there another" |
95d07a2c | 7597 | "failed sub-array.\n", |
c4acd1e5 | 7598 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
7599 | /* check if states of the other volumes allow for rebuild */ |
7600 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
7601 | if (i != inst) { | |
7602 | allowed = imsm_rebuild_allowed(a->container, | |
7603 | i, failed); | |
7604 | if (!allowed) | |
7605 | return NULL; | |
7606 | } | |
7607 | } | |
7608 | } | |
7609 | ||
88758e9d | 7610 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
7611 | for (i = 0; i < a->info.array.raid_disks; i++) { |
7612 | for (d = a->info.devs ; d ; d = d->next) | |
7613 | if (d->disk.raid_disk == i) | |
7614 | break; | |
7615 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
7616 | if (d && (d->state_fd >= 0)) | |
7617 | continue; | |
7618 | ||
272906ef | 7619 | /* |
a20d2ba5 DW |
7620 | * OK, this device needs recovery. Try to re-add the |
7621 | * previous occupant of this slot, if this fails see if | |
7622 | * we can continue the assimilation of a spare that was | |
7623 | * partially assimilated, finally try to activate a new | |
7624 | * spare. | |
272906ef DW |
7625 | */ |
7626 | dl = imsm_readd(super, i, a); | |
7627 | if (!dl) | |
b303fe21 | 7628 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 7629 | if (!dl) |
b303fe21 | 7630 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
7631 | if (!dl) |
7632 | continue; | |
7633 | ||
7634 | /* found a usable disk with enough space */ | |
503975b9 | 7635 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
7636 | |
7637 | /* dl->index will be -1 in the case we are activating a | |
7638 | * pristine spare. imsm_process_update() will create a | |
7639 | * new index in this case. Once a disk is found to be | |
7640 | * failed in all member arrays it is kicked from the | |
7641 | * metadata | |
7642 | */ | |
7643 | di->disk.number = dl->index; | |
d23fe947 | 7644 | |
272906ef DW |
7645 | /* (ab)use di->devs to store a pointer to the device |
7646 | * we chose | |
7647 | */ | |
7648 | di->devs = (struct mdinfo *) dl; | |
7649 | ||
7650 | di->disk.raid_disk = i; | |
7651 | di->disk.major = dl->major; | |
7652 | di->disk.minor = dl->minor; | |
7653 | di->disk.state = 0; | |
d23534e4 | 7654 | di->recovery_start = 0; |
5551b113 | 7655 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
7656 | di->component_size = a->info.component_size; |
7657 | di->container_member = inst; | |
148acb7b | 7658 | super->random = random32(); |
272906ef DW |
7659 | di->next = rv; |
7660 | rv = di; | |
7661 | num_spares++; | |
7662 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
7663 | i, di->data_offset); | |
88758e9d DW |
7664 | } |
7665 | ||
7666 | if (!rv) | |
7667 | /* No spares found */ | |
7668 | return rv; | |
7669 | /* Now 'rv' has a list of devices to return. | |
7670 | * Create a metadata_update record to update the | |
7671 | * disk_ord_tbl for the array | |
7672 | */ | |
503975b9 N |
7673 | mu = xmalloc(sizeof(*mu)); |
7674 | mu->buf = xcalloc(num_spares, | |
7675 | sizeof(struct imsm_update_activate_spare)); | |
88758e9d | 7676 | mu->space = NULL; |
cb23f1f4 | 7677 | mu->space_list = NULL; |
88758e9d DW |
7678 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
7679 | mu->next = *updates; | |
7680 | u = (struct imsm_update_activate_spare *) mu->buf; | |
7681 | ||
7682 | for (di = rv ; di ; di = di->next) { | |
7683 | u->type = update_activate_spare; | |
d23fe947 DW |
7684 | u->dl = (struct dl *) di->devs; |
7685 | di->devs = NULL; | |
88758e9d DW |
7686 | u->slot = di->disk.raid_disk; |
7687 | u->array = inst; | |
7688 | u->next = u + 1; | |
7689 | u++; | |
7690 | } | |
7691 | (u-1)->next = NULL; | |
7692 | *updates = mu; | |
7693 | ||
7694 | return rv; | |
7695 | } | |
7696 | ||
54c2c1ea | 7697 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 7698 | { |
54c2c1ea | 7699 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
7700 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7701 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
7702 | struct disk_info *inf = get_disk_info(u); |
7703 | struct imsm_disk *disk; | |
8273f55e DW |
7704 | int i; |
7705 | int j; | |
8273f55e | 7706 | |
54c2c1ea | 7707 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 7708 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
7709 | for (j = 0; j < new_map->num_members; j++) |
7710 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
7711 | return 1; |
7712 | } | |
7713 | ||
7714 | return 0; | |
7715 | } | |
7716 | ||
1a64be56 LM |
7717 | |
7718 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
7719 | { | |
7720 | struct dl *dl = NULL; | |
7721 | for (dl = super->disks; dl; dl = dl->next) | |
7722 | if ((dl->major == major) && (dl->minor == minor)) | |
7723 | return dl; | |
7724 | return NULL; | |
7725 | } | |
7726 | ||
7727 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7728 | { | |
7729 | struct dl *prev = NULL; | |
7730 | struct dl *dl; | |
7731 | ||
7732 | prev = NULL; | |
7733 | for (dl = super->disks; dl; dl = dl->next) { | |
7734 | if ((dl->major == major) && (dl->minor == minor)) { | |
7735 | /* remove */ | |
7736 | if (prev) | |
7737 | prev->next = dl->next; | |
7738 | else | |
7739 | super->disks = dl->next; | |
7740 | dl->next = NULL; | |
7741 | __free_imsm_disk(dl); | |
7742 | dprintf("%s: removed %x:%x\n", | |
7743 | __func__, major, minor); | |
7744 | break; | |
7745 | } | |
7746 | prev = dl; | |
7747 | } | |
7748 | return 0; | |
7749 | } | |
7750 | ||
f21e18ca | 7751 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7752 | |
1a64be56 LM |
7753 | static int add_remove_disk_update(struct intel_super *super) |
7754 | { | |
7755 | int check_degraded = 0; | |
7756 | struct dl *disk = NULL; | |
7757 | /* add/remove some spares to/from the metadata/contrainer */ | |
7758 | while (super->disk_mgmt_list) { | |
7759 | struct dl *disk_cfg; | |
7760 | ||
7761 | disk_cfg = super->disk_mgmt_list; | |
7762 | super->disk_mgmt_list = disk_cfg->next; | |
7763 | disk_cfg->next = NULL; | |
7764 | ||
7765 | if (disk_cfg->action == DISK_ADD) { | |
7766 | disk_cfg->next = super->disks; | |
7767 | super->disks = disk_cfg; | |
7768 | check_degraded = 1; | |
7769 | dprintf("%s: added %x:%x\n", | |
7770 | __func__, disk_cfg->major, | |
7771 | disk_cfg->minor); | |
7772 | } else if (disk_cfg->action == DISK_REMOVE) { | |
7773 | dprintf("Disk remove action processed: %x.%x\n", | |
7774 | disk_cfg->major, disk_cfg->minor); | |
7775 | disk = get_disk_super(super, | |
7776 | disk_cfg->major, | |
7777 | disk_cfg->minor); | |
7778 | if (disk) { | |
7779 | /* store action status */ | |
7780 | disk->action = DISK_REMOVE; | |
7781 | /* remove spare disks only */ | |
7782 | if (disk->index == -1) { | |
7783 | remove_disk_super(super, | |
7784 | disk_cfg->major, | |
7785 | disk_cfg->minor); | |
7786 | } | |
7787 | } | |
7788 | /* release allocate disk structure */ | |
7789 | __free_imsm_disk(disk_cfg); | |
7790 | } | |
7791 | } | |
7792 | return check_degraded; | |
7793 | } | |
7794 | ||
a29911da PC |
7795 | |
7796 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
7797 | struct intel_super *super, | |
7798 | void ***space_list) | |
7799 | { | |
7800 | struct intel_dev *id; | |
7801 | void **tofree = NULL; | |
7802 | int ret_val = 0; | |
7803 | ||
7804 | dprintf("apply_reshape_migration_update()\n"); | |
7805 | if ((u->subdev < 0) || | |
7806 | (u->subdev > 1)) { | |
7807 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7808 | return ret_val; | |
7809 | } | |
7810 | if ((space_list == NULL) || (*space_list == NULL)) { | |
7811 | dprintf("imsm: Error: Memory is not allocated\n"); | |
7812 | return ret_val; | |
7813 | } | |
7814 | ||
7815 | for (id = super->devlist ; id; id = id->next) { | |
7816 | if (id->index == (unsigned)u->subdev) { | |
7817 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7818 | struct imsm_map *map; | |
7819 | struct imsm_dev *new_dev = | |
7820 | (struct imsm_dev *)*space_list; | |
238c0a71 | 7821 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
7822 | int to_state; |
7823 | struct dl *new_disk; | |
7824 | ||
7825 | if (new_dev == NULL) | |
7826 | return ret_val; | |
7827 | *space_list = **space_list; | |
7828 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 7829 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
7830 | if (migr_map) { |
7831 | dprintf("imsm: Error: migration in progress"); | |
7832 | return ret_val; | |
7833 | } | |
7834 | ||
7835 | to_state = map->map_state; | |
7836 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
7837 | map->num_members++; | |
7838 | /* this should not happen */ | |
7839 | if (u->new_disks[0] < 0) { | |
7840 | map->failed_disk_num = | |
7841 | map->num_members - 1; | |
7842 | to_state = IMSM_T_STATE_DEGRADED; | |
7843 | } else | |
7844 | to_state = IMSM_T_STATE_NORMAL; | |
7845 | } | |
8e59f3d8 | 7846 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
7847 | if (u->new_level > -1) |
7848 | map->raid_level = u->new_level; | |
238c0a71 | 7849 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
7850 | if ((u->new_level == 5) && |
7851 | (migr_map->raid_level == 0)) { | |
7852 | int ord = map->num_members - 1; | |
7853 | migr_map->num_members--; | |
7854 | if (u->new_disks[0] < 0) | |
7855 | ord |= IMSM_ORD_REBUILD; | |
7856 | set_imsm_ord_tbl_ent(map, | |
7857 | map->num_members - 1, | |
7858 | ord); | |
7859 | } | |
7860 | id->dev = new_dev; | |
7861 | tofree = (void **)dev; | |
7862 | ||
4bba0439 PC |
7863 | /* update chunk size |
7864 | */ | |
7865 | if (u->new_chunksize > 0) | |
7866 | map->blocks_per_strip = | |
7867 | __cpu_to_le16(u->new_chunksize * 2); | |
7868 | ||
a29911da PC |
7869 | /* add disk |
7870 | */ | |
7871 | if ((u->new_level != 5) || | |
7872 | (migr_map->raid_level != 0) || | |
7873 | (migr_map->raid_level == map->raid_level)) | |
7874 | goto skip_disk_add; | |
7875 | ||
7876 | if (u->new_disks[0] >= 0) { | |
7877 | /* use passes spare | |
7878 | */ | |
7879 | new_disk = get_disk_super(super, | |
7880 | major(u->new_disks[0]), | |
7881 | minor(u->new_disks[0])); | |
7882 | dprintf("imsm: new disk for reshape is: %i:%i " | |
7883 | "(%p, index = %i)\n", | |
7884 | major(u->new_disks[0]), | |
7885 | minor(u->new_disks[0]), | |
7886 | new_disk, new_disk->index); | |
7887 | if (new_disk == NULL) | |
7888 | goto error_disk_add; | |
7889 | ||
7890 | new_disk->index = map->num_members - 1; | |
7891 | /* slot to fill in autolayout | |
7892 | */ | |
7893 | new_disk->raiddisk = new_disk->index; | |
7894 | new_disk->disk.status |= CONFIGURED_DISK; | |
7895 | new_disk->disk.status &= ~SPARE_DISK; | |
7896 | } else | |
7897 | goto error_disk_add; | |
7898 | ||
7899 | skip_disk_add: | |
7900 | *tofree = *space_list; | |
7901 | /* calculate new size | |
7902 | */ | |
f3871fdc | 7903 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
7904 | |
7905 | ret_val = 1; | |
7906 | } | |
7907 | } | |
7908 | ||
7909 | if (tofree) | |
7910 | *space_list = tofree; | |
7911 | return ret_val; | |
7912 | ||
7913 | error_disk_add: | |
7914 | dprintf("Error: imsm: Cannot find disk.\n"); | |
7915 | return ret_val; | |
7916 | } | |
7917 | ||
f3871fdc AK |
7918 | static int apply_size_change_update(struct imsm_update_size_change *u, |
7919 | struct intel_super *super) | |
7920 | { | |
7921 | struct intel_dev *id; | |
7922 | int ret_val = 0; | |
7923 | ||
7924 | dprintf("apply_size_change_update()\n"); | |
7925 | if ((u->subdev < 0) || | |
7926 | (u->subdev > 1)) { | |
7927 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7928 | return ret_val; | |
7929 | } | |
7930 | ||
7931 | for (id = super->devlist ; id; id = id->next) { | |
7932 | if (id->index == (unsigned)u->subdev) { | |
7933 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7934 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
7935 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
7936 | unsigned long long blocks_per_member; | |
7937 | ||
7938 | /* calculate new size | |
7939 | */ | |
7940 | blocks_per_member = u->new_size / used_disks; | |
7941 | dprintf("imsm: apply_size_change_update(size: %llu, " | |
7942 | "blocks per member: %llu)\n", | |
7943 | u->new_size, blocks_per_member); | |
7944 | set_blocks_per_member(map, blocks_per_member); | |
7945 | imsm_set_array_size(dev, u->new_size); | |
7946 | ||
7947 | ret_val = 1; | |
7948 | break; | |
7949 | } | |
7950 | } | |
7951 | ||
7952 | return ret_val; | |
7953 | } | |
7954 | ||
7955 | ||
061d7da3 | 7956 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 7957 | struct intel_super *super, |
061d7da3 LO |
7958 | struct active_array *active_array) |
7959 | { | |
7960 | struct imsm_super *mpb = super->anchor; | |
7961 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 7962 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
7963 | struct imsm_map *migr_map; |
7964 | struct active_array *a; | |
7965 | struct imsm_disk *disk; | |
7966 | __u8 to_state; | |
7967 | struct dl *dl; | |
7968 | unsigned int found; | |
7969 | int failed; | |
5961eeec | 7970 | int victim; |
061d7da3 | 7971 | int i; |
5961eeec | 7972 | int second_map_created = 0; |
061d7da3 | 7973 | |
5961eeec | 7974 | for (; u; u = u->next) { |
238c0a71 | 7975 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 7976 | |
5961eeec | 7977 | if (victim < 0) |
7978 | return 0; | |
061d7da3 | 7979 | |
5961eeec | 7980 | for (dl = super->disks; dl; dl = dl->next) |
7981 | if (dl == u->dl) | |
7982 | break; | |
061d7da3 | 7983 | |
5961eeec | 7984 | if (!dl) { |
7985 | fprintf(stderr, "error: imsm_activate_spare passed " | |
7986 | "an unknown disk (index: %d)\n", | |
7987 | u->dl->index); | |
7988 | return 0; | |
7989 | } | |
061d7da3 | 7990 | |
5961eeec | 7991 | /* count failures (excluding rebuilds and the victim) |
7992 | * to determine map[0] state | |
7993 | */ | |
7994 | failed = 0; | |
