]>
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 */ | |
98 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
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 DW |
104 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
105 | #define IMSM_DISK_FILLERS 4 | |
cdddbdbc DW |
106 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ |
107 | }; | |
108 | ||
3b451610 AK |
109 | /* map selector for map managment |
110 | */ | |
238c0a71 AK |
111 | #define MAP_0 0 |
112 | #define MAP_1 1 | |
113 | #define MAP_X -1 | |
3b451610 | 114 | |
cdddbdbc DW |
115 | /* RAID map configuration infos. */ |
116 | struct imsm_map { | |
117 | __u32 pba_of_lba0; /* start address of partition */ | |
118 | __u32 blocks_per_member;/* blocks per member */ | |
119 | __u32 num_data_stripes; /* number of data stripes */ | |
120 | __u16 blocks_per_strip; | |
121 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
122 | #define IMSM_T_STATE_NORMAL 0 | |
123 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
124 | #define IMSM_T_STATE_DEGRADED 2 |
125 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
126 | __u8 raid_level; |
127 | #define IMSM_T_RAID0 0 | |
128 | #define IMSM_T_RAID1 1 | |
129 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
130 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
131 | __u8 num_domains; /* number of parity domains */ |
132 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 133 | __u8 ddf; |
cdddbdbc | 134 | __u32 filler[7]; /* expansion area */ |
7eef0453 | 135 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 136 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
137 | * top byte contains some flags |
138 | */ | |
cdddbdbc DW |
139 | } __attribute__ ((packed)); |
140 | ||
141 | struct imsm_vol { | |
f8f603f1 | 142 | __u32 curr_migr_unit; |
fe7ed8cb | 143 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 144 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
145 | #define MIGR_INIT 0 |
146 | #define MIGR_REBUILD 1 | |
147 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
148 | #define MIGR_GEN_MIGR 3 | |
149 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 150 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
151 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
152 | __u8 dirty; | |
fe7ed8cb DW |
153 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
154 | __u16 verify_errors; /* number of mismatches */ | |
155 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
156 | __u32 filler[4]; | |
cdddbdbc DW |
157 | struct imsm_map map[1]; |
158 | /* here comes another one if migr_state */ | |
159 | } __attribute__ ((packed)); | |
160 | ||
161 | struct imsm_dev { | |
fe7ed8cb | 162 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
163 | __u32 size_low; |
164 | __u32 size_high; | |
fe7ed8cb DW |
165 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
166 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
167 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
168 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
169 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
170 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
171 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
172 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
173 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
174 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
175 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
176 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
177 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
178 | __u32 status; /* Persistent RaidDev status */ |
179 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
180 | __u8 migr_priority; |
181 | __u8 num_sub_vols; | |
182 | __u8 tid; | |
183 | __u8 cng_master_disk; | |
184 | __u16 cache_policy; | |
185 | __u8 cng_state; | |
186 | __u8 cng_sub_state; | |
187 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
188 | __u32 filler[IMSM_DEV_FILLERS]; |
189 | struct imsm_vol vol; | |
190 | } __attribute__ ((packed)); | |
191 | ||
192 | struct imsm_super { | |
193 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
194 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
195 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
196 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
197 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
198 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
199 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
200 | __u8 num_disks; /* 0x38 Number of configured disks */ |
201 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
202 | __u8 error_log_pos; /* 0x3A */ |
203 | __u8 fill[1]; /* 0x3B */ | |
204 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
205 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
206 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
207 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
208 | #define IMSM_FILLERS 35 | |
209 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
210 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
211 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 212 | /* here comes BBM logs */ |
cdddbdbc DW |
213 | } __attribute__ ((packed)); |
214 | ||
604b746f JD |
215 | #define BBM_LOG_MAX_ENTRIES 254 |
216 | ||
217 | struct bbm_log_entry { | |
218 | __u64 defective_block_start; | |
219 | #define UNREADABLE 0xFFFFFFFF | |
220 | __u32 spare_block_offset; | |
221 | __u16 remapped_marked_count; | |
222 | __u16 disk_ordinal; | |
223 | } __attribute__ ((__packed__)); | |
224 | ||
225 | struct bbm_log { | |
226 | __u32 signature; /* 0xABADB10C */ | |
227 | __u32 entry_count; | |
228 | __u32 reserved_spare_block_count; /* 0 */ | |
229 | __u32 reserved; /* 0xFFFF */ | |
230 | __u64 first_spare_lba; | |
231 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
232 | } __attribute__ ((__packed__)); | |
233 | ||
234 | ||
cdddbdbc DW |
235 | #ifndef MDASSEMBLE |
236 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
237 | #endif | |
238 | ||
8e59f3d8 AK |
239 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
240 | ||
241 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
242 | ||
243 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must | |
244 | * be recovered using srcMap */ | |
245 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
246 | * already been migrated and must | |
247 | * be recovered from checkpoint area */ | |
248 | struct migr_record { | |
249 | __u32 rec_status; /* Status used to determine how to restart | |
250 | * migration in case it aborts | |
251 | * in some fashion */ | |
252 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
253 | __u32 family_num; /* Family number of MPB | |
254 | * containing the RaidDev | |
255 | * that is migrating */ | |
256 | __u32 ascending_migr; /* True if migrating in increasing | |
257 | * order of lbas */ | |
258 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
259 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
260 | * member disk | |
261 | * advances per unit-of-operation */ | |
262 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
263 | __u32 dest_1st_member_lba; /* First member lba on first | |
264 | * stripe of destination */ | |
265 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
266 | __u32 post_migr_vol_cap; /* Size of volume after | |
267 | * migration completes */ | |
268 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
269 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
270 | * migration ckpt record was read from | |
271 | * (for recovered migrations) */ | |
272 | } __attribute__ ((__packed__)); | |
273 | ||
1484e727 DW |
274 | static __u8 migr_type(struct imsm_dev *dev) |
275 | { | |
276 | if (dev->vol.migr_type == MIGR_VERIFY && | |
277 | dev->status & DEV_VERIFY_AND_FIX) | |
278 | return MIGR_REPAIR; | |
279 | else | |
280 | return dev->vol.migr_type; | |
281 | } | |
282 | ||
283 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
284 | { | |
285 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
286 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
287 | */ | |
288 | if (migr_type == MIGR_REPAIR) { | |
289 | dev->vol.migr_type = MIGR_VERIFY; | |
290 | dev->status |= DEV_VERIFY_AND_FIX; | |
291 | } else { | |
292 | dev->vol.migr_type = migr_type; | |
293 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
294 | } | |
295 | } | |
296 | ||
87eb16df | 297 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 298 | { |
87eb16df DW |
299 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
300 | } | |
cdddbdbc | 301 | |
87eb16df DW |
302 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
303 | { | |
304 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
305 | } |
306 | ||
ba2de7ba DW |
307 | struct intel_dev { |
308 | struct imsm_dev *dev; | |
309 | struct intel_dev *next; | |
f21e18ca | 310 | unsigned index; |
ba2de7ba DW |
311 | }; |
312 | ||
88654014 LM |
313 | struct intel_hba { |
314 | enum sys_dev_type type; | |
315 | char *path; | |
316 | char *pci_id; | |
317 | struct intel_hba *next; | |
318 | }; | |
319 | ||
1a64be56 LM |
320 | enum action { |
321 | DISK_REMOVE = 1, | |
322 | DISK_ADD | |
323 | }; | |
cdddbdbc DW |
324 | /* internal representation of IMSM metadata */ |
325 | struct intel_super { | |
326 | union { | |
949c47a0 DW |
327 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
328 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 329 | }; |
8e59f3d8 AK |
330 | union { |
331 | void *migr_rec_buf; /* buffer for I/O operations */ | |
332 | struct migr_record *migr_rec; /* migration record */ | |
333 | }; | |
949c47a0 | 334 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
335 | void *next_buf; /* for realloc'ing buf from the manager */ |
336 | size_t next_len; | |
c2c087e6 | 337 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 338 | int current_vol; /* index of raid device undergoing creation */ |
0dcecb2e | 339 | __u32 create_offset; /* common start for 'current_vol' */ |
148acb7b | 340 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 341 | struct intel_dev *devlist; |
cdddbdbc DW |
342 | struct dl { |
343 | struct dl *next; | |
344 | int index; | |
345 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
346 | int major, minor; | |
347 | char *devname; | |
b9f594fe | 348 | struct imsm_disk disk; |
cdddbdbc | 349 | int fd; |
0dcecb2e DW |
350 | int extent_cnt; |
351 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 352 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 353 | enum action action; |
ca0748fa | 354 | } *disks, *current_disk; |
1a64be56 LM |
355 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
356 | active */ | |
47ee5a45 | 357 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 358 | struct bbm_log *bbm_log; |
88654014 | 359 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 360 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
361 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
362 | }; | |
363 | ||
364 | struct intel_disk { | |
365 | struct imsm_disk disk; | |
366 | #define IMSM_UNKNOWN_OWNER (-1) | |
367 | int owner; | |
368 | struct intel_disk *next; | |
cdddbdbc DW |
369 | }; |
370 | ||
c2c087e6 DW |
371 | struct extent { |
372 | unsigned long long start, size; | |
373 | }; | |
374 | ||
694575e7 KW |
375 | /* definitions of reshape process types */ |
376 | enum imsm_reshape_type { | |
377 | CH_TAKEOVER, | |
b5347799 | 378 | CH_MIGRATION, |
694575e7 KW |
379 | }; |
380 | ||
88758e9d DW |
381 | /* definition of messages passed to imsm_process_update */ |
382 | enum imsm_update_type { | |
383 | update_activate_spare, | |
8273f55e | 384 | update_create_array, |
33414a01 | 385 | update_kill_array, |
aa534678 | 386 | update_rename_array, |
1a64be56 | 387 | update_add_remove_disk, |
78b10e66 | 388 | update_reshape_container_disks, |
48c5303a | 389 | update_reshape_migration, |
2d40f3a1 AK |
390 | update_takeover, |
391 | update_general_migration_checkpoint, | |
88758e9d DW |
392 | }; |
393 | ||
394 | struct imsm_update_activate_spare { | |
395 | enum imsm_update_type type; | |
d23fe947 | 396 | struct dl *dl; |
88758e9d DW |
397 | int slot; |
398 | int array; | |
399 | struct imsm_update_activate_spare *next; | |
400 | }; | |
401 | ||
78b10e66 N |
402 | struct geo_params { |
403 | int dev_id; | |
404 | char *dev_name; | |
405 | long long size; | |
406 | int level; | |
407 | int layout; | |
408 | int chunksize; | |
409 | int raid_disks; | |
410 | }; | |
411 | ||
bb025c2f KW |
412 | enum takeover_direction { |
413 | R10_TO_R0, | |
414 | R0_TO_R10 | |
415 | }; | |
416 | struct imsm_update_takeover { | |
417 | enum imsm_update_type type; | |
418 | int subarray; | |
419 | enum takeover_direction direction; | |
420 | }; | |
78b10e66 N |
421 | |
422 | struct imsm_update_reshape { | |
423 | enum imsm_update_type type; | |
424 | int old_raid_disks; | |
425 | int new_raid_disks; | |
48c5303a PC |
426 | |
427 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
428 | }; | |
429 | ||
430 | struct imsm_update_reshape_migration { | |
431 | enum imsm_update_type type; | |
432 | int old_raid_disks; | |
433 | int new_raid_disks; | |
434 | /* fields for array migration changes | |
435 | */ | |
436 | int subdev; | |
437 | int new_level; | |
438 | int new_layout; | |
4bba0439 | 439 | int new_chunksize; |
48c5303a | 440 | |
d195167d | 441 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
442 | }; |
443 | ||
2d40f3a1 AK |
444 | struct imsm_update_general_migration_checkpoint { |
445 | enum imsm_update_type type; | |
446 | __u32 curr_migr_unit; | |
447 | }; | |
448 | ||
54c2c1ea DW |
449 | struct disk_info { |
450 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
451 | }; | |
452 | ||
8273f55e DW |
453 | struct imsm_update_create_array { |
454 | enum imsm_update_type type; | |
8273f55e | 455 | int dev_idx; |
6a3e913e | 456 | struct imsm_dev dev; |
8273f55e DW |
457 | }; |
458 | ||
33414a01 DW |
459 | struct imsm_update_kill_array { |
460 | enum imsm_update_type type; | |
461 | int dev_idx; | |
462 | }; | |
463 | ||
aa534678 DW |
464 | struct imsm_update_rename_array { |
465 | enum imsm_update_type type; | |
466 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
467 | int dev_idx; | |
468 | }; | |
469 | ||
1a64be56 | 470 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
471 | enum imsm_update_type type; |
472 | }; | |
473 | ||
88654014 LM |
474 | |
475 | static const char *_sys_dev_type[] = { | |
476 | [SYS_DEV_UNKNOWN] = "Unknown", | |
477 | [SYS_DEV_SAS] = "SAS", | |
478 | [SYS_DEV_SATA] = "SATA" | |
479 | }; | |
480 | ||
481 | const char *get_sys_dev_type(enum sys_dev_type type) | |
482 | { | |
483 | if (type >= SYS_DEV_MAX) | |
484 | type = SYS_DEV_UNKNOWN; | |
485 | ||
486 | return _sys_dev_type[type]; | |
487 | } | |
488 | ||
489 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
490 | { | |
491 | struct intel_hba *result = malloc(sizeof(*result)); | |
492 | if (result) { | |
493 | result->type = device->type; | |
494 | result->path = strdup(device->path); | |
495 | result->next = NULL; | |
496 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
497 | result->pci_id++; | |
498 | } | |
499 | return result; | |
500 | } | |
501 | ||
502 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
503 | { | |
504 | struct intel_hba *result=NULL; | |
505 | for (result = hba; result; result = result->next) { | |
506 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
507 | break; | |
508 | } | |
509 | return result; | |
510 | } | |
511 | ||
b4cf4cba | 512 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
513 | { |
514 | struct intel_hba *hba; | |
515 | ||
516 | /* check if disk attached to Intel HBA */ | |
517 | hba = find_intel_hba(super->hba, device); | |
518 | if (hba != NULL) | |
519 | return 1; | |
520 | /* Check if HBA is already attached to super */ | |
521 | if (super->hba == NULL) { | |
522 | super->hba = alloc_intel_hba(device); | |
523 | return 1; | |
524 | } | |
525 | ||
526 | hba = super->hba; | |
527 | /* Intel metadata allows for all disks attached to the same type HBA. | |
528 | * Do not sypport odf HBA types mixing | |
529 | */ | |
530 | if (device->type != hba->type) | |
531 | return 2; | |
532 | ||
533 | while (hba->next) | |
534 | hba = hba->next; | |
535 | ||
536 | hba->next = alloc_intel_hba(device); | |
537 | return 1; | |
538 | } | |
539 | ||
540 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
541 | { | |
542 | struct sys_dev *list, *elem, *prev; | |
543 | char *disk_path; | |
544 | ||
545 | if ((list = find_intel_devices()) == NULL) | |
546 | return 0; | |
547 | ||
548 | if (fd < 0) | |
549 | disk_path = (char *) devname; | |
550 | else | |
551 | disk_path = diskfd_to_devpath(fd); | |
552 | ||
553 | if (!disk_path) { | |
554 | free_sys_dev(&list); | |
555 | return 0; | |
556 | } | |
557 | ||
558 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
559 | if (path_attached_to_hba(disk_path, elem->path)) { | |
560 | if (prev == NULL) | |
561 | list = list->next; | |
562 | else | |
563 | prev->next = elem->next; | |
564 | elem->next = NULL; | |
565 | if (disk_path != devname) | |
566 | free(disk_path); | |
567 | free_sys_dev(&list); | |
568 | return elem; | |
569 | } | |
570 | } | |
571 | if (disk_path != devname) | |
572 | free(disk_path); | |
573 | free_sys_dev(&list); | |
574 | ||
575 | return NULL; | |
576 | } | |
577 | ||
578 | ||
d424212e N |
579 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
580 | char *devname); | |
f2f5c343 | 581 | |
cdddbdbc DW |
582 | static struct supertype *match_metadata_desc_imsm(char *arg) |
583 | { | |
584 | struct supertype *st; | |
585 | ||
586 | if (strcmp(arg, "imsm") != 0 && | |
587 | strcmp(arg, "default") != 0 | |
588 | ) | |
589 | return NULL; | |
590 | ||
591 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
592 | if (!st) |
593 | return NULL; | |
ef609477 | 594 | memset(st, 0, sizeof(*st)); |
d1d599ea | 595 | st->container_dev = NoMdDev; |
cdddbdbc DW |
596 | st->ss = &super_imsm; |
597 | st->max_devs = IMSM_MAX_DEVICES; | |
598 | st->minor_version = 0; | |
599 | st->sb = NULL; | |
600 | return st; | |
601 | } | |
602 | ||
0e600426 | 603 | #ifndef MDASSEMBLE |
cdddbdbc DW |
604 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
605 | { | |
606 | return &mpb->sig[MPB_SIG_LEN]; | |
607 | } | |
9e2d750d | 608 | #endif |
cdddbdbc | 609 | |
949c47a0 DW |
610 | /* retrieve a disk directly from the anchor when the anchor is known to be |
611 | * up-to-date, currently only at load time | |
612 | */ | |
613 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 614 | { |
949c47a0 | 615 | if (index >= mpb->num_disks) |
cdddbdbc DW |
616 | return NULL; |
617 | return &mpb->disk[index]; | |
618 | } | |
619 | ||
95d07a2c LM |
620 | /* retrieve the disk description based on a index of the disk |
621 | * in the sub-array | |
622 | */ | |
623 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 624 | { |
b9f594fe DW |
625 | struct dl *d; |
626 | ||
627 | for (d = super->disks; d; d = d->next) | |
628 | if (d->index == index) | |
95d07a2c LM |
629 | return d; |
630 | ||
631 | return NULL; | |
632 | } | |
633 | /* retrieve a disk from the parsed metadata */ | |
634 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
635 | { | |
636 | struct dl *dl; | |
637 | ||
638 | dl = get_imsm_dl_disk(super, index); | |
639 | if (dl) | |
640 | return &dl->disk; | |
641 | ||
b9f594fe | 642 | return NULL; |
949c47a0 DW |
643 | } |
644 | ||
645 | /* generate a checksum directly from the anchor when the anchor is known to be | |
646 | * up-to-date, currently only at load or write_super after coalescing | |
647 | */ | |
648 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
649 | { |
650 | __u32 end = mpb->mpb_size / sizeof(end); | |
651 | __u32 *p = (__u32 *) mpb; | |
652 | __u32 sum = 0; | |
653 | ||
97f734fd N |
654 | while (end--) { |
655 | sum += __le32_to_cpu(*p); | |
656 | p++; | |
657 | } | |
cdddbdbc DW |
658 | |
659 | return sum - __le32_to_cpu(mpb->check_sum); | |
660 | } | |
661 | ||
a965f303 DW |
662 | static size_t sizeof_imsm_map(struct imsm_map *map) |
663 | { | |
664 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
665 | } | |
666 | ||
667 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 668 | { |
5e7b0330 AK |
669 | /* A device can have 2 maps if it is in the middle of a migration. |
670 | * If second_map is: | |
238c0a71 AK |
671 | * MAP_0 - we return the first map |
672 | * MAP_1 - we return the second map if it exists, else NULL | |
673 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 674 | */ |
a965f303 | 675 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 676 | struct imsm_map *map2 = NULL; |
a965f303 | 677 | |
9535fc47 AK |
678 | if (dev->vol.migr_state) |
679 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 680 | |
9535fc47 | 681 | switch (second_map) { |
3b451610 | 682 | case MAP_0: |
9535fc47 | 683 | break; |
3b451610 | 684 | case MAP_1: |
9535fc47 AK |
685 | map = map2; |
686 | break; | |
238c0a71 | 687 | case MAP_X: |
9535fc47 AK |
688 | if (map2) |
689 | map = map2; | |
690 | break; | |
9535fc47 AK |
691 | default: |
692 | map = NULL; | |
693 | } | |
694 | return map; | |
5e7b0330 | 695 | |
a965f303 | 696 | } |
cdddbdbc | 697 | |
3393c6af DW |
698 | /* return the size of the device. |
699 | * migr_state increases the returned size if map[0] were to be duplicated | |
700 | */ | |
701 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
702 | { |
703 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 704 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
705 | |
706 | /* migrating means an additional map */ | |
a965f303 | 707 | if (dev->vol.migr_state) |
238c0a71 | 708 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 709 | else if (migr_state) |
238c0a71 | 710 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
711 | |
712 | return size; | |
713 | } | |
714 | ||
54c2c1ea DW |
715 | #ifndef MDASSEMBLE |
716 | /* retrieve disk serial number list from a metadata update */ | |
717 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
718 | { | |
719 | void *u = update; | |
720 | struct disk_info *inf; | |
721 | ||
722 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
723 | sizeof_imsm_dev(&update->dev, 0); | |
724 | ||
725 | return inf; | |
726 | } | |
727 | #endif | |
728 | ||
949c47a0 | 729 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
730 | { |
731 | int offset; | |
732 | int i; | |
733 | void *_mpb = mpb; | |
734 | ||
949c47a0 | 735 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
736 | return NULL; |
737 | ||
738 | /* devices start after all disks */ | |
739 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
740 | ||
741 | for (i = 0; i <= index; i++) | |
742 | if (i == index) | |
743 | return _mpb + offset; | |
744 | else | |
3393c6af | 745 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
746 | |
747 | return NULL; | |
748 | } | |
749 | ||
949c47a0 DW |
750 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
751 | { | |
ba2de7ba DW |
752 | struct intel_dev *dv; |
753 | ||
949c47a0 DW |
754 | if (index >= super->anchor->num_raid_devs) |
755 | return NULL; | |
ba2de7ba DW |
756 | for (dv = super->devlist; dv; dv = dv->next) |
757 | if (dv->index == index) | |
758 | return dv->dev; | |
759 | return NULL; | |
949c47a0 DW |
760 | } |
761 | ||
98130f40 AK |
762 | /* |
763 | * for second_map: | |
238c0a71 AK |
764 | * == MAP_0 get first map |
765 | * == MAP_1 get second map | |
766 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
767 | */ |
768 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
769 | int slot, | |
770 | int second_map) | |
7eef0453 DW |
771 | { |
772 | struct imsm_map *map; | |
773 | ||
5e7b0330 | 774 | map = get_imsm_map(dev, second_map); |
7eef0453 | 775 | |
ff077194 DW |
776 | /* top byte identifies disk under rebuild */ |
777 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
778 | } | |
779 | ||
780 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 781 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 782 | { |
98130f40 | 783 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
784 | |
785 | return ord_to_idx(ord); | |
7eef0453 DW |
786 | } |
787 | ||
be73972f DW |
788 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
789 | { | |
790 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
791 | } | |
792 | ||
f21e18ca | 793 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
794 | { |
795 | int slot; | |
796 | __u32 ord; | |
797 | ||
798 | for (slot = 0; slot < map->num_members; slot++) { | |
799 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
800 | if (ord_to_idx(ord) == idx) | |
801 | return slot; | |
802 | } | |
803 | ||
804 | return -1; | |
805 | } | |
806 | ||
cdddbdbc DW |
807 | static int get_imsm_raid_level(struct imsm_map *map) |
808 | { | |
809 | if (map->raid_level == 1) { | |
810 | if (map->num_members == 2) | |
811 | return 1; | |
812 | else | |
813 | return 10; | |
814 | } | |
815 | ||
816 | return map->raid_level; | |
817 | } | |
818 | ||
c2c087e6 DW |
819 | static int cmp_extent(const void *av, const void *bv) |
820 | { | |
821 | const struct extent *a = av; | |
822 | const struct extent *b = bv; | |
823 | if (a->start < b->start) | |
824 | return -1; | |
825 | if (a->start > b->start) | |
826 | return 1; | |
827 | return 0; | |
828 | } | |
829 | ||
0dcecb2e | 830 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 831 | { |
c2c087e6 | 832 | int memberships = 0; |
620b1713 | 833 | int i; |
c2c087e6 | 834 | |
949c47a0 DW |
835 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
836 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 837 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 838 | |
620b1713 DW |
839 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
840 | memberships++; | |
c2c087e6 | 841 | } |
0dcecb2e DW |
842 | |
843 | return memberships; | |
844 | } | |
845 | ||
b81221b7 CA |
846 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
847 | ||
0dcecb2e DW |
848 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
849 | { | |
850 | /* find a list of used extents on the given physical device */ | |
851 | struct extent *rv, *e; | |
620b1713 | 852 | int i; |
0dcecb2e | 853 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
854 | __u32 reservation; |
855 | ||
856 | /* trim the reserved area for spares, so they can join any array | |
857 | * regardless of whether the OROM has assigned sectors from the | |
858 | * IMSM_RESERVED_SECTORS region | |
859 | */ | |
860 | if (dl->index == -1) | |
b81221b7 | 861 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
862 | else |
863 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 864 | |
c2c087e6 DW |
865 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
866 | if (!rv) | |
867 | return NULL; | |
868 | e = rv; | |
869 | ||
949c47a0 DW |
870 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
871 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 872 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 873 | |
620b1713 DW |
874 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
875 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
876 | e->size = __le32_to_cpu(map->blocks_per_member); | |
877 | e++; | |
c2c087e6 DW |
878 | } |
879 | } | |
880 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
881 | ||
14e8215b DW |
882 | /* determine the start of the metadata |
883 | * when no raid devices are defined use the default | |
884 | * ...otherwise allow the metadata to truncate the value | |
885 | * as is the case with older versions of imsm | |
886 | */ | |
887 | if (memberships) { | |
888 | struct extent *last = &rv[memberships - 1]; | |
889 | __u32 remainder; | |
890 | ||
891 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
892 | (last->start + last->size); | |
dda5855f DW |
893 | /* round down to 1k block to satisfy precision of the kernel |
894 | * 'size' interface | |
895 | */ | |
896 | remainder &= ~1UL; | |
897 | /* make sure remainder is still sane */ | |
f21e18ca | 898 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 899 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
900 | if (reservation > remainder) |
901 | reservation = remainder; | |
902 | } | |
903 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
904 | e->size = 0; |
905 | return rv; | |
906 | } | |
907 | ||
14e8215b DW |
908 | /* try to determine how much space is reserved for metadata from |
909 | * the last get_extents() entry, otherwise fallback to the | |
910 | * default | |
911 | */ | |
912 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
913 | { | |
914 | struct extent *e; | |
915 | int i; | |
916 | __u32 rv; | |
917 | ||
918 | /* for spares just return a minimal reservation which will grow | |
919 | * once the spare is picked up by an array | |
920 | */ | |
921 | if (dl->index == -1) | |
922 | return MPB_SECTOR_CNT; | |
923 | ||
924 | e = get_extents(super, dl); | |
925 | if (!e) | |
926 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
927 | ||
928 | /* scroll to last entry */ | |
929 | for (i = 0; e[i].size; i++) | |
930 | continue; | |
931 | ||
932 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
933 | ||
934 | free(e); | |
935 | ||
936 | return rv; | |
937 | } | |
938 | ||
25ed7e59 DW |
939 | static int is_spare(struct imsm_disk *disk) |
940 | { | |
941 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
942 | } | |
943 | ||
944 | static int is_configured(struct imsm_disk *disk) | |
945 | { | |
946 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
947 | } | |
948 | ||
949 | static int is_failed(struct imsm_disk *disk) | |
950 | { | |
951 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
952 | } | |
953 | ||
b81221b7 CA |
954 | /* try to determine how much space is reserved for metadata from |
955 | * the last get_extents() entry on the smallest active disk, | |
956 | * otherwise fallback to the default | |
957 | */ | |
958 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
959 | { | |
960 | struct extent *e; | |
961 | int i; | |
962 | __u32 min_active, remainder; | |
963 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
964 | struct dl *dl, *dl_min = NULL; | |
965 | ||
966 | if (!super) | |
967 | return rv; | |
968 | ||
969 | min_active = 0; | |
970 | for (dl = super->disks; dl; dl = dl->next) { | |
971 | if (dl->index < 0) | |
972 | continue; | |
973 | if (dl->disk.total_blocks < min_active || min_active == 0) { | |
974 | dl_min = dl; | |
975 | min_active = dl->disk.total_blocks; | |
976 | } | |
977 | } | |
978 | if (!dl_min) | |
979 | return rv; | |
980 | ||
981 | /* find last lba used by subarrays on the smallest active disk */ | |
982 | e = get_extents(super, dl_min); | |
983 | if (!e) | |
984 | return rv; | |
985 | for (i = 0; e[i].size; i++) | |
986 | continue; | |
987 | ||
988 | remainder = min_active - e[i].start; | |
989 | free(e); | |
990 | ||
991 | /* to give priority to recovery we should not require full | |
992 | IMSM_RESERVED_SECTORS from the spare */ | |
993 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
994 | ||
995 | /* if real reservation is smaller use that value */ | |
996 | return (remainder < rv) ? remainder : rv; | |
997 | } | |
998 | ||
80e7f8c3 AC |
999 | /* Return minimum size of a spare that can be used in this array*/ |
1000 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1001 | { | |
1002 | struct intel_super *super = st->sb; | |
1003 | struct dl *dl; | |
1004 | struct extent *e; | |
1005 | int i; | |
1006 | unsigned long long rv = 0; | |
1007 | ||
1008 | if (!super) | |
1009 | return rv; | |
1010 | /* find first active disk in array */ | |
1011 | dl = super->disks; | |
1012 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1013 | dl = dl->next; | |
1014 | if (!dl) | |
1015 | return rv; | |
1016 | /* find last lba used by subarrays */ | |
1017 | e = get_extents(super, dl); | |
1018 | if (!e) | |
1019 | return rv; | |
1020 | for (i = 0; e[i].size; i++) | |
1021 | continue; | |
1022 | if (i > 0) | |
1023 | rv = e[i-1].start + e[i-1].size; | |
1024 | free(e); | |
b81221b7 | 1025 | |
80e7f8c3 | 1026 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1027 | rv = rv + imsm_min_reserved_sectors(super); |
1028 | ||
80e7f8c3 AC |
1029 | return rv * 512; |
1030 | } | |
1031 | ||
d1e02575 AK |
1032 | static int is_gen_migration(struct imsm_dev *dev); |
1033 | ||
1799c9e8 | 1034 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1035 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1036 | struct imsm_dev *dev); | |
1e5c6983 | 1037 | |
c47b0ff6 AK |
1038 | static void print_imsm_dev(struct intel_super *super, |
1039 | struct imsm_dev *dev, | |
1040 | char *uuid, | |
1041 | int disk_idx) | |
cdddbdbc DW |
1042 | { |
1043 | __u64 sz; | |
0d80bb2f | 1044 | int slot, i; |
238c0a71 AK |
1045 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1046 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1047 | __u32 ord; |
cdddbdbc DW |
1048 | |
1049 | printf("\n"); | |
1e7bc0ed | 1050 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1051 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1052 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1053 | if (map2) | |
1054 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1055 | printf("\n"); | |
1056 | printf(" Members : %d", map->num_members); | |
1057 | if (map2) | |
1058 | printf(" <-- %d", map2->num_members); | |
1059 | printf("\n"); | |
0d80bb2f DW |
1060 | printf(" Slots : ["); |
1061 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1062 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1063 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1064 | } | |
dd8bcb3b AK |
1065 | printf("]"); |
1066 | if (map2) { | |
1067 | printf(" <-- ["); | |
1068 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1069 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1070 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1071 | } | |
1072 | printf("]"); | |
1073 | } | |
1074 | printf("\n"); | |
7095bccb AK |
1075 | printf(" Failed disk : "); |
1076 | if (map->failed_disk_num == 0xff) | |
1077 | printf("none"); | |
1078 | else | |
1079 | printf("%i", map->failed_disk_num); | |
1080 | printf("\n"); | |
620b1713 DW |
1081 | slot = get_imsm_disk_slot(map, disk_idx); |
1082 | if (slot >= 0) { | |
238c0a71 | 1083 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1084 | printf(" This Slot : %d%s\n", slot, |
1085 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1086 | } else | |
cdddbdbc DW |
1087 | printf(" This Slot : ?\n"); |
1088 | sz = __le32_to_cpu(dev->size_high); | |
1089 | sz <<= 32; | |
1090 | sz += __le32_to_cpu(dev->size_low); | |
1091 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1092 | human_size(sz * 512)); | |
1093 | sz = __le32_to_cpu(map->blocks_per_member); | |
1094 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
1095 | human_size(sz * 512)); | |
1096 | printf(" Sector Offset : %u\n", | |
1097 | __le32_to_cpu(map->pba_of_lba0)); | |
1098 | printf(" Num Stripes : %u\n", | |
1099 | __le32_to_cpu(map->num_data_stripes)); | |
dd8bcb3b | 1100 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1101 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1102 | if (map2) |
1103 | printf(" <-- %u KiB", | |
1104 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1105 | printf("\n"); | |
cdddbdbc | 1106 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1107 | printf(" Migrate State : "); |
1484e727 DW |
1108 | if (dev->vol.migr_state) { |
1109 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1110 | printf("initialize\n"); |
1484e727 | 1111 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1112 | printf("rebuild\n"); |
1484e727 | 1113 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1114 | printf("check\n"); |
1484e727 | 1115 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1116 | printf("general migration\n"); |
1484e727 | 1117 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1118 | printf("state change\n"); |
1484e727 | 1119 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1120 | printf("repair\n"); |
1484e727 | 1121 | else |
8655a7b1 DW |
1122 | printf("<unknown:%d>\n", migr_type(dev)); |
1123 | } else | |
1124 | printf("idle\n"); | |
3393c6af DW |
1125 | printf(" Map State : %s", map_state_str[map->map_state]); |
1126 | if (dev->vol.migr_state) { | |
238c0a71 | 1127 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1128 | |
b10b37b8 | 1129 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1130 | printf("\n Checkpoint : %u ", |
1131 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