7995 | for (i = 0; i < map->num_members; i++) { | |
7996 | if (i == u->slot) | |
7997 | continue; | |
7998 | disk = get_imsm_disk(super, | |
238c0a71 | 7999 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 8000 | if (!disk || is_failed(disk)) |
8001 | failed++; | |
8002 | } | |
061d7da3 | 8003 | |
5961eeec | 8004 | /* adding a pristine spare, assign a new index */ |
8005 | if (dl->index < 0) { | |
8006 | dl->index = super->anchor->num_disks; | |
8007 | super->anchor->num_disks++; | |
8008 | } | |
8009 | disk = &dl->disk; | |
8010 | disk->status |= CONFIGURED_DISK; | |
8011 | disk->status &= ~SPARE_DISK; | |
8012 | ||
8013 | /* mark rebuild */ | |
238c0a71 | 8014 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 8015 | if (!second_map_created) { |
8016 | second_map_created = 1; | |
8017 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8018 | migrate(dev, super, to_state, MIGR_REBUILD); | |
8019 | } else | |
8020 | map->map_state = to_state; | |
238c0a71 | 8021 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 8022 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
8023 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
8024 | dl->index | IMSM_ORD_REBUILD); | |
8025 | ||
8026 | /* update the family_num to mark a new container | |
8027 | * generation, being careful to record the existing | |
8028 | * family_num in orig_family_num to clean up after | |
8029 | * earlier mdadm versions that neglected to set it. | |
8030 | */ | |
8031 | if (mpb->orig_family_num == 0) | |
8032 | mpb->orig_family_num = mpb->family_num; | |
8033 | mpb->family_num += super->random; | |
8034 | ||
8035 | /* count arrays using the victim in the metadata */ | |
8036 | found = 0; | |
8037 | for (a = active_array; a ; a = a->next) { | |
8038 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8039 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 8040 | |
5961eeec | 8041 | if (get_imsm_disk_slot(map, victim) >= 0) |
8042 | found++; | |
8043 | } | |
061d7da3 | 8044 | |
5961eeec | 8045 | /* delete the victim if it is no longer being |
8046 | * utilized anywhere | |
061d7da3 | 8047 | */ |
5961eeec | 8048 | if (!found) { |
8049 | struct dl **dlp; | |
061d7da3 | 8050 | |
5961eeec | 8051 | /* We know that 'manager' isn't touching anything, |
8052 | * so it is safe to delete | |
8053 | */ | |
8054 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
8055 | if ((*dlp)->index == victim) |
8056 | break; | |
5961eeec | 8057 | |
8058 | /* victim may be on the missing list */ | |
8059 | if (!*dlp) | |
8060 | for (dlp = &super->missing; *dlp; | |
8061 | dlp = &(*dlp)->next) | |
8062 | if ((*dlp)->index == victim) | |
8063 | break; | |
8064 | imsm_delete(super, dlp, victim); | |
8065 | } | |
061d7da3 LO |
8066 | } |
8067 | ||
8068 | return 1; | |
8069 | } | |
a29911da | 8070 | |
2e5dc010 N |
8071 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
8072 | struct intel_super *super, | |
8073 | void ***space_list) | |
8074 | { | |
8075 | struct dl *new_disk; | |
8076 | struct intel_dev *id; | |
8077 | int i; | |
8078 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 8079 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
8080 | void **tofree = NULL; |
8081 | int devices_to_reshape = 1; | |
8082 | struct imsm_super *mpb = super->anchor; | |
8083 | int ret_val = 0; | |
d098291a | 8084 | unsigned int dev_id; |
2e5dc010 | 8085 | |
ed7333bd | 8086 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
8087 | |
8088 | /* enable spares to use in array */ | |
8089 | for (i = 0; i < delta_disks; i++) { | |
8090 | new_disk = get_disk_super(super, | |
8091 | major(u->new_disks[i]), | |
8092 | minor(u->new_disks[i])); | |
ed7333bd AK |
8093 | dprintf("imsm: new disk for reshape is: %i:%i " |
8094 | "(%p, index = %i)\n", | |
2e5dc010 N |
8095 | major(u->new_disks[i]), minor(u->new_disks[i]), |
8096 | new_disk, new_disk->index); | |
8097 | if ((new_disk == NULL) || | |
8098 | ((new_disk->index >= 0) && | |
8099 | (new_disk->index < u->old_raid_disks))) | |
8100 | goto update_reshape_exit; | |
ee4beede | 8101 | new_disk->index = disk_count++; |
2e5dc010 N |
8102 | /* slot to fill in autolayout |
8103 | */ | |
8104 | new_disk->raiddisk = new_disk->index; | |
8105 | new_disk->disk.status |= | |
8106 | CONFIGURED_DISK; | |
8107 | new_disk->disk.status &= ~SPARE_DISK; | |
8108 | } | |
8109 | ||
ed7333bd AK |
8110 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
8111 | mpb->num_raid_devs); | |
2e5dc010 N |
8112 | /* manage changes in volume |
8113 | */ | |
d098291a | 8114 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
8115 | void **sp = *space_list; |
8116 | struct imsm_dev *newdev; | |
8117 | struct imsm_map *newmap, *oldmap; | |
8118 | ||
d098291a AK |
8119 | for (id = super->devlist ; id; id = id->next) { |
8120 | if (id->index == dev_id) | |
8121 | break; | |
8122 | } | |
8123 | if (id == NULL) | |
8124 | break; | |
2e5dc010 N |
8125 | if (!sp) |
8126 | continue; | |
8127 | *space_list = *sp; | |
8128 | newdev = (void*)sp; | |
8129 | /* Copy the dev, but not (all of) the map */ | |
8130 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
8131 | oldmap = get_imsm_map(id->dev, MAP_0); |
8132 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
8133 | /* Copy the current map */ |
8134 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
8135 | /* update one device only | |
8136 | */ | |
8137 | if (devices_to_reshape) { | |
ed7333bd AK |
8138 | dprintf("imsm: modifying subdev: %i\n", |
8139 | id->index); | |
2e5dc010 N |
8140 | devices_to_reshape--; |
8141 | newdev->vol.migr_state = 1; | |
8142 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8143 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8144 | newmap->num_members = u->new_raid_disks; |
8145 | for (i = 0; i < delta_disks; i++) { | |
8146 | set_imsm_ord_tbl_ent(newmap, | |
8147 | u->old_raid_disks + i, | |
8148 | u->old_raid_disks + i); | |
8149 | } | |
8150 | /* New map is correct, now need to save old map | |
8151 | */ | |
238c0a71 | 8152 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8153 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8154 | ||
f3871fdc | 8155 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
8156 | } |
8157 | ||
8158 | sp = (void **)id->dev; | |
8159 | id->dev = newdev; | |
8160 | *sp = tofree; | |
8161 | tofree = sp; | |
8e59f3d8 AK |
8162 | |
8163 | /* Clear migration record */ | |
8164 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8165 | } |
819bc634 AK |
8166 | if (tofree) |
8167 | *space_list = tofree; | |
2e5dc010 N |
8168 | ret_val = 1; |
8169 | ||
8170 | update_reshape_exit: | |
8171 | ||
8172 | return ret_val; | |
8173 | } | |
8174 | ||
bb025c2f | 8175 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8176 | struct intel_super *super, |
8177 | void ***space_list) | |
bb025c2f KW |
8178 | { |
8179 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8180 | struct intel_dev *dv; |
8181 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8182 | struct imsm_map *map; |
8183 | struct dl *dm, *du; | |
8ca6df95 | 8184 | int i; |
bb025c2f KW |
8185 | |
8186 | for (dv = super->devlist; dv; dv = dv->next) | |
8187 | if (dv->index == (unsigned int)u->subarray) { | |
8188 | dev = dv->dev; | |
8189 | break; | |
8190 | } | |
8191 | ||
8192 | if (dev == NULL) | |
8193 | return 0; | |
8194 | ||
238c0a71 | 8195 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8196 | |
8197 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 8198 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
8199 | if (imsm_count_failed(super, dev, MAP_0) != |
8200 | (map->num_members / 2)) | |
43d5ec18 KW |
8201 | return 0; |
8202 | ||
bb025c2f KW |
8203 | /* iterate through devices to mark removed disks as spare */ |
8204 | for (dm = super->disks; dm; dm = dm->next) { | |
8205 | if (dm->disk.status & FAILED_DISK) { | |
8206 | int idx = dm->index; | |
8207 | /* update indexes on the disk list */ | |
8208 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
8209 | the index values will end up being correct.... NB */ | |
8210 | for (du = super->disks; du; du = du->next) | |
8211 | if (du->index > idx) | |
8212 | du->index--; | |
8213 | /* mark as spare disk */ | |
a8619d23 | 8214 | mark_spare(dm); |
bb025c2f KW |
8215 | } |
8216 | } | |
bb025c2f KW |
8217 | /* update map */ |
8218 | map->num_members = map->num_members / 2; | |
8219 | map->map_state = IMSM_T_STATE_NORMAL; | |
8220 | map->num_domains = 1; | |
8221 | map->raid_level = 0; | |
8222 | map->failed_disk_num = -1; | |
8223 | } | |
8224 | ||
8ca6df95 KW |
8225 | if (u->direction == R0_TO_R10) { |
8226 | void **space; | |
8227 | /* update slots in current disk list */ | |
8228 | for (dm = super->disks; dm; dm = dm->next) { | |
8229 | if (dm->index >= 0) | |
8230 | dm->index *= 2; | |
8231 | } | |
8232 | /* create new *missing* disks */ | |
8233 | for (i = 0; i < map->num_members; i++) { | |
8234 | space = *space_list; | |
8235 | if (!space) | |
8236 | continue; | |
8237 | *space_list = *space; | |
8238 | du = (void *)space; | |
8239 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
8240 | du->fd = -1; |
8241 | du->minor = 0; | |
8242 | du->major = 0; | |
8243 | du->index = (i * 2) + 1; | |
8244 | sprintf((char *)du->disk.serial, | |
8245 | " MISSING_%d", du->index); | |
8246 | sprintf((char *)du->serial, | |
8247 | "MISSING_%d", du->index); | |
8248 | du->next = super->missing; | |
8249 | super->missing = du; | |
8250 | } | |
8251 | /* create new dev and map */ | |
8252 | space = *space_list; | |
8253 | if (!space) | |
8254 | return 0; | |
8255 | *space_list = *space; | |
8256 | dev_new = (void *)space; | |
8257 | memcpy(dev_new, dev, sizeof(*dev)); | |
8258 | /* update new map */ | |
238c0a71 | 8259 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 8260 | map->num_members = map->num_members * 2; |
1a2487c2 | 8261 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
8262 | map->num_domains = 2; |
8263 | map->raid_level = 1; | |
8264 | /* replace dev<->dev_new */ | |
8265 | dv->dev = dev_new; | |
8266 | } | |
bb025c2f KW |
8267 | /* update disk order table */ |
8268 | for (du = super->disks; du; du = du->next) | |
8269 | if (du->index >= 0) | |
8270 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 8271 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
8272 | if (du->index >= 0) { |
8273 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 8274 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 8275 | } |
bb025c2f KW |
8276 | |
8277 | return 1; | |
8278 | } | |
8279 | ||
e8319a19 DW |
8280 | static void imsm_process_update(struct supertype *st, |
8281 | struct metadata_update *update) | |
8282 | { | |
8283 | /** | |
8284 | * crack open the metadata_update envelope to find the update record | |
8285 | * update can be one of: | |
d195167d AK |
8286 | * update_reshape_container_disks - all the arrays in the container |
8287 | * are being reshaped to have more devices. We need to mark | |
8288 | * the arrays for general migration and convert selected spares | |
8289 | * into active devices. | |
8290 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
8291 | * device in an array, update the disk_ord_tbl. If this disk is |
8292 | * present in all member arrays then also clear the SPARE_DISK | |
8293 | * flag | |
d195167d AK |
8294 | * update_create_array |
8295 | * update_kill_array | |
8296 | * update_rename_array | |
8297 | * update_add_remove_disk | |
e8319a19 DW |
8298 | */ |
8299 | struct intel_super *super = st->sb; | |
4d7b1503 | 8300 | struct imsm_super *mpb; |
e8319a19 DW |
8301 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
8302 | ||
4d7b1503 DW |
8303 | /* update requires a larger buf but the allocation failed */ |
8304 | if (super->next_len && !super->next_buf) { | |
8305 | super->next_len = 0; | |
8306 | return; | |
8307 | } | |
8308 | ||
8309 | if (super->next_buf) { | |
8310 | memcpy(super->next_buf, super->buf, super->len); | |
8311 | free(super->buf); | |
8312 | super->len = super->next_len; | |
8313 | super->buf = super->next_buf; | |
8314 | ||
8315 | super->next_len = 0; | |
8316 | super->next_buf = NULL; | |
8317 | } | |
8318 | ||
8319 | mpb = super->anchor; | |
8320 | ||
e8319a19 | 8321 | switch (type) { |
0ec5d470 AK |
8322 | case update_general_migration_checkpoint: { |
8323 | struct intel_dev *id; | |
8324 | struct imsm_update_general_migration_checkpoint *u = | |
8325 | (void *)update->buf; | |
8326 | ||
8327 | dprintf("imsm: process_update() " | |
8328 | "for update_general_migration_checkpoint called\n"); | |
8329 | ||
8330 | /* find device under general migration */ | |
8331 | for (id = super->devlist ; id; id = id->next) { | |
8332 | if (is_gen_migration(id->dev)) { | |
8333 | id->dev->vol.curr_migr_unit = | |
8334 | __cpu_to_le32(u->curr_migr_unit); | |
8335 | super->updates_pending++; | |
8336 | } | |
8337 | } | |
8338 | break; | |
8339 | } | |
bb025c2f KW |
8340 | case update_takeover: { |
8341 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
8342 | if (apply_takeover_update(u, super, &update->space_list)) { |
8343 | imsm_update_version_info(super); | |
bb025c2f | 8344 | super->updates_pending++; |
1a2487c2 | 8345 | } |
bb025c2f KW |
8346 | break; |
8347 | } | |
8348 | ||
78b10e66 | 8349 | case update_reshape_container_disks: { |
d195167d | 8350 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
8351 | if (apply_reshape_container_disks_update( |
8352 | u, super, &update->space_list)) | |
8353 | super->updates_pending++; | |
78b10e66 N |
8354 | break; |
8355 | } | |
48c5303a | 8356 | case update_reshape_migration: { |
a29911da PC |
8357 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
8358 | if (apply_reshape_migration_update( | |
8359 | u, super, &update->space_list)) | |
8360 | super->updates_pending++; | |
48c5303a PC |
8361 | break; |
8362 | } | |
f3871fdc AK |
8363 | case update_size_change: { |
8364 | struct imsm_update_size_change *u = (void *)update->buf; | |
8365 | if (apply_size_change_update(u, super)) | |
8366 | super->updates_pending++; | |
8367 | break; | |
8368 | } | |
e8319a19 DW |
8369 | case update_activate_spare: { |
8370 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
061d7da3 LO |
8371 | if (apply_update_activate_spare(u, super, st->arrays)) |
8372 | super->updates_pending++; | |
8273f55e DW |
8373 | break; |
8374 | } | |
8375 | case update_create_array: { | |
8376 | /* someone wants to create a new array, we need to be aware of | |
8377 | * a few races/collisions: | |
8378 | * 1/ 'Create' called by two separate instances of mdadm | |
8379 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
8380 | * devices that have since been assimilated via | |
8381 | * activate_spare. | |
8382 | * In the event this update can not be carried out mdadm will | |
8383 | * (FIX ME) notice that its update did not take hold. | |
8384 | */ | |
8385 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8386 | struct intel_dev *dv; |
8273f55e DW |
8387 | struct imsm_dev *dev; |
8388 | struct imsm_map *map, *new_map; | |
8389 | unsigned long long start, end; | |
8390 | unsigned long long new_start, new_end; | |
8391 | int i; | |
54c2c1ea DW |
8392 | struct disk_info *inf; |
8393 | struct dl *dl; | |
8273f55e DW |
8394 | |
8395 | /* handle racing creates: first come first serve */ | |
8396 | if (u->dev_idx < mpb->num_raid_devs) { | |
8397 | dprintf("%s: subarray %d already defined\n", | |
8398 | __func__, u->dev_idx); | |
ba2de7ba | 8399 | goto create_error; |
8273f55e DW |
8400 | } |
8401 | ||
8402 | /* check update is next in sequence */ | |
8403 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
8404 | dprintf("%s: can not create array %d expected index %d\n", |
8405 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 8406 | goto create_error; |
8273f55e DW |
8407 | } |
8408 | ||
238c0a71 | 8409 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
8410 | new_start = pba_of_lba0(new_map); |
8411 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 8412 | inf = get_disk_info(u); |
8273f55e DW |
8413 | |
8414 | /* handle activate_spare versus create race: | |
8415 | * check to make sure that overlapping arrays do not include | |
8416 | * overalpping disks | |
8417 | */ | |
8418 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 8419 | dev = get_imsm_dev(super, i); |
238c0a71 | 8420 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
8421 | start = pba_of_lba0(map); |
8422 | end = start + blocks_per_member(map); | |
8273f55e DW |
8423 | if ((new_start >= start && new_start <= end) || |
8424 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
8425 | /* overlap */; |
8426 | else | |
8427 | continue; | |
8428 | ||
8429 | if (disks_overlap(super, i, u)) { | |
8273f55e | 8430 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 8431 | goto create_error; |
8273f55e DW |
8432 | } |
8433 | } | |
8273f55e | 8434 | |
949c47a0 DW |
8435 | /* check that prepare update was successful */ |
8436 | if (!update->space) { | |
8437 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 8438 | goto create_error; |
949c47a0 DW |
8439 | } |
8440 | ||
54c2c1ea DW |
8441 | /* check that all disks are still active before committing |
8442 | * changes. FIXME: could we instead handle this by creating a | |
8443 | * degraded array? That's probably not what the user expects, | |
8444 | * so better to drop this update on the floor. | |
8445 | */ | |
8446 | for (i = 0; i < new_map->num_members; i++) { | |
8447 | dl = serial_to_dl(inf[i].serial, super); | |
8448 | if (!dl) { | |
8449 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 8450 | goto create_error; |
54c2c1ea | 8451 | } |
949c47a0 DW |
8452 | } |
8453 | ||
8273f55e | 8454 | super->updates_pending++; |
54c2c1ea DW |
8455 | |
8456 | /* convert spares to members and fixup ord_tbl */ | |
8457 | for (i = 0; i < new_map->num_members; i++) { | |
8458 | dl = serial_to_dl(inf[i].serial, super); | |
8459 | if (dl->index == -1) { | |
8460 | dl->index = mpb->num_disks; | |
8461 | mpb->num_disks++; | |
8462 | dl->disk.status |= CONFIGURED_DISK; | |
8463 | dl->disk.status &= ~SPARE_DISK; | |
8464 | } | |
8465 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
8466 | } | |
8467 | ||
ba2de7ba DW |
8468 | dv = update->space; |
8469 | dev = dv->dev; | |
949c47a0 DW |
8470 | update->space = NULL; |
8471 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
8472 | dv->index = u->dev_idx; |
8473 | dv->next = super->devlist; | |
8474 | super->devlist = dv; | |
8273f55e | 8475 | mpb->num_raid_devs++; |
8273f55e | 8476 | |
4d1313e9 | 8477 | imsm_update_version_info(super); |
8273f55e | 8478 | break; |
ba2de7ba DW |
8479 | create_error: |
8480 | /* mdmon knows how to release update->space, but not | |
8481 | * ((struct intel_dev *) update->space)->dev | |
8482 | */ | |
8483 | if (update->space) { | |
8484 | dv = update->space; | |
8485 | free(dv->dev); | |
8486 | } | |
8273f55e | 8487 | break; |
e8319a19 | 8488 | } |
33414a01 DW |
8489 | case update_kill_array: { |
8490 | struct imsm_update_kill_array *u = (void *) update->buf; | |
8491 | int victim = u->dev_idx; | |
8492 | struct active_array *a; | |
8493 | struct intel_dev **dp; | |
8494 | struct imsm_dev *dev; | |
8495 | ||
8496 | /* sanity check that we are not affecting the uuid of | |
8497 | * active arrays, or deleting an active array | |
8498 | * | |
8499 | * FIXME when immutable ids are available, but note that | |
8500 | * we'll also need to fixup the invalidated/active | |
8501 | * subarray indexes in mdstat | |
8502 | */ | |
8503 | for (a = st->arrays; a; a = a->next) | |
8504 | if (a->info.container_member >= victim) | |
8505 | break; | |
8506 | /* by definition if mdmon is running at least one array | |
8507 | * is active in the container, so checking | |
8508 | * mpb->num_raid_devs is just extra paranoia | |
8509 | */ | |
8510 | dev = get_imsm_dev(super, victim); | |
8511 | if (a || !dev || mpb->num_raid_devs == 1) { | |
8512 | dprintf("failed to delete subarray-%d\n", victim); | |
8513 | break; | |
8514 | } | |
8515 | ||
8516 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 8517 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
8518 | *dp = (*dp)->next; |
8519 | } else { | |
f21e18ca | 8520 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
8521 | (*dp)->index--; |
8522 | dp = &(*dp)->next; | |
8523 | } | |
8524 | mpb->num_raid_devs--; | |
8525 | super->updates_pending++; | |
8526 | break; | |
8527 | } | |
aa534678 DW |
8528 | case update_rename_array: { |
8529 | struct imsm_update_rename_array *u = (void *) update->buf; | |
8530 | char name[MAX_RAID_SERIAL_LEN+1]; | |
8531 | int target = u->dev_idx; | |
8532 | struct active_array *a; | |
8533 | struct imsm_dev *dev; | |
8534 | ||
8535 | /* sanity check that we are not affecting the uuid of | |
8536 | * an active array | |
8537 | */ | |
8538 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
8539 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
8540 | for (a = st->arrays; a; a = a->next) | |
8541 | if (a->info.container_member == target) | |
8542 | break; | |
8543 | dev = get_imsm_dev(super, u->dev_idx); | |
8544 | if (a || !dev || !check_name(super, name, 1)) { | |
8545 | dprintf("failed to rename subarray-%d\n", target); | |
8546 | break; | |
8547 | } | |
8548 | ||
cdbe98cd | 8549 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
8550 | super->updates_pending++; |
8551 | break; | |
8552 | } | |
1a64be56 | 8553 | case update_add_remove_disk: { |
43dad3d6 | 8554 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
8555 | * being added, check teh status of add_remove_disk |
8556 | * if discs has been added. | |
8557 | */ | |
8558 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 8559 | struct active_array *a; |
072b727f DW |
8560 | |
8561 | super->updates_pending++; | |
1a64be56 | 8562 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
8563 | a->check_degraded = 1; |
8564 | } | |
43dad3d6 | 8565 | break; |
e8319a19 | 8566 | } |
1a64be56 LM |
8567 | default: |
8568 | fprintf(stderr, "error: unsuported process update type:" | |
8569 | "(type: %d)\n", type); | |
8570 | } | |
e8319a19 | 8571 | } |
88758e9d | 8572 | |
bc0b9d34 PC |
8573 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
8574 | ||
8273f55e DW |
8575 | static void imsm_prepare_update(struct supertype *st, |
8576 | struct metadata_update *update) | |
8577 | { | |
949c47a0 | 8578 | /** |
4d7b1503 DW |
8579 | * Allocate space to hold new disk entries, raid-device entries or a new |
8580 | * mpb if necessary. The manager synchronously waits for updates to | |
8581 | * complete in the monitor, so new mpb buffers allocated here can be | |
8582 | * integrated by the monitor thread without worrying about live pointers | |
8583 | * in the manager thread. | |
8273f55e | 8584 | */ |
949c47a0 | 8585 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
8586 | struct intel_super *super = st->sb; |
8587 | struct imsm_super *mpb = super->anchor; | |
8588 | size_t buf_len; | |
8589 | size_t len = 0; | |
949c47a0 DW |
8590 | |
8591 | switch (type) { | |
0ec5d470 AK |
8592 | case update_general_migration_checkpoint: |
8593 | dprintf("imsm: prepare_update() " | |
8594 | "for update_general_migration_checkpoint called\n"); | |
8595 | break; | |
abedf5fc KW |
8596 | case update_takeover: { |
8597 | struct imsm_update_takeover *u = (void *)update->buf; | |
8598 | if (u->direction == R0_TO_R10) { | |
8599 | void **tail = (void **)&update->space_list; | |
8600 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 8601 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
8602 | int num_members = map->num_members; |
8603 | void *space; | |
8604 | int size, i; | |
abedf5fc KW |
8605 | /* allocate memory for added disks */ |
8606 | for (i = 0; i < num_members; i++) { | |
8607 | size = sizeof(struct dl); | |
503975b9 | 8608 | space = xmalloc(size); |
abedf5fc KW |
8609 | *tail = space; |
8610 | tail = space; | |
8611 | *tail = NULL; | |
8612 | } | |
8613 | /* allocate memory for new device */ | |
8614 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
8615 | (num_members * sizeof(__u32)); | |
503975b9 N |
8616 | space = xmalloc(size); |
8617 | *tail = space; | |
8618 | tail = space; | |
8619 | *tail = NULL; | |
8620 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
8621 | } |
8622 | ||
8623 | break; | |
8624 | } | |
78b10e66 | 8625 | case update_reshape_container_disks: { |
d195167d AK |
8626 | /* Every raid device in the container is about to |
8627 | * gain some more devices, and we will enter a | |
8628 | * reconfiguration. | |
8629 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
8630 | * will now hold 2 of them. | |
8631 | * Thus we need new 'struct imsm_dev' allocations sized | |
8632 | * as sizeof_imsm_dev but with more devices in both maps. | |
8633 | */ | |
8634 | struct imsm_update_reshape *u = (void *)update->buf; | |
8635 | struct intel_dev *dl; | |
8636 | void **space_tail = (void**)&update->space_list; | |
8637 | ||
8638 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8639 | ||
8640 | for (dl = super->devlist; dl; dl = dl->next) { | |
8641 | int size = sizeof_imsm_dev(dl->dev, 1); | |
8642 | void *s; | |
d677e0b8 AK |
8643 | if (u->new_raid_disks > u->old_raid_disks) |
8644 | size += sizeof(__u32)*2* | |
8645 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8646 | s = xmalloc(size); |
d195167d AK |
8647 | *space_tail = s; |
8648 | space_tail = s; | |
8649 | *space_tail = NULL; | |
8650 | } | |
8651 | ||
8652 | len = disks_to_mpb_size(u->new_raid_disks); | |
8653 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
8654 | break; |
8655 | } | |
48c5303a | 8656 | case update_reshape_migration: { |
bc0b9d34 PC |
8657 | /* for migration level 0->5 we need to add disks |
8658 | * so the same as for container operation we will copy | |
8659 | * device to the bigger location. | |
8660 | * in memory prepared device and new disk area are prepared | |
8661 | * for usage in process update | |
8662 | */ | |
8663 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
8664 | struct intel_dev *id; | |
8665 | void **space_tail = (void **)&update->space_list; | |
8666 | int size; | |
8667 | void *s; | |
8668 | int current_level = -1; | |
8669 | ||
8670 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8671 | ||
8672 | /* add space for bigger array in update | |
8673 | */ | |
8674 | for (id = super->devlist; id; id = id->next) { | |
8675 | if (id->index == (unsigned)u->subdev) { | |
8676 | size = sizeof_imsm_dev(id->dev, 1); | |
8677 | if (u->new_raid_disks > u->old_raid_disks) | |
8678 | size += sizeof(__u32)*2* | |
8679 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8680 | s = xmalloc(size); |
bc0b9d34 PC |
8681 | *space_tail = s; |
8682 | space_tail = s; | |
8683 | *space_tail = NULL; | |
8684 | break; | |
8685 | } | |
8686 | } | |
8687 | if (update->space_list == NULL) | |
8688 | break; | |
8689 | ||
8690 | /* add space for disk in update | |
8691 | */ | |
8692 | size = sizeof(struct dl); | |
503975b9 | 8693 | s = xmalloc(size); |
bc0b9d34 PC |
8694 | *space_tail = s; |
8695 | space_tail = s; | |
8696 | *space_tail = NULL; | |
8697 | ||
8698 | /* add spare device to update | |
8699 | */ | |
8700 | for (id = super->devlist ; id; id = id->next) | |
8701 | if (id->index == (unsigned)u->subdev) { | |
8702 | struct imsm_dev *dev; | |
8703 | struct imsm_map *map; | |
8704 | ||
8705 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 8706 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
8707 | current_level = map->raid_level; |
8708 | break; | |
8709 | } | |
8710 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
8711 | struct mdinfo *spares; | |
8712 | ||
8713 | spares = get_spares_for_grow(st); | |
8714 | if (spares) { | |