1132 | if ((is_gen_migration(dev)) && (super->disks->index > 1)) | |
1133 | printf("(N/A)"); | |
1134 | else | |
1135 | printf("(%llu)", (unsigned long long) | |
1136 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1137 | } |
1138 | printf("\n"); | |
cdddbdbc | 1139 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1140 | } |
1141 | ||
0ec1f4e8 | 1142 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1143 | { |
1f24f035 | 1144 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1145 | __u64 sz; |
1146 | ||
0ec1f4e8 | 1147 | if (index < -1 || !disk) |
e9d82038 DW |
1148 | return; |
1149 | ||
cdddbdbc | 1150 | printf("\n"); |
1f24f035 | 1151 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1152 | if (index >= 0) |
1153 | printf(" Disk%02d Serial : %s\n", index, str); | |
1154 | else | |
1155 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1156 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1157 | is_configured(disk) ? " active" : "", | |
1158 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1159 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
14e8215b | 1160 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
1161 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1162 | human_size(sz * 512)); | |
1163 | } | |
1164 | ||
520e69e2 AK |
1165 | void examine_migr_rec_imsm(struct intel_super *super) |
1166 | { | |
1167 | struct migr_record *migr_rec = super->migr_rec; | |
1168 | struct imsm_super *mpb = super->anchor; | |
1169 | int i; | |
1170 | ||
1171 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1172 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1173 | if (is_gen_migration(dev) == 0) | |
1174 | continue; | |
1175 | ||
1176 | printf("\nMigration Record Information:"); | |
1177 | if (super->disks->index > 1) { | |
1178 | printf(" Empty\n "); | |
1179 | printf("Examine one of first two disks in array\n"); | |
1180 | break; | |
1181 | } | |
1182 | printf("\n Status : "); | |
1183 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1184 | printf("Normal\n"); | |
1185 | else | |
1186 | printf("Contains Data\n"); | |
1187 | printf(" Current Unit : %u\n", | |
1188 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1189 | printf(" Family : %u\n", | |
1190 | __le32_to_cpu(migr_rec->family_num)); | |
1191 | printf(" Ascending : %u\n", | |
1192 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1193 | printf(" Blocks Per Unit : %u\n", | |
1194 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1195 | printf(" Dest. Depth Per Unit : %u\n", | |
1196 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1197 | printf(" Checkpoint Area pba : %u\n", | |
1198 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1199 | printf(" First member lba : %u\n", | |
1200 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1201 | printf(" Total Number of Units : %u\n", | |
1202 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1203 | printf(" Size of volume : %u\n", | |
1204 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1205 | printf(" Expansion space for LBA64 : %u\n", | |
1206 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1207 | printf(" Record was read from : %u\n", | |
1208 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1209 | ||
1210 | break; | |
1211 | } | |
1212 | } | |
9e2d750d | 1213 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1214 | /******************************************************************************* |
1215 | * function: imsm_check_attributes | |
1216 | * Description: Function checks if features represented by attributes flags | |
1217 | * are supported by mdadm. | |
1218 | * Parameters: | |
1219 | * attributes - Attributes read from metadata | |
1220 | * Returns: | |
1221 | * 0 - passed attributes contains unsupported features flags | |
1222 | * 1 - all features are supported | |
1223 | ******************************************************************************/ | |
1224 | static int imsm_check_attributes(__u32 attributes) | |
1225 | { | |
1226 | int ret_val = 1; | |
418f9b36 N |
1227 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1228 | ||
1229 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1230 | |
1231 | not_supported &= attributes; | |
1232 | if (not_supported) { | |
418f9b36 N |
1233 | fprintf(stderr, Name "(IMSM): Unsupported attributes : %x\n", |
1234 | (unsigned)__le32_to_cpu(not_supported)); | |
19482bcc AK |
1235 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1236 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1237 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1238 | } | |
1239 | if (not_supported & MPB_ATTRIB_2TB) { | |
1240 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1241 | not_supported ^= MPB_ATTRIB_2TB; | |
1242 | } | |
1243 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1244 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1245 | not_supported ^= MPB_ATTRIB_RAID0; | |
1246 | } | |
1247 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1248 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1249 | not_supported ^= MPB_ATTRIB_RAID1; | |
1250 | } | |
1251 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1252 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1253 | not_supported ^= MPB_ATTRIB_RAID10; | |
1254 | } | |
1255 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1256 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1257 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1258 | } | |
1259 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1260 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1261 | not_supported ^= MPB_ATTRIB_RAID5; | |
1262 | } | |
1263 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1264 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1265 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1266 | } | |
1267 | if (not_supported & MPB_ATTRIB_BBM) { | |
1268 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1269 | not_supported ^= MPB_ATTRIB_BBM; | |
1270 | } | |
1271 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1272 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1273 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1274 | } | |
1275 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1276 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1277 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1278 | } | |
1279 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1280 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1281 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1282 | } | |
1283 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1284 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1285 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1286 | } | |
1287 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1288 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1289 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1290 | } | |
1291 | ||
1292 | if (not_supported) | |
1293 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1294 | ||
1295 | ret_val = 0; | |
1296 | } | |
1297 | ||
1298 | return ret_val; | |
1299 | } | |
1300 | ||
9e2d750d | 1301 | #ifndef MDASSEMBLE |
a5d85af7 | 1302 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1303 | |
cdddbdbc DW |
1304 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1305 | { | |
1306 | struct intel_super *super = st->sb; | |
949c47a0 | 1307 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1308 | char str[MAX_SIGNATURE_LENGTH]; |
1309 | int i; | |
27fd6274 DW |
1310 | struct mdinfo info; |
1311 | char nbuf[64]; | |
cdddbdbc | 1312 | __u32 sum; |
14e8215b | 1313 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1314 | struct dl *dl; |
27fd6274 | 1315 | |
cdddbdbc DW |
1316 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1317 | printf(" Magic : %s\n", str); | |
1318 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1319 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1320 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1321 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1322 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1323 | printf(" Attributes : "); |
1324 | if (imsm_check_attributes(mpb->attributes)) | |
1325 | printf("All supported\n"); | |
1326 | else | |
1327 | printf("not supported\n"); | |
a5d85af7 | 1328 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1329 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1330 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1331 | sum = __le32_to_cpu(mpb->check_sum); |
1332 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1333 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1334 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1335 | printf(" Disks : %d\n", mpb->num_disks); |
1336 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1337 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1338 | if (super->bbm_log) { |
1339 | struct bbm_log *log = super->bbm_log; | |
1340 | ||
1341 | printf("\n"); | |
1342 | printf("Bad Block Management Log:\n"); | |
1343 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1344 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1345 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1346 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1347 | printf(" First Spare : %llx\n", |
1348 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1349 | } |
44470971 DW |
1350 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1351 | struct mdinfo info; | |
1352 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1353 | ||
1354 | super->current_vol = i; | |
a5d85af7 | 1355 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1356 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1357 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1358 | } |
cdddbdbc DW |
1359 | for (i = 0; i < mpb->num_disks; i++) { |
1360 | if (i == super->disks->index) | |
1361 | continue; | |
0ec1f4e8 | 1362 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1363 | } |
94827db3 | 1364 | |
0ec1f4e8 DW |
1365 | for (dl = super->disks; dl; dl = dl->next) |
1366 | if (dl->index == -1) | |
1367 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1368 | |
1369 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1370 | } |
1371 | ||
061f2c6a | 1372 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1373 | { |
27fd6274 | 1374 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1375 | struct mdinfo info; |
1376 | char nbuf[64]; | |
1e7bc0ed | 1377 | struct intel_super *super = st->sb; |
1e7bc0ed | 1378 | |
0d5a423f DW |
1379 | if (!super->anchor->num_raid_devs) { |
1380 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1381 | return; |
0d5a423f | 1382 | } |
ff54de6e | 1383 | |
a5d85af7 | 1384 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1385 | fname_from_uuid(st, &info, nbuf, ':'); |
1386 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1387 | } | |
1388 | ||
1389 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1390 | { | |
1391 | /* We just write a generic IMSM ARRAY entry */ | |
1392 | struct mdinfo info; | |
1393 | char nbuf[64]; | |
1394 | char nbuf1[64]; | |
1395 | struct intel_super *super = st->sb; | |
1396 | int i; | |
1397 | ||
1398 | if (!super->anchor->num_raid_devs) | |
1399 | return; | |
1400 | ||
a5d85af7 | 1401 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1402 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1403 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1404 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1405 | ||
1406 | super->current_vol = i; | |
a5d85af7 | 1407 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1408 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1409 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1410 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1411 | } |
cdddbdbc DW |
1412 | } |
1413 | ||
9d84c8ea DW |
1414 | static void export_examine_super_imsm(struct supertype *st) |
1415 | { | |
1416 | struct intel_super *super = st->sb; | |
1417 | struct imsm_super *mpb = super->anchor; | |
1418 | struct mdinfo info; | |
1419 | char nbuf[64]; | |
1420 | ||
a5d85af7 | 1421 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1422 | fname_from_uuid(st, &info, nbuf, ':'); |
1423 | printf("MD_METADATA=imsm\n"); | |
1424 | printf("MD_LEVEL=container\n"); | |
1425 | printf("MD_UUID=%s\n", nbuf+5); | |
1426 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1427 | } | |
1428 | ||
cdddbdbc DW |
1429 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1430 | { | |
3ebe00a1 DW |
1431 | struct mdinfo info; |
1432 | char nbuf[64]; | |
1433 | ||
a5d85af7 | 1434 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1435 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1436 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1437 | } |
1438 | ||
1439 | static void brief_detail_super_imsm(struct supertype *st) | |
1440 | { | |
ff54de6e N |
1441 | struct mdinfo info; |
1442 | char nbuf[64]; | |
a5d85af7 | 1443 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1444 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1445 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1446 | } |
d665cc31 DW |
1447 | |
1448 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1449 | static void fd2devname(int fd, char *name); | |
1450 | ||
120dc887 | 1451 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1452 | { |
120dc887 LM |
1453 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1454 | * controller, as well as non-connected ports | |
d665cc31 DW |
1455 | */ |
1456 | int hba_len = strlen(hba_path) + 1; | |
1457 | struct dirent *ent; | |
1458 | DIR *dir; | |
1459 | char *path = NULL; | |
1460 | int err = 0; | |
1461 | unsigned long port_mask = (1 << port_count) - 1; | |
1462 | ||
f21e18ca | 1463 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1464 | if (verbose) |
1465 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1466 | return 2; | |
1467 | } | |
1468 | ||
1469 | /* scroll through /sys/dev/block looking for devices attached to | |
1470 | * this hba | |
1471 | */ | |
1472 | dir = opendir("/sys/dev/block"); | |
1473 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1474 | int fd; | |
1475 | char model[64]; | |
1476 | char vendor[64]; | |
1477 | char buf[1024]; | |
1478 | int major, minor; | |
1479 | char *device; | |
1480 | char *c; | |
1481 | int port; | |
1482 | int type; | |
1483 | ||
1484 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1485 | continue; | |
1486 | path = devt_to_devpath(makedev(major, minor)); | |
1487 | if (!path) | |
1488 | continue; | |
1489 | if (!path_attached_to_hba(path, hba_path)) { | |
1490 | free(path); | |
1491 | path = NULL; | |
1492 | continue; | |
1493 | } | |
1494 | ||
1495 | /* retrieve the scsi device type */ | |
1496 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1497 | if (verbose) | |
1498 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1499 | err = 2; | |
1500 | break; | |
1501 | } | |
1502 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1503 | if (load_sys(device, buf) != 0) { | |
1504 | if (verbose) | |
1505 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1506 | path); | |
1507 | err = 2; | |
1508 | free(device); | |
1509 | break; | |
1510 | } | |
1511 | type = strtoul(buf, NULL, 10); | |
1512 | ||
1513 | /* if it's not a disk print the vendor and model */ | |
1514 | if (!(type == 0 || type == 7 || type == 14)) { | |
1515 | vendor[0] = '\0'; | |
1516 | model[0] = '\0'; | |
1517 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1518 | if (load_sys(device, buf) == 0) { | |
1519 | strncpy(vendor, buf, sizeof(vendor)); | |
1520 | vendor[sizeof(vendor) - 1] = '\0'; | |
1521 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1522 | while (isspace(*c) || *c == '\0') | |
1523 | *c-- = '\0'; | |
1524 | ||
1525 | } | |
1526 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1527 | if (load_sys(device, buf) == 0) { | |
1528 | strncpy(model, buf, sizeof(model)); | |
1529 | model[sizeof(model) - 1] = '\0'; | |
1530 | c = (char *) &model[sizeof(model) - 1]; | |
1531 | while (isspace(*c) || *c == '\0') | |
1532 | *c-- = '\0'; | |
1533 | } | |
1534 | ||
1535 | if (vendor[0] && model[0]) | |
1536 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1537 | else | |
1538 | switch (type) { /* numbers from hald/linux/device.c */ | |
1539 | case 1: sprintf(buf, "tape"); break; | |
1540 | case 2: sprintf(buf, "printer"); break; | |
1541 | case 3: sprintf(buf, "processor"); break; | |
1542 | case 4: | |
1543 | case 5: sprintf(buf, "cdrom"); break; | |
1544 | case 6: sprintf(buf, "scanner"); break; | |
1545 | case 8: sprintf(buf, "media_changer"); break; | |
1546 | case 9: sprintf(buf, "comm"); break; | |
1547 | case 12: sprintf(buf, "raid"); break; | |
1548 | default: sprintf(buf, "unknown"); | |
1549 | } | |
1550 | } else | |
1551 | buf[0] = '\0'; | |
1552 | free(device); | |
1553 | ||
1554 | /* chop device path to 'host%d' and calculate the port number */ | |
1555 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1556 | if (!c) { |
1557 | if (verbose) | |
1558 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1559 | err = 2; | |
1560 | break; | |
1561 | } | |
d665cc31 DW |
1562 | *c = '\0'; |
1563 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1564 | port -= host_base; | |
1565 | else { | |
1566 | if (verbose) { | |
1567 | *c = '/'; /* repair the full string */ | |
1568 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1569 | path); | |
1570 | } | |
1571 | err = 2; | |
1572 | break; | |
1573 | } | |
1574 | ||
1575 | /* mark this port as used */ | |
1576 | port_mask &= ~(1 << port); | |
1577 | ||
1578 | /* print out the device information */ | |
1579 | if (buf[0]) { | |
1580 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1581 | continue; | |
1582 | } | |
1583 | ||
1584 | fd = dev_open(ent->d_name, O_RDONLY); | |
1585 | if (fd < 0) | |
1586 | printf(" Port%d : - disk info unavailable -\n", port); | |
1587 | else { | |
1588 | fd2devname(fd, buf); | |
1589 | printf(" Port%d : %s", port, buf); | |
1590 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1591 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1592 | else |
664d5325 | 1593 | printf(" ()\n"); |
4dab422a | 1594 | close(fd); |
d665cc31 | 1595 | } |
d665cc31 DW |
1596 | free(path); |
1597 | path = NULL; | |
1598 | } | |
1599 | if (path) | |
1600 | free(path); | |
1601 | if (dir) | |
1602 | closedir(dir); | |
1603 | if (err == 0) { | |
1604 | int i; | |
1605 | ||
1606 | for (i = 0; i < port_count; i++) | |
1607 | if (port_mask & (1 << i)) | |
1608 | printf(" Port%d : - no device attached -\n", i); | |
1609 | } | |
1610 | ||
1611 | return err; | |
1612 | } | |
1613 | ||
120dc887 LM |
1614 | static void print_found_intel_controllers(struct sys_dev *elem) |
1615 | { | |
1616 | for (; elem; elem = elem->next) { | |
1617 | fprintf(stderr, Name ": found Intel(R) "); | |
1618 | if (elem->type == SYS_DEV_SATA) | |
1619 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1620 | else if (elem->type == SYS_DEV_SAS) |
1621 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1622 | fprintf(stderr, "RAID controller"); |
1623 | if (elem->pci_id) | |
1624 | fprintf(stderr, " at %s", elem->pci_id); | |
1625 | fprintf(stderr, ".\n"); | |
1626 | } | |
1627 | fflush(stderr); | |
1628 | } | |
1629 | ||
120dc887 LM |
1630 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1631 | { | |
1632 | struct dirent *ent; | |
1633 | DIR *dir; | |
1634 | int host_base = -1; | |
1635 | ||
1636 | *port_count = 0; | |
1637 | if ((dir = opendir(hba_path)) == NULL) | |
1638 | return -1; | |
1639 | ||
1640 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1641 | int host; | |
1642 | ||
1643 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1644 | continue; | |
1645 | if (*port_count == 0) | |
1646 | host_base = host; | |
1647 | else if (host < host_base) | |
1648 | host_base = host; | |
1649 | ||
1650 | if (host + 1 > *port_count + host_base) | |
1651 | *port_count = host + 1 - host_base; | |
1652 | } | |
1653 | closedir(dir); | |
1654 | return host_base; | |
1655 | } | |
1656 | ||
a891a3c2 LM |
1657 | static void print_imsm_capability(const struct imsm_orom *orom) |
1658 | { | |
1659 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1660 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1661 | orom->hotfix_ver, orom->build); | |
1662 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1663 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1664 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1665 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1666 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1667 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1668 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1669 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1670 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1671 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1672 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1673 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1674 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1675 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1676 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1677 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1678 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1679 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1680 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1681 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1682 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1683 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1684 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
1685 | printf(" Max Disks : %d\n", orom->tds); | |
1686 | printf(" Max Volumes : %d\n", orom->vpa); | |
1687 | return; | |
1688 | } | |
1689 | ||
5615172f | 1690 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1691 | { |
1692 | /* There are two components to imsm platform support, the ahci SATA | |
1693 | * controller and the option-rom. To find the SATA controller we | |
1694 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1695 | * controller with the Intel vendor id is present. This approach | |
1696 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1697 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1698 | * detect platform raid capabilities. The option-rom resides in a | |
1699 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1700 | * platform capabilities. If raid support is disabled in the BIOS the | |
1701 | * option-rom capability structure will not be available. | |
1702 | */ | |
1703 | const struct imsm_orom *orom; | |
1704 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1705 | int host_base = 0; |
1706 | int port_count = 0; | |
120dc887 | 1707 | int result=0; |
d665cc31 | 1708 | |
5615172f | 1709 | if (enumerate_only) { |
a891a3c2 | 1710 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1711 | return 0; |
a891a3c2 LM |
1712 | list = find_intel_devices(); |
1713 | if (!list) | |
1714 | return 2; | |
1715 | for (hba = list; hba; hba = hba->next) { | |
1716 | orom = find_imsm_capability(hba->type); | |
1717 | if (!orom) { | |
1718 | result = 2; | |
1719 | break; | |
1720 | } | |
1721 | } | |
1722 | free_sys_dev(&list); | |
1723 | return result; | |
5615172f DW |
1724 | } |
1725 | ||
155cbb4c LM |
1726 | list = find_intel_devices(); |
1727 | if (!list) { | |
d665cc31 | 1728 | if (verbose) |
155cbb4c LM |
1729 | fprintf(stderr, Name ": no active Intel(R) RAID " |
1730 | "controller found.\n"); | |
d665cc31 DW |
1731 | free_sys_dev(&list); |
1732 | return 2; | |
1733 | } else if (verbose) | |
155cbb4c | 1734 | print_found_intel_controllers(list); |
d665cc31 | 1735 | |
a891a3c2 LM |
1736 | for (hba = list; hba; hba = hba->next) { |
1737 | orom = find_imsm_capability(hba->type); | |
1738 | if (!orom) | |
1739 | fprintf(stderr, Name ": imsm capabilities not found for controller: %s (type %s)\n", | |
1740 | hba->path, get_sys_dev_type(hba->type)); | |
1741 | else | |
1742 | print_imsm_capability(orom); | |
d665cc31 DW |
1743 | } |
1744 | ||
120dc887 LM |
1745 | for (hba = list; hba; hba = hba->next) { |
1746 | printf(" I/O Controller : %s (%s)\n", | |
1747 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1748 | |
120dc887 LM |
1749 | if (hba->type == SYS_DEV_SATA) { |
1750 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1751 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1752 | if (verbose) | |
1753 | fprintf(stderr, Name ": failed to enumerate " | |
1754 | "ports on SATA controller at %s.", hba->pci_id); | |
1755 | result |= 2; | |
1756 | } | |
1757 | } | |
d665cc31 | 1758 | } |
155cbb4c | 1759 | |
120dc887 LM |
1760 | free_sys_dev(&list); |
1761 | return result; | |
d665cc31 | 1762 | } |
cdddbdbc DW |
1763 | #endif |
1764 | ||
1765 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1766 | { | |
5115ca67 DW |
1767 | /* the imsm metadata format does not specify any host |
1768 | * identification information. We return -1 since we can never | |
1769 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1770 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1771 | * exclude member disks that do not belong, and we rely on |
1772 | * mdadm.conf to specify the arrays that should be assembled. | |
1773 | * Auto-assembly may still pick up "foreign" arrays. | |
1774 | */ | |
cdddbdbc | 1775 | |
9362c1c8 | 1776 | return -1; |
cdddbdbc DW |
1777 | } |
1778 | ||
1779 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1780 | { | |
51006d85 N |
1781 | /* The uuid returned here is used for: |
1782 | * uuid to put into bitmap file (Create, Grow) | |
1783 | * uuid for backup header when saving critical section (Grow) | |
1784 | * comparing uuids when re-adding a device into an array | |
1785 | * In these cases the uuid required is that of the data-array, | |
1786 | * not the device-set. | |
1787 | * uuid to recognise same set when adding a missing device back | |
1788 | * to an array. This is a uuid for the device-set. | |
1789 | * | |
1790 | * For each of these we can make do with a truncated | |
1791 | * or hashed uuid rather than the original, as long as | |
1792 | * everyone agrees. | |
1793 | * In each case the uuid required is that of the data-array, | |
1794 | * not the device-set. | |
43dad3d6 | 1795 | */ |
51006d85 N |
1796 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1797 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1798 | * - the orig_family_num of the container |
51006d85 N |
1799 | * - the index number of the volume |
1800 | * - the 'serial' number of the volume. | |
1801 | * Hopefully these are all constant. | |
1802 | */ | |
1803 | struct intel_super *super = st->sb; | |
43dad3d6 | 1804 | |
51006d85 N |
1805 | char buf[20]; |
1806 | struct sha1_ctx ctx; | |
1807 | struct imsm_dev *dev = NULL; | |
148acb7b | 1808 | __u32 family_num; |
51006d85 | 1809 | |
148acb7b DW |
1810 | /* some mdadm versions failed to set ->orig_family_num, in which |
1811 | * case fall back to ->family_num. orig_family_num will be | |
1812 | * fixed up with the first metadata update. | |
1813 | */ | |
1814 | family_num = super->anchor->orig_family_num; | |
1815 | if (family_num == 0) | |
1816 | family_num = super->anchor->family_num; | |
51006d85 | 1817 | sha1_init_ctx(&ctx); |
92bd8f8d | 1818 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1819 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1820 | if (super->current_vol >= 0) |
1821 | dev = get_imsm_dev(super, super->current_vol); | |
1822 | if (dev) { | |
1823 | __u32 vol = super->current_vol; | |
1824 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1825 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1826 | } | |
1827 | sha1_finish_ctx(&ctx, buf); | |
1828 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1829 | } |
1830 | ||
0d481d37 | 1831 | #if 0 |
4f5bc454 DW |
1832 | static void |
1833 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1834 | { |
cdddbdbc DW |
1835 | __u8 *v = get_imsm_version(mpb); |
1836 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1837 | char major[] = { 0, 0, 0 }; | |
1838 | char minor[] = { 0 ,0, 0 }; | |
1839 | char patch[] = { 0, 0, 0 }; | |
1840 | char *ver_parse[] = { major, minor, patch }; | |
1841 | int i, j; | |
1842 | ||
1843 | i = j = 0; | |
1844 | while (*v != '\0' && v < end) { | |
1845 | if (*v != '.' && j < 2) | |
1846 | ver_parse[i][j++] = *v; | |
1847 | else { | |
1848 | i++; | |
1849 | j = 0; | |
1850 | } | |
1851 | v++; | |
1852 | } | |
1853 | ||
4f5bc454 DW |
1854 | *m = strtol(minor, NULL, 0); |
1855 | *p = strtol(patch, NULL, 0); | |
1856 | } | |
0d481d37 | 1857 | #endif |
4f5bc454 | 1858 | |
1e5c6983 DW |
1859 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1860 | { | |
1861 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 1862 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1863 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1864 | ||
1865 | switch (get_imsm_raid_level(map)) { | |
1866 | case 5: | |
1867 | case 10: | |
1868 | return chunk; | |
1869 | default: | |
1870 | return 128*1024 >> 9; | |
1871 | } | |
1872 | } | |
1873 | ||
1874 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1875 | { | |
1876 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1877 | * this is different than migr_strip_size_resync(), but it's good | |
1878 | * to be compatible | |
1879 | */ | |
238c0a71 | 1880 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
1881 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1882 | ||
1883 | switch (get_imsm_raid_level(map)) { | |
1884 | case 1: | |
1885 | case 10: | |
1886 | if (map->num_members % map->num_domains == 0) | |
1887 | return 128*1024 >> 9; | |
1888 | else | |
1889 | return chunk; | |
1890 | case 5: | |
1891 | return max((__u32) 64*1024 >> 9, chunk); | |
1892 | default: | |
1893 | return 128*1024 >> 9; | |
1894 | } | |
1895 | } | |
1896 | ||
1897 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
1898 | { | |
238c0a71 AK |
1899 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1900 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
1901 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
1902 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
1903 | ||
1904 | return max((__u32) 1, hi_chunk / lo_chunk); | |
1905 | } | |
1906 | ||
1907 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
1908 | { | |
238c0a71 | 1909 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1910 | int level = get_imsm_raid_level(lo); |
1911 | ||
1912 | if (level == 1 || level == 10) { | |
238c0a71 | 1913 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
1914 | |
1915 | return hi->num_domains; | |
1916 | } else | |
1917 | return num_stripes_per_unit_resync(dev); | |
1918 | } | |
1919 | ||
98130f40 | 1920 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
1921 | { |
1922 | /* named 'imsm_' because raid0, raid1 and raid10 | |
1923 | * counter-intuitively have the same number of data disks | |
1924 | */ | |
98130f40 | 1925 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
1926 | |
1927 | switch (get_imsm_raid_level(map)) { | |
1928 | case 0: | |
1929 | case 1: | |
1930 | case 10: | |
1931 | return map->num_members; | |
1932 | case 5: | |
1933 | return map->num_members - 1; | |
1934 | default: | |
1935 | dprintf("%s: unsupported raid level\n", __func__); | |
1936 | return 0; | |
1937 | } | |
1938 | } | |
1939 | ||
1940 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
1941 | { | |
238c0a71 | 1942 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1943 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1944 | ||
1945 | switch(get_imsm_raid_level(map)) { | |
1946 | case 1: | |
1947 | case 10: | |
1948 | return chunk * map->num_domains; | |
1949 | case 5: | |
1950 | return chunk * map->num_members; | |
1951 | default: | |
1952 | return chunk; | |
1953 | } | |
1954 | } | |
1955 | ||
1956 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
1957 | { | |
238c0a71 | 1958 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
1959 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1960 | __u32 strip = block / chunk; | |
1961 | ||
1962 | switch (get_imsm_raid_level(map)) { | |
1963 | case 1: | |
1964 | case 10: { | |
1965 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
1966 | __u32 vol_stripe = vol_strip / map->num_members; | |
1967 | ||
1968 | return vol_stripe * chunk + block % chunk; | |
1969 | } case 5: { | |
1970 | __u32 stripe = strip / (map->num_members - 1); | |
1971 | ||
1972 | return stripe * chunk + block % chunk; | |
1973 | } | |
1974 | default: | |
1975 | return 0; | |
1976 | } | |
1977 | } | |
1978 | ||
c47b0ff6 AK |
1979 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1980 | struct imsm_dev *dev) | |
1e5c6983 DW |
1981 | { |
1982 | /* calculate the conversion factor between per member 'blocks' | |
1983 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
1984 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
1985 | */ | |
1986 | if (!dev->vol.migr_state) | |
1987 | return 0; | |
1988 | ||
1989 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
1990 | case MIGR_GEN_MIGR: { |
1991 | struct migr_record *migr_rec = super->migr_rec; | |
1992 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
1993 | } | |
1e5c6983 DW |
1994 | case MIGR_VERIFY: |
1995 | case MIGR_REPAIR: | |
1996 | case MIGR_INIT: { | |
238c0a71 | 1997 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1998 | __u32 stripes_per_unit; |
1999 | __u32 blocks_per_unit; | |
2000 | __u32 parity_depth; | |
2001 | __u32 migr_chunk; | |
2002 | __u32 block_map; | |
2003 | __u32 block_rel; | |
2004 | __u32 segment; | |
2005 | __u32 stripe; | |
2006 | __u8 disks; | |
2007 | ||
2008 | /* yes, this is really the translation of migr_units to | |
2009 | * per-member blocks in the 'resync' case | |
2010 | */ | |
2011 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2012 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2013 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2014 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2015 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2016 | segment = blocks_per_unit / stripe; |
2017 | block_rel = blocks_per_unit - segment * stripe; | |
2018 | parity_depth = parity_segment_depth(dev); | |
2019 | block_map = map_migr_block(dev, block_rel); | |
2020 | return block_map + parity_depth * segment; | |
2021 | } | |
2022 | case MIGR_REBUILD: { | |
2023 | __u32 stripes_per_unit; | |
2024 | __u32 migr_chunk; | |
2025 | ||
2026 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2027 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2028 | return migr_chunk * stripes_per_unit; | |
2029 | } | |
1e5c6983 DW |
2030 | case MIGR_STATE_CHANGE: |
2031 | default: | |
2032 | return 0; | |
2033 | } | |
2034 | } | |
2035 | ||
c2c087e6 DW |
2036 | static int imsm_level_to_layout(int level) |
2037 | { | |
2038 | switch (level) { | |
2039 | case 0: | |
2040 | case 1: | |
2041 | return 0; | |
2042 | case 5: | |
2043 | case 6: | |
a380c027 | 2044 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2045 | case 10: |
c92a2527 | 2046 | return 0x102; |
c2c087e6 | 2047 | } |
a18a888e | 2048 | return UnSet; |
c2c087e6 DW |
2049 | } |
2050 | ||
8e59f3d8 AK |
2051 | /******************************************************************************* |
2052 | * Function: read_imsm_migr_rec | |
2053 | * Description: Function reads imsm migration record from last sector of disk | |
2054 | * Parameters: | |
2055 | * fd : disk descriptor | |
2056 | * super : metadata info | |
2057 | * Returns: | |
2058 | * 0 : success, | |
2059 | * -1 : fail | |
2060 | ******************************************************************************/ | |
2061 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2062 | { | |
2063 | int ret_val = -1; | |
2064 | unsigned long long dsize; | |
2065 | ||
2066 | get_dev_size(fd, NULL, &dsize); | |
2067 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
2068 | fprintf(stderr, | |
2069 | Name ": Cannot seek to anchor block: %s\n", | |
2070 | strerror(errno)); | |
2071 | goto out; | |
2072 | } | |
2073 | if (read(fd, super->migr_rec_buf, 512) != 512) { | |
2074 | fprintf(stderr, | |
2075 | Name ": Cannot read migr record block: %s\n", | |
2076 | strerror(errno)); | |
2077 | goto out; | |
2078 | } | |
2079 | ret_val = 0; | |
2080 | ||
2081 | out: | |
2082 | return ret_val; | |
2083 | } | |
2084 | ||
2085 | /******************************************************************************* | |