8715 | struct dl *dl; | |
8716 | struct mdinfo *dev; | |
8717 | ||
8718 | dev = spares->devs; | |
8719 | if (dev) { | |
8720 | u->new_disks[0] = | |
8721 | makedev(dev->disk.major, | |
8722 | dev->disk.minor); | |
8723 | dl = get_disk_super(super, | |
8724 | dev->disk.major, | |
8725 | dev->disk.minor); | |
8726 | dl->index = u->old_raid_disks; | |
8727 | dev = dev->next; | |
8728 | } | |
8729 | sysfs_free(spares); | |
8730 | } | |
8731 | } | |
8732 | len = disks_to_mpb_size(u->new_raid_disks); | |
8733 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8734 | break; |
8735 | } | |
f3871fdc AK |
8736 | case update_size_change: { |
8737 | break; | |
8738 | } | |
949c47a0 DW |
8739 | case update_create_array: { |
8740 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8741 | struct intel_dev *dv; |
54c2c1ea | 8742 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8743 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8744 | struct dl *dl; |
8745 | struct disk_info *inf; | |
8746 | int i; | |
8747 | int activate = 0; | |
949c47a0 | 8748 | |
54c2c1ea DW |
8749 | inf = get_disk_info(u); |
8750 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 8751 | /* allocate a new super->devlist entry */ |
503975b9 N |
8752 | dv = xmalloc(sizeof(*dv)); |
8753 | dv->dev = xmalloc(len); | |
8754 | update->space = dv; | |
949c47a0 | 8755 | |
54c2c1ea DW |
8756 | /* count how many spares will be converted to members */ |
8757 | for (i = 0; i < map->num_members; i++) { | |
8758 | dl = serial_to_dl(inf[i].serial, super); | |
8759 | if (!dl) { | |
8760 | /* hmm maybe it failed?, nothing we can do about | |
8761 | * it here | |
8762 | */ | |
8763 | continue; | |
8764 | } | |
8765 | if (count_memberships(dl, super) == 0) | |
8766 | activate++; | |
8767 | } | |
8768 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
8769 | break; |
8770 | default: | |
8771 | break; | |
8772 | } | |
8773 | } | |
8273f55e | 8774 | |
4d7b1503 DW |
8775 | /* check if we need a larger metadata buffer */ |
8776 | if (super->next_buf) | |
8777 | buf_len = super->next_len; | |
8778 | else | |
8779 | buf_len = super->len; | |
8780 | ||
8781 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
8782 | /* ok we need a larger buf than what is currently allocated | |
8783 | * if this allocation fails process_update will notice that | |
8784 | * ->next_len is set and ->next_buf is NULL | |
8785 | */ | |
8786 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
8787 | if (super->next_buf) | |
8788 | free(super->next_buf); | |
8789 | ||
8790 | super->next_len = buf_len; | |
1f45a8ad DW |
8791 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
8792 | memset(super->next_buf, 0, buf_len); | |
8793 | else | |
4d7b1503 DW |
8794 | super->next_buf = NULL; |
8795 | } | |
8273f55e DW |
8796 | } |
8797 | ||
ae6aad82 | 8798 | /* must be called while manager is quiesced */ |
f21e18ca | 8799 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
8800 | { |
8801 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
8802 | struct dl *iter; |
8803 | struct imsm_dev *dev; | |
8804 | struct imsm_map *map; | |
24565c9a DW |
8805 | int i, j, num_members; |
8806 | __u32 ord; | |
ae6aad82 | 8807 | |
24565c9a DW |
8808 | dprintf("%s: deleting device[%d] from imsm_super\n", |
8809 | __func__, index); | |
ae6aad82 DW |
8810 | |
8811 | /* shift all indexes down one */ | |
8812 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 8813 | if (iter->index > (int)index) |
ae6aad82 | 8814 | iter->index--; |
47ee5a45 | 8815 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 8816 | if (iter->index > (int)index) |
47ee5a45 | 8817 | iter->index--; |
ae6aad82 DW |
8818 | |
8819 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8820 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8821 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
8822 | num_members = map->num_members; |
8823 | for (j = 0; j < num_members; j++) { | |
8824 | /* update ord entries being careful not to propagate | |
8825 | * ord-flags to the first map | |
8826 | */ | |
238c0a71 | 8827 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 8828 | |
24565c9a DW |
8829 | if (ord_to_idx(ord) <= index) |
8830 | continue; | |
ae6aad82 | 8831 | |
238c0a71 | 8832 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 8833 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 8834 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
8835 | if (map) |
8836 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
8837 | } |
8838 | } | |
8839 | ||
8840 | mpb->num_disks--; | |
8841 | super->updates_pending++; | |
24565c9a DW |
8842 | if (*dlp) { |
8843 | struct dl *dl = *dlp; | |
8844 | ||
8845 | *dlp = (*dlp)->next; | |
8846 | __free_imsm_disk(dl); | |
8847 | } | |
ae6aad82 | 8848 | } |
9e2d750d | 8849 | #endif /* MDASSEMBLE */ |
9a717282 AK |
8850 | |
8851 | static void close_targets(int *targets, int new_disks) | |
8852 | { | |
8853 | int i; | |
8854 | ||
8855 | if (!targets) | |
8856 | return; | |
8857 | ||
8858 | for (i = 0; i < new_disks; i++) { | |
8859 | if (targets[i] >= 0) { | |
8860 | close(targets[i]); | |
8861 | targets[i] = -1; | |
8862 | } | |
8863 | } | |
8864 | } | |
8865 | ||
8866 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
8867 | struct intel_super *super, | |
8868 | struct imsm_dev *dev) | |
8869 | { | |
8870 | switch (level) { | |
bf5cf7c7 | 8871 | case 1: |
9a717282 AK |
8872 | case 10:{ |
8873 | int ret_val = 0; | |
8874 | struct imsm_map *map; | |
8875 | int i; | |
8876 | ||
8877 | ret_val = raid_disks/2; | |
8878 | /* check map if all disks pairs not failed | |
8879 | * in both maps | |
8880 | */ | |
238c0a71 | 8881 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
8882 | for (i = 0; i < ret_val; i++) { |
8883 | int degradation = 0; | |
8884 | if (get_imsm_disk(super, i) == NULL) | |
8885 | degradation++; | |
8886 | if (get_imsm_disk(super, i + 1) == NULL) | |
8887 | degradation++; | |
8888 | if (degradation == 2) | |
8889 | return 0; | |
8890 | } | |
238c0a71 | 8891 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
8892 | /* if there is no second map |
8893 | * result can be returned | |
8894 | */ | |
8895 | if (map == NULL) | |
8896 | return ret_val; | |
8897 | /* check degradation in second map | |
8898 | */ | |
8899 | for (i = 0; i < ret_val; i++) { | |
8900 | int degradation = 0; | |
8901 | if (get_imsm_disk(super, i) == NULL) | |
8902 | degradation++; | |
8903 | if (get_imsm_disk(super, i + 1) == NULL) | |
8904 | degradation++; | |
8905 | if (degradation == 2) | |
8906 | return 0; | |
8907 | } | |
8908 | return ret_val; | |
8909 | } | |
8910 | case 5: | |
8911 | return 1; | |
8912 | case 6: | |
8913 | return 2; | |
8914 | default: | |
8915 | return 0; | |
8916 | } | |
8917 | } | |
8918 | ||
8919 | ||
687629c2 AK |
8920 | /******************************************************************************* |
8921 | * Function: open_backup_targets | |
8922 | * Description: Function opens file descriptors for all devices given in | |
8923 | * info->devs | |
8924 | * Parameters: | |
8925 | * info : general array info | |
8926 | * raid_disks : number of disks | |
8927 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
8928 | * super : intel super for raid10 degradation check |
8929 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
8930 | * Returns: |
8931 | * 0 : success | |
8932 | * -1 : fail | |
8933 | ******************************************************************************/ | |
9a717282 AK |
8934 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
8935 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
8936 | { |
8937 | struct mdinfo *sd; | |
f627f5ad | 8938 | int i; |
9a717282 | 8939 | int opened = 0; |
f627f5ad AK |
8940 | |
8941 | for (i = 0; i < raid_disks; i++) | |
8942 | raid_fds[i] = -1; | |
687629c2 AK |
8943 | |
8944 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8945 | char *dn; | |
8946 | ||
8947 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
8948 | dprintf("disk is faulty!!\n"); | |
8949 | continue; | |
8950 | } | |
8951 | ||
8952 | if ((sd->disk.raid_disk >= raid_disks) || | |
8953 | (sd->disk.raid_disk < 0)) | |
8954 | continue; | |
8955 | ||
8956 | dn = map_dev(sd->disk.major, | |
8957 | sd->disk.minor, 1); | |
8958 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
8959 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
8960 | fprintf(stderr, "cannot open component\n"); | |
9a717282 | 8961 | continue; |
687629c2 | 8962 | } |
9a717282 AK |
8963 | opened++; |
8964 | } | |
8965 | /* check if maximum array degradation level is not exceeded | |
8966 | */ | |
8967 | if ((raid_disks - opened) > | |
8968 | imsm_get_allowed_degradation(info->new_level, | |
8969 | raid_disks, | |
8970 | super, dev)) { | |
8971 | fprintf(stderr, "Not enough disks can be opened.\n"); | |
8972 | close_targets(raid_fds, raid_disks); | |
8973 | return -2; | |
687629c2 AK |
8974 | } |
8975 | return 0; | |
8976 | } | |
8977 | ||
9e2d750d | 8978 | #ifndef MDASSEMBLE |
687629c2 AK |
8979 | /******************************************************************************* |
8980 | * Function: init_migr_record_imsm | |
8981 | * Description: Function inits imsm migration record | |
8982 | * Parameters: | |
8983 | * super : imsm internal array info | |
8984 | * dev : device under migration | |
8985 | * info : general array info to find the smallest device | |
8986 | * Returns: | |
8987 | * none | |
8988 | ******************************************************************************/ | |
8989 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
8990 | struct mdinfo *info) | |
8991 | { | |
8992 | struct intel_super *super = st->sb; | |
8993 | struct migr_record *migr_rec = super->migr_rec; | |
8994 | int new_data_disks; | |
8995 | unsigned long long dsize, dev_sectors; | |
8996 | long long unsigned min_dev_sectors = -1LLU; | |
8997 | struct mdinfo *sd; | |
8998 | char nm[30]; | |
8999 | int fd; | |
238c0a71 AK |
9000 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
9001 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 9002 | unsigned long long num_migr_units; |
3ef4403c | 9003 | unsigned long long array_blocks; |
687629c2 AK |
9004 | |
9005 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
9006 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
9007 | ||
9008 | /* only ascending reshape supported now */ | |
9009 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
9010 | ||
9011 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
9012 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
9013 | migr_rec->dest_depth_per_unit *= |
9014 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 9015 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
9016 | migr_rec->blocks_per_unit = |
9017 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
9018 | migr_rec->dest_depth_per_unit = | |
9019 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 9020 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
9021 | num_migr_units = |
9022 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
9023 | ||
9024 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
9025 | num_migr_units++; | |
9026 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
9027 | ||
9028 | migr_rec->post_migr_vol_cap = dev->size_low; | |
9029 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
9030 | ||
9031 | ||
9032 | /* Find the smallest dev */ | |
9033 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9034 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
9035 | fd = dev_open(nm, O_RDONLY); | |
9036 | if (fd < 0) | |
9037 | continue; | |
9038 | get_dev_size(fd, NULL, &dsize); | |
9039 | dev_sectors = dsize / 512; | |
9040 | if (dev_sectors < min_dev_sectors) | |
9041 | min_dev_sectors = dev_sectors; | |
9042 | close(fd); | |
9043 | } | |
9044 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
9045 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
9046 | ||
9047 | write_imsm_migr_rec(st); | |
9048 | ||
9049 | return; | |
9050 | } | |
9051 | ||
9052 | /******************************************************************************* | |
9053 | * Function: save_backup_imsm | |
9054 | * Description: Function saves critical data stripes to Migration Copy Area | |
9055 | * and updates the current migration unit status. | |
9056 | * Use restore_stripes() to form a destination stripe, | |
9057 | * and to write it to the Copy Area. | |
9058 | * Parameters: | |
9059 | * st : supertype information | |
aea93171 | 9060 | * dev : imsm device that backup is saved for |
687629c2 AK |
9061 | * info : general array info |
9062 | * buf : input buffer | |
687629c2 AK |
9063 | * length : length of data to backup (blocks_per_unit) |
9064 | * Returns: | |
9065 | * 0 : success | |
9066 | *, -1 : fail | |
9067 | ******************************************************************************/ | |
9068 | int save_backup_imsm(struct supertype *st, | |
9069 | struct imsm_dev *dev, | |
9070 | struct mdinfo *info, | |
9071 | void *buf, | |
687629c2 AK |
9072 | int length) |
9073 | { | |
9074 | int rv = -1; | |
9075 | struct intel_super *super = st->sb; | |
9076 | unsigned long long *target_offsets = NULL; | |
9077 | int *targets = NULL; | |
9078 | int i; | |
238c0a71 | 9079 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 9080 | int new_disks = map_dest->num_members; |
ab724b98 AK |
9081 | int dest_layout = 0; |
9082 | int dest_chunk; | |
d1877f69 | 9083 | unsigned long long start; |
238c0a71 | 9084 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 9085 | |
503975b9 | 9086 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 9087 | |
7e45b550 AK |
9088 | for (i = 0; i < new_disks; i++) |
9089 | targets[i] = -1; | |
9090 | ||
503975b9 | 9091 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 9092 | |
d1877f69 | 9093 | start = info->reshape_progress * 512; |
687629c2 | 9094 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
9095 | target_offsets[i] = (unsigned long long) |
9096 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
9097 | /* move back copy area adderss, it will be moved forward |
9098 | * in restore_stripes() using start input variable | |
9099 | */ | |
9100 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
9101 | } |
9102 | ||
9a717282 AK |
9103 | if (open_backup_targets(info, new_disks, targets, |
9104 | super, dev)) | |
687629c2 AK |
9105 | goto abort; |
9106 | ||
68eb8bc6 | 9107 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
9108 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
9109 | ||
687629c2 AK |
9110 | if (restore_stripes(targets, /* list of dest devices */ |
9111 | target_offsets, /* migration record offsets */ | |
9112 | new_disks, | |
ab724b98 AK |
9113 | dest_chunk, |
9114 | map_dest->raid_level, | |
9115 | dest_layout, | |
9116 | -1, /* source backup file descriptor */ | |
9117 | 0, /* input buf offset | |
9118 | * always 0 buf is already offseted */ | |
d1877f69 | 9119 | start, |
687629c2 AK |
9120 | length, |
9121 | buf) != 0) { | |
e7b84f9d | 9122 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
9123 | goto abort; |
9124 | } | |
9125 | ||
9126 | rv = 0; | |
9127 | ||
9128 | abort: | |
9129 | if (targets) { | |
9a717282 | 9130 | close_targets(targets, new_disks); |
687629c2 AK |
9131 | free(targets); |
9132 | } | |
9133 | free(target_offsets); | |
9134 | ||
9135 | return rv; | |
9136 | } | |
9137 | ||
9138 | /******************************************************************************* | |
9139 | * Function: save_checkpoint_imsm | |
9140 | * Description: Function called for current unit status update | |
9141 | * in the migration record. It writes it to disk. | |
9142 | * Parameters: | |
9143 | * super : imsm internal array info | |
9144 | * info : general array info | |
9145 | * Returns: | |
9146 | * 0: success | |
9147 | * 1: failure | |
0228d92c AK |
9148 | * 2: failure, means no valid migration record |
9149 | * / no general migration in progress / | |
687629c2 AK |
9150 | ******************************************************************************/ |
9151 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
9152 | { | |
9153 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
9154 | unsigned long long blocks_per_unit; |
9155 | unsigned long long curr_migr_unit; | |
9156 | ||
2e062e82 AK |
9157 | if (load_imsm_migr_rec(super, info) != 0) { |
9158 | dprintf("imsm: ERROR: Cannot read migration record " | |
9159 | "for checkpoint save.\n"); | |
9160 | return 1; | |
9161 | } | |
9162 | ||
f8b72ef5 AK |
9163 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
9164 | if (blocks_per_unit == 0) { | |
0228d92c AK |
9165 | dprintf("imsm: no migration in progress.\n"); |
9166 | return 2; | |
687629c2 | 9167 | } |
f8b72ef5 AK |
9168 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
9169 | /* check if array is alligned to copy area | |
9170 | * if it is not alligned, add one to current migration unit value | |
9171 | * this can happend on array reshape finish only | |
9172 | */ | |
9173 | if (info->reshape_progress % blocks_per_unit) | |
9174 | curr_migr_unit++; | |
687629c2 AK |
9175 | |
9176 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 9177 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
9178 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9179 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
9180 | __cpu_to_le32(curr_migr_unit * |
9181 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
9182 | if (write_imsm_migr_rec(st) < 0) { |
9183 | dprintf("imsm: Cannot write migration record " | |
9184 | "outside backup area\n"); | |
9185 | return 1; | |
9186 | } | |
9187 | ||
9188 | return 0; | |
9189 | } | |
9190 | ||
276d77db AK |
9191 | /******************************************************************************* |
9192 | * Function: recover_backup_imsm | |
9193 | * Description: Function recovers critical data from the Migration Copy Area | |
9194 | * while assembling an array. | |
9195 | * Parameters: | |
9196 | * super : imsm internal array info | |
9197 | * info : general array info | |
9198 | * Returns: | |
9199 | * 0 : success (or there is no data to recover) | |
9200 | * 1 : fail | |
9201 | ******************************************************************************/ | |
9202 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
9203 | { | |
9204 | struct intel_super *super = st->sb; | |
9205 | struct migr_record *migr_rec = super->migr_rec; | |
9206 | struct imsm_map *map_dest = NULL; | |
9207 | struct intel_dev *id = NULL; | |
9208 | unsigned long long read_offset; | |
9209 | unsigned long long write_offset; | |
9210 | unsigned unit_len; | |
9211 | int *targets = NULL; | |
9212 | int new_disks, i, err; | |
9213 | char *buf = NULL; | |
9214 | int retval = 1; | |
9215 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
9216 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 9217 | char buffer[20]; |
6c3560c0 | 9218 | int skipped_disks = 0; |
276d77db AK |
9219 | |
9220 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
9221 | if (err < 1) | |
9222 | return 1; | |
9223 | ||
9224 | /* recover data only during assemblation */ | |
9225 | if (strncmp(buffer, "inactive", 8) != 0) | |
9226 | return 0; | |
9227 | /* no data to recover */ | |
9228 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
9229 | return 0; | |
9230 | if (curr_migr_unit >= num_migr_units) | |
9231 | return 1; | |
9232 | ||
9233 | /* find device during reshape */ | |
9234 | for (id = super->devlist; id; id = id->next) | |
9235 | if (is_gen_migration(id->dev)) | |
9236 | break; | |
9237 | if (id == NULL) | |
9238 | return 1; | |
9239 | ||
238c0a71 | 9240 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
9241 | new_disks = map_dest->num_members; |
9242 | ||
9243 | read_offset = (unsigned long long) | |
9244 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
9245 | ||
9246 | write_offset = ((unsigned long long) | |
9247 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 9248 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
9249 | |
9250 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9251 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
9252 | goto abort; | |
503975b9 | 9253 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 9254 | |
9a717282 | 9255 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 9256 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
9257 | goto abort; |
9258 | } | |
276d77db AK |
9259 | |
9260 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
9261 | if (targets[i] < 0) { |
9262 | skipped_disks++; | |
9263 | continue; | |
9264 | } | |
276d77db | 9265 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
9266 | pr_err("Cannot seek to block: %s\n", |
9267 | strerror(errno)); | |
137debce AK |
9268 | skipped_disks++; |
9269 | continue; | |
276d77db | 9270 | } |
9ec11d1a | 9271 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9272 | pr_err("Cannot read copy area block: %s\n", |
9273 | strerror(errno)); | |
137debce AK |
9274 | skipped_disks++; |
9275 | continue; | |
276d77db AK |
9276 | } |
9277 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
9278 | pr_err("Cannot seek to block: %s\n", |
9279 | strerror(errno)); | |
137debce AK |
9280 | skipped_disks++; |
9281 | continue; | |
276d77db | 9282 | } |
9ec11d1a | 9283 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9284 | pr_err("Cannot restore block: %s\n", |
9285 | strerror(errno)); | |
137debce AK |
9286 | skipped_disks++; |
9287 | continue; | |
276d77db AK |
9288 | } |
9289 | } | |
9290 | ||
137debce AK |
9291 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
9292 | new_disks, | |
9293 | super, | |
9294 | id->dev)) { | |
e7b84f9d N |
9295 | pr_err("Cannot restore data from backup." |
9296 | " Too many failed disks\n"); | |
6c3560c0 AK |
9297 | goto abort; |
9298 | } | |
9299 | ||
befb629b AK |
9300 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
9301 | /* ignore error == 2, this can mean end of reshape here | |
9302 | */ | |
9303 | dprintf("imsm: Cannot write checkpoint to " | |
9304 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
9305 | } else | |
276d77db | 9306 | retval = 0; |
276d77db AK |
9307 | |
9308 | abort: | |
9309 | if (targets) { | |
9310 | for (i = 0; i < new_disks; i++) | |
9311 | if (targets[i]) | |
9312 | close(targets[i]); | |
9313 | free(targets); | |
9314 | } | |
9315 | free(buf); | |
9316 | return retval; | |
9317 | } | |
9318 | ||
2cda7640 ML |
9319 | static char disk_by_path[] = "/dev/disk/by-path/"; |
9320 | ||
9321 | static const char *imsm_get_disk_controller_domain(const char *path) | |
9322 | { | |
2cda7640 | 9323 | char disk_path[PATH_MAX]; |
96234762 LM |
9324 | char *drv=NULL; |
9325 | struct stat st; | |
2cda7640 | 9326 | |
96234762 LM |
9327 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
9328 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
9329 | if (stat(disk_path, &st) == 0) { | |
9330 | struct sys_dev* hba; | |
9331 | char *path=NULL; | |
9332 | ||
9333 | path = devt_to_devpath(st.st_rdev); | |
9334 | if (path == NULL) | |
9335 | return "unknown"; | |
9336 | hba = find_disk_attached_hba(-1, path); | |
9337 | if (hba && hba->type == SYS_DEV_SAS) | |
9338 | drv = "isci"; | |
9339 | else if (hba && hba->type == SYS_DEV_SATA) | |
9340 | drv = "ahci"; | |
9341 | else | |
9342 | drv = "unknown"; | |
9343 | dprintf("path: %s hba: %s attached: %s\n", | |
9344 | path, (hba) ? hba->path : "NULL", drv); | |
9345 | free(path); | |
9346 | if (hba) | |
9347 | free_sys_dev(&hba); | |
2cda7640 | 9348 | } |
96234762 | 9349 | return drv; |
2cda7640 ML |
9350 | } |
9351 | ||
78b10e66 N |
9352 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
9353 | { | |
9354 | char subdev_name[20]; | |
9355 | struct mdstat_ent *mdstat; | |
9356 | ||
9357 | sprintf(subdev_name, "%d", subdev); | |
9358 | mdstat = mdstat_by_subdev(subdev_name, container); | |
9359 | if (!mdstat) | |
9360 | return -1; | |
9361 | ||
9362 | *minor = mdstat->devnum; | |
9363 | free_mdstat(mdstat); | |
9364 | return 0; | |
9365 | } | |
9366 | ||
9367 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
9368 | struct geo_params *geo, | |
fbf3d202 AK |
9369 | int *old_raid_disks, |
9370 | int direction) | |
78b10e66 | 9371 | { |
694575e7 KW |
9372 | /* currently we only support increasing the number of devices |
9373 | * for a container. This increases the number of device for each | |
9374 | * member array. They must all be RAID0 or RAID5. | |
9375 | */ | |
78b10e66 N |
9376 | int ret_val = 0; |
9377 | struct mdinfo *info, *member; | |
9378 | int devices_that_can_grow = 0; | |
9379 | ||
9380 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
9381 | "st->devnum = (%i)\n", | |
9382 | st->devnum); | |
9383 | ||
d04f65f4 | 9384 | if (geo->size > 0 || |
78b10e66 N |
9385 | geo->level != UnSet || |
9386 | geo->layout != UnSet || | |
9387 | geo->chunksize != 0 || | |
9388 | geo->raid_disks == UnSet) { | |
9389 | dprintf("imsm: Container operation is allowed for " | |
9390 | "raid disks number change only.\n"); | |
9391 | return ret_val; | |
9392 | } | |
9393 | ||
fbf3d202 AK |
9394 | if (direction == ROLLBACK_METADATA_CHANGES) { |
9395 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9396 | "container operation.\n"); | |
9397 | return ret_val; | |
9398 | } | |
9399 | ||
78b10e66 N |
9400 | info = container_content_imsm(st, NULL); |
9401 | for (member = info; member; member = member->next) { | |
9402 | int result; | |
9403 | int minor; | |
9404 | ||
9405 | dprintf("imsm: checking device_num: %i\n", | |
9406 | member->container_member); | |
9407 | ||
d7d205bd | 9408 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
9409 | /* we work on container for Online Capacity Expansion |
9410 | * only so raid_disks has to grow | |
9411 | */ | |
9412 | dprintf("imsm: for container operation raid disks " | |
9413 | "increase is required\n"); | |
9414 | break; | |
9415 | } | |
9416 | ||
9417 | if ((info->array.level != 0) && | |
9418 | (info->array.level != 5)) { | |
9419 | /* we cannot use this container with other raid level | |
9420 | */ | |
690aae1a | 9421 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
9422 | " raid level (%i) detected\n", |
9423 | info->array.level); | |
9424 | break; | |
9425 | } else { | |
9426 | /* check for platform support | |
9427 | * for this raid level configuration | |
9428 | */ | |
9429 | struct intel_super *super = st->sb; | |
9430 | if (!is_raid_level_supported(super->orom, | |
9431 | member->array.level, | |
9432 | geo->raid_disks)) { | |
690aae1a | 9433 | dprintf("platform does not support raid%d with" |
78b10e66 N |
9434 | " %d disk%s\n", |
9435 | info->array.level, | |
9436 | geo->raid_disks, | |
9437 | geo->raid_disks > 1 ? "s" : ""); | |
9438 | break; | |
9439 | } | |
2a4a08e7 AK |
9440 | /* check if component size is aligned to chunk size |
9441 | */ | |
9442 | if (info->component_size % | |
9443 | (info->array.chunk_size/512)) { | |
9444 | dprintf("Component size is not aligned to " | |
9445 | "chunk size\n"); | |
9446 | break; | |
9447 | } | |
78b10e66 N |
9448 | } |
9449 | ||
9450 | if (*old_raid_disks && | |
9451 | info->array.raid_disks != *old_raid_disks) | |
9452 | break; | |
9453 | *old_raid_disks = info->array.raid_disks; | |
9454 | ||
9455 | /* All raid5 and raid0 volumes in container | |
9456 | * have to be ready for Online Capacity Expansion | |
9457 | * so they need to be assembled. We have already | |
9458 | * checked that no recovery etc is happening. | |
9459 | */ | |
9460 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
9461 | st->container_dev, | |
9462 | &minor); | |
9463 | if (result < 0) { | |
9464 | dprintf("imsm: cannot find array\n"); | |
9465 | break; | |