2086 | * Function: load_imsm_migr_rec | |
2087 | * Description: Function reads imsm migration record (it is stored at the last | |
2088 | * sector of disk) | |
2089 | * Parameters: | |
2090 | * super : imsm internal array info | |
2091 | * info : general array info | |
2092 | * Returns: | |
2093 | * 0 : success | |
2094 | * -1 : fail | |
2095 | ******************************************************************************/ | |
2096 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2097 | { | |
2098 | struct mdinfo *sd; | |
2099 | struct dl *dl = NULL; | |
2100 | char nm[30]; | |
2101 | int retval = -1; | |
2102 | int fd = -1; | |
2103 | ||
2104 | if (info) { | |
2105 | for (sd = info->devs ; sd ; sd = sd->next) { | |
2106 | /* read only from one of the first two slots */ | |
2107 | if ((sd->disk.raid_disk > 1) || | |
2108 | (sd->disk.raid_disk < 0)) | |
2109 | continue; | |
2110 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
2111 | fd = dev_open(nm, O_RDONLY); | |
2112 | if (fd >= 0) | |
2113 | break; | |
2114 | } | |
2115 | } | |
2116 | if (fd < 0) { | |
2117 | for (dl = super->disks; dl; dl = dl->next) { | |
2118 | /* read only from one of the first two slots */ | |
2119 | if (dl->index > 1) | |
2120 | continue; | |
2121 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2122 | fd = dev_open(nm, O_RDONLY); | |
2123 | if (fd >= 0) | |
2124 | break; | |
2125 | } | |
2126 | } | |
2127 | if (fd < 0) | |
2128 | goto out; | |
2129 | retval = read_imsm_migr_rec(fd, super); | |
2130 | ||
2131 | out: | |
2132 | if (fd >= 0) | |
2133 | close(fd); | |
2134 | return retval; | |
2135 | } | |
2136 | ||
9e2d750d | 2137 | #ifndef MDASSEMBLE |
c17608ea AK |
2138 | /******************************************************************************* |
2139 | * function: imsm_create_metadata_checkpoint_update | |
2140 | * Description: It creates update for checkpoint change. | |
2141 | * Parameters: | |
2142 | * super : imsm internal array info | |
2143 | * u : pointer to prepared update | |
2144 | * Returns: | |
2145 | * Uptate length. | |
2146 | * If length is equal to 0, input pointer u contains no update | |
2147 | ******************************************************************************/ | |
2148 | static int imsm_create_metadata_checkpoint_update( | |
2149 | struct intel_super *super, | |
2150 | struct imsm_update_general_migration_checkpoint **u) | |
2151 | { | |
2152 | ||
2153 | int update_memory_size = 0; | |
2154 | ||
2155 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2156 | ||
2157 | if (u == NULL) | |
2158 | return 0; | |
2159 | *u = NULL; | |
2160 | ||
2161 | /* size of all update data without anchor */ | |
2162 | update_memory_size = | |
2163 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2164 | ||
2165 | *u = calloc(1, update_memory_size); | |
2166 | if (*u == NULL) { | |
2167 | dprintf("error: cannot get memory for " | |
2168 | "imsm_create_metadata_checkpoint_update update\n"); | |
2169 | return 0; | |
2170 | } | |
2171 | (*u)->type = update_general_migration_checkpoint; | |
2172 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2173 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2174 | (*u)->curr_migr_unit); | |
2175 | ||
2176 | return update_memory_size; | |
2177 | } | |
2178 | ||
2179 | ||
2180 | static void imsm_update_metadata_locally(struct supertype *st, | |
2181 | void *buf, int len); | |
2182 | ||
687629c2 AK |
2183 | /******************************************************************************* |
2184 | * Function: write_imsm_migr_rec | |
2185 | * Description: Function writes imsm migration record | |
2186 | * (at the last sector of disk) | |
2187 | * Parameters: | |
2188 | * super : imsm internal array info | |
2189 | * Returns: | |
2190 | * 0 : success | |
2191 | * -1 : if fail | |
2192 | ******************************************************************************/ | |
2193 | static int write_imsm_migr_rec(struct supertype *st) | |
2194 | { | |
2195 | struct intel_super *super = st->sb; | |
2196 | unsigned long long dsize; | |
2197 | char nm[30]; | |
2198 | int fd = -1; | |
2199 | int retval = -1; | |
2200 | struct dl *sd; | |
c17608ea AK |
2201 | int len; |
2202 | struct imsm_update_general_migration_checkpoint *u; | |
687629c2 AK |
2203 | |
2204 | for (sd = super->disks ; sd ; sd = sd->next) { | |
2205 | /* write to 2 first slots only */ | |
2206 | if ((sd->index < 0) || (sd->index > 1)) | |
2207 | continue; | |
2208 | sprintf(nm, "%d:%d", sd->major, sd->minor); | |
2209 | fd = dev_open(nm, O_RDWR); | |
2210 | if (fd < 0) | |
2211 | continue; | |
2212 | get_dev_size(fd, NULL, &dsize); | |
2213 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
2214 | fprintf(stderr, | |
2215 | Name ": Cannot seek to anchor block: %s\n", | |
2216 | strerror(errno)); | |
2217 | goto out; | |
2218 | } | |
2219 | if (write(fd, super->migr_rec_buf, 512) != 512) { | |
2220 | fprintf(stderr, | |
2221 | Name ": Cannot write migr record block: %s\n", | |
2222 | strerror(errno)); | |
2223 | goto out; | |
2224 | } | |
2225 | close(fd); | |
2226 | fd = -1; | |
2227 | } | |
c17608ea AK |
2228 | /* update checkpoint information in metadata */ |
2229 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2230 | ||
2231 | if (len <= 0) { | |
2232 | dprintf("imsm: Cannot prepare update\n"); | |
2233 | goto out; | |
2234 | } | |
2235 | /* update metadata locally */ | |
2236 | imsm_update_metadata_locally(st, u, len); | |
2237 | /* and possibly remotely */ | |
2238 | if (st->update_tail) { | |
2239 | append_metadata_update(st, u, len); | |
2240 | /* during reshape we do all work inside metadata handler | |
2241 | * manage_reshape(), so metadata update has to be triggered | |
2242 | * insida it | |
2243 | */ | |
2244 | flush_metadata_updates(st); | |
2245 | st->update_tail = &st->updates; | |
2246 | } else | |
2247 | free(u); | |
687629c2 AK |
2248 | |
2249 | retval = 0; | |
2250 | out: | |
2251 | if (fd >= 0) | |
2252 | close(fd); | |
2253 | return retval; | |
2254 | } | |
9e2d750d | 2255 | #endif /* MDASSEMBLE */ |
687629c2 | 2256 | |
e2962bfc AK |
2257 | /* spare/missing disks activations are not allowe when |
2258 | * array/container performs reshape operation, because | |
2259 | * all arrays in container works on the same disks set | |
2260 | */ | |
2261 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2262 | { | |
2263 | int rv = 0; | |
2264 | struct intel_dev *i_dev; | |
2265 | struct imsm_dev *dev; | |
2266 | ||
2267 | /* check whole container | |
2268 | */ | |
2269 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2270 | dev = i_dev->dev; | |
3ad25638 | 2271 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2272 | /* No repair during any migration in container |
2273 | */ | |
2274 | rv = 1; | |
2275 | break; | |
2276 | } | |
2277 | } | |
2278 | return rv; | |
2279 | } | |
2280 | ||
a5d85af7 | 2281 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2282 | { |
2283 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2284 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2285 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2286 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2287 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2288 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2289 | struct dl *dl; |
e207da2f | 2290 | char *devname; |
139dae11 | 2291 | unsigned int component_size_alligment; |
a5d85af7 | 2292 | int map_disks = info->array.raid_disks; |
bf5a934a | 2293 | |
95eeceeb | 2294 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2295 | if (prev_map) |
2296 | map_to_analyse = prev_map; | |
2297 | ||
ca0748fa | 2298 | dl = super->current_disk; |
9894ec0d | 2299 | |
bf5a934a | 2300 | info->container_member = super->current_vol; |
cd0430a1 | 2301 | info->array.raid_disks = map->num_members; |
b335e593 | 2302 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2303 | info->array.layout = imsm_level_to_layout(info->array.level); |
2304 | info->array.md_minor = -1; | |
2305 | info->array.ctime = 0; | |
2306 | info->array.utime = 0; | |
b335e593 AK |
2307 | info->array.chunk_size = |
2308 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2309 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2310 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2311 | info->custom_array_size <<= 32; | |
2312 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2313 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2314 | ||
3f510843 | 2315 | if (is_gen_migration(dev)) { |
3f83228a | 2316 | info->reshape_active = 1; |
b335e593 AK |
2317 | info->new_level = get_imsm_raid_level(map); |
2318 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2319 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2320 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2321 | if (info->delta_disks) { |
2322 | /* this needs to be applied to every array | |
2323 | * in the container. | |
2324 | */ | |
81219e70 | 2325 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2326 | } |
3f83228a N |
2327 | /* We shape information that we give to md might have to be |
2328 | * modify to cope with md's requirement for reshaping arrays. | |
2329 | * For example, when reshaping a RAID0, md requires it to be | |
2330 | * presented as a degraded RAID4. | |
2331 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2332 | * the array as already being RAID5, but the 'before' layout | |
2333 | * is a RAID4-like layout. | |
2334 | */ | |
2335 | switch (info->array.level) { | |
2336 | case 0: | |
2337 | switch(info->new_level) { | |
2338 | case 0: | |
2339 | /* conversion is happening as RAID4 */ | |
2340 | info->array.level = 4; | |
2341 | info->array.raid_disks += 1; | |
2342 | break; | |
2343 | case 5: | |
2344 | /* conversion is happening as RAID5 */ | |
2345 | info->array.level = 5; | |
2346 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2347 | info->delta_disks -= 1; |
2348 | break; | |
2349 | default: | |
2350 | /* FIXME error message */ | |
2351 | info->array.level = UnSet; | |
2352 | break; | |
2353 | } | |
2354 | break; | |
2355 | } | |
b335e593 AK |
2356 | } else { |
2357 | info->new_level = UnSet; | |
2358 | info->new_layout = UnSet; | |
2359 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2360 | info->delta_disks = 0; |
b335e593 | 2361 | } |
ca0748fa | 2362 | |
efb30e7f DW |
2363 | if (dl) { |
2364 | info->disk.major = dl->major; | |
2365 | info->disk.minor = dl->minor; | |
ca0748fa | 2366 | info->disk.number = dl->index; |
656b6b5a N |
2367 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2368 | dl->index); | |
efb30e7f | 2369 | } |
bf5a934a | 2370 | |
b335e593 AK |
2371 | info->data_offset = __le32_to_cpu(map_to_analyse->pba_of_lba0); |
2372 | info->component_size = | |
2373 | __le32_to_cpu(map_to_analyse->blocks_per_member); | |
139dae11 AK |
2374 | |
2375 | /* check component size aligment | |
2376 | */ | |
2377 | component_size_alligment = | |
2378 | info->component_size % (info->array.chunk_size/512); | |
2379 | ||
2380 | if (component_size_alligment && | |
2381 | (info->array.level != 1) && (info->array.level != UnSet)) { | |
2382 | dprintf("imsm: reported component size alligned from %llu ", | |
2383 | info->component_size); | |
2384 | info->component_size -= component_size_alligment; | |
2385 | dprintf("to %llu (%i).\n", | |
2386 | info->component_size, component_size_alligment); | |
2387 | } | |
2388 | ||
301406c9 | 2389 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2390 | info->recovery_start = MaxSector; |
bf5a934a | 2391 | |
d2e6d5d6 | 2392 | info->reshape_progress = 0; |
b6796ce1 | 2393 | info->resync_start = MaxSector; |
b9172665 AK |
2394 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2395 | dev->vol.dirty) && | |
2396 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2397 | info->resync_start = 0; |
b6796ce1 AK |
2398 | } |
2399 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2400 | switch (migr_type(dev)) { |
2401 | case MIGR_REPAIR: | |
2402 | case MIGR_INIT: { | |
c47b0ff6 AK |
2403 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2404 | dev); | |
1e5c6983 DW |
2405 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2406 | ||
2407 | info->resync_start = blocks_per_unit * units; | |
2408 | break; | |
2409 | } | |
d2e6d5d6 | 2410 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2411 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2412 | dev); | |
2413 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2414 | unsigned long long array_blocks; |
2415 | int used_disks; | |
d2e6d5d6 | 2416 | |
befb629b AK |
2417 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2418 | (units < | |
2419 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2420 | (super->migr_rec->rec_status == | |
2421 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2422 | units++; | |
2423 | ||
d2e6d5d6 | 2424 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2425 | |
d2e6d5d6 AK |
2426 | dprintf("IMSM: General Migration checkpoint : %llu " |
2427 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2428 | (unsigned long long)units, |
2429 | (unsigned long long)blocks_per_unit, | |
2430 | info->reshape_progress); | |
75156c46 | 2431 | |
238c0a71 | 2432 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 AK |
2433 | if (used_disks > 0) { |
2434 | array_blocks = map->blocks_per_member * | |
2435 | used_disks; | |
2436 | /* round array size down to closest MB | |
2437 | */ | |
2438 | info->custom_array_size = (array_blocks | |
2439 | >> SECT_PER_MB_SHIFT) | |
2440 | << SECT_PER_MB_SHIFT; | |
2441 | } | |
d2e6d5d6 | 2442 | } |
1e5c6983 DW |
2443 | case MIGR_VERIFY: |
2444 | /* we could emulate the checkpointing of | |
2445 | * 'sync_action=check' migrations, but for now | |
2446 | * we just immediately complete them | |
2447 | */ | |
2448 | case MIGR_REBUILD: | |
2449 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2450 | case MIGR_STATE_CHANGE: |
2451 | /* FIXME handle other migrations */ | |
2452 | default: | |
2453 | /* we are not dirty, so... */ | |
2454 | info->resync_start = MaxSector; | |
2455 | } | |
b6796ce1 | 2456 | } |
301406c9 DW |
2457 | |
2458 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2459 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2460 | |
f35f2525 N |
2461 | info->array.major_version = -1; |
2462 | info->array.minor_version = -2; | |
e207da2f AW |
2463 | devname = devnum2devname(st->container_dev); |
2464 | *info->text_version = '\0'; | |
2465 | if (devname) | |
2466 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2467 | free(devname); | |
a67dd8cc | 2468 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2469 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2470 | |
2471 | if (dmap) { | |
2472 | int i, j; | |
2473 | for (i=0; i<map_disks; i++) { | |
2474 | dmap[i] = 0; | |
2475 | if (i < info->array.raid_disks) { | |
2476 | struct imsm_disk *dsk; | |
238c0a71 | 2477 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2478 | dsk = get_imsm_disk(super, j); |
2479 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2480 | dmap[i] = 1; | |
2481 | } | |
2482 | } | |
2483 | } | |
81ac8b4d | 2484 | } |
bf5a934a | 2485 | |
3b451610 AK |
2486 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2487 | int failed, int look_in_map); | |
2488 | ||
2489 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2490 | int look_in_map); | |
2491 | ||
2492 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) | |
2493 | { | |
2494 | if (is_gen_migration(dev)) { | |
2495 | int failed; | |
2496 | __u8 map_state; | |
2497 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2498 | ||
2499 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2500 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2501 | if (map2->map_state != map_state) { |
2502 | map2->map_state = map_state; | |
2503 | super->updates_pending++; | |
2504 | } | |
2505 | } | |
2506 | } | |
97b4d0e9 DW |
2507 | |
2508 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2509 | { | |
2510 | struct dl *d; | |
2511 | ||
2512 | for (d = super->missing; d; d = d->next) | |
2513 | if (d->index == index) | |
2514 | return &d->disk; | |
2515 | return NULL; | |
2516 | } | |
2517 | ||
a5d85af7 | 2518 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2519 | { |
2520 | struct intel_super *super = st->sb; | |
4f5bc454 | 2521 | struct imsm_disk *disk; |
a5d85af7 | 2522 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2523 | int max_enough = -1; |
2524 | int i; | |
2525 | struct imsm_super *mpb; | |
4f5bc454 | 2526 | |
bf5a934a | 2527 | if (super->current_vol >= 0) { |
a5d85af7 | 2528 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2529 | return; |
2530 | } | |
95eeceeb | 2531 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2532 | |
2533 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2534 | * spares | |
2535 | */ | |
2536 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2537 | info->array.level = LEVEL_CONTAINER; |
2538 | info->array.layout = 0; | |
2539 | info->array.md_minor = -1; | |
c2c087e6 | 2540 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2541 | info->array.utime = 0; |
2542 | info->array.chunk_size = 0; | |
2543 | ||
2544 | info->disk.major = 0; | |
2545 | info->disk.minor = 0; | |
cdddbdbc | 2546 | info->disk.raid_disk = -1; |
c2c087e6 | 2547 | info->reshape_active = 0; |
f35f2525 N |
2548 | info->array.major_version = -1; |
2549 | info->array.minor_version = -2; | |
c2c087e6 | 2550 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2551 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2552 | info->disk.number = -1; |
2553 | info->disk.state = 0; | |
c5afc314 | 2554 | info->name[0] = 0; |
921d9e16 | 2555 | info->recovery_start = MaxSector; |
3ad25638 | 2556 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2557 | |
97b4d0e9 | 2558 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2559 | mpb = super->anchor; |
97b4d0e9 | 2560 | |
ab3cb6b3 N |
2561 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2562 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2563 | int failed, enough, j, missing = 0; | |
2564 | struct imsm_map *map; | |
2565 | __u8 state; | |
97b4d0e9 | 2566 | |
3b451610 AK |
2567 | failed = imsm_count_failed(super, dev, MAP_0); |
2568 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2569 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2570 | |
2571 | /* any newly missing disks? | |
2572 | * (catches single-degraded vs double-degraded) | |
2573 | */ | |
2574 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2575 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2576 | __u32 idx = ord_to_idx(ord); |
2577 | ||
2578 | if (!(ord & IMSM_ORD_REBUILD) && | |
2579 | get_imsm_missing(super, idx)) { | |
2580 | missing = 1; | |
2581 | break; | |
2582 | } | |
97b4d0e9 | 2583 | } |
ab3cb6b3 N |
2584 | |
2585 | if (state == IMSM_T_STATE_FAILED) | |
2586 | enough = -1; | |
2587 | else if (state == IMSM_T_STATE_DEGRADED && | |
2588 | (state != map->map_state || missing)) | |
2589 | enough = 0; | |
2590 | else /* we're normal, or already degraded */ | |
2591 | enough = 1; | |
2592 | ||
2593 | /* in the missing/failed disk case check to see | |
2594 | * if at least one array is runnable | |
2595 | */ | |
2596 | max_enough = max(max_enough, enough); | |
2597 | } | |
2598 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2599 | info->container_enough = max_enough; | |
97b4d0e9 | 2600 | |
4a04ec6c | 2601 | if (super->disks) { |
14e8215b DW |
2602 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2603 | ||
b9f594fe | 2604 | disk = &super->disks->disk; |
14e8215b DW |
2605 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
2606 | info->component_size = reserved; | |
25ed7e59 | 2607 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2608 | /* we don't change info->disk.raid_disk here because |
2609 | * this state will be finalized in mdmon after we have | |
2610 | * found the 'most fresh' version of the metadata | |
2611 | */ | |
25ed7e59 DW |
2612 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2613 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2614 | } |
a575e2a7 DW |
2615 | |
2616 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2617 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2618 | */ | |
2619 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2620 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2621 | else |
2622 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2623 | |
2624 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2625 | if (map) { | |
2626 | int i; | |
2627 | for (i = 0; i < map_disks; i++) | |
2628 | map[i] = 1; | |
2629 | } | |
2630 | ||
cdddbdbc DW |
2631 | } |
2632 | ||
5c4cd5da AC |
2633 | /* allocates memory and fills disk in mdinfo structure |
2634 | * for each disk in array */ | |
2635 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2636 | { | |
2637 | struct mdinfo *mddev = NULL; | |
2638 | struct intel_super *super = st->sb; | |
2639 | struct imsm_disk *disk; | |
2640 | int count = 0; | |
2641 | struct dl *dl; | |
2642 | if (!super || !super->disks) | |
2643 | return NULL; | |
2644 | dl = super->disks; | |
2645 | mddev = malloc(sizeof(*mddev)); | |
2646 | if (!mddev) { | |
2647 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2648 | return NULL; | |
2649 | } | |
2650 | memset(mddev, 0, sizeof(*mddev)); | |
2651 | while (dl) { | |
2652 | struct mdinfo *tmp; | |
2653 | disk = &dl->disk; | |
2654 | tmp = malloc(sizeof(*tmp)); | |
2655 | if (!tmp) { | |
2656 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2657 | if (mddev) | |
2658 | sysfs_free(mddev); | |
2659 | return NULL; | |
2660 | } | |
2661 | memset(tmp, 0, sizeof(*tmp)); | |
2662 | if (mddev->devs) | |
2663 | tmp->next = mddev->devs; | |
2664 | mddev->devs = tmp; | |
2665 | tmp->disk.number = count++; | |
2666 | tmp->disk.major = dl->major; | |
2667 | tmp->disk.minor = dl->minor; | |
2668 | tmp->disk.state = is_configured(disk) ? | |
2669 | (1 << MD_DISK_ACTIVE) : 0; | |
2670 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2671 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2672 | tmp->disk.raid_disk = -1; | |
2673 | dl = dl->next; | |
2674 | } | |
2675 | return mddev; | |
2676 | } | |
2677 | ||
cdddbdbc DW |
2678 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2679 | char *update, char *devname, int verbose, | |
2680 | int uuid_set, char *homehost) | |
2681 | { | |
f352c545 DW |
2682 | /* For 'assemble' and 'force' we need to return non-zero if any |
2683 | * change was made. For others, the return value is ignored. | |
2684 | * Update options are: | |
2685 | * force-one : This device looks a bit old but needs to be included, | |
2686 | * update age info appropriately. | |
2687 | * assemble: clear any 'faulty' flag to allow this device to | |
2688 | * be assembled. | |
2689 | * force-array: Array is degraded but being forced, mark it clean | |
2690 | * if that will be needed to assemble it. | |
2691 | * | |
2692 | * newdev: not used ???? | |
2693 | * grow: Array has gained a new device - this is currently for | |
2694 | * linear only | |
2695 | * resync: mark as dirty so a resync will happen. | |
2696 | * name: update the name - preserving the homehost | |
6e46bf34 | 2697 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2698 | * |
2699 | * Following are not relevant for this imsm: | |
2700 | * sparc2.2 : update from old dodgey metadata | |
2701 | * super-minor: change the preferred_minor number | |
2702 | * summaries: update redundant counters. | |
f352c545 DW |
2703 | * homehost: update the recorded homehost |
2704 | * _reshape_progress: record new reshape_progress position. | |
2705 | */ | |
6e46bf34 DW |
2706 | int rv = 1; |
2707 | struct intel_super *super = st->sb; | |
2708 | struct imsm_super *mpb; | |
f352c545 | 2709 | |
6e46bf34 DW |
2710 | /* we can only update container info */ |
2711 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2712 | return 1; | |
2713 | ||
2714 | mpb = super->anchor; | |
2715 | ||
2716 | if (strcmp(update, "uuid") == 0 && uuid_set && !info->update_private) | |
1e2b2765 | 2717 | rv = -1; |
6e46bf34 DW |
2718 | else if (strcmp(update, "uuid") == 0 && uuid_set && info->update_private) { |
2719 | mpb->orig_family_num = *((__u32 *) info->update_private); | |
2720 | rv = 0; | |
2721 | } else if (strcmp(update, "uuid") == 0) { | |
2722 | __u32 *new_family = malloc(sizeof(*new_family)); | |
2723 | ||
2724 | /* update orig_family_number with the incoming random | |
2725 | * data, report the new effective uuid, and store the | |
2726 | * new orig_family_num for future updates. | |
2727 | */ | |
2728 | if (new_family) { | |
2729 | memcpy(&mpb->orig_family_num, info->uuid, sizeof(__u32)); | |
2730 | uuid_from_super_imsm(st, info->uuid); | |
2731 | *new_family = mpb->orig_family_num; | |
2732 | info->update_private = new_family; | |
2733 | rv = 0; | |
2734 | } | |
2735 | } else if (strcmp(update, "assemble") == 0) | |
2736 | rv = 0; | |
2737 | else | |
1e2b2765 | 2738 | rv = -1; |
f352c545 | 2739 | |
6e46bf34 DW |
2740 | /* successful update? recompute checksum */ |
2741 | if (rv == 0) | |
2742 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2743 | |
2744 | return rv; | |
cdddbdbc DW |
2745 | } |
2746 | ||
c2c087e6 | 2747 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2748 | { |
c2c087e6 | 2749 | size_t size; |
cdddbdbc | 2750 | |
c2c087e6 DW |
2751 | size = sizeof(struct imsm_super); |
2752 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2753 | size += 2 * sizeof(struct imsm_dev); | |
2754 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2755 | size += (4 - 2) * sizeof(struct imsm_map); | |
2756 | /* 4 possible disk_ord_tbl's */ | |
2757 | size += 4 * (disks - 1) * sizeof(__u32); | |
2758 | ||
2759 | return size; | |
2760 | } | |
2761 | ||
2762 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
2763 | { | |
2764 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
2765 | return 0; | |
2766 | ||
2767 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
2768 | } |
2769 | ||
ba2de7ba DW |
2770 | static void free_devlist(struct intel_super *super) |
2771 | { | |
2772 | struct intel_dev *dv; | |
2773 | ||
2774 | while (super->devlist) { | |
2775 | dv = super->devlist->next; | |
2776 | free(super->devlist->dev); | |
2777 | free(super->devlist); | |
2778 | super->devlist = dv; | |
2779 | } | |
2780 | } | |
2781 | ||
2782 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
2783 | { | |
2784 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
2785 | } | |
2786 | ||
cdddbdbc DW |
2787 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
2788 | { | |
2789 | /* | |
2790 | * return: | |
2791 | * 0 same, or first was empty, and second was copied | |
2792 | * 1 second had wrong number | |
2793 | * 2 wrong uuid | |
2794 | * 3 wrong other info | |
2795 | */ | |
2796 | struct intel_super *first = st->sb; | |
2797 | struct intel_super *sec = tst->sb; | |
2798 | ||
2799 | if (!first) { | |
2800 | st->sb = tst->sb; | |
2801 | tst->sb = NULL; | |
2802 | return 0; | |
2803 | } | |
8603ea6f LM |
2804 | /* in platform dependent environment test if the disks |
2805 | * use the same Intel hba | |
2806 | */ | |
2807 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
2808 | if (!first->hba || !sec->hba || |
2809 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
2810 | fprintf(stderr, |
2811 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
2812 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
2813 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
2814 | return 3; |
2815 | } | |
2816 | } | |
cdddbdbc | 2817 | |
d23fe947 DW |
2818 | /* if an anchor does not have num_raid_devs set then it is a free |
2819 | * floating spare | |
2820 | */ | |
2821 | if (first->anchor->num_raid_devs > 0 && | |
2822 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
2823 | /* Determine if these disks might ever have been |
2824 | * related. Further disambiguation can only take place | |
2825 | * in load_super_imsm_all | |
2826 | */ | |
2827 | __u32 first_family = first->anchor->orig_family_num; | |
2828 | __u32 sec_family = sec->anchor->orig_family_num; | |
2829 | ||
f796af5d DW |
2830 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
2831 | MAX_SIGNATURE_LENGTH) != 0) | |
2832 | return 3; | |
2833 | ||
a2b97981 DW |
2834 | if (first_family == 0) |
2835 | first_family = first->anchor->family_num; | |
2836 | if (sec_family == 0) | |
2837 | sec_family = sec->anchor->family_num; | |
2838 | ||
2839 | if (first_family != sec_family) | |
d23fe947 | 2840 | return 3; |
f796af5d | 2841 | |
d23fe947 | 2842 | } |
cdddbdbc | 2843 | |
f796af5d | 2844 | |
3e372e5a DW |
2845 | /* if 'first' is a spare promote it to a populated mpb with sec's |
2846 | * family number | |
2847 | */ | |
2848 | if (first->anchor->num_raid_devs == 0 && | |
2849 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 2850 | int i; |
ba2de7ba DW |
2851 | struct intel_dev *dv; |
2852 | struct imsm_dev *dev; | |
78d30f94 DW |
2853 | |
2854 | /* we need to copy raid device info from sec if an allocation | |
2855 | * fails here we don't associate the spare | |
2856 | */ | |
2857 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
2858 | dv = malloc(sizeof(*dv)); |
2859 | if (!dv) | |
2860 | break; | |
2861 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
2862 | if (!dev) { | |
2863 | free(dv); | |
2864 | break; | |
78d30f94 | 2865 | } |
ba2de7ba DW |
2866 | dv->dev = dev; |
2867 | dv->index = i; | |
2868 | dv->next = first->devlist; | |
2869 | first->devlist = dv; | |
78d30f94 | 2870 | } |
709743c5 | 2871 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
2872 | /* allocation failure */ |
2873 | free_devlist(first); | |
2874 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
2875 | return 3; | |
78d30f94 | 2876 | } |
3e372e5a | 2877 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 2878 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 2879 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 2880 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
2881 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
2882 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
2883 | } |
2884 | ||
cdddbdbc DW |
2885 | return 0; |
2886 | } | |
2887 | ||
0030e8d6 DW |
2888 | static void fd2devname(int fd, char *name) |
2889 | { | |
2890 | struct stat st; | |
2891 | char path[256]; | |
33a6535d | 2892 | char dname[PATH_MAX]; |
0030e8d6 DW |
2893 | char *nm; |
2894 | int rv; | |
2895 | ||
2896 | name[0] = '\0'; | |
2897 | if (fstat(fd, &st) != 0) | |
2898 | return; | |
2899 | sprintf(path, "/sys/dev/block/%d:%d", | |
2900 | major(st.st_rdev), minor(st.st_rdev)); | |
2901 | ||
9cf014ec | 2902 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
2903 | if (rv <= 0) |
2904 | return; | |
2905 | ||
2906 | dname[rv] = '\0'; | |
2907 | nm = strrchr(dname, '/'); | |
7897de29 JS |
2908 | if (nm) { |
2909 | nm++; | |
2910 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
2911 | } | |
0030e8d6 DW |
2912 | } |
2913 | ||
cdddbdbc DW |
2914 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
2915 | ||
2916 | static int imsm_read_serial(int fd, char *devname, | |
2917 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
2918 | { | |
2919 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
2920 | int rv; |
2921 | int rsp_len; | |
1f24f035 | 2922 | int len; |
316e2bf4 DW |
2923 | char *dest; |
2924 | char *src; | |
2925 | char *rsp_buf; | |
2926 | int i; | |
cdddbdbc DW |
2927 | |
2928 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 2929 | |
f9ba0ff1 DW |
2930 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
2931 | ||
40ebbb9c | 2932 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
2933 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
2934 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
2935 | return 0; |
2936 | } | |
2937 | ||
cdddbdbc DW |
2938 | if (rv != 0) { |
2939 | if (devname) | |
2940 | fprintf(stderr, | |
2941 | Name ": Failed to retrieve serial for %s\n", | |
2942 | devname); | |
2943 | return rv; | |
2944 | } | |
2945 | ||
2946 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
2947 | if (!rsp_len) { |
2948 | if (devname) | |
2949 | fprintf(stderr, | |
2950 | Name ": Failed to retrieve serial for %s\n", | |
2951 | devname); | |
2952 | return 2; | |
2953 | } | |
1f24f035 | 2954 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 2955 | |
316e2bf4 DW |
2956 | /* trim all whitespace and non-printable characters and convert |
2957 | * ':' to ';' | |
2958 | */ | |
2959 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
2960 | src = &rsp_buf[i]; | |
2961 | if (*src > 0x20) { | |
2962 | /* ':' is reserved for use in placeholder serial | |
2963 | * numbers for missing disks | |
2964 | */ | |
2965 | if (*src == ':') | |
2966 | *dest++ = ';'; | |
2967 | else | |
2968 | *dest++ = *src; | |
2969 | } | |
2970 | } | |
2971 | len = dest - rsp_buf; | |
2972 | dest = rsp_buf; | |
2973 | ||
2974 | /* truncate leading characters */ | |
2975 | if (len > MAX_RAID_SERIAL_LEN) { | |
2976 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 2977 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 2978 | } |
5c3db629 | 2979 | |
5c3db629 | 2980 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 2981 | memcpy(serial, dest, len); |
cdddbdbc DW |
2982 | |
2983 | return 0; | |
2984 | } | |
2985 | ||
1f24f035 DW |
2986 | static int serialcmp(__u8 *s1, __u8 *s2) |
2987 | { | |
2988 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
2989 | } | |
2990 | ||
2991 | static void serialcpy(__u8 *dest, __u8 *src) | |
2992 | { | |
2993 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
2994 | } | |
2995 | ||
54c2c1ea DW |
2996 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
2997 | { | |
2998 | struct dl *dl; | |
2999 | ||
3000 | for (dl = super->disks; dl; dl = dl->next) | |
3001 | if (serialcmp(dl->serial, serial) == 0) | |
3002 | break; | |
3003 | ||
3004 | return dl; | |
3005 | } | |
3006 | ||
a2b97981 DW |
3007 | static struct imsm_disk * |
3008 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3009 | { | |
3010 | int i; | |
3011 | ||
3012 | for (i = 0; i < mpb->num_disks; i++) { | |
3013 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3014 | ||
3015 | if (serialcmp(disk->serial, serial) == 0) { | |
3016 | if (idx) | |
3017 | *idx = i; | |
3018 | return disk; | |
3019 | } | |
3020 | } | |
3021 | ||
3022 | return NULL; | |
3023 | } | |
3024 | ||
cdddbdbc DW |
3025 | static int |
3026 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3027 | { | |
a2b97981 | 3028 | struct imsm_disk *disk; |
cdddbdbc DW |
3029 | struct dl *dl; |
3030 | struct stat stb; | |
cdddbdbc | 3031 | int rv; |
a2b97981 | 3032 | char name[40]; |
d23fe947 DW |
3033 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3034 | ||
3035 | rv = imsm_read_serial(fd, devname, serial); | |
3036 | ||
3037 | if (rv != 0) | |
3038 | return 2; | |
3039 | ||
a2b97981 | 3040 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 3041 | if (!dl) { |
cdddbdbc DW |
3042 | if (devname) |
3043 | fprintf(stderr, | |
3044 | Name ": failed to allocate disk buffer for %s\n", | |
3045 | devname); | |
3046 | return 2; | |
3047 | } | |
cdddbdbc | 3048 | |
a2b97981 DW |
3049 | fstat(fd, &stb); |
3050 | dl->major = major(stb.st_rdev); | |
3051 | dl->minor = minor(stb.st_rdev); | |
3052 | dl->next = super->disks; | |
3053 | dl->fd = keep_fd ? fd : -1; | |
3054 | assert(super->disks == NULL); | |
3055 | super->disks = dl; | |
3056 | serialcpy(dl->serial, serial); | |
3057 | dl->index = -2; | |
3058 | dl->e = NULL; | |
3059 | fd2devname(fd, name); | |
3060 | if (devname) | |
3061 | dl->devname = strdup(devname); | |
3062 | else | |
3063 | dl->devname = strdup(name); | |
cdddbdbc | 3064 | |
d23fe947 | 3065 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3066 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3067 | if (disk) { | |
3068 | dl->disk = *disk; | |
3069 | /* only set index on disks that are a member of a | |
3070 | * populated contianer, i.e. one with raid_devs | |
3071 | */ | |
3072 | if (is_failed(&dl->disk)) | |
3f6efecc | 3073 | dl->index = -2; |
a2b97981 DW |
3074 | else if (is_spare(&dl->disk)) |
3075 | dl->index = -1; | |
3f6efecc DW |
3076 | } |
3077 | ||
949c47a0 DW |
3078 | return 0; |
3079 | } | |
3080 | ||
0e600426 | 3081 | #ifndef MDASSEMBLE |
0c046afd DW |
3082 | /* When migrating map0 contains the 'destination' state while map1 |
3083 | * contains the current state. When not migrating map0 contains the | |
3084 | * current state. This routine assumes that map[0].map_state is set to | |
3085 | * the current array state before being called. | |
3086 | * | |
3087 | * Migration is indicated by one of the following states | |
3088 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3089 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3090 | * map1state=unitialized) |
1484e727 | 3091 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3092 | * map1state=normal) |
e3bba0e0 | 3093 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3094 | * map1state=degraded) |
8e59f3d8 AK |
3095 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3096 | * map1state=normal) | |
0c046afd | 3097 | */ |
8e59f3d8 AK |
3098 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3099 | __u8 to_state, int migr_type) | |
3393c6af | 3100 | { |
0c046afd | 3101 | struct imsm_map *dest; |
238c0a71 | 3102 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3103 | |
0c046afd | 3104 | dev->vol.migr_state = 1; |
1484e727 | 3105 | set_migr_type(dev, migr_type); |
f8f603f1 | 3106 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3107 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3108 | |
0556e1a2 | 3109 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3110 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3111 | if ((migr_type == MIGR_REBUILD) || |
3112 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3113 | __u32 ord; |
3114 | int i; | |
3115 | ||
3116 | for (i = 0; i < src->num_members; i++) { | |
3117 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3118 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3119 | } | |
3120 | } | |
3121 | ||
8e59f3d8 AK |
3122 | if (migr_type == MIGR_GEN_MIGR) |
3123 | /* Clear migration record */ | |
3124 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3125 | ||
0c046afd | 3126 | src->map_state = to_state; |
949c47a0 | 3127 | } |
f8f603f1 | 3128 | |
809da78e AK |
3129 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3130 | __u8 map_state) | |
f8f603f1 | 3131 | { |
238c0a71 AK |
3132 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3133 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3134 | MAP_0 : MAP_1); | |
28bce06f | 3135 | int i, j; |
0556e1a2 DW |
3136 | |
3137 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3138 | * completed in the last migration. | |
3139 | * | |
28bce06f | 3140 | * FIXME add support for raid-level-migration |
0556e1a2 | 3141 | */ |
809da78e AK |
3142 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3143 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3144 | /* when final map state is other than expected | |
3145 | * merge maps (not for migration) | |
3146 | */ | |
3147 | int failed; | |
3148 | ||
3149 | for (i = 0; i < prev->num_members; i++) | |
3150 | for (j = 0; j < map->num_members; j++) | |
3151 | /* during online capacity expansion | |
3152 | * disks position can be changed | |
3153 | * if takeover is used | |
3154 | */ | |
3155 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3156 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3157 | map->disk_ord_tbl[j] |= | |
3158 | prev->disk_ord_tbl[i]; | |
3159 | break; | |
3160 | } | |
3161 | failed = imsm_count_failed(super, dev, MAP_0); | |
3162 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3163 | } | |
f8f603f1 DW |
3164 | |
3165 | dev->vol.migr_state = 0; | |
ea672ee1 | 3166 | set_migr_type(dev, 0); |
f8f603f1 DW |
3167 | dev->vol.curr_migr_unit = 0; |
3168 | map->map_state = map_state; | |
3169 | } | |
0e600426 | 3170 | #endif |
949c47a0 DW |
3171 | |
3172 | static int parse_raid_devices(struct intel_super *super) | |
3173 | { | |
3174 | int i; | |
3175 | struct imsm_dev *dev_new; | |
4d7b1503 | 3176 | size_t len, len_migr; |
401d313b | 3177 | size_t max_len = 0; |
4d7b1503 DW |
3178 | size_t space_needed = 0; |
3179 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3180 | |
3181 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3182 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3183 | struct intel_dev *dv; |
949c47a0 | 3184 | |
4d7b1503 DW |
3185 | len = sizeof_imsm_dev(dev_iter, 0); |
3186 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3187 | if (len_migr > len) | |
3188 | space_needed += len_migr - len; | |
3189 | ||
ba2de7ba DW |
3190 | dv = malloc(sizeof(*dv)); |
3191 | if (!dv) | |
3192 | return 1; | |
401d313b AK |
3193 | if (max_len < len_migr) |
3194 | max_len = len_migr; | |
3195 | if (max_len > len_migr) | |
3196 | space_needed += max_len - len_migr; | |
3197 | dev_new = malloc(max_len); | |
ba2de7ba DW |
3198 | if (!dev_new) { |
3199 | free(dv); | |
949c47a0 | 3200 | return 1; |
ba2de7ba | 3201 | } |
949c47a0 | 3202 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3203 | dv->dev = dev_new; |
3204 | dv->index = i; | |
3205 | dv->next = super->devlist; | |
3206 | super->devlist = dv; | |
949c47a0 | 3207 | } |
cdddbdbc | 3208 | |
4d7b1503 DW |
3209 | /* ensure that super->buf is large enough when all raid devices |
3210 | * are migrating | |
3211 | */ | |
3212 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3213 | void *buf; | |
3214 | ||
3215 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3216 | if (posix_memalign(&buf, 512, len) != 0) | |
3217 | return 1; | |
3218 | ||
1f45a8ad DW |
3219 | memcpy(buf, super->buf, super->len); |
3220 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3221 | free(super->buf); |
3222 | super->buf = buf; | |
3223 | super->len = len; | |
3224 | } | |
3225 | ||
cdddbdbc DW |
3226 | return 0; |
3227 | } | |
3228 | ||
604b746f JD |
3229 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3230 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3231 | { | |
3232 | void *ptr = NULL; | |
3233 | ||
3234 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3235 | ptr = mpb; | |
3236 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3237 | } | |
3238 | ||
3239 | return ptr; | |
3240 | } | |
3241 | ||
e2f41b2c AK |
3242 | /******************************************************************************* |
3243 | * Function: check_mpb_migr_compatibility | |
3244 | * Description: Function checks for unsupported migration features: | |
3245 | * - migration optimization area (pba_of_lba0) | |
3246 | * - descending reshape (ascending_migr) | |
3247 | * Parameters: | |
3248 | * super : imsm metadata information | |
3249 | * Returns: | |
3250 | * 0 : migration is compatible | |
3251 | * -1 : migration is not compatible | |
3252 | ******************************************************************************/ | |
3253 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3254 | { | |
3255 | struct imsm_map *map0, *map1; | |
3256 | struct migr_record *migr_rec = super->migr_rec; | |
3257 | int i; | |
3258 | ||
3259 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3260 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3261 | ||
3262 | if (dev_iter && | |
3263 | dev_iter->vol.migr_state == 1 && | |
3264 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3265 | /* This device is migrating */ | |
238c0a71 AK |
3266 | map0 = get_imsm_map(dev_iter, MAP_0); |
3267 | map1 = get_imsm_map(dev_iter, MAP_1); | |
e2f41b2c AK |
3268 | if (map0->pba_of_lba0 != map1->pba_of_lba0) |
3269 | /* migration optimization area was used */ | |
3270 | return -1; | |
3271 | if (migr_rec->ascending_migr == 0 | |
3272 | && migr_rec->dest_depth_per_unit > 0) | |
3273 | /* descending reshape not supported yet */ | |
3274 | return -1; | |
3275 | } | |
3276 | } | |
3277 | return 0; | |
3278 | } | |
3279 | ||
d23fe947 | 3280 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3281 | |
cdddbdbc | 3282 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3283 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3284 | */ |
3285 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3286 | { | |
3287 | unsigned long long dsize; | |
cdddbdbc DW |
3288 | unsigned long long sectors; |
3289 | struct stat; | |
6416d527 | 3290 | struct imsm_super *anchor; |
cdddbdbc DW |
3291 | __u32 check_sum; |
3292 | ||
cdddbdbc | 3293 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3294 | if (dsize < 1024) { |
3295 | if (devname) | |
3296 | fprintf(stderr, | |
3297 | Name ": %s: device to small for imsm\n", | |
3298 | devname); | |
3299 | return 1; | |
3300 | } | |
cdddbdbc DW |
3301 | |
3302 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3303 | if (devname) | |
2e062e82 AK |
3304 | fprintf(stderr, Name |
3305 | ": Cannot seek to anchor block on %s: %s\n", | |
cdddbdbc DW |
3306 | devname, strerror(errno)); |
3307 | return 1; | |
3308 | } | |
3309 | ||
949c47a0 | 3310 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
3311 | if (devname) |
3312 | fprintf(stderr, | |
3313 | Name ": Failed to allocate imsm anchor buffer" | |
3314 | " on %s\n", devname); | |
3315 | return 1; | |
3316 | } | |
949c47a0 | 3317 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
3318 | if (devname) |
3319 | fprintf(stderr, | |
3320 | Name ": Cannot read anchor block on %s: %s\n", | |
3321 | devname, strerror(errno)); | |
6416d527 | 3322 | free(anchor); |
cdddbdbc DW |
3323 | return 1; |
3324 | } | |
3325 | ||
6416d527 | 3326 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
3327 | if (devname) |
3328 | fprintf(stderr, | |
3329 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 3330 | free(anchor); |
cdddbdbc DW |
3331 | return 2; |
3332 | } | |
3333 | ||
d23fe947 | 3334 | __free_imsm(super, 0); |
f2f5c343 LM |
3335 | /* reload capability and hba */ |
3336 | ||
3337 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3338 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3339 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3340 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
3341 | if (devname) |
3342 | fprintf(stderr, | |
3343 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 3344 | super->len); |
6416d527 | 3345 | free(anchor); |
cdddbdbc DW |
3346 | return 2; |
3347 | } | |
949c47a0 | 3348 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3349 | |
6416d527 NB |
3350 | sectors = mpb_sectors(anchor) - 1; |
3351 | free(anchor); | |
8e59f3d8 AK |
3352 | |
3353 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { | |
3354 | fprintf(stderr, Name | |
3355 | ": %s could not allocate migr_rec buffer\n", __func__); | |
3356 | free(super->buf); | |
3357 | return 2; | |
3358 | } | |
3359 | ||
949c47a0 | 3360 | if (!sectors) { |
ecf45690 DW |
3361 | check_sum = __gen_imsm_checksum(super->anchor); |
3362 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3363 | if (devname) | |
3364 | fprintf(stderr, | |
3365 | Name ": IMSM checksum %x != %x on %s\n", | |
3366 | check_sum, | |
3367 | __le32_to_cpu(super->anchor->check_sum), | |
3368 | devname); | |
3369 | return 2; | |
3370 | } | |
3371 | ||
a2b97981 | 3372 | return 0; |
949c47a0 | 3373 | } |
cdddbdbc DW |
3374 | |
3375 | /* read the extended mpb */ | |
3376 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3377 | if (devname) | |
3378 | fprintf(stderr, | |
3379 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
3380 | devname, strerror(errno)); | |
3381 | return 1; | |
3382 | } | |
3383 | ||
f21e18ca | 3384 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
3385 | if (devname) |
3386 | fprintf(stderr, | |
3387 | Name ": Cannot read extended mpb on %s: %s\n", | |
3388 | devname, strerror(errno)); | |
3389 | return 2; | |
3390 | } | |
3391 | ||
949c47a0 DW |
3392 | check_sum = __gen_imsm_checksum(super->anchor); |
3393 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
3394 | if (devname) |
3395 | fprintf(stderr, | |
3396 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 3397 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 3398 | devname); |
db575f3b | 3399 | return 3; |
cdddbdbc DW |
3400 | } |
3401 | ||
604b746f JD |
3402 | /* FIXME the BBM log is disk specific so we cannot use this global |
3403 | * buffer for all disks. Ok for now since we only look at the global | |
3404 | * bbm_log_size parameter to gate assembly | |
3405 | */ | |
3406 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3407 | ||
a2b97981 DW |
3408 | return 0; |
3409 | } | |
3410 | ||
8e59f3d8 AK |
3411 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3412 | ||
a2b97981 DW |
3413 | static int |
3414 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3415 | { | |
3416 | int err; | |
3417 | ||
3418 | err = load_imsm_mpb(fd, super, devname); | |
3419 | if (err) | |
3420 | return err; | |
3421 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3422 | if (err) | |
3423 | return err; | |
3424 | err = parse_raid_devices(super); | |
4d7b1503 | 3425 | |
a2b97981 | 3426 | return err; |
cdddbdbc DW |
3427 | } |
3428 | ||
ae6aad82 DW |
3429 | static void __free_imsm_disk(struct dl *d) |
3430 | { | |
3431 | if (d->fd >= 0) | |
3432 | close(d->fd); | |
3433 | if (d->devname) | |
3434 | free(d->devname); | |
0dcecb2e DW |
3435 | if (d->e) |
3436 | free(d->e); | |
ae6aad82 DW |
3437 | free(d); |
3438 | ||
3439 | } | |
1a64be56 | 3440 | |
cdddbdbc DW |
3441 | static void free_imsm_disks(struct intel_super *super) |
3442 | { | |
47ee5a45 | 3443 | struct dl *d; |
cdddbdbc | 3444 | |
47ee5a45 DW |
3445 | while (super->disks) { |
3446 | d = super->disks; | |
cdddbdbc | 3447 | super->disks = d->next; |
ae6aad82 | 3448 | __free_imsm_disk(d); |
cdddbdbc | 3449 | } |
cb82edca AK |
3450 | while (super->disk_mgmt_list) { |
3451 | d = super->disk_mgmt_list; | |
3452 | super->disk_mgmt_list = d->next; | |
3453 | __free_imsm_disk(d); | |
3454 | } | |
47ee5a45 DW |
3455 | while (super->missing) { |
3456 | d = super->missing; | |
3457 | super->missing = d->next; | |
3458 | __free_imsm_disk(d); | |
3459 | } | |
3460 | ||
cdddbdbc DW |
3461 | } |
3462 | ||
9ca2c81c | 3463 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3464 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3465 | { |
88654014 LM |
3466 | struct intel_hba *elem, *next; |
3467 | ||
9ca2c81c | 3468 | if (super->buf) { |
949c47a0 | 3469 | free(super->buf); |
9ca2c81c DW |
3470 | super->buf = NULL; |
3471 | } | |
f2f5c343 LM |
3472 | /* unlink capability description */ |
3473 | super->orom = NULL; | |
8e59f3d8 AK |
3474 | if (super->migr_rec_buf) { |
3475 | free(super->migr_rec_buf); | |
3476 | super->migr_rec_buf = NULL; | |
3477 | } | |
d23fe947 DW |
3478 | if (free_disks) |
3479 | free_imsm_disks(super); | |
ba2de7ba | 3480 | free_devlist(super); |
88654014 LM |
3481 | elem = super->hba; |
3482 | while (elem) { | |
3483 | if (elem->path) | |
3484 | free((void *)elem->path); | |
3485 | next = elem->next; | |
3486 | free(elem); | |
3487 | elem = next; | |
88c32bb1 | 3488 | } |
88654014 | 3489 | super->hba = NULL; |
cdddbdbc DW |
3490 | } |
3491 | ||
9ca2c81c DW |
3492 | static void free_imsm(struct intel_super *super) |
3493 | { | |
d23fe947 | 3494 | __free_imsm(super, 1); |
9ca2c81c DW |
3495 | free(super); |
3496 | } | |
cdddbdbc DW |
3497 | |
3498 | static void free_super_imsm(struct supertype *st) | |
3499 | { | |
3500 | struct intel_super *super = st->sb; | |
3501 | ||
3502 | if (!super) | |
3503 | return; | |
3504 | ||
3505 | free_imsm(super); | |
3506 | st->sb = NULL; | |
3507 | } | |
3508 | ||
49133e57 | 3509 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
3510 | { |
3511 | struct intel_super *super = malloc(sizeof(*super)); | |
3512 | ||
3513 | if (super) { | |
3514 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 3515 | super->current_vol = -1; |
0dcecb2e | 3516 | super->create_offset = ~((__u32 ) 0); |
c2c087e6 | 3517 | } |
c2c087e6 DW |
3518 | return super; |
3519 | } | |
3520 | ||
f0f5a016 LM |
3521 | /* |
3522 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3523 | */ | |
d424212e | 3524 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3525 | { |
3526 | struct sys_dev *hba_name; | |
3527 | int rv = 0; | |
3528 | ||
3529 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3530 | super->orom = NULL; |
f0f5a016 LM |
3531 | super->hba = NULL; |
3532 | return 0; | |
3533 | } | |
3534 | hba_name = find_disk_attached_hba(fd, NULL); | |
3535 | if (!hba_name) { | |
d424212e | 3536 | if (devname) |
f0f5a016 LM |
3537 | fprintf(stderr, |
3538 | Name ": %s is not attached to Intel(R) RAID controller.\n", | |
d424212e | 3539 | devname); |
f0f5a016 LM |
3540 | return 1; |
3541 | } | |
3542 | rv = attach_hba_to_super(super, hba_name); | |
3543 | if (rv == 2) { | |
d424212e N |
3544 | if (devname) { |
3545 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3546 | |
f0f5a016 LM |
3547 | fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID " |
3548 | "controller (%s),\n" | |
3549 | " but the container is assigned to Intel(R) " | |
3550 | "%s RAID controller (", | |
d424212e | 3551 | devname, |
f0f5a016 LM |
3552 | hba_name->path, |
3553 | hba_name->pci_id ? : "Err!", | |
3554 | get_sys_dev_type(hba_name->type)); | |
3555 | ||
f0f5a016 LM |
3556 | while (hba) { |
3557 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3558 | if (hba->next) | |
3559 | fprintf(stderr, ", "); | |
3560 | hba = hba->next; | |
3561 | } | |
3562 | ||
3563 | fprintf(stderr, ").\n" | |
3564 | " Mixing devices attached to different controllers " | |
3565 | "is not allowed.\n"); | |
3566 | } | |
3567 | free_sys_dev(&hba_name); | |
3568 | return 2; | |
3569 | } | |
f2f5c343 | 3570 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3571 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3572 | if (!super->orom) |
3573 | return 3; | |
f0f5a016 LM |
3574 | return 0; |
3575 | } | |
3576 | ||
47ee5a45 DW |
3577 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3578 | * disks that have disappeared from the system. This routine relies on | |
3579 | * the mpb being uptodate, which it is at load time. | |
3580 | */ | |
3581 | static int find_missing(struct intel_super *super) | |
3582 | { | |
3583 | int i; | |
3584 | struct imsm_super *mpb = super->anchor; | |
3585 | struct dl *dl; | |
3586 | struct imsm_disk *disk; | |
47ee5a45 DW |
3587 | |
3588 | for (i = 0; i < mpb->num_disks; i++) { | |
3589 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3590 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3591 | if (dl) |
3592 | continue; | |
47ee5a45 DW |
3593 | |
3594 | dl = malloc(sizeof(*dl)); | |
3595 | if (!dl) | |
3596 | return 1; | |
3597 | dl->major = 0; | |
3598 | dl->minor = 0; | |
3599 | dl->fd = -1; | |
3600 | dl->devname = strdup("missing"); | |
3601 | dl->index = i; | |
3602 | serialcpy(dl->serial, disk->serial); | |
3603 | dl->disk = *disk; | |
689c9bf3 | 3604 | dl->e = NULL; |
47ee5a45 DW |
3605 | dl->next = super->missing; |
3606 | super->missing = dl; | |
3607 | } | |
3608 | ||
3609 | return 0; | |
3610 | } | |
3611 | ||
3960e579 | 3612 | #ifndef MDASSEMBLE |
a2b97981 DW |
3613 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3614 | { | |
3615 | struct intel_disk *idisk = disk_list; | |
3616 | ||
3617 | while (idisk) { | |
3618 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3619 | break; | |
3620 | idisk = idisk->next; | |
3621 | } | |
3622 | ||
3623 | return idisk; | |
3624 | } | |
3625 | ||
3626 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3627 | struct intel_super *super, | |
3628 | struct intel_disk **disk_list) | |
3629 | { | |
3630 | struct imsm_disk *d = &super->disks->disk; | |
3631 | struct imsm_super *mpb = super->anchor; | |
3632 | int i, j; | |
3633 | ||
3634 | for (i = 0; i < tbl_size; i++) { | |
3635 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3636 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3637 | ||
3638 | if (tbl_mpb->family_num == mpb->family_num) { | |
3639 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3640 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3641 | __func__, super->disks->major, | |
3642 | super->disks->minor, | |
3643 | table[i]->disks->major, | |
3644 | table[i]->disks->minor); | |
3645 | break; | |
3646 | } | |
3647 | ||
3648 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3649 | is_configured(d) == is_configured(tbl_d)) && | |
3650 | tbl_mpb->generation_num < mpb->generation_num) { | |
3651 | /* current version of the mpb is a | |
3652 | * better candidate than the one in | |
3653 | * super_table, but copy over "cross | |
3654 | * generational" status | |
3655 | */ | |
3656 | struct intel_disk *idisk; | |
3657 | ||
3658 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3659 | __func__, super->disks->major, | |
3660 | super->disks->minor, | |
3661 | table[i]->disks->major, | |
3662 | table[i]->disks->minor); | |
3663 | ||
3664 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3665 | if (idisk && is_failed(&idisk->disk)) | |
3666 | tbl_d->status |= FAILED_DISK; | |
3667 | break; | |
3668 | } else { | |
3669 | struct intel_disk *idisk; | |
3670 | struct imsm_disk *disk; | |
3671 | ||
3672 | /* tbl_mpb is more up to date, but copy | |
3673 | * over cross generational status before | |
3674 | * returning | |
3675 | */ | |
3676 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3677 | if (disk && is_failed(disk)) | |
3678 | d->status |= FAILED_DISK; | |
3679 | ||
3680 | idisk = disk_list_get(d->serial, *disk_list); | |
3681 | if (idisk) { | |
3682 | idisk->owner = i; | |
3683 | if (disk && is_configured(disk)) | |
3684 | idisk->disk.status |= CONFIGURED_DISK; | |
3685 | } | |
3686 | ||
3687 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3688 | __func__, super->disks->major, | |
3689 | super->disks->minor, | |
3690 | table[i]->disks->major, | |
3691 | table[i]->disks->minor); | |
3692 | ||
3693 | return tbl_size; | |
3694 | } | |
3695 | } | |
3696 | } | |
3697 | ||
3698 | if (i >= tbl_size) | |
3699 | table[tbl_size++] = super; | |
3700 | else | |
3701 | table[i] = super; | |
3702 | ||
3703 | /* update/extend the merged list of imsm_disk records */ | |
3704 | for (j = 0; j < mpb->num_disks; j++) { | |
3705 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3706 | struct intel_disk *idisk; | |
3707 | ||
3708 | idisk = disk_list_get(disk->serial, *disk_list); | |
3709 | if (idisk) { | |
3710 | idisk->disk.status |= disk->status; | |
3711 | if (is_configured(&idisk->disk) || | |
3712 | is_failed(&idisk->disk)) | |
3713 | idisk->disk.status &= ~(SPARE_DISK); | |
3714 | } else { | |
3715 | idisk = calloc(1, sizeof(*idisk)); | |
3716 | if (!idisk) | |
3717 | return -1; | |
3718 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
3719 | idisk->disk = *disk; | |
3720 | idisk->next = *disk_list; | |
3721 | *disk_list = idisk; | |
3722 | } | |
3723 | ||
3724 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3725 | idisk->owner = i; | |
3726 | } | |
3727 | ||
3728 | return tbl_size; | |
3729 | } | |
3730 | ||
3731 | static struct intel_super * | |
3732 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3733 | const int owner) | |
3734 | { | |
3735 | struct imsm_super *mpb = super->anchor; | |
3736 | int ok_count = 0; | |
3737 | int i; | |
3738 | ||
3739 | for (i = 0; i < mpb->num_disks; i++) { | |
3740 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3741 | struct intel_disk *idisk; | |
3742 | ||
3743 | idisk = disk_list_get(disk->serial, disk_list); | |
3744 | if (idisk) { | |
3745 | if (idisk->owner == owner || | |
3746 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3747 | ok_count++; | |
3748 | else | |
3749 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3750 | __func__, disk->serial, idisk->owner, | |
3751 | owner); | |
3752 | } else { | |
3753 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
3754 | __func__, __le32_to_cpu(mpb->family_num), i, | |
3755 | disk->serial); | |
3756 | break; | |
3757 | } | |
3758 | } | |
3759 | ||
3760 | if (ok_count == mpb->num_disks) | |
3761 | return super; | |
3762 | return NULL; | |
3763 | } | |
3764 | ||
3765 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
3766 | { | |
3767 | struct intel_super *s; | |
3768 | ||
3769 | for (s = super_list; s; s = s->next) { | |
3770 | if (family_num != s->anchor->family_num) | |
3771 | continue; | |
3772 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
3773 | __le32_to_cpu(family_num), s->disks->devname); | |
3774 | } | |
3775 | } | |
3776 | ||
3777 | static struct intel_super * | |
3778 | imsm_thunderdome(struct intel_super **super_list, int len) | |
3779 | { | |
3780 | struct intel_super *super_table[len]; | |
3781 | struct intel_disk *disk_list = NULL; | |
3782 | struct intel_super *champion, *spare; | |
3783 | struct intel_super *s, **del; | |
3784 | int tbl_size = 0; | |
3785 | int conflict; | |
3786 | int i; | |
3787 | ||
3788 | memset(super_table, 0, sizeof(super_table)); | |
3789 | for (s = *super_list; s; s = s->next) | |
3790 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
3791 | ||
3792 | for (i = 0; i < tbl_size; i++) { | |
3793 | struct imsm_disk *d; | |
3794 | struct intel_disk *idisk; | |
3795 | struct imsm_super *mpb = super_table[i]->anchor; | |
3796 | ||
3797 | s = super_table[i]; | |
3798 | d = &s->disks->disk; | |
3799 | ||
3800 | /* 'd' must appear in merged disk list for its | |
3801 | * configuration to be valid | |
3802 | */ | |
3803 | idisk = disk_list_get(d->serial, disk_list); | |
3804 | if (idisk && idisk->owner == i) | |
3805 | s = validate_members(s, disk_list, i); | |
3806 | else | |
3807 | s = NULL; | |
3808 | ||
3809 | if (!s) | |
3810 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
3811 | __func__, mpb->family_num, | |
3812 | super_table[i]->disks->major, | |
3813 | super_table[i]->disks->minor); | |
3814 | super_table[i] = s; | |
3815 | } | |
3816 | ||
3817 | /* This is where the mdadm implementation differs from the Windows | |
3818 | * driver which has no strict concept of a container. We can only | |
3819 | * assemble one family from a container, so when returning a prodigal | |
3820 | * array member to this system the code will not be able to disambiguate | |
3821 | * the container contents that should be assembled ("foreign" versus | |
3822 | * "local"). It requires user intervention to set the orig_family_num | |
3823 | * to a new value to establish a new container. The Windows driver in | |
3824 | * this situation fixes up the volume name in place and manages the | |
3825 | * foreign array as an independent entity. | |
3826 | */ | |
3827 | s = NULL; | |
3828 | spare = NULL; | |
3829 | conflict = 0; | |
3830 | for (i = 0; i < tbl_size; i++) { | |
3831 | struct intel_super *tbl_ent = super_table[i]; | |
3832 | int is_spare = 0; | |
3833 | ||
3834 | if (!tbl_ent) | |
3835 | continue; | |
3836 | ||
3837 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
3838 | spare = tbl_ent; | |
3839 | is_spare = 1; | |
3840 | } | |
3841 | ||
3842 | if (s && !is_spare) { | |
3843 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
3844 | conflict++; | |
3845 | } else if (!s && !is_spare) | |
3846 | s = tbl_ent; | |
3847 | } | |
3848 | ||
3849 | if (!s) | |
3850 | s = spare; | |
3851 | if (!s) { | |
3852 | champion = NULL; | |
3853 | goto out; | |
3854 | } | |
3855 | champion = s; | |
3856 | ||
3857 | if (conflict) | |
3858 | fprintf(stderr, "Chose family %#x on '%s', " | |
3859 | "assemble conflicts to new container with '--update=uuid'\n", | |
3860 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
3861 | ||
3862 | /* collect all dl's onto 'champion', and update them to | |
3863 | * champion's version of the status | |
3864 | */ | |
3865 | for (s = *super_list; s; s = s->next) { | |
3866 | struct imsm_super *mpb = champion->anchor; | |
3867 | struct dl *dl = s->disks; | |
3868 | ||
3869 | if (s == champion) | |
3870 | continue; | |
3871 | ||
3872 | for (i = 0; i < mpb->num_disks; i++) { | |
3873 | struct imsm_disk *disk; | |
3874 | ||
3875 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
3876 | if (disk) { | |
3877 | dl->disk = *disk; | |
3878 | /* only set index on disks that are a member of | |
3879 | * a populated contianer, i.e. one with | |
3880 | * raid_devs | |
3881 | */ | |
3882 | if (is_failed(&dl->disk)) | |
3883 | dl->index = -2; | |
3884 | else if (is_spare(&dl->disk)) | |
3885 | dl->index = -1; | |
3886 | break; | |
3887 | } | |
3888 | } | |
3889 | ||
3890 | if (i >= mpb->num_disks) { | |
3891 | struct intel_disk *idisk; | |
3892 | ||
3893 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 3894 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
3895 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
3896 | dl->index = -1; | |
3897 | else { | |
3898 | dl->index = -2; | |
3899 | continue; | |
3900 | } | |
3901 | } | |
3902 | ||
3903 | dl->next = champion->disks; | |
3904 | champion->disks = dl; | |
3905 | s->disks = NULL; | |
3906 | } | |
3907 | ||
3908 | /* delete 'champion' from super_list */ | |
3909 | for (del = super_list; *del; ) { | |
3910 | if (*del == champion) { | |
3911 | *del = (*del)->next; | |
3912 | break; | |
3913 | } else | |
3914 | del = &(*del)->next; | |
3915 | } | |
3916 | champion->next = NULL; | |
3917 | ||
3918 | out: | |
3919 | while (disk_list) { | |
3920 | struct intel_disk *idisk = disk_list; | |
3921 | ||
3922 | disk_list = disk_list->next; | |
3923 | free(idisk); | |
3924 | } | |
3925 | ||
3926 | return champion; | |
3927 | } | |
3928 | ||
cdddbdbc | 3929 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
e1902a7b | 3930 | char *devname) |
cdddbdbc DW |
3931 | { |
3932 | struct mdinfo *sra; | |
a2b97981 DW |
3933 | struct intel_super *super_list = NULL; |
3934 | struct intel_super *super = NULL; | |
db575f3b | 3935 | int devnum = fd2devnum(fd); |
a2b97981 | 3936 | struct mdinfo *sd; |
db575f3b | 3937 | int retry; |
a2b97981 DW |
3938 | int err = 0; |
3939 | int i; | |
dab4a513 DW |
3940 | |
3941 | /* check if 'fd' an opened container */ | |
b526e52d | 3942 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
cdddbdbc DW |
3943 | if (!sra) |
3944 | return 1; | |
3945 | ||
3946 | if (sra->array.major_version != -1 || | |
3947 | sra->array.minor_version != -2 || | |
1602d52c AW |
3948 | strcmp(sra->text_version, "imsm") != 0) { |
3949 | err = 1; | |
3950 | goto error; | |
3951 | } | |
a2b97981 DW |
3952 | /* load all mpbs */ |
3953 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
49133e57 | 3954 | struct intel_super *s = alloc_super(); |
7a6ecd55 | 3955 | char nm[32]; |
a2b97981 | 3956 | int dfd; |
f2f5c343 | 3957 | int rv; |
a2b97981 DW |
3958 | |
3959 | err = 1; | |
3960 | if (!s) | |
3961 | goto error; | |
3962 | s->next = super_list; | |
3963 | super_list = s; | |
cdddbdbc | 3964 | |
a2b97981 | 3965 | err = 2; |
cdddbdbc | 3966 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3967 | dfd = dev_open(nm, O_RDWR); |
a2b97981 DW |
3968 | if (dfd < 0) |
3969 | goto error; | |
3970 | ||
d424212e | 3971 | rv = find_intel_hba_capability(dfd, s, devname); |
f2f5c343 LM |
3972 | /* no orom/efi or non-intel hba of the disk */ |
3973 | if (rv != 0) | |
3974 | goto error; | |
3975 | ||
e1902a7b | 3976 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
db575f3b DW |
3977 | |
3978 | /* retry the load if we might have raced against mdmon */ | |
a2b97981 | 3979 | if (err == 3 && mdmon_running(devnum)) |
db575f3b DW |
3980 | for (retry = 0; retry < 3; retry++) { |
3981 | usleep(3000); | |
e1902a7b | 3982 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
a2b97981 | 3983 | if (err != 3) |
db575f3b DW |
3984 | break; |
3985 | } | |
a2b97981 DW |
3986 | if (err) |
3987 | goto error; | |
cdddbdbc DW |
3988 | } |
3989 | ||
a2b97981 DW |
3990 | /* all mpbs enter, maybe one leaves */ |
3991 | super = imsm_thunderdome(&super_list, i); | |
3992 | if (!super) { | |
3993 | err = 1; | |
3994 | goto error; | |
cdddbdbc DW |
3995 | } |
3996 | ||
47ee5a45 DW |
3997 | if (find_missing(super) != 0) { |
3998 | free_imsm(super); | |
a2b97981 DW |
3999 | err = 2; |
4000 | goto error; | |
47ee5a45 | 4001 | } |
8e59f3d8 AK |
4002 | |
4003 | /* load migration record */ | |
4004 | err = load_imsm_migr_rec(super, NULL); | |
4005 | if (err) { | |
4006 | err = 4; | |
4007 | goto error; | |
4008 | } | |
e2f41b2c AK |
4009 | |
4010 | /* Check migration compatibility */ | |
4011 | if (check_mpb_migr_compatibility(super) != 0) { | |
4012 | fprintf(stderr, Name ": Unsupported migration detected"); | |
4013 | if (devname) | |
4014 | fprintf(stderr, " on %s\n", devname); | |
4015 | else | |
4016 | fprintf(stderr, " (IMSM).\n"); | |
4017 | ||
4018 | err = 5; | |
4019 | goto error; | |
4020 | } | |
4021 | ||
a2b97981 DW |
4022 | err = 0; |
4023 | ||
4024 | error: | |
4025 | while (super_list) { | |
4026 | struct intel_super *s = super_list; | |
4027 | ||
4028 | super_list = super_list->next; | |
4029 | free_imsm(s); | |
4030 | } | |
1602d52c | 4031 | sysfs_free(sra); |
a2b97981 DW |
4032 | |
4033 | if (err) | |
4034 | return err; | |
f7e7067b | 4035 | |
cdddbdbc | 4036 | *sbp = super; |
db575f3b | 4037 | st->container_dev = devnum; |
a2b97981 | 4038 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4039 | st->ss = &super_imsm; |
cdddbdbc DW |
4040 | st->minor_version = 0; |
4041 | st->max_devs = IMSM_MAX_DEVICES; | |
4042 | } | |
cdddbdbc DW |
4043 | return 0; |
4044 | } | |
2b959fbf N |
4045 | |
4046 | static int load_container_imsm(struct supertype *st, int fd, char *devname) | |
4047 | { | |
4048 | return load_super_imsm_all(st, fd, &st->sb, devname); | |
4049 | } | |
cdddbdbc DW |
4050 | #endif |
4051 | ||
4052 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4053 | { | |
4054 | struct intel_super *super; | |
4055 | int rv; | |
4056 | ||
691c6ee1 N |
4057 | if (test_partition(fd)) |
4058 | /* IMSM not allowed on partitions */ | |
4059 | return 1; | |
4060 | ||
37424f13 DW |
4061 | free_super_imsm(st); |
4062 | ||
49133e57 | 4063 | super = alloc_super(); |
cdddbdbc DW |
4064 | if (!super) { |
4065 | fprintf(stderr, | |
4066 | Name ": malloc of %zu failed.\n", | |
4067 | sizeof(*super)); | |
4068 | return 1; | |
4069 | } | |
ea2bc72b LM |
4070 | /* Load hba and capabilities if they exist. |
4071 | * But do not preclude loading metadata in case capabilities or hba are | |
4072 | * non-compliant and ignore_hw_compat is set. | |
4073 | */ | |
d424212e | 4074 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4075 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4076 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 LM |
4077 | if (devname) |
4078 | fprintf(stderr, | |
4079 | Name ": No OROM/EFI properties for %s\n", devname); | |
4080 | free_imsm(super); | |
4081 | return 2; | |
4082 | } | |
a2b97981 | 4083 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
4084 | |
4085 | if (rv) { | |
4086 | if (devname) | |
4087 | fprintf(stderr, | |
4088 | Name ": Failed to load all information " | |
4089 | "sections on %s\n", devname); | |
4090 | free_imsm(super); | |
4091 | return rv; | |
4092 | } | |
4093 | ||
4094 | st->sb = super; | |
4095 | if (st->ss == NULL) { | |
4096 | st->ss = &super_imsm; | |
4097 | st->minor_version = 0; | |
4098 | st->max_devs = IMSM_MAX_DEVICES; | |
4099 | } | |
8e59f3d8 AK |
4100 | |
4101 | /* load migration record */ | |
2e062e82 AK |
4102 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4103 | /* Check for unsupported migration features */ | |
4104 | if (check_mpb_migr_compatibility(super) != 0) { | |
4105 | fprintf(stderr, | |
4106 | Name ": Unsupported migration detected"); | |
4107 | if (devname) | |
4108 | fprintf(stderr, " on %s\n", devname); | |
4109 | else | |
4110 | fprintf(stderr, " (IMSM).\n"); | |
4111 | return 3; | |
4112 | } | |
e2f41b2c AK |
4113 | } |
4114 | ||
cdddbdbc DW |
4115 | return 0; |
4116 | } | |
4117 | ||
ef6ffade DW |
4118 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4119 | { | |
4120 | if (info->level == 1) | |
4121 | return 128; | |
4122 | return info->chunk_size >> 9; | |
4123 | } | |
4124 | ||
ff596308 | 4125 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains) |
ef6ffade DW |
4126 | { |
4127 | __u32 num_stripes; | |
4128 | ||
4129 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
ff596308 | 4130 | num_stripes /= num_domains; |
ef6ffade DW |
4131 | |
4132 | return num_stripes; | |
4133 | } | |
4134 | ||
fcfd9599 DW |
4135 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
4136 | { | |
4025c288 DW |
4137 | if (info->level == 1) |
4138 | return info->size * 2; | |
4139 | else | |
4140 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
fcfd9599 DW |
4141 | } |
4142 | ||
4d1313e9 DW |
4143 | static void imsm_update_version_info(struct intel_super *super) |
4144 | { | |
4145 | /* update the version and attributes */ | |
4146 | struct imsm_super *mpb = super->anchor; | |
4147 | char *version; | |
4148 | struct imsm_dev *dev; | |
4149 | struct imsm_map *map; | |