9466 | } | |
9467 | devices_that_can_grow++; | |
9468 | } | |
9469 | sysfs_free(info); | |
9470 | if (!member && devices_that_can_grow) | |
9471 | ret_val = 1; | |
9472 | ||
9473 | if (ret_val) | |
9474 | dprintf("\tContainer operation allowed\n"); | |
9475 | else | |
9476 | dprintf("\tError: %i\n", ret_val); | |
9477 | ||
9478 | return ret_val; | |
9479 | } | |
9480 | ||
9481 | /* Function: get_spares_for_grow | |
9482 | * Description: Allocates memory and creates list of spare devices | |
9483 | * avaliable in container. Checks if spare drive size is acceptable. | |
9484 | * Parameters: Pointer to the supertype structure | |
9485 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
9486 | * NULL if fail | |
9487 | */ | |
9488 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
9489 | { | |
78b10e66 | 9490 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 9491 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
9492 | } |
9493 | ||
9494 | /****************************************************************************** | |
9495 | * function: imsm_create_metadata_update_for_reshape | |
9496 | * Function creates update for whole IMSM container. | |
9497 | * | |
9498 | ******************************************************************************/ | |
9499 | static int imsm_create_metadata_update_for_reshape( | |
9500 | struct supertype *st, | |
9501 | struct geo_params *geo, | |
9502 | int old_raid_disks, | |
9503 | struct imsm_update_reshape **updatep) | |
9504 | { | |
9505 | struct intel_super *super = st->sb; | |
9506 | struct imsm_super *mpb = super->anchor; | |
9507 | int update_memory_size = 0; | |
9508 | struct imsm_update_reshape *u = NULL; | |
9509 | struct mdinfo *spares = NULL; | |
9510 | int i; | |
9511 | int delta_disks = 0; | |
bbd24d86 | 9512 | struct mdinfo *dev; |
78b10e66 N |
9513 | |
9514 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
9515 | geo->raid_disks); | |
9516 | ||
9517 | delta_disks = geo->raid_disks - old_raid_disks; | |
9518 | ||
9519 | /* size of all update data without anchor */ | |
9520 | update_memory_size = sizeof(struct imsm_update_reshape); | |
9521 | ||
9522 | /* now add space for spare disks that we need to add. */ | |
9523 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
9524 | ||
503975b9 | 9525 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
9526 | u->type = update_reshape_container_disks; |
9527 | u->old_raid_disks = old_raid_disks; | |
9528 | u->new_raid_disks = geo->raid_disks; | |
9529 | ||
9530 | /* now get spare disks list | |
9531 | */ | |
9532 | spares = get_spares_for_grow(st); | |
9533 | ||
9534 | if (spares == NULL | |
9535 | || delta_disks > spares->array.spare_disks) { | |
e7b84f9d | 9536 | pr_err("imsm: ERROR: Cannot get spare devices " |
e14e5960 | 9537 | "for %s.\n", geo->dev_name); |
e4c72d1d | 9538 | i = -1; |
78b10e66 N |
9539 | goto abort; |
9540 | } | |
9541 | ||
9542 | /* we have got spares | |
9543 | * update disk list in imsm_disk list table in anchor | |
9544 | */ | |
9545 | dprintf("imsm: %i spares are available.\n\n", | |
9546 | spares->array.spare_disks); | |
9547 | ||
bbd24d86 | 9548 | dev = spares->devs; |
78b10e66 | 9549 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
9550 | struct dl *dl; |
9551 | ||
bbd24d86 AK |
9552 | if (dev == NULL) |
9553 | break; | |
78b10e66 N |
9554 | u->new_disks[i] = makedev(dev->disk.major, |
9555 | dev->disk.minor); | |
9556 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
9557 | dl->index = mpb->num_disks; |
9558 | mpb->num_disks++; | |
bbd24d86 | 9559 | dev = dev->next; |
78b10e66 | 9560 | } |
78b10e66 N |
9561 | |
9562 | abort: | |
9563 | /* free spares | |
9564 | */ | |
9565 | sysfs_free(spares); | |
9566 | ||
d677e0b8 | 9567 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 9568 | if (i == delta_disks) { |
d677e0b8 | 9569 | dprintf(" OK\n"); |
78b10e66 N |
9570 | *updatep = u; |
9571 | return update_memory_size; | |
9572 | } | |
9573 | free(u); | |
d677e0b8 | 9574 | dprintf(" Error\n"); |
78b10e66 N |
9575 | |
9576 | return 0; | |
9577 | } | |
9578 | ||
f3871fdc AK |
9579 | |
9580 | /****************************************************************************** | |
9581 | * function: imsm_create_metadata_update_for_size_change() | |
9582 | * Creates update for IMSM array for array size change. | |
9583 | * | |
9584 | ******************************************************************************/ | |
9585 | static int imsm_create_metadata_update_for_size_change( | |
9586 | struct supertype *st, | |
9587 | struct geo_params *geo, | |
9588 | struct imsm_update_size_change **updatep) | |
9589 | { | |
9590 | struct intel_super *super = st->sb; | |
9591 | int update_memory_size = 0; | |
9592 | struct imsm_update_size_change *u = NULL; | |
9593 | ||
9594 | dprintf("imsm_create_metadata_update_for_size_change(enter)" | |
9595 | " New size = %llu\n", geo->size); | |
9596 | ||
9597 | /* size of all update data without anchor */ | |
9598 | update_memory_size = sizeof(struct imsm_update_size_change); | |
9599 | ||
503975b9 | 9600 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
9601 | u->type = update_size_change; |
9602 | u->subdev = super->current_vol; | |
9603 | u->new_size = geo->size; | |
9604 | ||
9605 | dprintf("imsm: reshape update preparation : OK\n"); | |
9606 | *updatep = u; | |
9607 | ||
9608 | return update_memory_size; | |
9609 | } | |
9610 | ||
48c5303a PC |
9611 | /****************************************************************************** |
9612 | * function: imsm_create_metadata_update_for_migration() | |
9613 | * Creates update for IMSM array. | |
9614 | * | |
9615 | ******************************************************************************/ | |
9616 | static int imsm_create_metadata_update_for_migration( | |
9617 | struct supertype *st, | |
9618 | struct geo_params *geo, | |
9619 | struct imsm_update_reshape_migration **updatep) | |
9620 | { | |
9621 | struct intel_super *super = st->sb; | |
9622 | int update_memory_size = 0; | |
9623 | struct imsm_update_reshape_migration *u = NULL; | |
9624 | struct imsm_dev *dev; | |
9625 | int previous_level = -1; | |
9626 | ||
9627 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
9628 | " New Level = %i\n", geo->level); | |
9629 | ||
9630 | /* size of all update data without anchor */ | |
9631 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
9632 | ||
503975b9 | 9633 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
9634 | u->type = update_reshape_migration; |
9635 | u->subdev = super->current_vol; | |
9636 | u->new_level = geo->level; | |
9637 | u->new_layout = geo->layout; | |
9638 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
9639 | u->new_disks[0] = -1; | |
4bba0439 | 9640 | u->new_chunksize = -1; |
48c5303a PC |
9641 | |
9642 | dev = get_imsm_dev(super, u->subdev); | |
9643 | if (dev) { | |
9644 | struct imsm_map *map; | |
9645 | ||
238c0a71 | 9646 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
9647 | if (map) { |
9648 | int current_chunk_size = | |
9649 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
9650 | ||
9651 | if (geo->chunksize != current_chunk_size) { | |
9652 | u->new_chunksize = geo->chunksize / 1024; | |
9653 | dprintf("imsm: " | |
9654 | "chunk size change from %i to %i\n", | |
9655 | current_chunk_size, u->new_chunksize); | |
9656 | } | |
48c5303a | 9657 | previous_level = map->raid_level; |
4bba0439 | 9658 | } |
48c5303a PC |
9659 | } |
9660 | if ((geo->level == 5) && (previous_level == 0)) { | |
9661 | struct mdinfo *spares = NULL; | |
9662 | ||
9663 | u->new_raid_disks++; | |
9664 | spares = get_spares_for_grow(st); | |
9665 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
9666 | free(u); | |
9667 | sysfs_free(spares); | |
9668 | update_memory_size = 0; | |
9669 | dprintf("error: cannot get spare device " | |
9670 | "for requested migration"); | |
9671 | return 0; | |
9672 | } | |
9673 | sysfs_free(spares); | |
9674 | } | |
9675 | dprintf("imsm: reshape update preparation : OK\n"); | |
9676 | *updatep = u; | |
9677 | ||
9678 | return update_memory_size; | |
9679 | } | |
9680 | ||
8dd70bce AK |
9681 | static void imsm_update_metadata_locally(struct supertype *st, |
9682 | void *buf, int len) | |
9683 | { | |
9684 | struct metadata_update mu; | |
9685 | ||
9686 | mu.buf = buf; | |
9687 | mu.len = len; | |
9688 | mu.space = NULL; | |
9689 | mu.space_list = NULL; | |
9690 | mu.next = NULL; | |
9691 | imsm_prepare_update(st, &mu); | |
9692 | imsm_process_update(st, &mu); | |
9693 | ||
9694 | while (mu.space_list) { | |
9695 | void **space = mu.space_list; | |
9696 | mu.space_list = *space; | |
9697 | free(space); | |
9698 | } | |
9699 | } | |
78b10e66 | 9700 | |
471bceb6 | 9701 | /*************************************************************************** |
694575e7 | 9702 | * Function: imsm_analyze_change |
471bceb6 KW |
9703 | * Description: Function analyze change for single volume |
9704 | * and validate if transition is supported | |
fbf3d202 AK |
9705 | * Parameters: Geometry parameters, supertype structure, |
9706 | * metadata change direction (apply/rollback) | |
694575e7 | 9707 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
9708 | ****************************************************************************/ |
9709 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
9710 | struct geo_params *geo, |
9711 | int direction) | |
694575e7 | 9712 | { |
471bceb6 KW |
9713 | struct mdinfo info; |
9714 | int change = -1; | |
9715 | int check_devs = 0; | |
c21e737b | 9716 | int chunk; |
67a2db32 AK |
9717 | /* number of added/removed disks in operation result */ |
9718 | int devNumChange = 0; | |
9719 | /* imsm compatible layout value for array geometry verification */ | |
9720 | int imsm_layout = -1; | |
7abc9871 AK |
9721 | int data_disks; |
9722 | struct imsm_dev *dev; | |
9723 | struct intel_super *super; | |
d04f65f4 | 9724 | unsigned long long current_size; |
65d38cca | 9725 | unsigned long long free_size; |
d04f65f4 | 9726 | unsigned long long max_size; |
65d38cca | 9727 | int rv; |
471bceb6 KW |
9728 | |
9729 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
9730 | if ((geo->level != info.array.level) && |
9731 | (geo->level >= 0) && | |
9732 | (geo->level != UnSet)) { | |
9733 | switch (info.array.level) { | |
9734 | case 0: | |
9735 | if (geo->level == 5) { | |
b5347799 | 9736 | change = CH_MIGRATION; |
e13ce846 | 9737 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
e7b84f9d N |
9738 | pr_err("Error. Requested Layout " |
9739 | "not supported (left-asymmetric layout " | |
9740 | "is supported only)!\n"); | |
e13ce846 AK |
9741 | change = -1; |
9742 | goto analyse_change_exit; | |
9743 | } | |
67a2db32 | 9744 | imsm_layout = geo->layout; |
471bceb6 | 9745 | check_devs = 1; |
e91a3bad LM |
9746 | devNumChange = 1; /* parity disk added */ |
9747 | } else if (geo->level == 10) { | |
471bceb6 KW |
9748 | change = CH_TAKEOVER; |
9749 | check_devs = 1; | |
e91a3bad | 9750 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 9751 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 9752 | } |
dfe77a9e KW |
9753 | break; |
9754 | case 1: | |
471bceb6 KW |
9755 | case 10: |
9756 | if (geo->level == 0) { | |
9757 | change = CH_TAKEOVER; | |
9758 | check_devs = 1; | |
e91a3bad | 9759 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 9760 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
9761 | } |
9762 | break; | |
9763 | } | |
9764 | if (change == -1) { | |
e7b84f9d N |
9765 | pr_err("Error. Level Migration from %d to %d " |
9766 | "not supported!\n", | |
9767 | info.array.level, geo->level); | |
471bceb6 KW |
9768 | goto analyse_change_exit; |
9769 | } | |
9770 | } else | |
9771 | geo->level = info.array.level; | |
9772 | ||
9773 | if ((geo->layout != info.array.layout) | |
9774 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 9775 | change = CH_MIGRATION; |
471bceb6 KW |
9776 | if ((info.array.layout == 0) |
9777 | && (info.array.level == 5) | |
9778 | && (geo->layout == 5)) { | |
9779 | /* reshape 5 -> 4 */ | |
9780 | } else if ((info.array.layout == 5) | |
9781 | && (info.array.level == 5) | |
9782 | && (geo->layout == 0)) { | |
9783 | /* reshape 4 -> 5 */ | |
9784 | geo->layout = 0; | |
9785 | geo->level = 5; | |
9786 | } else { | |
e7b84f9d N |
9787 | pr_err("Error. Layout Migration from %d to %d " |
9788 | "not supported!\n", | |
9789 | info.array.layout, geo->layout); | |
471bceb6 KW |
9790 | change = -1; |
9791 | goto analyse_change_exit; | |
9792 | } | |
67a2db32 | 9793 | } else { |
471bceb6 | 9794 | geo->layout = info.array.layout; |
67a2db32 AK |
9795 | if (imsm_layout == -1) |
9796 | imsm_layout = info.array.layout; | |
9797 | } | |
471bceb6 KW |
9798 | |
9799 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
9800 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 9801 | change = CH_MIGRATION; |
471bceb6 KW |
9802 | else |
9803 | geo->chunksize = info.array.chunk_size; | |
9804 | ||
c21e737b | 9805 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
9806 | |
9807 | super = st->sb; | |
9808 | dev = get_imsm_dev(super, super->current_vol); | |
9809 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 9810 | /* compute current size per disk member |
7abc9871 | 9811 | */ |
c41e00b2 AK |
9812 | current_size = info.custom_array_size / data_disks; |
9813 | ||
d5d2c614 | 9814 | if ((geo->size > 0) && (geo->size != MAX_SIZE)) { |
c41e00b2 AK |
9815 | /* align component size |
9816 | */ | |
9817 | geo->size = imsm_component_size_aligment_check( | |
9818 | get_imsm_raid_level(dev->vol.map), | |
9819 | chunk * 1024, | |
9820 | geo->size * 2); | |
65d0b4ce LD |
9821 | if (geo->size == 0) { |
9822 | pr_err("Error. Size expansion is " \ | |
9823 | "supported only (current size is %llu, " \ | |
9824 | "requested size /rounded/ is 0).\n", | |
9825 | current_size); | |