4150 | int i; | |
4151 | ||
4152 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4153 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4154 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4155 | if (__le32_to_cpu(dev->size_high) > 0) |
4156 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4157 | ||
4158 | /* FIXME detect when an array spans a port multiplier */ | |
4159 | #if 0 | |
4160 | mpb->attributes |= MPB_ATTRIB_PM; | |
4161 | #endif | |
4162 | ||
4163 | if (mpb->num_raid_devs > 1 || | |
4164 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4165 | version = MPB_VERSION_ATTRIBS; | |
4166 | switch (get_imsm_raid_level(map)) { | |
4167 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4168 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4169 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4170 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4171 | } | |
4172 | } else { | |
4173 | if (map->num_members >= 5) | |
4174 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4175 | else if (dev->status == DEV_CLONE_N_GO) | |
4176 | version = MPB_VERSION_CNG; | |
4177 | else if (get_imsm_raid_level(map) == 5) | |
4178 | version = MPB_VERSION_RAID5; | |
4179 | else if (map->num_members >= 3) | |
4180 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4181 | else if (get_imsm_raid_level(map) == 1) | |
4182 | version = MPB_VERSION_RAID1; | |
4183 | else | |
4184 | version = MPB_VERSION_RAID0; | |
4185 | } | |
4186 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4187 | } | |
4188 | } | |
4189 | ||
aa534678 DW |
4190 | static int check_name(struct intel_super *super, char *name, int quiet) |
4191 | { | |
4192 | struct imsm_super *mpb = super->anchor; | |
4193 | char *reason = NULL; | |
4194 | int i; | |
4195 | ||
4196 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4197 | reason = "must be 16 characters or less"; | |
4198 | ||
4199 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4200 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4201 | ||
4202 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4203 | reason = "already exists"; | |
4204 | break; | |
4205 | } | |
4206 | } | |
4207 | ||
4208 | if (reason && !quiet) | |
4209 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
4210 | ||
4211 | return !reason; | |
4212 | } | |
4213 | ||
8b353278 DW |
4214 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4215 | unsigned long long size, char *name, | |
4216 | char *homehost, int *uuid) | |
cdddbdbc | 4217 | { |
c2c087e6 DW |
4218 | /* We are creating a volume inside a pre-existing container. |
4219 | * so st->sb is already set. | |
4220 | */ | |
4221 | struct intel_super *super = st->sb; | |
949c47a0 | 4222 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4223 | struct intel_dev *dv; |
c2c087e6 DW |
4224 | struct imsm_dev *dev; |
4225 | struct imsm_vol *vol; | |
4226 | struct imsm_map *map; | |
4227 | int idx = mpb->num_raid_devs; | |
4228 | int i; | |
4229 | unsigned long long array_blocks; | |
2c092cad | 4230 | size_t size_old, size_new; |
ff596308 | 4231 | __u32 num_data_stripes; |
cdddbdbc | 4232 | |
88c32bb1 | 4233 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 4234 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 4235 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4236 | return 0; |
4237 | } | |
4238 | ||
2c092cad DW |
4239 | /* ensure the mpb is large enough for the new data */ |
4240 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4241 | size_new = disks_to_mpb_size(info->nr_disks); | |
4242 | if (size_new > size_old) { | |
4243 | void *mpb_new; | |
4244 | size_t size_round = ROUND_UP(size_new, 512); | |
4245 | ||
4246 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
4247 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
4248 | return 0; | |
4249 | } | |
8e59f3d8 AK |
4250 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
4251 | fprintf(stderr, Name | |
4252 | ": %s could not allocate migr_rec buffer\n", | |
4253 | __func__); | |
4254 | free(super->buf); | |
4255 | free(super); | |
ea944c8f | 4256 | free(mpb_new); |
8e59f3d8 AK |
4257 | return 0; |
4258 | } | |
2c092cad DW |
4259 | memcpy(mpb_new, mpb, size_old); |
4260 | free(mpb); | |
4261 | mpb = mpb_new; | |
949c47a0 | 4262 | super->anchor = mpb_new; |
2c092cad DW |
4263 | mpb->mpb_size = __cpu_to_le32(size_new); |
4264 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4265 | } | |
bf5a934a | 4266 | super->current_vol = idx; |
3960e579 DW |
4267 | |
4268 | /* handle 'failed_disks' by either: | |
4269 | * a) create dummy disk entries in the table if this the first | |
4270 | * volume in the array. We add them here as this is the only | |
4271 | * opportunity to add them. add_to_super_imsm_volume() | |
4272 | * handles the non-failed disks and continues incrementing | |
4273 | * mpb->num_disks. | |
4274 | * b) validate that 'failed_disks' matches the current number | |
4275 | * of missing disks if the container is populated | |
d23fe947 | 4276 | */ |
3960e579 | 4277 | if (super->current_vol == 0) { |
d23fe947 | 4278 | mpb->num_disks = 0; |
3960e579 DW |
4279 | for (i = 0; i < info->failed_disks; i++) { |
4280 | struct imsm_disk *disk; | |
4281 | ||
4282 | mpb->num_disks++; | |
4283 | disk = __get_imsm_disk(mpb, i); | |
4284 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4285 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4286 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4287 | "missing:%d", i); | |
4288 | } | |
4289 | find_missing(super); | |
4290 | } else { | |
4291 | int missing = 0; | |
4292 | struct dl *d; | |
4293 | ||
4294 | for (d = super->missing; d; d = d->next) | |
4295 | missing++; | |
4296 | if (info->failed_disks > missing) { | |
4297 | fprintf(stderr, Name": unable to add 'missing' disk to container\n"); | |
4298 | return 0; | |
4299 | } | |
4300 | } | |
5a038140 | 4301 | |
aa534678 DW |
4302 | if (!check_name(super, name, 0)) |
4303 | return 0; | |
ba2de7ba DW |
4304 | dv = malloc(sizeof(*dv)); |
4305 | if (!dv) { | |
4306 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
4307 | return 0; | |
4308 | } | |
1a2487c2 | 4309 | dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
949c47a0 | 4310 | if (!dev) { |
ba2de7ba | 4311 | free(dv); |
949c47a0 DW |
4312 | fprintf(stderr, Name": could not allocate raid device\n"); |
4313 | return 0; | |
4314 | } | |
1a2487c2 | 4315 | |
c2c087e6 | 4316 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
03bcbc65 DW |
4317 | if (info->level == 1) |
4318 | array_blocks = info_to_blocks_per_member(info); | |
4319 | else | |
4320 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
4321 | info->layout, info->chunk_size, | |
4322 | info->size*2); | |
979d38be DW |
4323 | /* round array size down to closest MB */ |
4324 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4325 | ||
c2c087e6 DW |
4326 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4327 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4328 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4329 | vol = &dev->vol; |
4330 | vol->migr_state = 0; | |
1484e727 | 4331 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4332 | vol->dirty = !info->state; |
f8f603f1 | 4333 | vol->curr_migr_unit = 0; |
238c0a71 | 4334 | map = get_imsm_map(dev, MAP_0); |
0dcecb2e | 4335 | map->pba_of_lba0 = __cpu_to_le32(super->create_offset); |
fcfd9599 | 4336 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade | 4337 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4338 | map->failed_disk_num = ~0; |
bf4442ab AK |
4339 | if (info->level > 0) |
4340 | map->map_state = IMSM_T_STATE_UNINITIALIZED; | |
4341 | else | |
4342 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4343 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4344 | map->ddf = 1; |
ef6ffade DW |
4345 | |
4346 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4347 | free(dev); |
4348 | free(dv); | |
ef6ffade DW |
4349 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
4350 | "in a raid1 volume\n"); | |
4351 | return 0; | |
4352 | } | |
81062a36 DW |
4353 | |
4354 | map->raid_level = info->level; | |
4d1313e9 | 4355 | if (info->level == 10) { |
c2c087e6 | 4356 | map->raid_level = 1; |
4d1313e9 | 4357 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4358 | } else if (info->level == 1) |
4359 | map->num_domains = info->raid_disks; | |
4360 | else | |
ff596308 | 4361 | map->num_domains = 1; |
81062a36 | 4362 | |
ff596308 DW |
4363 | num_data_stripes = info_to_num_data_stripes(info, map->num_domains); |
4364 | map->num_data_stripes = __cpu_to_le32(num_data_stripes); | |
ef6ffade | 4365 | |
c2c087e6 DW |
4366 | map->num_members = info->raid_disks; |
4367 | for (i = 0; i < map->num_members; i++) { | |
4368 | /* initialized in add_to_super */ | |
4eb26970 | 4369 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4370 | } |
949c47a0 | 4371 | mpb->num_raid_devs++; |
ba2de7ba DW |
4372 | |
4373 | dv->dev = dev; | |
4374 | dv->index = super->current_vol; | |
4375 | dv->next = super->devlist; | |
4376 | super->devlist = dv; | |
c2c087e6 | 4377 | |
4d1313e9 DW |
4378 | imsm_update_version_info(super); |
4379 | ||
c2c087e6 | 4380 | return 1; |
cdddbdbc DW |
4381 | } |
4382 | ||
bf5a934a DW |
4383 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4384 | unsigned long long size, char *name, | |
4385 | char *homehost, int *uuid) | |
4386 | { | |
4387 | /* This is primarily called by Create when creating a new array. | |
4388 | * We will then get add_to_super called for each component, and then | |
4389 | * write_init_super called to write it out to each device. | |
4390 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4391 | * array. | |
4392 | * To create on a pre-existing array a different method will be called. | |
4393 | * This one is just for fresh drives. | |
4394 | */ | |
4395 | struct intel_super *super; | |
4396 | struct imsm_super *mpb; | |
4397 | size_t mpb_size; | |
4d1313e9 | 4398 | char *version; |
bf5a934a | 4399 | |
bf5a934a | 4400 | if (st->sb) |
e683ca88 DW |
4401 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
4402 | ||
4403 | if (info) | |
4404 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4405 | else | |
4406 | mpb_size = 512; | |
bf5a934a | 4407 | |
49133e57 | 4408 | super = alloc_super(); |
e683ca88 | 4409 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4410 | free(super); |
e683ca88 DW |
4411 | super = NULL; |
4412 | } | |
4413 | if (!super) { | |
4414 | fprintf(stderr, Name | |
4415 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
4416 | return 0; |
4417 | } | |
8e59f3d8 AK |
4418 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
4419 | fprintf(stderr, Name | |
4420 | ": %s could not allocate migr_rec buffer\n", __func__); | |
4421 | free(super->buf); | |
4422 | free(super); | |
4423 | return 0; | |
4424 | } | |
e683ca88 | 4425 | memset(super->buf, 0, mpb_size); |
ef649044 | 4426 | mpb = super->buf; |
e683ca88 DW |
4427 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4428 | st->sb = super; | |
4429 | ||
4430 | if (info == NULL) { | |
4431 | /* zeroing superblock */ | |
4432 | return 0; | |
4433 | } | |
bf5a934a | 4434 | |
4d1313e9 DW |
4435 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4436 | ||
4437 | version = (char *) mpb->sig; | |
4438 | strcpy(version, MPB_SIGNATURE); | |
4439 | version += strlen(MPB_SIGNATURE); | |
4440 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4441 | |
bf5a934a DW |
4442 | return 1; |
4443 | } | |
4444 | ||
0e600426 | 4445 | #ifndef MDASSEMBLE |
f20c3968 | 4446 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4447 | int fd, char *devname) |
4448 | { | |
4449 | struct intel_super *super = st->sb; | |
d23fe947 | 4450 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4451 | struct imsm_disk *_disk; |
bf5a934a DW |
4452 | struct imsm_dev *dev; |
4453 | struct imsm_map *map; | |
3960e579 | 4454 | struct dl *dl, *df; |
4eb26970 | 4455 | int slot; |
bf5a934a | 4456 | |
949c47a0 | 4457 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4458 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4459 | |
208933a7 N |
4460 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
4461 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
4462 | devname); | |
4463 | return 1; | |
4464 | } | |
4465 | ||
efb30e7f DW |
4466 | if (fd == -1) { |
4467 | /* we're doing autolayout so grab the pre-marked (in | |
4468 | * validate_geometry) raid_disk | |
4469 | */ | |
4470 | for (dl = super->disks; dl; dl = dl->next) | |
4471 | if (dl->raiddisk == dk->raid_disk) | |
4472 | break; | |
4473 | } else { | |
4474 | for (dl = super->disks; dl ; dl = dl->next) | |
4475 | if (dl->major == dk->major && | |
4476 | dl->minor == dk->minor) | |
4477 | break; | |
4478 | } | |
d23fe947 | 4479 | |
208933a7 N |
4480 | if (!dl) { |
4481 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 4482 | return 1; |
208933a7 | 4483 | } |
bf5a934a | 4484 | |
d23fe947 DW |
4485 | /* add a pristine spare to the metadata */ |
4486 | if (dl->index < 0) { | |
4487 | dl->index = super->anchor->num_disks; | |
4488 | super->anchor->num_disks++; | |
4489 | } | |
4eb26970 DW |
4490 | /* Check the device has not already been added */ |
4491 | slot = get_imsm_disk_slot(map, dl->index); | |
4492 | if (slot >= 0 && | |
238c0a71 | 4493 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
4494 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
4495 | devname); | |
4496 | return 1; | |
4497 | } | |
656b6b5a | 4498 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4499 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4500 | |
3960e579 DW |
4501 | /* update size of 'missing' disks to be at least as large as the |
4502 | * largest acitve member (we only have dummy missing disks when | |
4503 | * creating the first volume) | |
4504 | */ | |
4505 | if (super->current_vol == 0) { | |
4506 | for (df = super->missing; df; df = df->next) { | |
4507 | if (dl->disk.total_blocks > df->disk.total_blocks) | |
4508 | df->disk.total_blocks = dl->disk.total_blocks; | |
4509 | _disk = __get_imsm_disk(mpb, df->index); | |
4510 | *_disk = df->disk; | |
4511 | } | |
4512 | } | |
4513 | ||
4514 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
4515 | for (df = super->missing; df; df = df->next) | |
4516 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 4517 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
4518 | |
4519 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
4520 | continue; | |
4521 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 4522 | if (is_gen_migration(dev)) { |
238c0a71 AK |
4523 | struct imsm_map *map2 = get_imsm_map(dev, |
4524 | MAP_1); | |
0a108d63 AK |
4525 | int slot2 = get_imsm_disk_slot(map2, df->index); |
4526 | if ((slot2 < map2->num_members) && | |
4527 | (slot2 >= 0)) { | |
1ace8403 | 4528 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
4529 | slot2, |
4530 | MAP_1); | |
1ace8403 AK |
4531 | if ((unsigned)df->index == |
4532 | ord_to_idx(ord2)) | |
4533 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 4534 | slot2, |
1ace8403 AK |
4535 | df->index | |
4536 | IMSM_ORD_REBUILD); | |
4537 | } | |
4538 | } | |
3960e579 DW |
4539 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
4540 | break; | |
4541 | } | |
4542 | ||
d23fe947 DW |
4543 | /* if we are creating the first raid device update the family number */ |
4544 | if (super->current_vol == 0) { | |
4545 | __u32 sum; | |
4546 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 4547 | |
3960e579 | 4548 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a AW |
4549 | if (!_dev || !_disk) { |
4550 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
4551 | return 1; | |
4552 | } | |
d23fe947 DW |
4553 | *_dev = *dev; |
4554 | *_disk = dl->disk; | |
148acb7b DW |
4555 | sum = random32(); |
4556 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4557 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4558 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4559 | } |
ca0748fa | 4560 | super->current_disk = dl; |
f20c3968 | 4561 | return 0; |
bf5a934a DW |
4562 | } |
4563 | ||
a8619d23 AK |
4564 | /* mark_spare() |
4565 | * Function marks disk as spare and restores disk serial | |
4566 | * in case it was previously marked as failed by takeover operation | |
4567 | * reruns: | |
4568 | * -1 : critical error | |
4569 | * 0 : disk is marked as spare but serial is not set | |
4570 | * 1 : success | |
4571 | */ | |
4572 | int mark_spare(struct dl *disk) | |
4573 | { | |
4574 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
4575 | int ret_val = -1; | |
4576 | ||
4577 | if (!disk) | |
4578 | return ret_val; | |
4579 | ||
4580 | ret_val = 0; | |
4581 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
4582 | /* Restore disk serial number, because takeover marks disk | |
4583 | * as failed and adds to serial ':0' before it becomes | |
4584 | * a spare disk. | |
4585 | */ | |
4586 | serialcpy(disk->serial, serial); | |
4587 | serialcpy(disk->disk.serial, serial); | |
4588 | ret_val = 1; | |
4589 | } | |
4590 | disk->disk.status = SPARE_DISK; | |
4591 | disk->index = -1; | |
4592 | ||
4593 | return ret_val; | |
4594 | } | |
88654014 | 4595 | |
f20c3968 | 4596 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4597 | int fd, char *devname) |
cdddbdbc | 4598 | { |
c2c087e6 | 4599 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4600 | struct dl *dd; |
4601 | unsigned long long size; | |
f2f27e63 | 4602 | __u32 id; |
c2c087e6 DW |
4603 | int rv; |
4604 | struct stat stb; | |
4605 | ||
88654014 LM |
4606 | /* If we are on an RAID enabled platform check that the disk is |
4607 | * attached to the raid controller. | |
4608 | * We do not need to test disks attachment for container based additions, | |
4609 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4610 | */ |
d424212e | 4611 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4612 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4613 | if (rv != 0) { |
4614 | dprintf("capability: %p fd: %d ret: %d\n", | |
4615 | super->orom, fd, rv); | |
4616 | return 1; | |
88c32bb1 DW |
4617 | } |
4618 | ||
f20c3968 DW |
4619 | if (super->current_vol >= 0) |
4620 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4621 | |
c2c087e6 DW |
4622 | fstat(fd, &stb); |
4623 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 4624 | if (!dd) { |
c2c087e6 DW |
4625 | fprintf(stderr, |
4626 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 4627 | return 1; |
c2c087e6 DW |
4628 | } |
4629 | memset(dd, 0, sizeof(*dd)); | |
4630 | dd->major = major(stb.st_rdev); | |
4631 | dd->minor = minor(stb.st_rdev); | |
c2c087e6 | 4632 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 4633 | dd->fd = fd; |
689c9bf3 | 4634 | dd->e = NULL; |
1a64be56 | 4635 | dd->action = DISK_ADD; |
c2c087e6 | 4636 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4637 | if (rv) { |
c2c087e6 | 4638 | fprintf(stderr, |
0030e8d6 | 4639 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4640 | free(dd); |
0030e8d6 | 4641 | abort(); |
c2c087e6 DW |
4642 | } |
4643 | ||
c2c087e6 DW |
4644 | get_dev_size(fd, NULL, &size); |
4645 | size /= 512; | |
1f24f035 | 4646 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 4647 | dd->disk.total_blocks = __cpu_to_le32(size); |
a8619d23 | 4648 | mark_spare(dd); |
c2c087e6 | 4649 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4650 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4651 | else |
b9f594fe | 4652 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4653 | |
4654 | if (st->update_tail) { | |
1a64be56 LM |
4655 | dd->next = super->disk_mgmt_list; |
4656 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4657 | } else { |
4658 | dd->next = super->disks; | |
4659 | super->disks = dd; | |
ceaf0ee1 | 4660 | super->updates_pending++; |
43dad3d6 | 4661 | } |
f20c3968 DW |
4662 | |
4663 | return 0; | |
cdddbdbc DW |
4664 | } |
4665 | ||
1a64be56 LM |
4666 | |
4667 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4668 | { | |
4669 | struct intel_super *super = st->sb; | |
4670 | struct dl *dd; | |
4671 | ||
4672 | /* remove from super works only in mdmon - for communication | |
4673 | * manager - monitor. Check if communication memory buffer | |
4674 | * is prepared. | |
4675 | */ | |
4676 | if (!st->update_tail) { | |
4677 | fprintf(stderr, | |
4678 | Name ": %s shall be used in mdmon context only" | |
4679 | "(line %d).\n", __func__, __LINE__); | |
4680 | return 1; | |
4681 | } | |
4682 | dd = malloc(sizeof(*dd)); | |
4683 | if (!dd) { | |
4684 | fprintf(stderr, | |
4685 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
4686 | return 1; | |
4687 | } | |
4688 | memset(dd, 0, sizeof(*dd)); | |
4689 | dd->major = dk->major; | |
4690 | dd->minor = dk->minor; | |
1a64be56 | 4691 | dd->fd = -1; |
a8619d23 | 4692 | mark_spare(dd); |
1a64be56 LM |
4693 | dd->action = DISK_REMOVE; |
4694 | ||
4695 | dd->next = super->disk_mgmt_list; | |
4696 | super->disk_mgmt_list = dd; | |
4697 | ||
4698 | ||
4699 | return 0; | |
4700 | } | |
4701 | ||
f796af5d DW |
4702 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
4703 | ||
4704 | static union { | |
4705 | char buf[512]; | |
4706 | struct imsm_super anchor; | |
4707 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 4708 | |
d23fe947 DW |
4709 | /* spare records have their own family number and do not have any defined raid |
4710 | * devices | |
4711 | */ | |
4712 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
4713 | { | |
d23fe947 | 4714 | struct imsm_super *mpb = super->anchor; |
f796af5d | 4715 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
4716 | __u32 sum; |
4717 | struct dl *d; | |
4718 | ||
f796af5d DW |
4719 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
4720 | spare->generation_num = __cpu_to_le32(1UL), | |
4721 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
4722 | spare->num_disks = 1, | |
4723 | spare->num_raid_devs = 0, | |
4724 | spare->cache_size = mpb->cache_size, | |
4725 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
4726 | ||
4727 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
4728 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
4729 | |
4730 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 4731 | if (d->index != -1) |
d23fe947 DW |
4732 | continue; |
4733 | ||
f796af5d DW |
4734 | spare->disk[0] = d->disk; |
4735 | sum = __gen_imsm_checksum(spare); | |
4736 | spare->family_num = __cpu_to_le32(sum); | |
4737 | spare->orig_family_num = 0; | |
4738 | sum = __gen_imsm_checksum(spare); | |
4739 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 4740 | |
f796af5d | 4741 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
4742 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4743 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 4744 | return 1; |
d23fe947 DW |
4745 | } |
4746 | if (doclose) { | |
4747 | close(d->fd); | |
4748 | d->fd = -1; | |
4749 | } | |
4750 | } | |
4751 | ||
e74255d9 | 4752 | return 0; |
d23fe947 DW |
4753 | } |
4754 | ||
36988a3d | 4755 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 4756 | { |
36988a3d | 4757 | struct intel_super *super = st->sb; |
949c47a0 | 4758 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4759 | struct dl *d; |
4760 | __u32 generation; | |
4761 | __u32 sum; | |
d23fe947 | 4762 | int spares = 0; |
949c47a0 | 4763 | int i; |
a48ac0a8 | 4764 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 4765 | int num_disks = 0; |
146c6260 | 4766 | int clear_migration_record = 1; |
cdddbdbc | 4767 | |
c2c087e6 DW |
4768 | /* 'generation' is incremented everytime the metadata is written */ |
4769 | generation = __le32_to_cpu(mpb->generation_num); | |
4770 | generation++; | |
4771 | mpb->generation_num = __cpu_to_le32(generation); | |
4772 | ||
148acb7b DW |
4773 | /* fix up cases where previous mdadm releases failed to set |
4774 | * orig_family_num | |
4775 | */ | |
4776 | if (mpb->orig_family_num == 0) | |
4777 | mpb->orig_family_num = mpb->family_num; | |
4778 | ||
d23fe947 | 4779 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 4780 | if (d->index == -1) |
d23fe947 | 4781 | spares++; |
36988a3d | 4782 | else { |
d23fe947 | 4783 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4784 | num_disks++; |
4785 | } | |
d23fe947 | 4786 | } |
36988a3d | 4787 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 4788 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4789 | num_disks++; |
4790 | } | |
4791 | mpb->num_disks = num_disks; | |
4792 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 4793 | |
949c47a0 DW |
4794 | for (i = 0; i < mpb->num_raid_devs; i++) { |
4795 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
4796 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
4797 | if (dev && dev2) { | |
4798 | imsm_copy_dev(dev, dev2); | |
4799 | mpb_size += sizeof_imsm_dev(dev, 0); | |
4800 | } | |
146c6260 AK |
4801 | if (is_gen_migration(dev2)) |
4802 | clear_migration_record = 0; | |
949c47a0 | 4803 | } |
a48ac0a8 DW |
4804 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
4805 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 4806 | |
c2c087e6 | 4807 | /* recalculate checksum */ |
949c47a0 | 4808 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
4809 | mpb->check_sum = __cpu_to_le32(sum); |
4810 | ||
146c6260 AK |
4811 | if (clear_migration_record) |
4812 | memset(super->migr_rec_buf, 0, 512); | |
4813 | ||
d23fe947 | 4814 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 4815 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 4816 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 4817 | continue; |
f796af5d | 4818 | if (store_imsm_mpb(d->fd, mpb)) |
c2c087e6 DW |
4819 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4820 | __func__, d->major, d->minor, strerror(errno)); | |
146c6260 AK |
4821 | if (clear_migration_record) { |
4822 | unsigned long long dsize; | |
4823 | ||
4824 | get_dev_size(d->fd, NULL, &dsize); | |
4825 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
9e2d750d N |
4826 | if (write(d->fd, super->migr_rec_buf, 512) != 512) |
4827 | perror("Write migr_rec failed"); | |
146c6260 AK |
4828 | } |
4829 | } | |
c2c087e6 DW |
4830 | if (doclose) { |
4831 | close(d->fd); | |
4832 | d->fd = -1; | |
4833 | } | |
4834 | } | |
4835 | ||
d23fe947 DW |
4836 | if (spares) |
4837 | return write_super_imsm_spares(super, doclose); | |
4838 | ||
e74255d9 | 4839 | return 0; |
c2c087e6 DW |
4840 | } |
4841 | ||
0e600426 | 4842 | |
9b1fb677 | 4843 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
4844 | { |
4845 | size_t len; | |
4846 | struct imsm_update_create_array *u; | |
4847 | struct intel_super *super = st->sb; | |
9b1fb677 | 4848 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 4849 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
4850 | struct disk_info *inf; |
4851 | struct imsm_disk *disk; | |
4852 | int i; | |
43dad3d6 | 4853 | |
54c2c1ea DW |
4854 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
4855 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
4856 | u = malloc(len); |
4857 | if (!u) { | |
4858 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4859 | __func__); | |
4860 | return 1; | |
4861 | } | |
4862 | ||
4863 | u->type = update_create_array; | |
9b1fb677 | 4864 | u->dev_idx = dev_idx; |
43dad3d6 | 4865 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
4866 | inf = get_disk_info(u); |
4867 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 4868 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 4869 | |
54c2c1ea DW |
4870 | disk = get_imsm_disk(super, idx); |
4871 | serialcpy(inf[i].serial, disk->serial); | |
4872 | } | |
43dad3d6 DW |
4873 | append_metadata_update(st, u, len); |
4874 | ||
4875 | return 0; | |
4876 | } | |
4877 | ||
1a64be56 | 4878 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
4879 | { |
4880 | struct intel_super *super = st->sb; | |
4881 | size_t len; | |
1a64be56 | 4882 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 4883 | |
1a64be56 | 4884 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
4885 | return 0; |
4886 | ||
4887 | len = sizeof(*u); | |
4888 | u = malloc(len); | |
4889 | if (!u) { | |
4890 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4891 | __func__); | |
4892 | return 1; | |
4893 | } | |
4894 | ||
1a64be56 | 4895 | u->type = update_add_remove_disk; |
43dad3d6 DW |
4896 | append_metadata_update(st, u, len); |
4897 | ||
4898 | return 0; | |
4899 | } | |
4900 | ||
c2c087e6 DW |
4901 | static int write_init_super_imsm(struct supertype *st) |
4902 | { | |
9b1fb677 DW |
4903 | struct intel_super *super = st->sb; |
4904 | int current_vol = super->current_vol; | |
4905 | ||
4906 | /* we are done with current_vol reset it to point st at the container */ | |
4907 | super->current_vol = -1; | |
4908 | ||
8273f55e | 4909 | if (st->update_tail) { |
43dad3d6 DW |
4910 | /* queue the recently created array / added disk |
4911 | * as a metadata update */ | |
43dad3d6 | 4912 | int rv; |
8273f55e | 4913 | |
43dad3d6 | 4914 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 4915 | if (current_vol < 0) { |
1a64be56 LM |
4916 | /* in the mgmt (add/remove) disk case we are running |
4917 | * in mdmon context, so don't close fd's | |
43dad3d6 | 4918 | */ |
1a64be56 | 4919 | return mgmt_disk(st); |
43dad3d6 | 4920 | } else |
9b1fb677 | 4921 | rv = create_array(st, current_vol); |
8273f55e | 4922 | |
43dad3d6 | 4923 | return rv; |
d682f344 N |
4924 | } else { |
4925 | struct dl *d; | |
4926 | for (d = super->disks; d; d = d->next) | |
4927 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 4928 | return write_super_imsm(st, 1); |
d682f344 | 4929 | } |
cdddbdbc | 4930 | } |
0e600426 | 4931 | #endif |
cdddbdbc | 4932 | |
e683ca88 | 4933 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 4934 | { |
e683ca88 DW |
4935 | struct intel_super *super = st->sb; |
4936 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 4937 | |
e683ca88 | 4938 | if (!mpb) |
ad97895e DW |
4939 | return 1; |
4940 | ||
1799c9e8 | 4941 | #ifndef MDASSEMBLE |
e683ca88 | 4942 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
4943 | #else |
4944 | return 1; | |
4945 | #endif | |
cdddbdbc DW |
4946 | } |
4947 | ||
0e600426 N |
4948 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
4949 | { | |
4950 | return __le32_to_cpu(mpb->bbm_log_size); | |
4951 | } | |
4952 | ||
4953 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
4954 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
4955 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 4956 | unsigned long long size, char *dev, |
2c514b71 NB |
4957 | unsigned long long *freesize, |
4958 | int verbose) | |
cdddbdbc | 4959 | { |
c2c087e6 DW |
4960 | int fd; |
4961 | unsigned long long ldsize; | |
f2f5c343 LM |
4962 | struct intel_super *super=NULL; |
4963 | int rv = 0; | |
cdddbdbc | 4964 | |
c2c087e6 DW |
4965 | if (level != LEVEL_CONTAINER) |
4966 | return 0; | |
4967 | if (!dev) | |
4968 | return 1; | |
4969 | ||
4970 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4971 | if (fd < 0) { | |
2c514b71 NB |
4972 | if (verbose) |
4973 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
4974 | dev, strerror(errno)); | |
c2c087e6 DW |
4975 | return 0; |
4976 | } | |
4977 | if (!get_dev_size(fd, dev, &ldsize)) { | |
4978 | close(fd); | |
4979 | return 0; | |
4980 | } | |
f2f5c343 LM |
4981 | |
4982 | /* capabilities retrieve could be possible | |
4983 | * note that there is no fd for the disks in array. | |
4984 | */ | |
4985 | super = alloc_super(); | |
4986 | if (!super) { | |
4987 | fprintf(stderr, | |
4988 | Name ": malloc of %zu failed.\n", | |
4989 | sizeof(*super)); | |
4990 | close(fd); | |
4991 | return 0; | |
4992 | } | |
4993 | ||
d424212e | 4994 | rv = find_intel_hba_capability(fd, super, verbose ? dev : NULL); |
f2f5c343 LM |
4995 | if (rv != 0) { |
4996 | #if DEBUG | |
4997 | char str[256]; | |
4998 | fd2devname(fd, str); | |
4999 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
5000 | fd, str, super->orom, rv, raiddisks); | |
5001 | #endif | |
5002 | /* no orom/efi or non-intel hba of the disk */ | |
5003 | close(fd); | |
5004 | free_imsm(super); | |
5005 | return 0; | |
5006 | } | |
c2c087e6 | 5007 | close(fd); |
f2f5c343 LM |
5008 | if (super->orom && raiddisks > super->orom->tds) { |
5009 | if (verbose) | |
5010 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
5011 | " platform supported disks: %d\n", | |
5012 | raiddisks, super->orom->tds); | |
5013 | ||
5014 | free_imsm(super); | |
5015 | return 0; | |
5016 | } | |
c2c087e6 DW |
5017 | |
5018 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 5019 | free_imsm(super); |
c2c087e6 DW |
5020 | |
5021 | return 1; | |
cdddbdbc DW |
5022 | } |
5023 | ||
0dcecb2e DW |
5024 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5025 | { | |
5026 | const unsigned long long base_start = e[*idx].start; | |
5027 | unsigned long long end = base_start + e[*idx].size; | |
5028 | int i; | |
5029 | ||
5030 | if (base_start == end) | |
5031 | return 0; | |
5032 | ||
5033 | *idx = *idx + 1; | |
5034 | for (i = *idx; i < num_extents; i++) { | |
5035 | /* extend overlapping extents */ | |
5036 | if (e[i].start >= base_start && | |
5037 | e[i].start <= end) { | |
5038 | if (e[i].size == 0) | |
5039 | return 0; | |
5040 | if (e[i].start + e[i].size > end) | |
5041 | end = e[i].start + e[i].size; | |
5042 | } else if (e[i].start > end) { | |
5043 | *idx = i; | |
5044 | break; | |
5045 | } | |
5046 | } | |
5047 | ||
5048 | return end - base_start; | |
5049 | } | |
5050 | ||
5051 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5052 | { | |
5053 | /* build a composite disk with all known extents and generate a new | |
5054 | * 'maxsize' given the "all disks in an array must share a common start | |
5055 | * offset" constraint | |
5056 | */ | |
5057 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
5058 | struct dl *dl; | |
5059 | int i, j; | |
5060 | int start_extent; | |
5061 | unsigned long long pos; | |
b9d77223 | 5062 | unsigned long long start = 0; |
0dcecb2e DW |
5063 | unsigned long long maxsize; |
5064 | unsigned long reserve; | |
5065 | ||
5066 | if (!e) | |
a7dd165b | 5067 | return 0; |
0dcecb2e DW |
5068 | |
5069 | /* coalesce and sort all extents. also, check to see if we need to | |
5070 | * reserve space between member arrays | |
5071 | */ | |
5072 | j = 0; | |
5073 | for (dl = super->disks; dl; dl = dl->next) { | |
5074 | if (!dl->e) | |
5075 | continue; | |
5076 | for (i = 0; i < dl->extent_cnt; i++) | |
5077 | e[j++] = dl->e[i]; | |
5078 | } | |
5079 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5080 | ||
5081 | /* merge extents */ | |
5082 | i = 0; | |
5083 | j = 0; | |
5084 | while (i < sum_extents) { | |
5085 | e[j].start = e[i].start; | |
5086 | e[j].size = find_size(e, &i, sum_extents); | |
5087 | j++; | |
5088 | if (e[j-1].size == 0) | |
5089 | break; | |
5090 | } | |
5091 | ||
5092 | pos = 0; | |
5093 | maxsize = 0; | |
5094 | start_extent = 0; | |
5095 | i = 0; | |
5096 | do { | |
5097 | unsigned long long esize; | |
5098 | ||
5099 | esize = e[i].start - pos; | |
5100 | if (esize >= maxsize) { | |
5101 | maxsize = esize; | |
5102 | start = pos; | |
5103 | start_extent = i; | |
5104 | } | |
5105 | pos = e[i].start + e[i].size; | |
5106 | i++; | |
5107 | } while (e[i-1].size); | |
5108 | free(e); | |
5109 | ||
a7dd165b DW |
5110 | if (maxsize == 0) |
5111 | return 0; | |
5112 | ||
5113 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5114 | * container | |
5115 | */ | |
0dcecb2e DW |
5116 | if (start_extent > 0) |
5117 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5118 | else | |
5119 | reserve = 0; | |
5120 | ||
5121 | if (maxsize < reserve) | |
a7dd165b | 5122 | return 0; |
0dcecb2e DW |
5123 | |
5124 | super->create_offset = ~((__u32) 0); | |
5125 | if (start + reserve > super->create_offset) | |
a7dd165b | 5126 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5127 | super->create_offset = start + reserve; |
5128 | ||
5129 | return maxsize - reserve; | |
5130 | } | |
5131 | ||
88c32bb1 DW |
5132 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5133 | { | |
5134 | if (level < 0 || level == 6 || level == 4) | |
5135 | return 0; | |
5136 | ||
5137 | /* if we have an orom prevent invalid raid levels */ | |
5138 | if (orom) | |
5139 | switch (level) { | |
5140 | case 0: return imsm_orom_has_raid0(orom); | |
5141 | case 1: | |
5142 | if (raiddisks > 2) | |
5143 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5144 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5145 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5146 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5147 | } |
5148 | else | |
5149 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5150 | ||
5151 | return 0; | |
5152 | } | |
5153 | ||
cd9d1ac7 DW |
5154 | static int imsm_default_chunk(const struct imsm_orom *orom) |
5155 | { | |
5156 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
5157 | * 128 if no platform detected | |
5158 | */ | |
5159 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
5160 | ||
5161 | return min(512, (1 << fs)); | |
5162 | } | |
73408129 | 5163 | |
35f81cbb | 5164 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) |
6592ce37 DW |
5165 | static int |
5166 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
c21e737b | 5167 | int raiddisks, int *chunk, int verbose) |
6592ce37 | 5168 | { |
660260d0 DW |
5169 | /* check/set platform and metadata limits/defaults */ |
5170 | if (super->orom && raiddisks > super->orom->dpa) { | |
5171 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
5172 | super->orom->dpa); | |
73408129 LM |
5173 | return 0; |
5174 | } | |
5175 | ||
5176 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
660260d0 | 5177 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
5178 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
5179 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
5180 | return 0; | |
5181 | } | |
cd9d1ac7 DW |
5182 | |
5183 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
5184 | *chunk = imsm_default_chunk(super->orom); | |
5185 | ||
5186 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
5187 | pr_vrb(": platform does not support a chunk size of: " | |
5188 | "%d\n", *chunk); | |
5189 | return 0; | |
6592ce37 | 5190 | } |
cd9d1ac7 | 5191 | |
6592ce37 DW |
5192 | if (layout != imsm_level_to_layout(level)) { |
5193 | if (level == 5) | |
5194 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
5195 | else if (level == 10) | |
5196 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
5197 | else | |
5198 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
5199 | layout, level); | |
5200 | return 0; | |
5201 | } | |
6592ce37 DW |
5202 | return 1; |
5203 | } | |
5204 | ||
c2c087e6 DW |
5205 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
5206 | * FIX ME add ahci details | |
5207 | */ | |
8b353278 | 5208 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 5209 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 5210 | unsigned long long size, char *dev, |
2c514b71 NB |
5211 | unsigned long long *freesize, |
5212 | int verbose) | |
cdddbdbc | 5213 | { |
c2c087e6 DW |
5214 | struct stat stb; |
5215 | struct intel_super *super = st->sb; | |
b2916f25 | 5216 | struct imsm_super *mpb; |
c2c087e6 DW |
5217 | struct dl *dl; |
5218 | unsigned long long pos = 0; | |
5219 | unsigned long long maxsize; | |
5220 | struct extent *e; | |
5221 | int i; | |
cdddbdbc | 5222 | |
88c32bb1 DW |
5223 | /* We must have the container info already read in. */ |
5224 | if (!super) | |
c2c087e6 DW |
5225 | return 0; |
5226 | ||
b2916f25 JS |
5227 | mpb = super->anchor; |
5228 | ||
e7cb06c8 LO |
5229 | if (mpb->num_raid_devs > 0 && mpb->num_disks != raiddisks) { |
5230 | fprintf(stderr, Name ": the option-rom requires all " | |
5231 | "member disks to be a member of all volumes.\n"); | |
5232 | return 0; | |
5233 | } | |
5234 | ||
d54559f0 LM |
5235 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, verbose)) { |
5236 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
5237 | "Cannot proceed with the action(s).\n"); | |
c2c087e6 | 5238 | return 0; |
d54559f0 | 5239 | } |
c2c087e6 DW |
5240 | if (!dev) { |
5241 | /* General test: make sure there is space for | |
2da8544a DW |
5242 | * 'raiddisks' device extents of size 'size' at a given |
5243 | * offset | |
c2c087e6 | 5244 | */ |
e46273eb | 5245 | unsigned long long minsize = size; |
b7528a20 | 5246 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5247 | int dcnt = 0; |
5248 | if (minsize == 0) | |
5249 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5250 | for (dl = super->disks; dl ; dl = dl->next) { | |
5251 | int found = 0; | |
5252 | ||
bf5a934a | 5253 | pos = 0; |
c2c087e6 DW |
5254 | i = 0; |
5255 | e = get_extents(super, dl); | |
5256 | if (!e) continue; | |
5257 | do { | |
5258 | unsigned long long esize; | |
5259 | esize = e[i].start - pos; | |
5260 | if (esize >= minsize) | |
5261 | found = 1; | |
b7528a20 | 5262 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5263 | start_offset = pos; |
5264 | break; | |
5265 | } else if (found && pos != start_offset) { | |
5266 | found = 0; | |
5267 | break; | |
5268 | } | |
c2c087e6 DW |
5269 | pos = e[i].start + e[i].size; |
5270 | i++; | |
5271 | } while (e[i-1].size); | |
5272 | if (found) | |
5273 | dcnt++; | |
5274 | free(e); | |
5275 | } | |
5276 | if (dcnt < raiddisks) { | |
2c514b71 NB |
5277 | if (verbose) |
5278 | fprintf(stderr, Name ": imsm: Not enough " | |
5279 | "devices with space for this array " | |
5280 | "(%d < %d)\n", | |
5281 | dcnt, raiddisks); | |
c2c087e6 DW |
5282 | return 0; |
5283 | } | |
5284 | return 1; | |
5285 | } | |
0dcecb2e | 5286 | |
c2c087e6 DW |
5287 | /* This device must be a member of the set */ |
5288 | if (stat(dev, &stb) < 0) | |
5289 | return 0; | |
5290 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
5291 | return 0; | |
5292 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
5293 | if (dl->major == (int)major(stb.st_rdev) && |
5294 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
5295 | break; |
5296 | } | |
5297 | if (!dl) { | |
2c514b71 NB |
5298 | if (verbose) |
5299 | fprintf(stderr, Name ": %s is not in the " | |
5300 | "same imsm set\n", dev); | |
c2c087e6 | 5301 | return 0; |
a20d2ba5 DW |
5302 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
5303 | /* If a volume is present then the current creation attempt | |
5304 | * cannot incorporate new spares because the orom may not | |
5305 | * understand this configuration (all member disks must be | |
5306 | * members of each array in the container). | |
5307 | */ | |
5308 | fprintf(stderr, Name ": %s is a spare and a volume" | |
5309 | " is already defined for this container\n", dev); | |
5310 | fprintf(stderr, Name ": The option-rom requires all member" | |
5311 | " disks to be a member of all volumes\n"); | |
5312 | return 0; | |
c2c087e6 | 5313 | } |
0dcecb2e DW |
5314 | |
5315 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
5316 | e = get_extents(super, dl); |
5317 | maxsize = 0; | |
5318 | i = 0; | |
0dcecb2e DW |
5319 | if (e) { |
5320 | do { | |
5321 | unsigned long long esize; | |
5322 | ||
5323 | esize = e[i].start - pos; | |
5324 | if (esize >= maxsize) | |
5325 | maxsize = esize; | |
5326 | pos = e[i].start + e[i].size; | |
5327 | i++; | |
5328 | } while (e[i-1].size); | |
5329 | dl->e = e; | |
5330 | dl->extent_cnt = i; | |
5331 | } else { | |
5332 | if (verbose) | |
5333 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
5334 | dev); | |
5335 | return 0; | |
5336 | } | |
5337 | if (maxsize < size) { | |
5338 | if (verbose) | |
5339 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
5340 | dev, maxsize, size); | |
5341 | return 0; | |
5342 | } | |
5343 | ||
5344 | /* count total number of extents for merge */ | |
5345 | i = 0; | |
5346 | for (dl = super->disks; dl; dl = dl->next) | |
5347 | if (dl->e) | |
5348 | i += dl->extent_cnt; | |
5349 | ||
5350 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
5351 | |
5352 | if (!check_env("IMSM_NO_PLATFORM") && | |
5353 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
5354 | fprintf(stderr, Name ": attempting to create a second " | |
5355 | "volume with size less then remaining space. " | |
5356 | "Aborting...\n"); | |
5357 | return 0; | |
5358 | } | |
5359 | ||
a7dd165b | 5360 | if (maxsize < size || maxsize == 0) { |
0dcecb2e DW |
5361 | if (verbose) |
5362 | fprintf(stderr, Name ": not enough space after merge (%llu < %llu)\n", | |
5363 | maxsize, size); | |
5364 | return 0; | |
0dcecb2e DW |
5365 | } |
5366 | ||
c2c087e6 DW |
5367 | *freesize = maxsize; |
5368 | ||
5369 | return 1; | |
cdddbdbc DW |
5370 | } |
5371 | ||
efb30e7f DW |
5372 | static int reserve_space(struct supertype *st, int raiddisks, |
5373 | unsigned long long size, int chunk, | |
5374 | unsigned long long *freesize) | |
5375 | { | |
5376 | struct intel_super *super = st->sb; | |
5377 | struct imsm_super *mpb = super->anchor; | |
5378 | struct dl *dl; | |
5379 | int i; | |
5380 | int extent_cnt; | |
5381 | struct extent *e; | |
5382 | unsigned long long maxsize; | |
5383 | unsigned long long minsize; | |
5384 | int cnt; | |
5385 | int used; | |
5386 | ||
5387 | /* find the largest common start free region of the possible disks */ | |
5388 | used = 0; | |
5389 | extent_cnt = 0; | |
5390 | cnt = 0; | |
5391 | for (dl = super->disks; dl; dl = dl->next) { | |
5392 | dl->raiddisk = -1; | |
5393 | ||
5394 | if (dl->index >= 0) | |
5395 | used++; | |
5396 | ||
5397 | /* don't activate new spares if we are orom constrained | |
5398 | * and there is already a volume active in the container | |
5399 | */ | |
5400 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
5401 | continue; | |
5402 | ||
5403 | e = get_extents(super, dl); | |
5404 | if (!e) | |
5405 | continue; | |
5406 | for (i = 1; e[i-1].size; i++) | |
5407 | ; | |
5408 | dl->e = e; | |
5409 | dl->extent_cnt = i; | |
5410 | extent_cnt += i; | |
5411 | cnt++; | |
5412 | } | |
5413 | ||
5414 | maxsize = merge_extents(super, extent_cnt); | |
5415 | minsize = size; | |
5416 | if (size == 0) | |
612e59d8 CA |
5417 | /* chunk is in K */ |
5418 | minsize = chunk * 2; | |
efb30e7f DW |
5419 | |
5420 | if (cnt < raiddisks || | |
5421 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
5422 | maxsize < minsize || |
5423 | maxsize == 0) { | |
efb30e7f DW |
5424 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
5425 | return 0; /* No enough free spaces large enough */ | |
5426 | } | |
5427 | ||
5428 | if (size == 0) { | |
5429 | size = maxsize; | |
5430 | if (chunk) { | |
612e59d8 CA |
5431 | size /= 2 * chunk; |
5432 | size *= 2 * chunk; | |
efb30e7f DW |
5433 | } |
5434 | } | |
5435 | ||
5436 | cnt = 0; | |
5437 | for (dl = super->disks; dl; dl = dl->next) | |
5438 | if (dl->e) | |
5439 | dl->raiddisk = cnt++; | |
5440 | ||
5441 | *freesize = size; | |
5442 | ||
5443 | return 1; | |
5444 | } | |
5445 | ||
bf5a934a | 5446 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 5447 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
5448 | char *dev, unsigned long long *freesize, |
5449 | int verbose) | |
5450 | { | |
5451 | int fd, cfd; | |
5452 | struct mdinfo *sra; | |
20cbe8d2 | 5453 | int is_member = 0; |
bf5a934a | 5454 | |
d54559f0 LM |
5455 | /* load capability |
5456 | * if given unused devices create a container | |
bf5a934a DW |
5457 | * if given given devices in a container create a member volume |
5458 | */ | |
5459 | if (level == LEVEL_CONTAINER) { | |
5460 | /* Must be a fresh device to add to a container */ | |
5461 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
5462 | raiddisks, |
5463 | chunk?*chunk:0, size, | |
bf5a934a DW |
5464 | dev, freesize, |
5465 | verbose); | |
5466 | } | |
5467 | ||
8592f29d | 5468 | if (!dev) { |
e91a3bad LM |
5469 | if (st->sb) { |
5470 | if (!validate_geometry_imsm_orom(st->sb, level, layout, | |
5471 | raiddisks, chunk, | |
5472 | verbose)) | |
5473 | return 0; | |
efb30e7f DW |
5474 | /* we are being asked to automatically layout a |
5475 | * new volume based on the current contents of | |
5476 | * the container. If the the parameters can be | |
5477 | * satisfied reserve_space will record the disks, | |
5478 | * start offset, and size of the volume to be | |
5479 | * created. add_to_super and getinfo_super | |
5480 | * detect when autolayout is in progress. | |
5481 | */ | |
e91a3bad LM |
5482 | if (freesize) |
5483 | return reserve_space(st, raiddisks, size, | |
5484 | chunk?*chunk:0, freesize); | |
8592f29d N |
5485 | } |
5486 | return 1; | |
5487 | } | |
bf5a934a DW |
5488 | if (st->sb) { |
5489 | /* creating in a given container */ | |
5490 | return validate_geometry_imsm_volume(st, level, layout, | |
5491 | raiddisks, chunk, size, | |
5492 | dev, freesize, verbose); | |
5493 | } | |
5494 | ||
bf5a934a DW |
5495 | /* This device needs to be a device in an 'imsm' container */ |
5496 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5497 | if (fd >= 0) { | |
5498 | if (verbose) | |
5499 | fprintf(stderr, | |
5500 | Name ": Cannot create this array on device %s\n", | |
5501 | dev); | |
5502 | close(fd); | |
5503 | return 0; | |
5504 | } | |
5505 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
5506 | if (verbose) | |
5507 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
5508 | dev, strerror(errno)); | |
5509 | return 0; | |
5510 | } | |
5511 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
5512 | cfd = open_container(fd); | |
20cbe8d2 | 5513 | close(fd); |
bf5a934a | 5514 | if (cfd < 0) { |
bf5a934a DW |
5515 | if (verbose) |
5516 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
5517 | dev); | |
5518 | return 0; | |
5519 | } | |
5520 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 5521 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
5522 | strcmp(sra->text_version, "imsm") == 0) |
5523 | is_member = 1; | |
5524 | sysfs_free(sra); | |
5525 | if (is_member) { | |
bf5a934a DW |
5526 | /* This is a member of a imsm container. Load the container |
5527 | * and try to create a volume | |
5528 | */ | |
5529 | struct intel_super *super; | |
5530 | ||
e1902a7b | 5531 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL) == 0) { |
bf5a934a DW |
5532 | st->sb = super; |
5533 | st->container_dev = fd2devnum(cfd); | |
5534 | close(cfd); | |
5535 | return validate_geometry_imsm_volume(st, level, layout, | |
5536 | raiddisks, chunk, | |
5537 | size, dev, | |
ecbd9e81 N |
5538 | freesize, 1) |
5539 | ? 1 : -1; | |
bf5a934a | 5540 | } |
20cbe8d2 | 5541 | } |
bf5a934a | 5542 | |
20cbe8d2 AW |
5543 | if (verbose) |
5544 | fprintf(stderr, Name ": failed container membership check\n"); | |
5545 | ||
5546 | close(cfd); | |
5547 | return 0; | |
bf5a934a | 5548 | } |
0bd16cf2 | 5549 | |
30f58b22 | 5550 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
5551 | { |
5552 | struct intel_super *super = st->sb; | |
5553 | ||
30f58b22 DW |
5554 | if (level && *level == UnSet) |
5555 | *level = LEVEL_CONTAINER; | |
5556 | ||
5557 | if (level && layout && *layout == UnSet) | |
5558 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 5559 | |
cd9d1ac7 DW |
5560 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
5561 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
5562 | } |
5563 | ||
33414a01 DW |
5564 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
5565 | ||
5566 | static int kill_subarray_imsm(struct supertype *st) | |
5567 | { | |
5568 | /* remove the subarray currently referenced by ->current_vol */ | |
5569 | __u8 i; | |
5570 | struct intel_dev **dp; | |
5571 | struct intel_super *super = st->sb; | |
5572 | __u8 current_vol = super->current_vol; | |
5573 | struct imsm_super *mpb = super->anchor; | |
5574 | ||
5575 | if (super->current_vol < 0) | |
5576 | return 2; | |
5577 | super->current_vol = -1; /* invalidate subarray cursor */ | |
5578 | ||
5579 | /* block deletions that would change the uuid of active subarrays | |
5580 | * | |
5581 | * FIXME when immutable ids are available, but note that we'll | |
5582 | * also need to fixup the invalidated/active subarray indexes in | |
5583 | * mdstat | |
5584 | */ | |
5585 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5586 | char subarray[4]; | |
5587 | ||
5588 | if (i < current_vol) | |
5589 | continue; | |
5590 | sprintf(subarray, "%u", i); | |
5591 | if (is_subarray_active(subarray, st->devname)) { | |
5592 | fprintf(stderr, | |
5593 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
5594 | current_vol, i); | |
5595 | ||
5596 | return 2; | |
5597 | } | |
5598 | } | |
5599 | ||
5600 | if (st->update_tail) { | |
5601 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
5602 | ||
5603 | if (!u) | |
5604 | return 2; | |
5605 | u->type = update_kill_array; | |
5606 | u->dev_idx = current_vol; | |
5607 | append_metadata_update(st, u, sizeof(*u)); | |
5608 | ||
5609 | return 0; | |
5610 | } | |
5611 | ||
5612 | for (dp = &super->devlist; *dp;) | |
5613 | if ((*dp)->index == current_vol) { | |
5614 | *dp = (*dp)->next; | |
5615 | } else { | |
5616 | handle_missing(super, (*dp)->dev); | |
5617 | if ((*dp)->index > current_vol) | |
5618 | (*dp)->index--; | |
5619 | dp = &(*dp)->next; | |
5620 | } | |
5621 | ||
5622 | /* no more raid devices, all active components are now spares, | |
5623 | * but of course failed are still failed | |
5624 | */ | |
5625 | if (--mpb->num_raid_devs == 0) { | |
5626 | struct dl *d; | |
5627 | ||
5628 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
5629 | if (d->index > -2) |
5630 | mark_spare(d); | |
33414a01 DW |
5631 | } |
5632 | ||
5633 | super->updates_pending++; | |
5634 | ||
5635 | return 0; | |
5636 | } | |
aa534678 | 5637 | |
a951a4f7 | 5638 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 5639 | char *update, struct mddev_ident *ident) |
aa534678 DW |
5640 | { |
5641 | /* update the subarray currently referenced by ->current_vol */ | |
5642 | struct intel_super *super = st->sb; | |
5643 | struct imsm_super *mpb = super->anchor; | |
5644 | ||
aa534678 DW |
5645 | if (strcmp(update, "name") == 0) { |
5646 | char *name = ident->name; | |
a951a4f7 N |
5647 | char *ep; |
5648 | int vol; | |
aa534678 | 5649 | |
a951a4f7 | 5650 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
5651 | fprintf(stderr, |
5652 | Name ": Unable to update name of active subarray\n"); | |
5653 | return 2; | |
5654 | } | |
5655 | ||
5656 | if (!check_name(super, name, 0)) | |
5657 | return 2; | |
5658 | ||
a951a4f7 N |
5659 | vol = strtoul(subarray, &ep, 10); |
5660 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
5661 | return 2; | |
5662 | ||
aa534678 DW |
5663 | if (st->update_tail) { |
5664 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
5665 | ||
5666 | if (!u) | |
5667 | return 2; | |
5668 | u->type = update_rename_array; | |
a951a4f7 | 5669 | u->dev_idx = vol; |
aa534678 DW |
5670 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
5671 | append_metadata_update(st, u, sizeof(*u)); | |
5672 | } else { | |
5673 | struct imsm_dev *dev; | |
5674 | int i; | |
5675 | ||
a951a4f7 | 5676 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
5677 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
5678 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5679 | dev = get_imsm_dev(super, i); | |
5680 | handle_missing(super, dev); | |
5681 | } | |
5682 | super->updates_pending++; | |
5683 | } | |
5684 | } else | |
5685 | return 2; | |
5686 | ||
5687 | return 0; | |
5688 | } | |
d1e02575 | 5689 | #endif /* MDASSEMBLE */ |
bf5a934a | 5690 | |
28bce06f AK |
5691 | static int is_gen_migration(struct imsm_dev *dev) |
5692 | { | |
7534230b AK |
5693 | if (dev == NULL) |
5694 | return 0; | |
5695 | ||
28bce06f AK |
5696 | if (!dev->vol.migr_state) |
5697 | return 0; | |
5698 | ||
5699 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
5700 | return 1; | |
5701 | ||
5702 | return 0; | |
5703 | } | |
5704 | ||
1e5c6983 DW |
5705 | static int is_rebuilding(struct imsm_dev *dev) |
5706 | { | |
5707 | struct imsm_map *migr_map; | |
5708 | ||
5709 | if (!dev->vol.migr_state) | |
5710 | return 0; | |
5711 | ||
5712 | if (migr_type(dev) != MIGR_REBUILD) | |
5713 | return 0; | |
5714 | ||
238c0a71 | 5715 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
5716 | |
5717 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
5718 | return 1; | |
5719 | else | |
5720 | return 0; | |
5721 | } | |
5722 | ||
6ce1fbf1 AK |
5723 | static int is_initializing(struct imsm_dev *dev) |
5724 | { | |
5725 | struct imsm_map *migr_map; | |
5726 | ||
5727 | if (!dev->vol.migr_state) | |
5728 | return 0; | |
5729 | ||
5730 | if (migr_type(dev) != MIGR_INIT) | |
5731 | return 0; | |
5732 | ||
238c0a71 | 5733 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
5734 | |
5735 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
5736 | return 1; | |
5737 | ||
5738 | return 0; | |
5739 | ||
5740 | } | |
5741 | ||
c47b0ff6 AK |
5742 | static void update_recovery_start(struct intel_super *super, |
5743 | struct imsm_dev *dev, | |
5744 | struct mdinfo *array) | |
1e5c6983 DW |
5745 | { |
5746 | struct mdinfo *rebuild = NULL; | |
5747 | struct mdinfo *d; | |
5748 | __u32 units; | |
5749 | ||
5750 | if (!is_rebuilding(dev)) | |
5751 | return; | |
5752 | ||
5753 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
5754 | for (d = array->devs; d; d = d->next) | |
5755 | if (d->recovery_start == 0) { | |
5756 | if (rebuild) | |
5757 | return; | |
5758 | rebuild = d; | |
5759 | } | |
5760 | ||
4363fd80 DW |
5761 | if (!rebuild) { |
5762 | /* (?) none of the disks are marked with | |
5763 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
5764 | * disk_ord_tbl was not correctly updated | |
5765 | */ | |
5766 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
5767 | return; | |
5768 | } | |
5769 | ||
1e5c6983 | 5770 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 5771 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
5772 | } |
5773 | ||
9e2d750d | 5774 | #ifndef MDASSEMBLE |
276d77db | 5775 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 5776 | #endif |
1e5c6983 | 5777 | |
00bbdbda | 5778 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 5779 | { |
4f5bc454 DW |
5780 | /* Given a container loaded by load_super_imsm_all, |
5781 | * extract information about all the arrays into | |
5782 | * an mdinfo tree. | |
00bbdbda | 5783 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
5784 | * |
5785 | * For each imsm_dev create an mdinfo, fill it in, | |
5786 | * then look for matching devices in super->disks | |
5787 | * and create appropriate device mdinfo. | |
5788 | */ | |
5789 | struct intel_super *super = st->sb; | |
949c47a0 | 5790 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 5791 | struct mdinfo *rest = NULL; |
00bbdbda | 5792 | unsigned int i; |
81219e70 | 5793 | int sb_errors = 0; |
abef11a3 AK |
5794 | struct dl *d; |
5795 | int spare_disks = 0; | |
cdddbdbc | 5796 | |
19482bcc AK |
5797 | /* do not assemble arrays when not all attributes are supported */ |
5798 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 LM |
5799 | sb_errors = 1; |
5800 | fprintf(stderr, Name ": Unsupported attributes in IMSM metadata." | |
5801 | "Arrays activation is blocked.\n"); | |
19482bcc AK |
5802 | } |
5803 | ||
a06d022d | 5804 | /* check for bad blocks */ |
81219e70 LM |
5805 | if (imsm_bbm_log_size(super->anchor)) { |
5806 | fprintf(stderr, Name ": BBM log found in IMSM metadata." | |
5807 | "Arrays activation is blocked.\n"); | |
5808 | sb_errors = 1; | |
5809 | } | |
5810 | ||
604b746f | 5811 | |
abef11a3 AK |
5812 | /* count spare devices, not used in maps |
5813 | */ | |
5814 | for (d = super->disks; d; d = d->next) | |
5815 | if (d->index == -1) | |
5816 | spare_disks++; | |
5817 | ||
4f5bc454 | 5818 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
5819 | struct imsm_dev *dev; |
5820 | struct imsm_map *map; | |
86e3692b | 5821 | struct imsm_map *map2; |
4f5bc454 | 5822 | struct mdinfo *this; |
2db86302 | 5823 | int slot, chunk; |
00bbdbda N |
5824 | char *ep; |
5825 | ||
5826 | if (subarray && | |
5827 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
5828 | continue; | |
5829 | ||
5830 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
5831 | map = get_imsm_map(dev, MAP_0); |
5832 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 5833 | |
1ce0101c DW |
5834 | /* do not publish arrays that are in the middle of an |
5835 | * unsupported migration | |
5836 | */ | |
5837 | if (dev->vol.migr_state && | |
28bce06f | 5838 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
5839 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
5840 | " unsupported migration in progress\n", | |
5841 | dev->volume); | |
5842 | continue; | |
5843 | } | |
2db86302 LM |
5844 | /* do not publish arrays that are not support by controller's |
5845 | * OROM/EFI | |
5846 | */ | |
1ce0101c | 5847 | |
2db86302 | 5848 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
4f5bc454 | 5849 | this = malloc(sizeof(*this)); |
0fbd635c | 5850 | if (!this) { |
cf1be220 | 5851 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
5852 | sizeof(*this)); |
5853 | break; | |
5854 | } | |
4f5bc454 | 5855 | |
301406c9 | 5856 | super->current_vol = i; |
a5d85af7 | 5857 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 5858 | this->next = rest; |
81219e70 LM |
5859 | #ifndef MDASSEMBLE |
5860 | /* mdadm does not support all metadata features- set the bit in all arrays state */ | |
5861 | if (!validate_geometry_imsm_orom(super, | |
5862 | get_imsm_raid_level(map), /* RAID level */ | |
5863 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
5864 | map->num_members, /* raid disks */ | |
5865 | &chunk, | |
5866 | 1 /* verbose */)) { | |
446894ea N |
5867 | fprintf(stderr, Name ": IMSM RAID geometry validation" |
5868 | " failed. Array %s activation is blocked.\n", | |
81219e70 LM |
5869 | dev->volume); |
5870 | this->array.state |= | |
5871 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
5872 | (1<<MD_SB_BLOCK_VOLUME); | |
5873 | } | |
5874 | #endif | |
5875 | ||
5876 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
5877 | if (sb_errors) | |
5878 | this->array.state |= | |
5879 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
5880 | (1<<MD_SB_BLOCK_VOLUME); | |
5881 | ||
4f5bc454 | 5882 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 5883 | unsigned long long recovery_start; |
4f5bc454 DW |
5884 | struct mdinfo *info_d; |
5885 | struct dl *d; | |
5886 | int idx; | |
9a1608e5 | 5887 | int skip; |
7eef0453 | 5888 | __u32 ord; |
4f5bc454 | 5889 | |
9a1608e5 | 5890 | skip = 0; |
238c0a71 AK |
5891 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
5892 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
5893 | for (d = super->disks; d ; d = d->next) |
5894 | if (d->index == idx) | |
0fbd635c | 5895 | break; |
4f5bc454 | 5896 | |
1e5c6983 | 5897 | recovery_start = MaxSector; |
4f5bc454 | 5898 | if (d == NULL) |
9a1608e5 | 5899 | skip = 1; |
25ed7e59 | 5900 | if (d && is_failed(&d->disk)) |
9a1608e5 | 5901 | skip = 1; |
7eef0453 | 5902 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 5903 | recovery_start = 0; |
9a1608e5 DW |
5904 | |
5905 | /* | |
5906 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
5907 | * reset resync start to avoid a dirty-degraded |
5908 | * situation when performing the intial sync | |
9a1608e5 DW |
5909 | * |
5910 | * FIXME handle dirty degraded | |
5911 | */ | |
1e5c6983 | 5912 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 5913 | this->resync_start = MaxSector; |
9a1608e5 DW |
5914 | if (skip) |
5915 | continue; | |
4f5bc454 | 5916 | |
1e5c6983 | 5917 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
5918 | if (!info_d) { |
5919 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 5920 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
5921 | info_d = this->devs; |
5922 | while (info_d) { | |
5923 | struct mdinfo *d = info_d->next; | |
5924 | ||
5925 | free(info_d); | |
5926 | info_d = d; | |
5927 | } | |
9a1608e5 DW |
5928 | free(this); |
5929 | this = rest; | |
5930 | break; | |
5931 | } | |
4f5bc454 DW |
5932 | info_d->next = this->devs; |
5933 | this->devs = info_d; | |
5934 | ||
4f5bc454 DW |
5935 | info_d->disk.number = d->index; |
5936 | info_d->disk.major = d->major; | |
5937 | info_d->disk.minor = d->minor; | |
5938 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 5939 | info_d->recovery_start = recovery_start; |
86e3692b AK |
5940 | if (map2) { |
5941 | if (slot < map2->num_members) | |
5942 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5943 | else |
5944 | this->array.spare_disks++; | |
86e3692b AK |
5945 | } else { |
5946 | if (slot < map->num_members) | |
5947 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5948 | else |
5949 | this->array.spare_disks++; | |
86e3692b | 5950 | } |
1e5c6983 DW |
5951 | if (info_d->recovery_start == MaxSector) |
5952 | this->array.working_disks++; | |
4f5bc454 DW |
5953 | |
5954 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5955 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
5956 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
4f5bc454 | 5957 | } |
1e5c6983 | 5958 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 5959 | update_recovery_start(super, dev, this); |
abef11a3 | 5960 | this->array.spare_disks += spare_disks; |
276d77db | 5961 | |
9e2d750d | 5962 | #ifndef MDASSEMBLE |
276d77db AK |
5963 | /* check for reshape */ |
5964 | if (this->reshape_active == 1) | |
5965 | recover_backup_imsm(st, this); | |
9e2d750d | 5966 | #endif |
9a1608e5 | 5967 | rest = this; |
4f5bc454 DW |
5968 | } |
5969 | ||
5970 | return rest; | |
cdddbdbc DW |
5971 | } |
5972 | ||
845dea95 | 5973 | |
3b451610 AK |
5974 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
5975 | int failed, int look_in_map) | |
c2a1e7da | 5976 | { |
3b451610 AK |
5977 | struct imsm_map *map; |
5978 | ||
5979 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
5980 | |
5981 | if (!failed) | |
3393c6af DW |
5982 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
5983 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
5984 | |
5985 | switch (get_imsm_raid_level(map)) { | |
5986 | case 0: | |
5987 | return IMSM_T_STATE_FAILED; | |
5988 | break; | |
5989 | case 1: | |
5990 | if (failed < map->num_members) | |
5991 | return IMSM_T_STATE_DEGRADED; | |
5992 | else | |
5993 | return IMSM_T_STATE_FAILED; | |
5994 | break; | |
5995 | case 10: | |
5996 | { | |
5997 | /** | |
c92a2527 DW |
5998 | * check to see if any mirrors have failed, otherwise we |
5999 | * are degraded. Even numbered slots are mirrored on | |
6000 | * slot+1 | |
c2a1e7da | 6001 | */ |
c2a1e7da | 6002 | int i; |
d9b420a5 N |
6003 | /* gcc -Os complains that this is unused */ |
6004 | int insync = insync; | |
c2a1e7da DW |
6005 | |
6006 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6007 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6008 | int idx = ord_to_idx(ord); |
6009 | struct imsm_disk *disk; | |
c2a1e7da | 6010 | |
c92a2527 DW |
6011 | /* reset the potential in-sync count on even-numbered |
6012 | * slots. num_copies is always 2 for imsm raid10 | |
6013 | */ | |
6014 | if ((i & 1) == 0) | |
6015 | insync = 2; | |
c2a1e7da | 6016 | |
c92a2527 | 6017 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6018 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6019 | insync--; |
c2a1e7da | 6020 | |
c92a2527 DW |
6021 | /* no in-sync disks left in this mirror the |
6022 | * array has failed | |
6023 | */ | |
6024 | if (insync == 0) | |
6025 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6026 | } |
6027 | ||
6028 | return IMSM_T_STATE_DEGRADED; | |
6029 | } | |
6030 | case 5: | |
6031 | if (failed < 2) | |
6032 | return IMSM_T_STATE_DEGRADED; | |
6033 | else | |
6034 | return IMSM_T_STATE_FAILED; | |
6035 | break; | |
6036 | default: | |
6037 | break; | |
6038 | } | |
6039 | ||
6040 | return map->map_state; | |
6041 | } | |
6042 | ||
3b451610 AK |
6043 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6044 | int look_in_map) | |
c2a1e7da DW |
6045 | { |
6046 | int i; | |
6047 | int failed = 0; | |
6048 | struct imsm_disk *disk; | |
d5985138 AK |
6049 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6050 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6051 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6052 | __u32 ord; |
6053 | int idx; | |
d5985138 | 6054 | int idx_1; |
c2a1e7da | 6055 | |
0556e1a2 DW |
6056 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6057 | * disks that are being rebuilt. New failures are recorded to | |
6058 | * map[0]. So we look through all the disks we started with and | |
6059 | * see if any failures are still present, or if any new ones | |
6060 | * have arrived | |
0556e1a2 | 6061 | */ |
d5985138 AK |
6062 | map_for_loop = map; |
6063 | if (prev && (map->num_members < prev->num_members)) | |
6064 | map_for_loop = prev; | |
68fe4598 LD |
6065 | |
6066 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6067 | idx_1 = -255; |
238c0a71 AK |
6068 | /* when MAP_X is passed both maps failures are counted |
6069 | */ | |
d5985138 | 6070 | if (prev && |
238c0a71 AK |
6071 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6072 | (i < prev->num_members)) { | |
d5985138 AK |
6073 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6074 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6075 | |
d5985138 AK |
6076 | disk = get_imsm_disk(super, idx_1); |
6077 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6078 | failed++; | |
6079 | } | |
238c0a71 AK |
6080 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6081 | (i < map->num_members)) { | |
d5985138 AK |
6082 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6083 | idx = ord_to_idx(ord); | |
6084 | ||
6085 | if (idx != idx_1) { | |
6086 | disk = get_imsm_disk(super, idx); | |
6087 | if (!disk || is_failed(disk) || | |
6088 | ord & IMSM_ORD_REBUILD) | |
6089 | failed++; | |
6090 | } | |
6091 | } | |
c2a1e7da DW |
6092 | } |
6093 | ||
6094 | return failed; | |
845dea95 NB |
6095 | } |
6096 | ||
97b4d0e9 DW |
6097 | #ifndef MDASSEMBLE |
6098 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
6099 | char *inst) | |
6100 | { | |
6101 | struct intel_super *super = c->sb; | |
6102 | struct imsm_super *mpb = super->anchor; | |
6103 | ||
6104 | if (atoi(inst) >= mpb->num_raid_devs) { | |
6105 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
6106 | __func__, atoi(inst)); | |
6107 | return -ENODEV; | |
6108 | } | |
6109 | ||
6110 | dprintf("imsm: open_new %s\n", inst); | |
6111 | a->info.container_member = atoi(inst); | |
6112 | return 0; | |
6113 | } | |
6114 | ||
0c046afd DW |
6115 | static int is_resyncing(struct imsm_dev *dev) |
6116 | { | |
6117 | struct imsm_map *migr_map; | |
6118 | ||
6119 | if (!dev->vol.migr_state) | |
6120 | return 0; | |
6121 | ||
1484e727 DW |
6122 | if (migr_type(dev) == MIGR_INIT || |
6123 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
6124 | return 1; |
6125 | ||
4c9bc37b AK |
6126 | if (migr_type(dev) == MIGR_GEN_MIGR) |
6127 | return 0; | |
6128 | ||
238c0a71 | 6129 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 6130 | |
4c9bc37b AK |
6131 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
6132 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
6133 | return 1; |
6134 | else | |
6135 | return 0; | |
6136 | } | |
6137 | ||
0556e1a2 DW |
6138 | /* return true if we recorded new information */ |
6139 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 6140 | { |
0556e1a2 DW |
6141 | __u32 ord; |
6142 | int slot; | |
6143 | struct imsm_map *map; | |
86c54047 DW |
6144 | char buf[MAX_RAID_SERIAL_LEN+3]; |
6145 | unsigned int len, shift = 0; | |
0556e1a2 DW |
6146 | |
6147 | /* new failures are always set in map[0] */ | |
238c0a71 | 6148 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
6149 | |
6150 | slot = get_imsm_disk_slot(map, idx); | |
6151 | if (slot < 0) | |
6152 | return 0; | |
6153 | ||
6154 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 6155 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
6156 | return 0; |
6157 | ||
7d0c5e24 LD |
6158 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
6159 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
6160 | strcat(buf, ":0"); | |
86c54047 DW |
6161 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
6162 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
6163 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
6164 | ||
f2f27e63 | 6165 | disk->status |= FAILED_DISK; |
0556e1a2 | 6166 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
6167 | /* mark failures in second map if second map exists and this disk |
6168 | * in this slot. | |
6169 | * This is valid for migration, initialization and rebuild | |
6170 | */ | |
6171 | if (dev->vol.migr_state) { | |
238c0a71 | 6172 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
6173 | int slot2 = get_imsm_disk_slot(map2, idx); |
6174 | ||
6175 | if ((slot2 < map2->num_members) && | |
6176 | (slot2 >= 0)) | |
6177 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