9826 | goto analyse_change_exit; | |
9827 | } | |
c41e00b2 | 9828 | } |
7abc9871 | 9829 | |
d04f65f4 | 9830 | if ((current_size != geo->size) && (geo->size > 0)) { |
7abc9871 | 9831 | if (change != -1) { |
e7b84f9d | 9832 | pr_err("Error. Size change should be the only " |
7abc9871 AK |
9833 | "one at a time.\n"); |
9834 | change = -1; | |
9835 | goto analyse_change_exit; | |
9836 | } | |
9837 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
e7b84f9d N |
9838 | pr_err("Error. The last volume in container " |
9839 | "can be expanded only (%i/%i).\n", | |
9840 | super->current_vol, st->devnum); | |
7abc9871 AK |
9841 | goto analyse_change_exit; |
9842 | } | |
65d38cca LD |
9843 | /* check the maximum available size |
9844 | */ | |
9845 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
9846 | 0, chunk, &free_size); | |
9847 | if (rv == 0) | |
9848 | /* Cannot find maximum available space | |
9849 | */ | |
9850 | max_size = 0; | |
9851 | else { | |
9852 | max_size = free_size + current_size; | |
9853 | /* align component size | |
9854 | */ | |
9855 | max_size = imsm_component_size_aligment_check( | |
9856 | get_imsm_raid_level(dev->vol.map), | |
9857 | chunk * 1024, | |
9858 | max_size); | |
9859 | } | |
d04f65f4 | 9860 | if (geo->size == MAX_SIZE) { |
b130333f AK |
9861 | /* requested size change to the maximum available size |
9862 | */ | |
65d38cca | 9863 | if (max_size == 0) { |
e7b84f9d | 9864 | pr_err("Error. Cannot find " |
b130333f AK |
9865 | "maximum available space.\n"); |
9866 | change = -1; | |
9867 | goto analyse_change_exit; | |
65d38cca LD |
9868 | } else |
9869 | geo->size = max_size; | |
c41e00b2 | 9870 | } |
b130333f | 9871 | |
fbf3d202 AK |
9872 | if ((direction == ROLLBACK_METADATA_CHANGES)) { |
9873 | /* accept size for rollback only | |
9874 | */ | |
9875 | } else { | |
9876 | /* round size due to metadata compatibility | |
9877 | */ | |
9878 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
9879 | << SECT_PER_MB_SHIFT; | |
9880 | dprintf("Prepare update for size change to %llu\n", | |
9881 | geo->size ); | |
9882 | if (current_size >= geo->size) { | |
e7b84f9d N |
9883 | pr_err("Error. Size expansion is " |
9884 | "supported only (current size is %llu, " | |
9885 | "requested size /rounded/ is %llu).\n", | |
9886 | current_size, geo->size); | |
fbf3d202 AK |
9887 | goto analyse_change_exit; |
9888 | } | |
65d38cca | 9889 | if (max_size && geo->size > max_size) { |
e7b84f9d N |
9890 | pr_err("Error. Requested size is larger " |
9891 | "than maximum available size (maximum " | |
9892 | "available size is %llu, " | |
9893 | "requested size /rounded/ is %llu).\n", | |
9894 | max_size, geo->size); | |
65d38cca LD |
9895 | goto analyse_change_exit; |
9896 | } | |
7abc9871 AK |
9897 | } |
9898 | geo->size *= data_disks; | |
9899 | geo->raid_disks = dev->vol.map->num_members; | |
9900 | change = CH_ARRAY_SIZE; | |
9901 | } | |
471bceb6 KW |
9902 | if (!validate_geometry_imsm(st, |
9903 | geo->level, | |
67a2db32 | 9904 | imsm_layout, |
e91a3bad | 9905 | geo->raid_disks + devNumChange, |
c21e737b | 9906 | &chunk, |
471bceb6 KW |
9907 | geo->size, |
9908 | 0, 0, 1)) | |
9909 | change = -1; | |
9910 | ||
9911 | if (check_devs) { | |
9912 | struct intel_super *super = st->sb; | |
9913 | struct imsm_super *mpb = super->anchor; | |
9914 | ||
9915 | if (mpb->num_raid_devs > 1) { | |
e7b84f9d N |
9916 | pr_err("Error. Cannot perform operation on %s" |
9917 | "- for this operation it MUST be single " | |
9918 | "array in container\n", | |
9919 | geo->dev_name); | |
471bceb6 KW |
9920 | change = -1; |
9921 | } | |
9922 | } | |
9923 | ||
9924 | analyse_change_exit: | |
fbf3d202 AK |
9925 | if ((direction == ROLLBACK_METADATA_CHANGES) && |
9926 | ((change == CH_MIGRATION) || (change == CH_TAKEOVER))) { | |
9927 | dprintf("imsm: Metadata changes rollback is not supported for " | |
9928 | "migration and takeover operations.\n"); | |
9929 | change = -1; | |
9930 | } | |
471bceb6 | 9931 | return change; |
694575e7 KW |
9932 | } |
9933 | ||
bb025c2f KW |
9934 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
9935 | { | |
9936 | struct intel_super *super = st->sb; | |
9937 | struct imsm_update_takeover *u; | |
9938 | ||
503975b9 | 9939 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
9940 | |
9941 | u->type = update_takeover; | |
9942 | u->subarray = super->current_vol; | |
9943 | ||
9944 | /* 10->0 transition */ | |
9945 | if (geo->level == 0) | |
9946 | u->direction = R10_TO_R0; | |
9947 | ||
0529c688 KW |
9948 | /* 0->10 transition */ |
9949 | if (geo->level == 10) | |
9950 | u->direction = R0_TO_R10; | |
9951 | ||
bb025c2f KW |
9952 | /* update metadata locally */ |
9953 | imsm_update_metadata_locally(st, u, | |
9954 | sizeof(struct imsm_update_takeover)); | |
9955 | /* and possibly remotely */ | |
9956 | if (st->update_tail) | |
9957 | append_metadata_update(st, u, | |
9958 | sizeof(struct imsm_update_takeover)); | |
9959 | else | |
9960 | free(u); | |
9961 | ||
9962 | return 0; | |
9963 | } | |
9964 | ||
d04f65f4 N |
9965 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
9966 | int level, | |
78b10e66 | 9967 | int layout, int chunksize, int raid_disks, |
41784c88 | 9968 | int delta_disks, char *backup, char *dev, |
016e00f5 | 9969 | int direction, int verbose) |
78b10e66 | 9970 | { |
78b10e66 N |
9971 | int ret_val = 1; |
9972 | struct geo_params geo; | |
9973 | ||
9974 | dprintf("imsm: reshape_super called.\n"); | |
9975 | ||
71204a50 | 9976 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
9977 | |
9978 | geo.dev_name = dev; | |
694575e7 | 9979 | geo.dev_id = st->devnum; |
78b10e66 N |
9980 | geo.size = size; |
9981 | geo.level = level; | |
9982 | geo.layout = layout; | |
9983 | geo.chunksize = chunksize; | |
9984 | geo.raid_disks = raid_disks; | |
41784c88 AK |
9985 | if (delta_disks != UnSet) |
9986 | geo.raid_disks += delta_disks; | |
78b10e66 N |
9987 | |
9988 | dprintf("\tfor level : %i\n", geo.level); | |
9989 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
9990 | ||
9991 | if (experimental() == 0) | |
9992 | return ret_val; | |
9993 | ||
78b10e66 | 9994 | if (st->container_dev == st->devnum) { |
694575e7 KW |
9995 | /* On container level we can only increase number of devices. */ |
9996 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 9997 | int old_raid_disks = 0; |
6dc0be30 | 9998 | |
78b10e66 | 9999 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 10000 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
10001 | struct imsm_update_reshape *u = NULL; |
10002 | int len; | |
10003 | ||
10004 | len = imsm_create_metadata_update_for_reshape( | |
10005 | st, &geo, old_raid_disks, &u); | |
10006 | ||
ed08d51c AK |
10007 | if (len <= 0) { |
10008 | dprintf("imsm: Cannot prepare update\n"); | |
10009 | goto exit_imsm_reshape_super; | |
10010 | } | |
10011 | ||
8dd70bce AK |
10012 | ret_val = 0; |
10013 | /* update metadata locally */ | |
10014 | imsm_update_metadata_locally(st, u, len); | |
10015 | /* and possibly remotely */ | |
10016 | if (st->update_tail) | |
10017 | append_metadata_update(st, u, len); | |
10018 | else | |
ed08d51c | 10019 | free(u); |
8dd70bce | 10020 | |
694575e7 | 10021 | } else { |
e7b84f9d | 10022 | pr_err("(imsm) Operation " |
e7ff7e40 | 10023 | "is not allowed on this container\n"); |
694575e7 KW |
10024 | } |
10025 | } else { | |
10026 | /* On volume level we support following operations | |
471bceb6 KW |
10027 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
10028 | * - chunk size migration | |
10029 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
10030 | */ | |
10031 | struct intel_super *super = st->sb; | |
10032 | struct intel_dev *dev = super->devlist; | |
10033 | int change, devnum; | |
694575e7 | 10034 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
10035 | /* find requested device */ |
10036 | while (dev) { | |
19986c72 MB |
10037 | if (imsm_find_array_minor_by_subdev( |
10038 | dev->index, st->container_dev, &devnum) == 0 | |
10039 | && devnum == geo.dev_id) | |
471bceb6 KW |
10040 | break; |
10041 | dev = dev->next; | |
10042 | } | |
10043 | if (dev == NULL) { | |
e7b84f9d | 10044 | pr_err("Cannot find %s (%i) subarray\n", |
471bceb6 KW |
10045 | geo.dev_name, geo.dev_id); |
10046 | goto exit_imsm_reshape_super; | |
10047 | } | |
10048 | super->current_vol = dev->index; | |
fbf3d202 | 10049 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 10050 | switch (change) { |
471bceb6 | 10051 | case CH_TAKEOVER: |
bb025c2f | 10052 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 10053 | break; |
48c5303a PC |
10054 | case CH_MIGRATION: { |
10055 | struct imsm_update_reshape_migration *u = NULL; | |
10056 | int len = | |
10057 | imsm_create_metadata_update_for_migration( | |
10058 | st, &geo, &u); | |
10059 | if (len < 1) { | |
10060 | dprintf("imsm: " | |
10061 | "Cannot prepare update\n"); | |
10062 | break; | |
10063 | } | |
471bceb6 | 10064 | ret_val = 0; |
48c5303a PC |
10065 | /* update metadata locally */ |
10066 | imsm_update_metadata_locally(st, u, len); | |
10067 | /* and possibly remotely */ | |
10068 | if (st->update_tail) | |
10069 | append_metadata_update(st, u, len); | |
10070 | else | |
10071 | free(u); | |
10072 | } | |
10073 | break; | |
7abc9871 | 10074 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
10075 | struct imsm_update_size_change *u = NULL; |
10076 | int len = | |
10077 | imsm_create_metadata_update_for_size_change( | |
10078 | st, &geo, &u); | |
10079 | if (len < 1) { | |
10080 | dprintf("imsm: " | |
10081 | "Cannot prepare update\n"); | |
10082 | break; | |
10083 | } | |
10084 | ret_val = 0; | |
10085 | /* update metadata locally */ | |
10086 | imsm_update_metadata_locally(st, u, len); | |
10087 | /* and possibly remotely */ | |
10088 | if (st->update_tail) | |
10089 | append_metadata_update(st, u, len); | |
10090 | else | |
10091 | free(u); | |
7abc9871 AK |
10092 | } |
10093 | break; | |
471bceb6 KW |
10094 | default: |
10095 | ret_val = 1; | |
694575e7 | 10096 | } |
694575e7 | 10097 | } |
78b10e66 | 10098 | |
ed08d51c | 10099 | exit_imsm_reshape_super: |
78b10e66 N |
10100 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
10101 | return ret_val; | |
10102 | } | |
2cda7640 | 10103 | |
eee67a47 AK |
10104 | /******************************************************************************* |
10105 | * Function: wait_for_reshape_imsm | |
10106 | * Description: Function writes new sync_max value and waits until | |
10107 | * reshape process reach new position | |
10108 | * Parameters: | |
10109 | * sra : general array info | |
eee67a47 AK |
10110 | * ndata : number of disks in new array's layout |
10111 | * Returns: | |
10112 | * 0 : success, | |
10113 | * 1 : there is no reshape in progress, | |
10114 | * -1 : fail | |
10115 | ******************************************************************************/ | |
ae9f01f8 | 10116 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
10117 | { |
10118 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
10119 | unsigned long long completed; | |
ae9f01f8 AK |
10120 | /* to_complete : new sync_max position */ |
10121 | unsigned long long to_complete = sra->reshape_progress; | |
10122 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 10123 | |
ae9f01f8 AK |
10124 | if (fd < 0) { |
10125 | dprintf("imsm: wait_for_reshape_imsm() " | |
10126 | "cannot open reshape_position\n"); | |
eee67a47 | 10127 | return 1; |
ae9f01f8 | 10128 | } |
eee67a47 | 10129 | |
ae9f01f8 AK |
10130 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
10131 | dprintf("imsm: wait_for_reshape_imsm() " | |
10132 | "cannot read reshape_position (no reshape in progres)\n"); | |
10133 | close(fd); | |
10134 | return 0; | |
10135 | } | |
eee67a47 | 10136 | |
ae9f01f8 AK |
10137 | if (completed > to_complete) { |
10138 | dprintf("imsm: wait_for_reshape_imsm() " | |
10139 | "wrong next position to set %llu (%llu)\n", | |
10140 | to_complete, completed); | |
10141 | close(fd); | |
10142 | return -1; | |
10143 | } | |
10144 | dprintf("Position set: %llu\n", position_to_set); | |
10145 | if (sysfs_set_num(sra, NULL, "sync_max", | |
10146 | position_to_set) != 0) { | |
10147 | dprintf("imsm: wait_for_reshape_imsm() " | |
10148 | "cannot set reshape position to %llu\n", | |
10149 | position_to_set); | |
10150 | close(fd); | |
10151 | return -1; | |
eee67a47 AK |
10152 | } |
10153 | ||
eee67a47 AK |
10154 | do { |
10155 | char action[20]; | |
10156 | fd_set rfds; | |
10157 | FD_ZERO(&rfds); | |
10158 | FD_SET(fd, &rfds); | |
a47e44fb AK |
10159 | select(fd+1, &rfds, NULL, NULL, NULL); |
10160 | if (sysfs_get_str(sra, NULL, "sync_action", | |
10161 | action, 20) > 0 && | |
10162 | strncmp(action, "reshape", 7) != 0) | |
10163 | break; | |
eee67a47 | 10164 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
10165 | dprintf("imsm: wait_for_reshape_imsm() " |
10166 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
10167 | close(fd); |
10168 | return 1; | |
10169 | } | |
eee67a47 AK |
10170 | } while (completed < to_complete); |
10171 | close(fd); | |
10172 | return 0; | |
10173 | ||
10174 | } | |
10175 | ||
b915c95f AK |
10176 | /******************************************************************************* |
10177 | * Function: check_degradation_change | |
10178 | * Description: Check that array hasn't become failed. | |
10179 | * Parameters: | |
10180 | * info : for sysfs access | |
10181 | * sources : source disks descriptors | |
10182 | * degraded: previous degradation level | |
10183 | * Returns: | |
10184 | * degradation level | |
10185 | ******************************************************************************/ | |
10186 | int check_degradation_change(struct mdinfo *info, | |
10187 | int *sources, | |
10188 | int degraded) | |
10189 | { | |
10190 | unsigned long long new_degraded; | |
e1993023 LD |
10191 | int rv; |