6178 | idx | IMSM_ORD_REBUILD); |
6179 | } | |
f21e18ca | 6180 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
6181 | map->failed_disk_num = slot; |
6182 | return 1; | |
6183 | } | |
6184 | ||
6185 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
6186 | { | |
6187 | mark_failure(dev, disk, idx); | |
6188 | ||
6189 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
6190 | return; | |
6191 | ||
47ee5a45 DW |
6192 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
6193 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
6194 | } | |
6195 | ||
33414a01 DW |
6196 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
6197 | { | |
33414a01 | 6198 | struct dl *dl; |
33414a01 DW |
6199 | |
6200 | if (!super->missing) | |
6201 | return; | |
33414a01 DW |
6202 | |
6203 | dprintf("imsm: mark missing\n"); | |
3d59f0c0 AK |
6204 | /* end process for initialization and rebuild only |
6205 | */ | |
6206 | if (is_gen_migration(dev) == 0) { | |
6207 | __u8 map_state; | |
6208 | int failed; | |
6209 | ||
6210 | failed = imsm_count_failed(super, dev, MAP_0); | |
6211 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
6212 | ||
6213 | end_migration(dev, super, map_state); | |
6214 | } | |
33414a01 DW |
6215 | for (dl = super->missing; dl; dl = dl->next) |
6216 | mark_missing(dev, &dl->disk, dl->index); | |
6217 | super->updates_pending++; | |
6218 | } | |
6219 | ||
70bdf0dc AK |
6220 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev) |
6221 | { | |
238c0a71 | 6222 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
6223 | unsigned long long array_blocks; |
6224 | struct imsm_map *map; | |
6225 | ||
6226 | if (used_disks == 0) { | |
6227 | /* when problems occures | |
6228 | * return current array_blocks value | |
6229 | */ | |
6230 | array_blocks = __le32_to_cpu(dev->size_high); | |
6231 | array_blocks = array_blocks << 32; | |
6232 | array_blocks += __le32_to_cpu(dev->size_low); | |
6233 | ||
6234 | return array_blocks; | |
6235 | } | |
6236 | ||
6237 | /* set array size in metadata | |
6238 | */ | |
238c0a71 | 6239 | map = get_imsm_map(dev, MAP_0); |
70bdf0dc AK |
6240 | array_blocks = map->blocks_per_member * used_disks; |
6241 | ||
6242 | /* round array size down to closest MB | |
6243 | */ | |
6244 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
6245 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
6246 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
6247 | ||
6248 | return array_blocks; | |
6249 | } | |
6250 | ||
28bce06f AK |
6251 | static void imsm_set_disk(struct active_array *a, int n, int state); |
6252 | ||
0e2d1a4e AK |
6253 | static void imsm_progress_container_reshape(struct intel_super *super) |
6254 | { | |
6255 | /* if no device has a migr_state, but some device has a | |
6256 | * different number of members than the previous device, start | |
6257 | * changing the number of devices in this device to match | |
6258 | * previous. | |
6259 | */ | |
6260 | struct imsm_super *mpb = super->anchor; | |
6261 | int prev_disks = -1; | |
6262 | int i; | |
1dfaa380 | 6263 | int copy_map_size; |
0e2d1a4e AK |
6264 | |
6265 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6266 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 6267 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
6268 | struct imsm_map *map2; |
6269 | int prev_num_members; | |
0e2d1a4e AK |
6270 | |
6271 | if (dev->vol.migr_state) | |
6272 | return; | |
6273 | ||
6274 | if (prev_disks == -1) | |
6275 | prev_disks = map->num_members; | |
6276 | if (prev_disks == map->num_members) | |
6277 | continue; | |
6278 | ||
6279 | /* OK, this array needs to enter reshape mode. | |
6280 | * i.e it needs a migr_state | |
6281 | */ | |
6282 | ||
1dfaa380 | 6283 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
6284 | prev_num_members = map->num_members; |
6285 | map->num_members = prev_disks; | |
6286 | dev->vol.migr_state = 1; | |
6287 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 6288 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
6289 | for (i = prev_num_members; |
6290 | i < map->num_members; i++) | |
6291 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 6292 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 6293 | /* Copy the current map */ |
1dfaa380 | 6294 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
6295 | map2->num_members = prev_num_members; |
6296 | ||
70bdf0dc | 6297 | imsm_set_array_size(dev); |
0e2d1a4e AK |
6298 | super->updates_pending++; |
6299 | } | |
6300 | } | |
6301 | ||
aad6f216 | 6302 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
6303 | * states are handled in imsm_set_disk() with one exception, when a |
6304 | * resync is stopped due to a new failure this routine will set the | |
6305 | * 'degraded' state for the array. | |
6306 | */ | |
01f157d7 | 6307 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
6308 | { |
6309 | int inst = a->info.container_member; | |
6310 | struct intel_super *super = a->container->sb; | |
949c47a0 | 6311 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 6312 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
6313 | int failed = imsm_count_failed(super, dev, MAP_0); |
6314 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 6315 | __u32 blocks_per_unit; |
a862209d | 6316 | |
1af97990 AK |
6317 | if (dev->vol.migr_state && |
6318 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
6319 | /* array state change is blocked due to reshape action | |
aad6f216 N |
6320 | * We might need to |
6321 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
6322 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
6323 | * - update curr_migr_unit | |
1af97990 | 6324 | */ |
aad6f216 N |
6325 | if (a->curr_action == reshape) { |
6326 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 6327 | goto mark_checkpoint; |
aad6f216 N |
6328 | } else { |
6329 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
6330 | /* for some reason we aborted the reshape. | |
b66e591b AK |
6331 | * |
6332 | * disable automatic metadata rollback | |
6333 | * user action is required to recover process | |
aad6f216 | 6334 | */ |
b66e591b | 6335 | if (0) { |
238c0a71 AK |
6336 | struct imsm_map *map2 = |
6337 | get_imsm_map(dev, MAP_1); | |
6338 | dev->vol.migr_state = 0; | |
6339 | set_migr_type(dev, 0); | |
6340 | dev->vol.curr_migr_unit = 0; | |
6341 | memcpy(map, map2, | |
6342 | sizeof_imsm_map(map2)); | |
6343 | super->updates_pending++; | |
b66e591b | 6344 | } |
aad6f216 N |
6345 | } |
6346 | if (a->last_checkpoint >= a->info.component_size) { | |
6347 | unsigned long long array_blocks; | |
6348 | int used_disks; | |
e154ced3 | 6349 | struct mdinfo *mdi; |
aad6f216 | 6350 | |
238c0a71 | 6351 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
6352 | if (used_disks > 0) { |
6353 | array_blocks = | |
6354 | map->blocks_per_member * | |
6355 | used_disks; | |
6356 | /* round array size down to closest MB | |
6357 | */ | |
6358 | array_blocks = (array_blocks | |
6359 | >> SECT_PER_MB_SHIFT) | |
6360 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
6361 | a->info.custom_array_size = array_blocks; |
6362 | /* encourage manager to update array | |
6363 | * size | |
6364 | */ | |
e154ced3 | 6365 | |
d55adef9 | 6366 | a->check_reshape = 1; |
633b5610 | 6367 | } |
e154ced3 AK |
6368 | /* finalize online capacity expansion/reshape */ |
6369 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
6370 | imsm_set_disk(a, | |
6371 | mdi->disk.raid_disk, | |
6372 | mdi->curr_state); | |
6373 | ||
0e2d1a4e | 6374 | imsm_progress_container_reshape(super); |
e154ced3 | 6375 | } |
aad6f216 | 6376 | } |
1af97990 AK |
6377 | } |
6378 | ||
47ee5a45 | 6379 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
6380 | if (consistent == 2) |
6381 | handle_missing(super, dev); | |
1e5c6983 | 6382 | |
0c046afd | 6383 | if (consistent == 2 && |
b7941fd6 | 6384 | (!is_resync_complete(&a->info) || |
0c046afd DW |
6385 | map_state != IMSM_T_STATE_NORMAL || |
6386 | dev->vol.migr_state)) | |
01f157d7 | 6387 | consistent = 0; |
272906ef | 6388 | |
b7941fd6 | 6389 | if (is_resync_complete(&a->info)) { |
0c046afd | 6390 | /* complete intialization / resync, |
0556e1a2 DW |
6391 | * recovery and interrupted recovery is completed in |
6392 | * ->set_disk | |
0c046afd DW |
6393 | */ |
6394 | if (is_resyncing(dev)) { | |
6395 | dprintf("imsm: mark resync done\n"); | |
809da78e | 6396 | end_migration(dev, super, map_state); |
115c3803 | 6397 | super->updates_pending++; |
484240d8 | 6398 | a->last_checkpoint = 0; |
115c3803 | 6399 | } |
b9172665 AK |
6400 | } else if ((!is_resyncing(dev) && !failed) && |
6401 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 6402 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 6403 | dprintf("imsm: mark resync start\n"); |
1484e727 | 6404 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 6405 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 6406 | else |
8e59f3d8 | 6407 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 6408 | super->updates_pending++; |
115c3803 | 6409 | } |
a862209d | 6410 | |
633b5610 | 6411 | mark_checkpoint: |
5b83bacf AK |
6412 | /* skip checkpointing for general migration, |
6413 | * it is controlled in mdadm | |
6414 | */ | |
6415 | if (is_gen_migration(dev)) | |
6416 | goto skip_mark_checkpoint; | |
6417 | ||
1e5c6983 | 6418 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 6419 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 6420 | if (blocks_per_unit) { |
1e5c6983 DW |
6421 | __u32 units32; |
6422 | __u64 units; | |
6423 | ||
4f0a7acc | 6424 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
6425 | units32 = units; |
6426 | ||
6427 | /* check that we did not overflow 32-bits, and that | |
6428 | * curr_migr_unit needs updating | |
6429 | */ | |
6430 | if (units32 == units && | |
bfd80a56 | 6431 | units32 != 0 && |
1e5c6983 DW |
6432 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
6433 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
6434 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
6435 | super->updates_pending++; | |
6436 | } | |
6437 | } | |
f8f603f1 | 6438 | |
5b83bacf | 6439 | skip_mark_checkpoint: |
3393c6af | 6440 | /* mark dirty / clean */ |
0c046afd | 6441 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 6442 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
6443 | if (consistent) |
6444 | dev->vol.dirty = 0; | |
6445 | else | |
6446 | dev->vol.dirty = 1; | |
a862209d DW |
6447 | super->updates_pending++; |
6448 | } | |
28bce06f | 6449 | |
01f157d7 | 6450 | return consistent; |
a862209d DW |
6451 | } |
6452 | ||
8d45d196 | 6453 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 6454 | { |
8d45d196 DW |
6455 | int inst = a->info.container_member; |
6456 | struct intel_super *super = a->container->sb; | |
949c47a0 | 6457 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 6458 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 6459 | struct imsm_disk *disk; |
0c046afd | 6460 | int failed; |
b10b37b8 | 6461 | __u32 ord; |
0c046afd | 6462 | __u8 map_state; |
8d45d196 DW |
6463 | |
6464 | if (n > map->num_members) | |
6465 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
6466 | n, map->num_members - 1); | |
6467 | ||
6468 | if (n < 0) | |
6469 | return; | |
6470 | ||
4e6e574a | 6471 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 6472 | |
238c0a71 | 6473 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 6474 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 6475 | |
5802a811 | 6476 | /* check for new failures */ |
0556e1a2 DW |
6477 | if (state & DS_FAULTY) { |
6478 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
6479 | super->updates_pending++; | |
8d45d196 | 6480 | } |
47ee5a45 | 6481 | |
19859edc | 6482 | /* check if in_sync */ |
0556e1a2 | 6483 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 6484 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
6485 | |
6486 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
6487 | super->updates_pending++; |
6488 | } | |
8d45d196 | 6489 | |
3b451610 AK |
6490 | failed = imsm_count_failed(super, dev, MAP_0); |
6491 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 6492 | |
0c046afd | 6493 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
6494 | dprintf("imsm: Detected transition to state "); |
6495 | switch (map_state) { | |
6496 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
6497 | dprintf("normal: "); | |
6498 | if (is_rebuilding(dev)) { | |
6499 | dprintf("while rebuilding"); | |
6500 | end_migration(dev, super, map_state); | |
238c0a71 | 6501 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
6502 | map->failed_disk_num = ~0; |
6503 | super->updates_pending++; | |
6504 | a->last_checkpoint = 0; | |
6505 | break; | |
6506 | } | |
6507 | if (is_gen_migration(dev)) { | |
6508 | dprintf("while general migration"); | |
bf2f0071 | 6509 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 6510 | end_migration(dev, super, map_state); |
94002678 AK |
6511 | else |
6512 | map->map_state = map_state; | |
238c0a71 | 6513 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 6514 | map->failed_disk_num = ~0; |
94002678 | 6515 | super->updates_pending++; |
bf2f0071 | 6516 | break; |
94002678 AK |
6517 | } |
6518 | break; | |
6519 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
6520 | dprintf("degraded: "); | |
6521 | if ((map->map_state != map_state) && | |
6522 | !dev->vol.migr_state) { | |
6523 | dprintf("mark degraded"); | |
6524 | map->map_state = map_state; | |
6525 | super->updates_pending++; | |
6526 | a->last_checkpoint = 0; | |
6527 | break; | |
6528 | } | |
6529 | if (is_rebuilding(dev)) { | |
6530 | dprintf("while rebuilding."); | |
6531 | if (map->map_state != map_state) { | |
6532 | dprintf(" Map state change"); | |
6533 | end_migration(dev, super, map_state); | |
6534 | super->updates_pending++; | |
6535 | } | |
6536 | break; | |
6537 | } | |
6538 | if (is_gen_migration(dev)) { | |
6539 | dprintf("while general migration"); | |
bf2f0071 | 6540 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 6541 | end_migration(dev, super, map_state); |
94002678 AK |
6542 | else { |
6543 | map->map_state = map_state; | |
3b451610 | 6544 | manage_second_map(super, dev); |
94002678 AK |
6545 | } |
6546 | super->updates_pending++; | |
bf2f0071 | 6547 | break; |
28bce06f | 6548 | } |
6ce1fbf1 AK |
6549 | if (is_initializing(dev)) { |
6550 | dprintf("while initialization."); | |
6551 | map->map_state = map_state; | |
6552 | super->updates_pending++; | |
6553 | break; | |
6554 | } | |
94002678 AK |
6555 | break; |
6556 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
6557 | dprintf("failed: "); | |
6558 | if (is_gen_migration(dev)) { | |
6559 | dprintf("while general migration"); | |
6560 | map->map_state = map_state; | |
6561 | super->updates_pending++; | |
6562 | break; | |
6563 | } | |
6564 | if (map->map_state != map_state) { | |
6565 | dprintf("mark failed"); | |
6566 | end_migration(dev, super, map_state); | |
6567 | super->updates_pending++; | |
6568 | a->last_checkpoint = 0; | |
6569 | break; | |
6570 | } | |
6571 | break; | |
6572 | default: | |
6573 | dprintf("state %i\n", map_state); | |
5802a811 | 6574 | } |
94002678 AK |
6575 | dprintf("\n"); |
6576 | ||
845dea95 NB |
6577 | } |
6578 | ||
f796af5d | 6579 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 6580 | { |
f796af5d | 6581 | void *buf = mpb; |
c2a1e7da DW |
6582 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
6583 | unsigned long long dsize; | |
6584 | unsigned long long sectors; | |
6585 | ||
6586 | get_dev_size(fd, NULL, &dsize); | |
6587 | ||
272f648f DW |
6588 | if (mpb_size > 512) { |
6589 | /* -1 to account for anchor */ | |
6590 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 6591 | |
272f648f DW |
6592 | /* write the extended mpb to the sectors preceeding the anchor */ |
6593 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
6594 | return 1; | |
c2a1e7da | 6595 | |
f21e18ca N |
6596 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
6597 | != 512 * sectors) | |
272f648f DW |
6598 | return 1; |
6599 | } | |
c2a1e7da | 6600 | |
272f648f DW |
6601 | /* first block is stored on second to last sector of the disk */ |
6602 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
6603 | return 1; |
6604 | ||
f796af5d | 6605 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
6606 | return 1; |
6607 | ||
c2a1e7da DW |
6608 | return 0; |
6609 | } | |
6610 | ||
2e735d19 | 6611 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 6612 | { |
2e735d19 | 6613 | struct intel_super *super = container->sb; |
c2a1e7da | 6614 | |
1a64be56 | 6615 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
6616 | if (!super->updates_pending) |
6617 | return; | |
6618 | ||
36988a3d | 6619 | write_super_imsm(container, 0); |
c2a1e7da DW |
6620 | |
6621 | super->updates_pending = 0; | |
845dea95 NB |
6622 | } |
6623 | ||
272906ef DW |
6624 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
6625 | { | |
6626 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 6627 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
6628 | struct dl *dl; |
6629 | ||
6630 | for (dl = super->disks; dl; dl = dl->next) | |
6631 | if (dl->index == i) | |
6632 | break; | |
6633 | ||
25ed7e59 | 6634 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
6635 | dl = NULL; |
6636 | ||
6637 | if (dl) | |
6638 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
6639 | ||
6640 | return dl; | |
6641 | } | |
6642 | ||
a20d2ba5 | 6643 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
6644 | struct active_array *a, int activate_new, |
6645 | struct mdinfo *additional_test_list) | |
272906ef DW |
6646 | { |
6647 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 6648 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
6649 | struct imsm_super *mpb = super->anchor; |
6650 | struct imsm_map *map; | |
272906ef DW |
6651 | unsigned long long pos; |
6652 | struct mdinfo *d; | |
6653 | struct extent *ex; | |
a20d2ba5 | 6654 | int i, j; |
272906ef | 6655 | int found; |
569cc43f DW |
6656 | __u32 array_start = 0; |
6657 | __u32 array_end = 0; | |
272906ef | 6658 | struct dl *dl; |
6c932028 | 6659 | struct mdinfo *test_list; |
272906ef DW |
6660 | |
6661 | for (dl = super->disks; dl; dl = dl->next) { | |
6662 | /* If in this array, skip */ | |
6663 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
6664 | if (d->state_fd >= 0 && |
6665 | d->disk.major == dl->major && | |
272906ef | 6666 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
6667 | dprintf("%x:%x already in array\n", |
6668 | dl->major, dl->minor); | |
272906ef DW |
6669 | break; |
6670 | } | |
6671 | if (d) | |
6672 | continue; | |
6c932028 AK |
6673 | test_list = additional_test_list; |
6674 | while (test_list) { | |
6675 | if (test_list->disk.major == dl->major && | |
6676 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
6677 | dprintf("%x:%x already in additional test list\n", |
6678 | dl->major, dl->minor); | |
6679 | break; | |
6680 | } | |
6c932028 | 6681 | test_list = test_list->next; |
8ba77d32 | 6682 | } |
6c932028 | 6683 | if (test_list) |
8ba77d32 | 6684 | continue; |
272906ef | 6685 | |
e553d2a4 | 6686 | /* skip in use or failed drives */ |
25ed7e59 | 6687 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
6688 | dl->index == -2) { |
6689 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 6690 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
6691 | continue; |
6692 | } | |
6693 | ||
a20d2ba5 DW |
6694 | /* skip pure spares when we are looking for partially |
6695 | * assimilated drives | |
6696 | */ | |
6697 | if (dl->index == -1 && !activate_new) | |
6698 | continue; | |
6699 | ||
272906ef | 6700 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 6701 | * It needs to be able to cover all member volumes |
272906ef DW |
6702 | */ |
6703 | ex = get_extents(super, dl); | |
6704 | if (!ex) { | |
6705 | dprintf("cannot get extents\n"); | |
6706 | continue; | |
6707 | } | |
a20d2ba5 DW |
6708 | for (i = 0; i < mpb->num_raid_devs; i++) { |
6709 | dev = get_imsm_dev(super, i); | |
238c0a71 | 6710 | map = get_imsm_map(dev, MAP_0); |
272906ef | 6711 | |
a20d2ba5 DW |
6712 | /* check if this disk is already a member of |
6713 | * this array | |
272906ef | 6714 | */ |
620b1713 | 6715 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
6716 | continue; |
6717 | ||
6718 | found = 0; | |
6719 | j = 0; | |
6720 | pos = 0; | |
6721 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
329c8278 DW |
6722 | array_end = array_start + |
6723 | __le32_to_cpu(map->blocks_per_member) - 1; | |
a20d2ba5 DW |
6724 | |
6725 | do { | |
6726 | /* check that we can start at pba_of_lba0 with | |
6727 | * blocks_per_member of space | |
6728 | */ | |
329c8278 | 6729 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
6730 | found = 1; |
6731 | break; | |
6732 | } | |
6733 | pos = ex[j].start + ex[j].size; | |
6734 | j++; | |
6735 | } while (ex[j-1].size); | |
6736 | ||
6737 | if (!found) | |
272906ef | 6738 | break; |
a20d2ba5 | 6739 | } |
272906ef DW |
6740 | |
6741 | free(ex); | |
a20d2ba5 | 6742 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
6743 | dprintf("%x:%x does not have %u to %u available\n", |
6744 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
6745 | /* No room */ |
6746 | continue; | |
a20d2ba5 DW |
6747 | } |
6748 | return dl; | |
272906ef DW |
6749 | } |
6750 | ||
6751 | return dl; | |
6752 | } | |
6753 | ||
95d07a2c LM |
6754 | |
6755 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
6756 | { | |
6757 | struct imsm_dev *dev2; | |
6758 | struct imsm_map *map; | |
6759 | struct dl *idisk; | |
6760 | int slot; | |
6761 | int idx; | |
6762 | __u8 state; | |
6763 | ||
6764 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
6765 | if (dev2) { | |
238c0a71 | 6766 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 6767 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 6768 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
6769 | if (!map) |
6770 | return 1; | |
6771 | for (slot = 0; slot < map->num_members; slot++) { | |
6772 | /* | |
6773 | * Check if failed disks are deleted from intel | |
6774 | * disk list or are marked to be deleted | |
6775 | */ | |
238c0a71 | 6776 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
6777 | idisk = get_imsm_dl_disk(cont->sb, idx); |
6778 | /* | |
6779 | * Do not rebuild the array if failed disks | |
6780 | * from failed sub-array are not removed from | |
6781 | * container. | |
6782 | */ | |
6783 | if (idisk && | |
6784 | is_failed(&idisk->disk) && | |
6785 | (idisk->action != DISK_REMOVE)) | |
6786 | return 0; | |
6787 | } | |
6788 | } | |
6789 | } | |
6790 | return 1; | |
6791 | } | |
6792 | ||
88758e9d DW |
6793 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
6794 | struct metadata_update **updates) | |
6795 | { | |
6796 | /** | |
d23fe947 DW |
6797 | * Find a device with unused free space and use it to replace a |
6798 | * failed/vacant region in an array. We replace failed regions one a | |
6799 | * array at a time. The result is that a new spare disk will be added | |
6800 | * to the first failed array and after the monitor has finished | |
6801 | * propagating failures the remainder will be consumed. | |
88758e9d | 6802 | * |
d23fe947 DW |
6803 | * FIXME add a capability for mdmon to request spares from another |
6804 | * container. | |
88758e9d DW |
6805 | */ |
6806 | ||
6807 | struct intel_super *super = a->container->sb; | |
88758e9d | 6808 | int inst = a->info.container_member; |
949c47a0 | 6809 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 6810 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
6811 | int failed = a->info.array.raid_disks; |
6812 | struct mdinfo *rv = NULL; | |
6813 | struct mdinfo *d; | |
6814 | struct mdinfo *di; | |
6815 | struct metadata_update *mu; | |
6816 | struct dl *dl; | |
6817 | struct imsm_update_activate_spare *u; | |
6818 | int num_spares = 0; | |
6819 | int i; | |
95d07a2c | 6820 | int allowed; |
88758e9d DW |
6821 | |
6822 | for (d = a->info.devs ; d ; d = d->next) { | |
6823 | if ((d->curr_state & DS_FAULTY) && | |
6824 | d->state_fd >= 0) | |
6825 | /* wait for Removal to happen */ | |
6826 | return NULL; | |
6827 | if (d->state_fd >= 0) | |
6828 | failed--; | |
6829 | } | |
6830 | ||
6831 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
6832 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 6833 | |
e2962bfc AK |
6834 | if (imsm_reshape_blocks_arrays_changes(super)) |
6835 | return NULL; | |
1af97990 | 6836 | |
fc8ca064 AK |
6837 | /* Cannot activate another spare if rebuild is in progress already |
6838 | */ | |
6839 | if (is_rebuilding(dev)) { | |
6840 | dprintf("imsm: No spare activation allowed. " | |
6841 | "Rebuild in progress already.\n"); | |
6842 | return NULL; | |
6843 | } | |
6844 | ||
89c67882 AK |
6845 | if (a->info.array.level == 4) |
6846 | /* No repair for takeovered array | |
6847 | * imsm doesn't support raid4 | |
6848 | */ | |
6849 | return NULL; | |
6850 | ||
3b451610 AK |
6851 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
6852 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
6853 | return NULL; |
6854 | ||
95d07a2c LM |
6855 | /* |
6856 | * If there are any failed disks check state of the other volume. | |
6857 | * Block rebuild if the another one is failed until failed disks | |
6858 | * are removed from container. | |
6859 | */ | |
6860 | if (failed) { | |
c4acd1e5 | 6861 | dprintf("found failed disks in %.*s, check if there another" |
95d07a2c | 6862 | "failed sub-array.\n", |
c4acd1e5 | 6863 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
6864 | /* check if states of the other volumes allow for rebuild */ |
6865 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
6866 | if (i != inst) { | |
6867 | allowed = imsm_rebuild_allowed(a->container, | |
6868 | i, failed); | |
6869 | if (!allowed) | |
6870 | return NULL; | |
6871 | } | |
6872 | } | |
6873 | } | |
6874 | ||
88758e9d | 6875 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
6876 | for (i = 0; i < a->info.array.raid_disks; i++) { |
6877 | for (d = a->info.devs ; d ; d = d->next) | |
6878 | if (d->disk.raid_disk == i) | |
6879 | break; | |
6880 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
6881 | if (d && (d->state_fd >= 0)) | |
6882 | continue; | |
6883 | ||
272906ef | 6884 | /* |
a20d2ba5 DW |
6885 | * OK, this device needs recovery. Try to re-add the |
6886 | * previous occupant of this slot, if this fails see if | |
6887 | * we can continue the assimilation of a spare that was | |
6888 | * partially assimilated, finally try to activate a new | |
6889 | * spare. | |
272906ef DW |
6890 | */ |
6891 | dl = imsm_readd(super, i, a); | |
6892 | if (!dl) | |
b303fe21 | 6893 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 6894 | if (!dl) |
b303fe21 | 6895 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
6896 | if (!dl) |
6897 | continue; | |
6898 | ||
6899 | /* found a usable disk with enough space */ | |
6900 | di = malloc(sizeof(*di)); | |
79244939 DW |
6901 | if (!di) |
6902 | continue; | |
272906ef DW |
6903 | memset(di, 0, sizeof(*di)); |
6904 | ||
6905 | /* dl->index will be -1 in the case we are activating a | |
6906 | * pristine spare. imsm_process_update() will create a | |
6907 | * new index in this case. Once a disk is found to be | |
6908 | * failed in all member arrays it is kicked from the | |
6909 | * metadata | |
6910 | */ | |
6911 | di->disk.number = dl->index; | |
d23fe947 | 6912 | |
272906ef DW |
6913 | /* (ab)use di->devs to store a pointer to the device |
6914 | * we chose | |
6915 | */ | |
6916 | di->devs = (struct mdinfo *) dl; | |
6917 | ||
6918 | di->disk.raid_disk = i; | |
6919 | di->disk.major = dl->major; | |
6920 | di->disk.minor = dl->minor; | |
6921 | di->disk.state = 0; | |
d23534e4 | 6922 | di->recovery_start = 0; |
272906ef DW |
6923 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); |
6924 | di->component_size = a->info.component_size; | |
6925 | di->container_member = inst; | |
148acb7b | 6926 | super->random = random32(); |
272906ef DW |
6927 | di->next = rv; |
6928 | rv = di; | |
6929 | num_spares++; | |
6930 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
6931 | i, di->data_offset); | |
88758e9d DW |
6932 | } |
6933 | ||
6934 | if (!rv) | |
6935 | /* No spares found */ | |
6936 | return rv; | |
6937 | /* Now 'rv' has a list of devices to return. | |
6938 | * Create a metadata_update record to update the | |
6939 | * disk_ord_tbl for the array | |
6940 | */ | |
6941 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
6942 | if (mu) { |
6943 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
6944 | if (mu->buf == NULL) { | |
6945 | free(mu); | |
6946 | mu = NULL; | |
6947 | } | |
6948 | } | |
6949 | if (!mu) { | |
6950 | while (rv) { | |
6951 | struct mdinfo *n = rv->next; | |
6952 | ||
6953 | free(rv); | |
6954 | rv = n; | |
6955 | } | |
6956 | return NULL; | |
6957 | } | |
6958 | ||
88758e9d | 6959 | mu->space = NULL; |
cb23f1f4 | 6960 | mu->space_list = NULL; |
88758e9d DW |
6961 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
6962 | mu->next = *updates; | |
6963 | u = (struct imsm_update_activate_spare *) mu->buf; | |
6964 | ||
6965 | for (di = rv ; di ; di = di->next) { | |
6966 | u->type = update_activate_spare; | |
d23fe947 DW |
6967 | u->dl = (struct dl *) di->devs; |
6968 | di->devs = NULL; | |
88758e9d DW |
6969 | u->slot = di->disk.raid_disk; |
6970 | u->array = inst; | |
6971 | u->next = u + 1; | |
6972 | u++; | |
6973 | } | |
6974 | (u-1)->next = NULL; | |
6975 | *updates = mu; | |
6976 | ||
6977 | return rv; | |
6978 | } | |
6979 | ||
54c2c1ea | 6980 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 6981 | { |
54c2c1ea | 6982 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
6983 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6984 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
6985 | struct disk_info *inf = get_disk_info(u); |
6986 | struct imsm_disk *disk; | |
8273f55e DW |
6987 | int i; |
6988 | int j; | |
8273f55e | 6989 | |
54c2c1ea | 6990 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 6991 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
6992 | for (j = 0; j < new_map->num_members; j++) |
6993 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
6994 | return 1; |
6995 | } | |
6996 | ||
6997 | return 0; | |
6998 | } | |
6999 | ||
1a64be56 LM |
7000 | |
7001 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
7002 | { | |
7003 | struct dl *dl = NULL; | |
7004 | for (dl = super->disks; dl; dl = dl->next) | |
7005 | if ((dl->major == major) && (dl->minor == minor)) | |
7006 | return dl; | |
7007 | return NULL; | |
7008 | } | |
7009 | ||
7010 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7011 | { | |
7012 | struct dl *prev = NULL; | |
7013 | struct dl *dl; | |
7014 | ||
7015 | prev = NULL; | |
7016 | for (dl = super->disks; dl; dl = dl->next) { | |
7017 | if ((dl->major == major) && (dl->minor == minor)) { | |
7018 | /* remove */ | |
7019 | if (prev) | |
7020 | prev->next = dl->next; | |
7021 | else | |
7022 | super->disks = dl->next; | |
7023 | dl->next = NULL; | |
7024 | __free_imsm_disk(dl); | |
7025 | dprintf("%s: removed %x:%x\n", | |
7026 | __func__, major, minor); | |
7027 | break; | |
7028 | } | |
7029 | prev = dl; | |
7030 | } | |
7031 | return 0; | |
7032 | } | |
7033 | ||
f21e18ca | 7034 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7035 | |
1a64be56 LM |
7036 | static int add_remove_disk_update(struct intel_super *super) |
7037 | { | |
7038 | int check_degraded = 0; | |
7039 | struct dl *disk = NULL; | |
7040 | /* add/remove some spares to/from the metadata/contrainer */ | |
7041 | while (super->disk_mgmt_list) { | |
7042 | struct dl *disk_cfg; | |
7043 | ||
7044 | disk_cfg = super->disk_mgmt_list; | |
7045 | super->disk_mgmt_list = disk_cfg->next; | |
7046 | disk_cfg->next = NULL; | |
7047 | ||
7048 | if (disk_cfg->action == DISK_ADD) { | |
7049 | disk_cfg->next = super->disks; | |
7050 | super->disks = disk_cfg; | |
7051 | check_degraded = 1; | |
7052 | dprintf("%s: added %x:%x\n", | |
7053 | __func__, disk_cfg->major, | |
7054 | disk_cfg->minor); | |
7055 | } else if (disk_cfg->action == DISK_REMOVE) { | |
7056 | dprintf("Disk remove action processed: %x.%x\n", | |
7057 | disk_cfg->major, disk_cfg->minor); | |
7058 | disk = get_disk_super(super, | |
7059 | disk_cfg->major, | |
7060 | disk_cfg->minor); | |
7061 | if (disk) { | |
7062 | /* store action status */ | |
7063 | disk->action = DISK_REMOVE; | |
7064 | /* remove spare disks only */ | |
7065 | if (disk->index == -1) { | |
7066 | remove_disk_super(super, | |
7067 | disk_cfg->major, | |
7068 | disk_cfg->minor); | |
7069 | } | |
7070 | } | |
7071 | /* release allocate disk structure */ | |
7072 | __free_imsm_disk(disk_cfg); | |
7073 | } | |
7074 | } | |
7075 | return check_degraded; | |
7076 | } | |
7077 | ||
a29911da PC |
7078 | |
7079 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
7080 | struct intel_super *super, | |
7081 | void ***space_list) | |
7082 | { | |
7083 | struct intel_dev *id; | |
7084 | void **tofree = NULL; | |
7085 | int ret_val = 0; | |
7086 | ||
7087 | dprintf("apply_reshape_migration_update()\n"); | |
7088 | if ((u->subdev < 0) || | |
7089 | (u->subdev > 1)) { | |
7090 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7091 | return ret_val; | |
7092 | } | |
7093 | if ((space_list == NULL) || (*space_list == NULL)) { | |
7094 | dprintf("imsm: Error: Memory is not allocated\n"); | |
7095 | return ret_val; | |
7096 | } | |
7097 | ||
7098 | for (id = super->devlist ; id; id = id->next) { | |
7099 | if (id->index == (unsigned)u->subdev) { | |
7100 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7101 | struct imsm_map *map; | |
7102 | struct imsm_dev *new_dev = | |
7103 | (struct imsm_dev *)*space_list; | |
238c0a71 | 7104 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
7105 | int to_state; |
7106 | struct dl *new_disk; | |
7107 | ||
7108 | if (new_dev == NULL) | |
7109 | return ret_val; | |
7110 | *space_list = **space_list; | |
7111 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 7112 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
7113 | if (migr_map) { |
7114 | dprintf("imsm: Error: migration in progress"); | |
7115 | return ret_val; | |
7116 | } | |
7117 | ||
7118 | to_state = map->map_state; | |
7119 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
7120 | map->num_members++; | |
7121 | /* this should not happen */ | |
7122 | if (u->new_disks[0] < 0) { | |
7123 | map->failed_disk_num = | |
7124 | map->num_members - 1; | |
7125 | to_state = IMSM_T_STATE_DEGRADED; | |
7126 | } else | |
7127 | to_state = IMSM_T_STATE_NORMAL; | |
7128 | } | |
8e59f3d8 | 7129 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
7130 | if (u->new_level > -1) |
7131 | map->raid_level = u->new_level; | |
238c0a71 | 7132 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
7133 | if ((u->new_level == 5) && |
7134 | (migr_map->raid_level == 0)) { | |
7135 | int ord = map->num_members - 1; | |
7136 | migr_map->num_members--; | |
7137 | if (u->new_disks[0] < 0) | |
7138 | ord |= IMSM_ORD_REBUILD; | |
7139 | set_imsm_ord_tbl_ent(map, | |
7140 | map->num_members - 1, | |
7141 | ord); | |
7142 | } | |
7143 | id->dev = new_dev; | |
7144 | tofree = (void **)dev; | |
7145 | ||
4bba0439 PC |
7146 | /* update chunk size |
7147 | */ | |
7148 | if (u->new_chunksize > 0) | |
7149 | map->blocks_per_strip = | |
7150 | __cpu_to_le16(u->new_chunksize * 2); | |
7151 | ||
a29911da PC |
7152 | /* add disk |
7153 | */ | |
7154 | if ((u->new_level != 5) || | |
7155 | (migr_map->raid_level != 0) || | |
7156 | (migr_map->raid_level == map->raid_level)) | |
7157 | goto skip_disk_add; | |
7158 | ||
7159 | if (u->new_disks[0] >= 0) { | |
7160 | /* use passes spare | |
7161 | */ | |
7162 | new_disk = get_disk_super(super, | |
7163 | major(u->new_disks[0]), | |
7164 | minor(u->new_disks[0])); | |
7165 | dprintf("imsm: new disk for reshape is: %i:%i " | |
7166 | "(%p, index = %i)\n", | |
7167 | major(u->new_disks[0]), | |
7168 | minor(u->new_disks[0]), | |
7169 | new_disk, new_disk->index); | |
7170 | if (new_disk == NULL) | |
7171 | goto error_disk_add; | |
7172 | ||
7173 | new_disk->index = map->num_members - 1; | |
7174 | /* slot to fill in autolayout | |
7175 | */ | |
7176 | new_disk->raiddisk = new_disk->index; | |
7177 | new_disk->disk.status |= CONFIGURED_DISK; | |
7178 | new_disk->disk.status &= ~SPARE_DISK; | |
7179 | } else | |