10192 | ||
10193 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
10194 | if ((rv == -1) || (new_degraded != (unsigned long long)degraded)) { | |
b915c95f AK |
10195 | /* check each device to ensure it is still working */ |
10196 | struct mdinfo *sd; | |
10197 | new_degraded = 0; | |
10198 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10199 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
10200 | continue; | |
10201 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
10202 | char sbuf[20]; | |
10203 | if (sysfs_get_str(info, | |
10204 | sd, "state", sbuf, 20) < 0 || | |
10205 | strstr(sbuf, "faulty") || | |
10206 | strstr(sbuf, "in_sync") == NULL) { | |
10207 | /* this device is dead */ | |
10208 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
10209 | if (sd->disk.raid_disk >= 0 && | |
10210 | sources[sd->disk.raid_disk] >= 0) { | |
10211 | close(sources[ | |
10212 | sd->disk.raid_disk]); | |
10213 | sources[sd->disk.raid_disk] = | |
10214 | -1; | |
10215 | } | |
10216 | new_degraded++; | |
10217 | } | |
10218 | } | |
10219 | } | |
10220 | } | |
10221 | ||
10222 | return new_degraded; | |
10223 | } | |
10224 | ||
10f22854 AK |
10225 | /******************************************************************************* |
10226 | * Function: imsm_manage_reshape | |
10227 | * Description: Function finds array under reshape and it manages reshape | |
10228 | * process. It creates stripes backups (if required) and sets | |
10229 | * checheckpoits. | |
10230 | * Parameters: | |
10231 | * afd : Backup handle (nattive) - not used | |
10232 | * sra : general array info | |
10233 | * reshape : reshape parameters - not used | |
10234 | * st : supertype structure | |
10235 | * blocks : size of critical section [blocks] | |
10236 | * fds : table of source device descriptor | |
10237 | * offsets : start of array (offest per devices) | |
10238 | * dests : not used | |
10239 | * destfd : table of destination device descriptor | |
10240 | * destoffsets : table of destination offsets (per device) | |
10241 | * Returns: | |
10242 | * 1 : success, reshape is done | |
10243 | * 0 : fail | |
10244 | ******************************************************************************/ | |
999b4972 N |
10245 | static int imsm_manage_reshape( |
10246 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 10247 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
10248 | int *fds, unsigned long long *offsets, |
10249 | int dests, int *destfd, unsigned long long *destoffsets) | |
10250 | { | |
10f22854 AK |
10251 | int ret_val = 0; |
10252 | struct intel_super *super = st->sb; | |
10253 | struct intel_dev *dv = NULL; | |
10254 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 10255 | struct imsm_map *map_src; |
10f22854 AK |
10256 | int migr_vol_qan = 0; |
10257 | int ndata, odata; /* [bytes] */ | |
10258 | int chunk; /* [bytes] */ | |
10259 | struct migr_record *migr_rec; | |
10260 | char *buf = NULL; | |
10261 | unsigned int buf_size; /* [bytes] */ | |
10262 | unsigned long long max_position; /* array size [bytes] */ | |
10263 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
10264 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
10265 | unsigned long long start_src; /* [bytes] */ |
10266 | unsigned long long start; /* [bytes] */ | |
10267 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 10268 | int degraded = 0; |
ab724b98 | 10269 | int source_layout = 0; |
10f22854 | 10270 | |
1ab242d8 | 10271 | if (!fds || !offsets || !sra) |
10f22854 AK |
10272 | goto abort; |
10273 | ||
10274 | /* Find volume during the reshape */ | |
10275 | for (dv = super->devlist; dv; dv = dv->next) { | |
10276 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
10277 | && dv->dev->vol.migr_state == 1) { | |
10278 | dev = dv->dev; | |
10279 | migr_vol_qan++; | |
10280 | } | |
10281 | } | |
10282 | /* Only one volume can migrate at the same time */ | |
10283 | if (migr_vol_qan != 1) { | |
e7b84f9d | 10284 | pr_err(": %s", migr_vol_qan ? |
10f22854 AK |
10285 | "Number of migrating volumes greater than 1\n" : |
10286 | "There is no volume during migrationg\n"); | |
10287 | goto abort; | |
10288 | } | |
10289 | ||
238c0a71 | 10290 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
10291 | if (map_src == NULL) |
10292 | goto abort; | |
10f22854 | 10293 | |
238c0a71 AK |
10294 | ndata = imsm_num_data_members(dev, MAP_0); |
10295 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 10296 | |
7b1ab482 | 10297 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
10298 | old_data_stripe_length = odata * chunk; |
10299 | ||
10300 | migr_rec = super->migr_rec; | |
10301 | ||
10f22854 AK |
10302 | /* initialize migration record for start condition */ |
10303 | if (sra->reshape_progress == 0) | |
10304 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
10305 | else { |
10306 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
10307 | dprintf("imsm: cannot restart migration when data " | |
10308 | "are present in copy area.\n"); | |
10309 | goto abort; | |
10310 | } | |
6a75c8ca AK |
10311 | /* Save checkpoint to update migration record for current |
10312 | * reshape position (in md). It can be farther than current | |
10313 | * reshape position in metadata. | |
10314 | */ | |
10315 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
10316 | /* ignore error == 2, this can mean end of reshape here | |
10317 | */ | |
10318 | dprintf("imsm: Cannot write checkpoint to " | |
10319 | "migration record (UNIT_SRC_NORMAL, " | |
10320 | "initial save)\n"); | |
10321 | goto abort; | |
10322 | } | |
b2c59438 | 10323 | } |
10f22854 AK |
10324 | |
10325 | /* size for data */ | |
10326 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
10327 | /* extend buffer size for parity disk */ | |
10328 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
10329 | /* add space for stripe aligment */ | |
10330 | buf_size += old_data_stripe_length; | |
10331 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
10332 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
10333 | goto abort; | |
10334 | } | |
10335 | ||
3ef4403c | 10336 | max_position = sra->component_size * ndata; |
68eb8bc6 | 10337 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
10338 | |
10339 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
10340 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
10341 | /* current reshape position [blocks] */ | |
10342 | unsigned long long current_position = | |
10343 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
10344 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
10345 | unsigned long long border; | |
10346 | ||
b915c95f AK |
10347 | /* Check that array hasn't become failed. |
10348 | */ | |
10349 | degraded = check_degradation_change(sra, fds, degraded); | |
10350 | if (degraded > 1) { | |
10351 | dprintf("imsm: Abort reshape due to degradation" | |
10352 | " level (%i)\n", degraded); | |
10353 | goto abort; | |
10354 | } | |
10355 | ||
10f22854 AK |
10356 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
10357 | ||
10358 | if ((current_position + next_step) > max_position) | |
10359 | next_step = max_position - current_position; | |
10360 | ||
92144abf | 10361 | start = current_position * 512; |
10f22854 AK |
10362 | |
10363 | /* allign reading start to old geometry */ | |
10364 | start_buf_shift = start % old_data_stripe_length; | |
10365 | start_src = start - start_buf_shift; | |
10366 | ||
10367 | border = (start_src / odata) - (start / ndata); | |
10368 | border /= 512; | |
10369 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
10370 | /* save critical stripes to buf | |
10371 | * start - start address of current unit | |
10372 | * to backup [bytes] | |
10373 | * start_src - start address of current unit | |
10374 | * to backup alligned to source array | |
10375 | * [bytes] | |
10376 | */ | |
10377 | unsigned long long next_step_filler = 0; | |
10378 | unsigned long long copy_length = next_step * 512; | |
10379 | ||
10380 | /* allign copy area length to stripe in old geometry */ | |
10381 | next_step_filler = ((copy_length + start_buf_shift) | |
10382 | % old_data_stripe_length); | |
10383 | if (next_step_filler) | |
10384 | next_step_filler = (old_data_stripe_length | |
10385 | - next_step_filler); | |
10386 | dprintf("save_stripes() parameters: start = %llu," | |
10387 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
10388 | "\tstart_in_buf_shift = %llu," | |
10389 | "\tnext_step_filler = %llu\n", | |
10390 | start, start_src, copy_length, | |
10391 | start_buf_shift, next_step_filler); | |
10392 | ||
10393 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
10394 | chunk, map_src->raid_level, |
10395 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
10396 | copy_length + |
10397 | next_step_filler + start_buf_shift, | |
10398 | buf)) { | |
10399 | dprintf("imsm: Cannot save stripes" | |
10400 | " to buffer\n"); | |
10401 | goto abort; | |
10402 | } | |
10403 | /* Convert data to destination format and store it | |
10404 | * in backup general migration area | |
10405 | */ | |
10406 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 10407 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
10408 | dprintf("imsm: Cannot save stripes to " |
10409 | "target devices\n"); | |
10410 | goto abort; | |
10411 | } | |
10412 | if (save_checkpoint_imsm(st, sra, | |
10413 | UNIT_SRC_IN_CP_AREA)) { | |
10414 | dprintf("imsm: Cannot write checkpoint to " | |
10415 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
10416 | goto abort; | |
10417 | } | |
8016a6d4 AK |
10418 | } else { |
10419 | /* set next step to use whole border area */ | |
10420 | border /= next_step; | |
10421 | if (border > 1) | |
10422 | next_step *= border; | |
10f22854 AK |
10423 | } |
10424 | /* When data backed up, checkpoint stored, | |
10425 | * kick the kernel to reshape unit of data | |
10426 | */ | |
10427 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
10428 | /* limit next step to array max position */ |
10429 | if (next_step > max_position) | |
10430 | next_step = max_position; | |
10f22854 AK |
10431 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
10432 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 10433 | sra->reshape_progress = next_step; |
10f22854 AK |
10434 | |
10435 | /* wait until reshape finish */ | |
ae9f01f8 | 10436 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
10437 | dprintf("wait_for_reshape_imsm returned error!\n"); |
10438 | goto abort; | |
10439 | } | |
10f22854 | 10440 | |
0228d92c AK |
10441 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
10442 | /* ignore error == 2, this can mean end of reshape here | |
10443 | */ | |
10f22854 AK |
10444 | dprintf("imsm: Cannot write checkpoint to " |
10445 | "migration record (UNIT_SRC_NORMAL)\n"); | |
10446 | goto abort; | |
10447 | } | |
10448 | ||
10449 | } | |
10450 | ||
10451 | /* return '1' if done */ | |
10452 | ret_val = 1; | |
10453 | abort: | |
10454 | free(buf); | |
10455 | abort_reshape(sra); | |
10456 | ||
10457 | return ret_val; | |
999b4972 | 10458 | } |
71204a50 | 10459 | #endif /* MDASSEMBLE */ |
999b4972 | 10460 | |
cdddbdbc DW |
10461 | struct superswitch super_imsm = { |
10462 | #ifndef MDASSEMBLE | |
10463 | .examine_super = examine_super_imsm, | |
10464 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 10465 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 10466 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
10467 | .detail_super = detail_super_imsm, |
10468 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 10469 | .write_init_super = write_init_super_imsm, |
0e600426 N |
10470 | .validate_geometry = validate_geometry_imsm, |
10471 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 10472 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 10473 | .detail_platform = detail_platform_imsm, |
e50cf220 | 10474 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 10475 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 10476 | .update_subarray = update_subarray_imsm, |
2b959fbf | 10477 | .load_container = load_container_imsm, |
71204a50 N |
10478 | .default_geometry = default_geometry_imsm, |
10479 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
10480 | .reshape_super = imsm_reshape_super, | |
10481 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 10482 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
10483 | #endif |
10484 | .match_home = match_home_imsm, | |
10485 | .uuid_from_super= uuid_from_super_imsm, | |
10486 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 10487 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
10488 | .update_super = update_super_imsm, |
10489 | ||
10490 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 10491 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
10492 | |
10493 | .compare_super = compare_super_imsm, | |
10494 | ||
10495 | .load_super = load_super_imsm, | |
bf5a934a | 10496 | .init_super = init_super_imsm, |
e683ca88 | 10497 | .store_super = store_super_imsm, |
cdddbdbc DW |
10498 | .free_super = free_super_imsm, |
10499 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 10500 | .container_content = container_content_imsm, |
cdddbdbc | 10501 | |
276d77db | 10502 | |
cdddbdbc | 10503 | .external = 1, |
4cce4069 | 10504 | .name = "imsm", |
845dea95 | 10505 | |
0e600426 | 10506 | #ifndef MDASSEMBLE |
845dea95 NB |
10507 | /* for mdmon */ |
10508 | .open_new = imsm_open_new, | |
ed9d66aa | 10509 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
10510 | .set_disk = imsm_set_disk, |
10511 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 10512 | .activate_spare = imsm_activate_spare, |
e8319a19 | 10513 | .process_update = imsm_process_update, |
8273f55e | 10514 | .prepare_update = imsm_prepare_update, |
0e600426 | 10515 | #endif /* MDASSEMBLE */ |
cdddbdbc | 10516 | }; |