7180 | goto error_disk_add; | |
7181 | ||
7182 | skip_disk_add: | |
7183 | *tofree = *space_list; | |
7184 | /* calculate new size | |
7185 | */ | |
7186 | imsm_set_array_size(new_dev); | |
7187 | ||
7188 | ret_val = 1; | |
7189 | } | |
7190 | } | |
7191 | ||
7192 | if (tofree) | |
7193 | *space_list = tofree; | |
7194 | return ret_val; | |
7195 | ||
7196 | error_disk_add: | |
7197 | dprintf("Error: imsm: Cannot find disk.\n"); | |
7198 | return ret_val; | |
7199 | } | |
7200 | ||
061d7da3 LO |
7201 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
7202 | struct intel_super *super, | |
7203 | struct active_array *active_array) | |
7204 | { | |
7205 | struct imsm_super *mpb = super->anchor; | |
7206 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 7207 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
7208 | struct imsm_map *migr_map; |
7209 | struct active_array *a; | |
7210 | struct imsm_disk *disk; | |
7211 | __u8 to_state; | |
7212 | struct dl *dl; | |
7213 | unsigned int found; | |
7214 | int failed; | |
5961eeec | 7215 | int victim; |
061d7da3 | 7216 | int i; |
5961eeec | 7217 | int second_map_created = 0; |
061d7da3 | 7218 | |
5961eeec | 7219 | for (; u; u = u->next) { |
238c0a71 | 7220 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 7221 | |
5961eeec | 7222 | if (victim < 0) |
7223 | return 0; | |
061d7da3 | 7224 | |
5961eeec | 7225 | for (dl = super->disks; dl; dl = dl->next) |
7226 | if (dl == u->dl) | |
7227 | break; | |
061d7da3 | 7228 | |
5961eeec | 7229 | if (!dl) { |
7230 | fprintf(stderr, "error: imsm_activate_spare passed " | |
7231 | "an unknown disk (index: %d)\n", | |
7232 | u->dl->index); | |
7233 | return 0; | |
7234 | } | |
061d7da3 | 7235 | |
5961eeec | 7236 | /* count failures (excluding rebuilds and the victim) |
7237 | * to determine map[0] state | |
7238 | */ | |
7239 | failed = 0; | |
7240 | for (i = 0; i < map->num_members; i++) { | |
7241 | if (i == u->slot) | |
7242 | continue; | |
7243 | disk = get_imsm_disk(super, | |
238c0a71 | 7244 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 7245 | if (!disk || is_failed(disk)) |
7246 | failed++; | |
7247 | } | |
061d7da3 | 7248 | |
5961eeec | 7249 | /* adding a pristine spare, assign a new index */ |
7250 | if (dl->index < 0) { | |
7251 | dl->index = super->anchor->num_disks; | |
7252 | super->anchor->num_disks++; | |
7253 | } | |
7254 | disk = &dl->disk; | |
7255 | disk->status |= CONFIGURED_DISK; | |
7256 | disk->status &= ~SPARE_DISK; | |
7257 | ||
7258 | /* mark rebuild */ | |
238c0a71 | 7259 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 7260 | if (!second_map_created) { |
7261 | second_map_created = 1; | |
7262 | map->map_state = IMSM_T_STATE_DEGRADED; | |
7263 | migrate(dev, super, to_state, MIGR_REBUILD); | |
7264 | } else | |
7265 | map->map_state = to_state; | |
238c0a71 | 7266 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 7267 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
7268 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
7269 | dl->index | IMSM_ORD_REBUILD); | |
7270 | ||
7271 | /* update the family_num to mark a new container | |
7272 | * generation, being careful to record the existing | |
7273 | * family_num in orig_family_num to clean up after | |
7274 | * earlier mdadm versions that neglected to set it. | |
7275 | */ | |
7276 | if (mpb->orig_family_num == 0) | |
7277 | mpb->orig_family_num = mpb->family_num; | |
7278 | mpb->family_num += super->random; | |
7279 | ||
7280 | /* count arrays using the victim in the metadata */ | |
7281 | found = 0; | |
7282 | for (a = active_array; a ; a = a->next) { | |
7283 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7284 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 7285 | |
5961eeec | 7286 | if (get_imsm_disk_slot(map, victim) >= 0) |
7287 | found++; | |
7288 | } | |
061d7da3 | 7289 | |
5961eeec | 7290 | /* delete the victim if it is no longer being |
7291 | * utilized anywhere | |
061d7da3 | 7292 | */ |
5961eeec | 7293 | if (!found) { |
7294 | struct dl **dlp; | |
061d7da3 | 7295 | |
5961eeec | 7296 | /* We know that 'manager' isn't touching anything, |
7297 | * so it is safe to delete | |
7298 | */ | |
7299 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
7300 | if ((*dlp)->index == victim) |
7301 | break; | |
5961eeec | 7302 | |
7303 | /* victim may be on the missing list */ | |
7304 | if (!*dlp) | |
7305 | for (dlp = &super->missing; *dlp; | |
7306 | dlp = &(*dlp)->next) | |
7307 | if ((*dlp)->index == victim) | |
7308 | break; | |
7309 | imsm_delete(super, dlp, victim); | |
7310 | } | |
061d7da3 LO |
7311 | } |
7312 | ||
7313 | return 1; | |
7314 | } | |
a29911da | 7315 | |
2e5dc010 N |
7316 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
7317 | struct intel_super *super, | |
7318 | void ***space_list) | |
7319 | { | |
7320 | struct dl *new_disk; | |
7321 | struct intel_dev *id; | |
7322 | int i; | |
7323 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 7324 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
7325 | void **tofree = NULL; |
7326 | int devices_to_reshape = 1; | |
7327 | struct imsm_super *mpb = super->anchor; | |
7328 | int ret_val = 0; | |
d098291a | 7329 | unsigned int dev_id; |
2e5dc010 | 7330 | |
ed7333bd | 7331 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
7332 | |
7333 | /* enable spares to use in array */ | |
7334 | for (i = 0; i < delta_disks; i++) { | |
7335 | new_disk = get_disk_super(super, | |
7336 | major(u->new_disks[i]), | |
7337 | minor(u->new_disks[i])); | |
ed7333bd AK |
7338 | dprintf("imsm: new disk for reshape is: %i:%i " |
7339 | "(%p, index = %i)\n", | |
2e5dc010 N |
7340 | major(u->new_disks[i]), minor(u->new_disks[i]), |
7341 | new_disk, new_disk->index); | |
7342 | if ((new_disk == NULL) || | |
7343 | ((new_disk->index >= 0) && | |
7344 | (new_disk->index < u->old_raid_disks))) | |
7345 | goto update_reshape_exit; | |
ee4beede | 7346 | new_disk->index = disk_count++; |
2e5dc010 N |
7347 | /* slot to fill in autolayout |
7348 | */ | |
7349 | new_disk->raiddisk = new_disk->index; | |
7350 | new_disk->disk.status |= | |
7351 | CONFIGURED_DISK; | |
7352 | new_disk->disk.status &= ~SPARE_DISK; | |
7353 | } | |
7354 | ||
ed7333bd AK |
7355 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
7356 | mpb->num_raid_devs); | |
2e5dc010 N |
7357 | /* manage changes in volume |
7358 | */ | |
d098291a | 7359 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
7360 | void **sp = *space_list; |
7361 | struct imsm_dev *newdev; | |
7362 | struct imsm_map *newmap, *oldmap; | |
7363 | ||
d098291a AK |
7364 | for (id = super->devlist ; id; id = id->next) { |
7365 | if (id->index == dev_id) | |
7366 | break; | |
7367 | } | |
7368 | if (id == NULL) | |
7369 | break; | |
2e5dc010 N |
7370 | if (!sp) |
7371 | continue; | |
7372 | *space_list = *sp; | |
7373 | newdev = (void*)sp; | |
7374 | /* Copy the dev, but not (all of) the map */ | |
7375 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
7376 | oldmap = get_imsm_map(id->dev, MAP_0); |
7377 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
7378 | /* Copy the current map */ |
7379 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
7380 | /* update one device only | |
7381 | */ | |
7382 | if (devices_to_reshape) { | |
ed7333bd AK |
7383 | dprintf("imsm: modifying subdev: %i\n", |
7384 | id->index); | |
2e5dc010 N |
7385 | devices_to_reshape--; |
7386 | newdev->vol.migr_state = 1; | |
7387 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7388 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
7389 | newmap->num_members = u->new_raid_disks; |
7390 | for (i = 0; i < delta_disks; i++) { | |
7391 | set_imsm_ord_tbl_ent(newmap, | |
7392 | u->old_raid_disks + i, | |
7393 | u->old_raid_disks + i); | |
7394 | } | |
7395 | /* New map is correct, now need to save old map | |
7396 | */ | |
238c0a71 | 7397 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
7398 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
7399 | ||
70bdf0dc | 7400 | imsm_set_array_size(newdev); |
2e5dc010 N |
7401 | } |
7402 | ||
7403 | sp = (void **)id->dev; | |
7404 | id->dev = newdev; | |
7405 | *sp = tofree; | |
7406 | tofree = sp; | |
8e59f3d8 AK |
7407 | |
7408 | /* Clear migration record */ | |
7409 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 7410 | } |
819bc634 AK |
7411 | if (tofree) |
7412 | *space_list = tofree; | |
2e5dc010 N |
7413 | ret_val = 1; |
7414 | ||
7415 | update_reshape_exit: | |
7416 | ||
7417 | return ret_val; | |
7418 | } | |
7419 | ||
bb025c2f | 7420 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
7421 | struct intel_super *super, |
7422 | void ***space_list) | |
bb025c2f KW |
7423 | { |
7424 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
7425 | struct intel_dev *dv; |
7426 | struct imsm_dev *dev_new; | |
bb025c2f KW |
7427 | struct imsm_map *map; |
7428 | struct dl *dm, *du; | |
8ca6df95 | 7429 | int i; |
bb025c2f KW |
7430 | |
7431 | for (dv = super->devlist; dv; dv = dv->next) | |
7432 | if (dv->index == (unsigned int)u->subarray) { | |
7433 | dev = dv->dev; | |
7434 | break; | |
7435 | } | |
7436 | ||
7437 | if (dev == NULL) | |
7438 | return 0; | |
7439 | ||
238c0a71 | 7440 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
7441 | |
7442 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 7443 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
7444 | if (imsm_count_failed(super, dev, MAP_0) != |
7445 | (map->num_members / 2)) | |
43d5ec18 KW |
7446 | return 0; |
7447 | ||
bb025c2f KW |
7448 | /* iterate through devices to mark removed disks as spare */ |
7449 | for (dm = super->disks; dm; dm = dm->next) { | |
7450 | if (dm->disk.status & FAILED_DISK) { | |
7451 | int idx = dm->index; | |
7452 | /* update indexes on the disk list */ | |
7453 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
7454 | the index values will end up being correct.... NB */ | |
7455 | for (du = super->disks; du; du = du->next) | |
7456 | if (du->index > idx) | |
7457 | du->index--; | |
7458 | /* mark as spare disk */ | |
a8619d23 | 7459 | mark_spare(dm); |
bb025c2f KW |
7460 | } |
7461 | } | |
bb025c2f KW |
7462 | /* update map */ |
7463 | map->num_members = map->num_members / 2; | |
7464 | map->map_state = IMSM_T_STATE_NORMAL; | |
7465 | map->num_domains = 1; | |
7466 | map->raid_level = 0; | |
7467 | map->failed_disk_num = -1; | |
7468 | } | |
7469 | ||
8ca6df95 KW |
7470 | if (u->direction == R0_TO_R10) { |
7471 | void **space; | |
7472 | /* update slots in current disk list */ | |
7473 | for (dm = super->disks; dm; dm = dm->next) { | |
7474 | if (dm->index >= 0) | |
7475 | dm->index *= 2; | |
7476 | } | |
7477 | /* create new *missing* disks */ | |
7478 | for (i = 0; i < map->num_members; i++) { | |
7479 | space = *space_list; | |
7480 | if (!space) | |
7481 | continue; | |
7482 | *space_list = *space; | |
7483 | du = (void *)space; | |
7484 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
7485 | du->fd = -1; |
7486 | du->minor = 0; | |
7487 | du->major = 0; | |
7488 | du->index = (i * 2) + 1; | |
7489 | sprintf((char *)du->disk.serial, | |
7490 | " MISSING_%d", du->index); | |
7491 | sprintf((char *)du->serial, | |
7492 | "MISSING_%d", du->index); | |
7493 | du->next = super->missing; | |
7494 | super->missing = du; | |
7495 | } | |
7496 | /* create new dev and map */ | |
7497 | space = *space_list; | |
7498 | if (!space) | |
7499 | return 0; | |
7500 | *space_list = *space; | |
7501 | dev_new = (void *)space; | |
7502 | memcpy(dev_new, dev, sizeof(*dev)); | |
7503 | /* update new map */ | |
238c0a71 | 7504 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 7505 | map->num_members = map->num_members * 2; |
1a2487c2 | 7506 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
7507 | map->num_domains = 2; |
7508 | map->raid_level = 1; | |
7509 | /* replace dev<->dev_new */ | |
7510 | dv->dev = dev_new; | |
7511 | } | |
bb025c2f KW |
7512 | /* update disk order table */ |
7513 | for (du = super->disks; du; du = du->next) | |
7514 | if (du->index >= 0) | |
7515 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 7516 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
7517 | if (du->index >= 0) { |
7518 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 7519 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 7520 | } |
bb025c2f KW |
7521 | |
7522 | return 1; | |
7523 | } | |
7524 | ||
e8319a19 DW |
7525 | static void imsm_process_update(struct supertype *st, |
7526 | struct metadata_update *update) | |
7527 | { | |
7528 | /** | |
7529 | * crack open the metadata_update envelope to find the update record | |
7530 | * update can be one of: | |
d195167d AK |
7531 | * update_reshape_container_disks - all the arrays in the container |
7532 | * are being reshaped to have more devices. We need to mark | |
7533 | * the arrays for general migration and convert selected spares | |
7534 | * into active devices. | |
7535 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
7536 | * device in an array, update the disk_ord_tbl. If this disk is |
7537 | * present in all member arrays then also clear the SPARE_DISK | |
7538 | * flag | |
d195167d AK |
7539 | * update_create_array |
7540 | * update_kill_array | |
7541 | * update_rename_array | |
7542 | * update_add_remove_disk | |
e8319a19 DW |
7543 | */ |
7544 | struct intel_super *super = st->sb; | |
4d7b1503 | 7545 | struct imsm_super *mpb; |
e8319a19 DW |
7546 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
7547 | ||
4d7b1503 DW |
7548 | /* update requires a larger buf but the allocation failed */ |
7549 | if (super->next_len && !super->next_buf) { | |
7550 | super->next_len = 0; | |
7551 | return; | |
7552 | } | |
7553 | ||
7554 | if (super->next_buf) { | |
7555 | memcpy(super->next_buf, super->buf, super->len); | |
7556 | free(super->buf); | |
7557 | super->len = super->next_len; | |
7558 | super->buf = super->next_buf; | |
7559 | ||
7560 | super->next_len = 0; | |
7561 | super->next_buf = NULL; | |
7562 | } | |
7563 | ||
7564 | mpb = super->anchor; | |
7565 | ||
e8319a19 | 7566 | switch (type) { |
0ec5d470 AK |
7567 | case update_general_migration_checkpoint: { |
7568 | struct intel_dev *id; | |
7569 | struct imsm_update_general_migration_checkpoint *u = | |
7570 | (void *)update->buf; | |
7571 | ||
7572 | dprintf("imsm: process_update() " | |
7573 | "for update_general_migration_checkpoint called\n"); | |
7574 | ||
7575 | /* find device under general migration */ | |
7576 | for (id = super->devlist ; id; id = id->next) { | |
7577 | if (is_gen_migration(id->dev)) { | |
7578 | id->dev->vol.curr_migr_unit = | |
7579 | __cpu_to_le32(u->curr_migr_unit); | |
7580 | super->updates_pending++; | |
7581 | } | |
7582 | } | |
7583 | break; | |
7584 | } | |
bb025c2f KW |
7585 | case update_takeover: { |
7586 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
7587 | if (apply_takeover_update(u, super, &update->space_list)) { |
7588 | imsm_update_version_info(super); | |
bb025c2f | 7589 | super->updates_pending++; |
1a2487c2 | 7590 | } |
bb025c2f KW |
7591 | break; |
7592 | } | |
7593 | ||
78b10e66 | 7594 | case update_reshape_container_disks: { |
d195167d | 7595 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
7596 | if (apply_reshape_container_disks_update( |
7597 | u, super, &update->space_list)) | |
7598 | super->updates_pending++; | |
78b10e66 N |
7599 | break; |
7600 | } | |
48c5303a | 7601 | case update_reshape_migration: { |
a29911da PC |
7602 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
7603 | if (apply_reshape_migration_update( | |
7604 | u, super, &update->space_list)) | |
7605 | super->updates_pending++; | |
48c5303a PC |
7606 | break; |
7607 | } | |
e8319a19 DW |
7608 | case update_activate_spare: { |
7609 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
061d7da3 LO |
7610 | if (apply_update_activate_spare(u, super, st->arrays)) |
7611 | super->updates_pending++; | |
8273f55e DW |
7612 | break; |
7613 | } | |
7614 | case update_create_array: { | |
7615 | /* someone wants to create a new array, we need to be aware of | |
7616 | * a few races/collisions: | |
7617 | * 1/ 'Create' called by two separate instances of mdadm | |
7618 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
7619 | * devices that have since been assimilated via | |
7620 | * activate_spare. | |
7621 | * In the event this update can not be carried out mdadm will | |
7622 | * (FIX ME) notice that its update did not take hold. | |
7623 | */ | |
7624 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 7625 | struct intel_dev *dv; |
8273f55e DW |
7626 | struct imsm_dev *dev; |
7627 | struct imsm_map *map, *new_map; | |
7628 | unsigned long long start, end; | |
7629 | unsigned long long new_start, new_end; | |
7630 | int i; | |
54c2c1ea DW |
7631 | struct disk_info *inf; |
7632 | struct dl *dl; | |
8273f55e DW |
7633 | |
7634 | /* handle racing creates: first come first serve */ | |
7635 | if (u->dev_idx < mpb->num_raid_devs) { | |
7636 | dprintf("%s: subarray %d already defined\n", | |
7637 | __func__, u->dev_idx); | |
ba2de7ba | 7638 | goto create_error; |
8273f55e DW |
7639 | } |
7640 | ||
7641 | /* check update is next in sequence */ | |
7642 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
7643 | dprintf("%s: can not create array %d expected index %d\n", |
7644 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 7645 | goto create_error; |
8273f55e DW |
7646 | } |
7647 | ||
238c0a71 | 7648 | new_map = get_imsm_map(&u->dev, MAP_0); |
8273f55e DW |
7649 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
7650 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
54c2c1ea | 7651 | inf = get_disk_info(u); |
8273f55e DW |
7652 | |
7653 | /* handle activate_spare versus create race: | |
7654 | * check to make sure that overlapping arrays do not include | |
7655 | * overalpping disks | |
7656 | */ | |
7657 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 7658 | dev = get_imsm_dev(super, i); |
238c0a71 | 7659 | map = get_imsm_map(dev, MAP_0); |
8273f55e DW |
7660 | start = __le32_to_cpu(map->pba_of_lba0); |
7661 | end = start + __le32_to_cpu(map->blocks_per_member); | |
7662 | if ((new_start >= start && new_start <= end) || | |
7663 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
7664 | /* overlap */; |
7665 | else | |
7666 | continue; | |
7667 | ||
7668 | if (disks_overlap(super, i, u)) { | |
8273f55e | 7669 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 7670 | goto create_error; |
8273f55e DW |
7671 | } |
7672 | } | |
8273f55e | 7673 | |
949c47a0 DW |
7674 | /* check that prepare update was successful */ |
7675 | if (!update->space) { | |
7676 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 7677 | goto create_error; |
949c47a0 DW |
7678 | } |
7679 | ||
54c2c1ea DW |
7680 | /* check that all disks are still active before committing |
7681 | * changes. FIXME: could we instead handle this by creating a | |
7682 | * degraded array? That's probably not what the user expects, | |
7683 | * so better to drop this update on the floor. | |
7684 | */ | |
7685 | for (i = 0; i < new_map->num_members; i++) { | |
7686 | dl = serial_to_dl(inf[i].serial, super); | |
7687 | if (!dl) { | |
7688 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 7689 | goto create_error; |
54c2c1ea | 7690 | } |
949c47a0 DW |
7691 | } |
7692 | ||
8273f55e | 7693 | super->updates_pending++; |
54c2c1ea DW |
7694 | |
7695 | /* convert spares to members and fixup ord_tbl */ | |
7696 | for (i = 0; i < new_map->num_members; i++) { | |
7697 | dl = serial_to_dl(inf[i].serial, super); | |
7698 | if (dl->index == -1) { | |
7699 | dl->index = mpb->num_disks; | |
7700 | mpb->num_disks++; | |
7701 | dl->disk.status |= CONFIGURED_DISK; | |
7702 | dl->disk.status &= ~SPARE_DISK; | |
7703 | } | |
7704 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
7705 | } | |
7706 | ||
ba2de7ba DW |
7707 | dv = update->space; |
7708 | dev = dv->dev; | |
949c47a0 DW |
7709 | update->space = NULL; |
7710 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
7711 | dv->index = u->dev_idx; |
7712 | dv->next = super->devlist; | |
7713 | super->devlist = dv; | |
8273f55e | 7714 | mpb->num_raid_devs++; |
8273f55e | 7715 | |
4d1313e9 | 7716 | imsm_update_version_info(super); |
8273f55e | 7717 | break; |
ba2de7ba DW |
7718 | create_error: |
7719 | /* mdmon knows how to release update->space, but not | |
7720 | * ((struct intel_dev *) update->space)->dev | |
7721 | */ | |
7722 | if (update->space) { | |
7723 | dv = update->space; | |
7724 | free(dv->dev); | |
7725 | } | |
8273f55e | 7726 | break; |
e8319a19 | 7727 | } |
33414a01 DW |
7728 | case update_kill_array: { |
7729 | struct imsm_update_kill_array *u = (void *) update->buf; | |
7730 | int victim = u->dev_idx; | |
7731 | struct active_array *a; | |
7732 | struct intel_dev **dp; | |
7733 | struct imsm_dev *dev; | |
7734 | ||
7735 | /* sanity check that we are not affecting the uuid of | |
7736 | * active arrays, or deleting an active array | |
7737 | * | |
7738 | * FIXME when immutable ids are available, but note that | |
7739 | * we'll also need to fixup the invalidated/active | |
7740 | * subarray indexes in mdstat | |
7741 | */ | |
7742 | for (a = st->arrays; a; a = a->next) | |
7743 | if (a->info.container_member >= victim) | |
7744 | break; | |
7745 | /* by definition if mdmon is running at least one array | |
7746 | * is active in the container, so checking | |
7747 | * mpb->num_raid_devs is just extra paranoia | |
7748 | */ | |
7749 | dev = get_imsm_dev(super, victim); | |
7750 | if (a || !dev || mpb->num_raid_devs == 1) { | |
7751 | dprintf("failed to delete subarray-%d\n", victim); | |
7752 | break; | |
7753 | } | |
7754 | ||
7755 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 7756 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
7757 | *dp = (*dp)->next; |
7758 | } else { | |
f21e18ca | 7759 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
7760 | (*dp)->index--; |
7761 | dp = &(*dp)->next; | |
7762 | } | |
7763 | mpb->num_raid_devs--; | |
7764 | super->updates_pending++; | |
7765 | break; | |
7766 | } | |
aa534678 DW |
7767 | case update_rename_array: { |
7768 | struct imsm_update_rename_array *u = (void *) update->buf; | |
7769 | char name[MAX_RAID_SERIAL_LEN+1]; | |
7770 | int target = u->dev_idx; | |
7771 | struct active_array *a; | |
7772 | struct imsm_dev *dev; | |
7773 | ||
7774 | /* sanity check that we are not affecting the uuid of | |
7775 | * an active array | |
7776 | */ | |
7777 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
7778 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
7779 | for (a = st->arrays; a; a = a->next) | |
7780 | if (a->info.container_member == target) | |
7781 | break; | |
7782 | dev = get_imsm_dev(super, u->dev_idx); | |
7783 | if (a || !dev || !check_name(super, name, 1)) { | |
7784 | dprintf("failed to rename subarray-%d\n", target); | |
7785 | break; | |
7786 | } | |
7787 | ||
cdbe98cd | 7788 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
7789 | super->updates_pending++; |
7790 | break; | |
7791 | } | |
1a64be56 | 7792 | case update_add_remove_disk: { |
43dad3d6 | 7793 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
7794 | * being added, check teh status of add_remove_disk |
7795 | * if discs has been added. | |
7796 | */ | |
7797 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 7798 | struct active_array *a; |
072b727f DW |
7799 | |
7800 | super->updates_pending++; | |
1a64be56 | 7801 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
7802 | a->check_degraded = 1; |
7803 | } | |
43dad3d6 | 7804 | break; |
e8319a19 | 7805 | } |
1a64be56 LM |
7806 | default: |
7807 | fprintf(stderr, "error: unsuported process update type:" | |
7808 | "(type: %d)\n", type); | |
7809 | } | |
e8319a19 | 7810 | } |
88758e9d | 7811 | |
bc0b9d34 PC |
7812 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
7813 | ||
8273f55e DW |
7814 | static void imsm_prepare_update(struct supertype *st, |
7815 | struct metadata_update *update) | |
7816 | { | |
949c47a0 | 7817 | /** |
4d7b1503 DW |
7818 | * Allocate space to hold new disk entries, raid-device entries or a new |
7819 | * mpb if necessary. The manager synchronously waits for updates to | |
7820 | * complete in the monitor, so new mpb buffers allocated here can be | |
7821 | * integrated by the monitor thread without worrying about live pointers | |
7822 | * in the manager thread. | |
8273f55e | 7823 | */ |
949c47a0 | 7824 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
7825 | struct intel_super *super = st->sb; |
7826 | struct imsm_super *mpb = super->anchor; | |
7827 | size_t buf_len; | |
7828 | size_t len = 0; | |
949c47a0 DW |
7829 | |
7830 | switch (type) { | |
0ec5d470 AK |
7831 | case update_general_migration_checkpoint: |
7832 | dprintf("imsm: prepare_update() " | |
7833 | "for update_general_migration_checkpoint called\n"); | |
7834 | break; | |
abedf5fc KW |
7835 | case update_takeover: { |
7836 | struct imsm_update_takeover *u = (void *)update->buf; | |
7837 | if (u->direction == R0_TO_R10) { | |
7838 | void **tail = (void **)&update->space_list; | |
7839 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 7840 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
7841 | int num_members = map->num_members; |
7842 | void *space; | |
7843 | int size, i; | |
7844 | int err = 0; | |
7845 | /* allocate memory for added disks */ | |
7846 | for (i = 0; i < num_members; i++) { | |
7847 | size = sizeof(struct dl); | |
7848 | space = malloc(size); | |
7849 | if (!space) { | |
7850 | err++; | |
7851 | break; | |
7852 | } | |
7853 | *tail = space; | |
7854 | tail = space; | |
7855 | *tail = NULL; | |
7856 | } | |
7857 | /* allocate memory for new device */ | |
7858 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
7859 | (num_members * sizeof(__u32)); | |
7860 | space = malloc(size); | |
7861 | if (!space) | |
7862 | err++; | |
7863 | else { | |
7864 | *tail = space; | |
7865 | tail = space; | |
7866 | *tail = NULL; | |
7867 | } | |
7868 | if (!err) { | |
7869 | len = disks_to_mpb_size(num_members * 2); | |
7870 | } else { | |
7871 | /* if allocation didn't success, free buffer */ | |
7872 | while (update->space_list) { | |
7873 | void **sp = update->space_list; | |
7874 | update->space_list = *sp; | |
7875 | free(sp); | |
7876 | } | |
7877 | } | |
7878 | } | |
7879 | ||
7880 | break; | |
7881 | } | |
78b10e66 | 7882 | case update_reshape_container_disks: { |
d195167d AK |
7883 | /* Every raid device in the container is about to |
7884 | * gain some more devices, and we will enter a | |
7885 | * reconfiguration. | |
7886 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
7887 | * will now hold 2 of them. | |
7888 | * Thus we need new 'struct imsm_dev' allocations sized | |
7889 | * as sizeof_imsm_dev but with more devices in both maps. | |
7890 | */ | |
7891 | struct imsm_update_reshape *u = (void *)update->buf; | |
7892 | struct intel_dev *dl; | |
7893 | void **space_tail = (void**)&update->space_list; | |
7894 | ||
7895 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7896 | ||
7897 | for (dl = super->devlist; dl; dl = dl->next) { | |
7898 | int size = sizeof_imsm_dev(dl->dev, 1); | |
7899 | void *s; | |
d677e0b8 AK |
7900 | if (u->new_raid_disks > u->old_raid_disks) |
7901 | size += sizeof(__u32)*2* | |
7902 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
7903 | s = malloc(size); |
7904 | if (!s) | |
7905 | break; | |
7906 | *space_tail = s; | |
7907 | space_tail = s; | |
7908 | *space_tail = NULL; | |
7909 | } | |
7910 | ||
7911 | len = disks_to_mpb_size(u->new_raid_disks); | |
7912 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
7913 | break; |
7914 | } | |
48c5303a | 7915 | case update_reshape_migration: { |
bc0b9d34 PC |
7916 | /* for migration level 0->5 we need to add disks |
7917 | * so the same as for container operation we will copy | |
7918 | * device to the bigger location. | |
7919 | * in memory prepared device and new disk area are prepared | |
7920 | * for usage in process update | |
7921 | */ | |
7922 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
7923 | struct intel_dev *id; | |
7924 | void **space_tail = (void **)&update->space_list; | |
7925 | int size; | |
7926 | void *s; | |
7927 | int current_level = -1; | |
7928 | ||
7929 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7930 | ||
7931 | /* add space for bigger array in update | |
7932 | */ | |
7933 | for (id = super->devlist; id; id = id->next) { | |
7934 | if (id->index == (unsigned)u->subdev) { | |
7935 | size = sizeof_imsm_dev(id->dev, 1); | |
7936 | if (u->new_raid_disks > u->old_raid_disks) | |
7937 | size += sizeof(__u32)*2* | |
7938 | (u->new_raid_disks - u->old_raid_disks); | |
7939 | s = malloc(size); | |
7940 | if (!s) | |
7941 | break; | |
7942 | *space_tail = s; | |
7943 | space_tail = s; | |
7944 | *space_tail = NULL; | |
7945 | break; | |
7946 | } | |
7947 | } | |
7948 | if (update->space_list == NULL) | |
7949 | break; | |
7950 | ||
7951 | /* add space for disk in update | |
7952 | */ | |
7953 | size = sizeof(struct dl); | |
7954 | s = malloc(size); | |
7955 | if (!s) { | |
7956 | free(update->space_list); | |
7957 | update->space_list = NULL; | |
7958 | break; | |
7959 | } | |
7960 | *space_tail = s; | |
7961 | space_tail = s; | |
7962 | *space_tail = NULL; | |
7963 | ||
7964 | /* add spare device to update | |
7965 | */ | |
7966 | for (id = super->devlist ; id; id = id->next) | |
7967 | if (id->index == (unsigned)u->subdev) { | |
7968 | struct imsm_dev *dev; | |
7969 | struct imsm_map *map; | |
7970 | ||
7971 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 7972 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
7973 | current_level = map->raid_level; |
7974 | break; | |
7975 | } | |
7976 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
7977 | struct mdinfo *spares; | |
7978 | ||
7979 | spares = get_spares_for_grow(st); | |
7980 | if (spares) { | |
7981 | struct dl *dl; | |
7982 | struct mdinfo *dev; | |
7983 | ||
7984 | dev = spares->devs; | |
7985 | if (dev) { | |
7986 | u->new_disks[0] = | |
7987 | makedev(dev->disk.major, | |
7988 | dev->disk.minor); | |
7989 | dl = get_disk_super(super, | |
7990 | dev->disk.major, | |
7991 | dev->disk.minor); | |
7992 | dl->index = u->old_raid_disks; | |
7993 | dev = dev->next; | |
7994 | } | |
7995 | sysfs_free(spares); | |
7996 | } | |
7997 | } | |
7998 | len = disks_to_mpb_size(u->new_raid_disks); | |
7999 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8000 | break; |
8001 | } | |
949c47a0 DW |
8002 | case update_create_array: { |
8003 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8004 | struct intel_dev *dv; |
54c2c1ea | 8005 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8006 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8007 | struct dl *dl; |
8008 | struct disk_info *inf; | |
8009 | int i; | |
8010 | int activate = 0; | |
949c47a0 | 8011 | |
54c2c1ea DW |
8012 | inf = get_disk_info(u); |
8013 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
8014 | /* allocate a new super->devlist entry */ |
8015 | dv = malloc(sizeof(*dv)); | |
8016 | if (dv) { | |
8017 | dv->dev = malloc(len); | |
8018 | if (dv->dev) | |
8019 | update->space = dv; | |
8020 | else { | |
8021 | free(dv); | |
8022 | update->space = NULL; | |
8023 | } | |
8024 | } | |
949c47a0 | 8025 | |
54c2c1ea DW |
8026 | /* count how many spares will be converted to members */ |
8027 | for (i = 0; i < map->num_members; i++) { | |
8028 | dl = serial_to_dl(inf[i].serial, super); | |
8029 | if (!dl) { | |
8030 | /* hmm maybe it failed?, nothing we can do about | |
8031 | * it here | |
8032 | */ | |
8033 | continue; | |
8034 | } | |
8035 | if (count_memberships(dl, super) == 0) | |
8036 | activate++; | |
8037 | } | |
8038 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
8039 | break; |
8040 | default: | |
8041 | break; | |
8042 | } | |
8043 | } | |
8273f55e | 8044 | |
4d7b1503 DW |
8045 | /* check if we need a larger metadata buffer */ |
8046 | if (super->next_buf) | |
8047 | buf_len = super->next_len; | |
8048 | else | |
8049 | buf_len = super->len; | |
8050 | ||
8051 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
8052 | /* ok we need a larger buf than what is currently allocated | |
8053 | * if this allocation fails process_update will notice that | |
8054 | * ->next_len is set and ->next_buf is NULL | |
8055 | */ | |
8056 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
8057 | if (super->next_buf) | |
8058 | free(super->next_buf); | |
8059 | ||
8060 | super->next_len = buf_len; | |
1f45a8ad DW |
8061 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
8062 | memset(super->next_buf, 0, buf_len); | |
8063 | else | |
4d7b1503 DW |
8064 | super->next_buf = NULL; |
8065 | } | |
8273f55e DW |
8066 | } |
8067 | ||
ae6aad82 | 8068 | /* must be called while manager is quiesced */ |
f21e18ca | 8069 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
8070 | { |
8071 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
8072 | struct dl *iter; |
8073 | struct imsm_dev *dev; | |
8074 | struct imsm_map *map; | |
24565c9a DW |
8075 | int i, j, num_members; |
8076 | __u32 ord; | |
ae6aad82 | 8077 | |
24565c9a DW |
8078 | dprintf("%s: deleting device[%d] from imsm_super\n", |
8079 | __func__, index); | |
ae6aad82 DW |
8080 | |
8081 | /* shift all indexes down one */ | |
8082 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 8083 | if (iter->index > (int)index) |
ae6aad82 | 8084 | iter->index--; |
47ee5a45 | 8085 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 8086 | if (iter->index > (int)index) |
47ee5a45 | 8087 | iter->index--; |
ae6aad82 DW |
8088 | |
8089 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8090 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8091 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
8092 | num_members = map->num_members; |
8093 | for (j = 0; j < num_members; j++) { | |
8094 | /* update ord entries being careful not to propagate | |
8095 | * ord-flags to the first map | |
8096 | */ | |
238c0a71 | 8097 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 8098 | |
24565c9a DW |
8099 | if (ord_to_idx(ord) <= index) |
8100 | continue; | |
ae6aad82 | 8101 | |
238c0a71 | 8102 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 8103 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 8104 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
8105 | if (map) |
8106 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
8107 | } |
8108 | } | |
8109 | ||
8110 | mpb->num_disks--; | |
8111 | super->updates_pending++; | |
24565c9a DW |
8112 | if (*dlp) { |
8113 | struct dl *dl = *dlp; | |
8114 | ||
8115 | *dlp = (*dlp)->next; | |
8116 | __free_imsm_disk(dl); | |
8117 | } | |
ae6aad82 | 8118 | } |
9e2d750d | 8119 | #endif /* MDASSEMBLE */ |
9a717282 AK |
8120 | |
8121 | static void close_targets(int *targets, int new_disks) | |
8122 | { | |
8123 | int i; | |
8124 | ||
8125 | if (!targets) | |
8126 | return; | |
8127 | ||
8128 | for (i = 0; i < new_disks; i++) { | |
8129 | if (targets[i] >= 0) { | |
8130 | close(targets[i]); | |
8131 | targets[i] = -1; | |
8132 | } | |
8133 | } | |
8134 | } | |
8135 | ||
8136 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
8137 | struct intel_super *super, | |
8138 | struct imsm_dev *dev) | |
8139 | { | |
8140 | switch (level) { | |
8141 | case 10:{ | |
8142 | int ret_val = 0; | |
8143 | struct imsm_map *map; | |
8144 | int i; | |
8145 | ||
8146 | ret_val = raid_disks/2; | |
8147 | /* check map if all disks pairs not failed | |
8148 | * in both maps | |
8149 | */ | |
238c0a71 | 8150 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
8151 | for (i = 0; i < ret_val; i++) { |
8152 | int degradation = 0; | |
8153 | if (get_imsm_disk(super, i) == NULL) | |
8154 | degradation++; | |
8155 | if (get_imsm_disk(super, i + 1) == NULL) | |
8156 | degradation++; | |
8157 | if (degradation == 2) | |
8158 | return 0; | |
8159 | } | |
238c0a71 | 8160 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
8161 | /* if there is no second map |
8162 | * result can be returned | |
8163 | */ | |
8164 | if (map == NULL) | |
8165 | return ret_val; | |
8166 | /* check degradation in second map | |
8167 | */ | |
8168 | for (i = 0; i < ret_val; i++) { | |
8169 | int degradation = 0; | |
8170 | if (get_imsm_disk(super, i) == NULL) | |
8171 | degradation++; | |
8172 | if (get_imsm_disk(super, i + 1) == NULL) | |
8173 | degradation++; | |
8174 | if (degradation == 2) | |
8175 | return 0; | |
8176 | } | |
8177 | return ret_val; | |
8178 | } | |
8179 | case 5: | |
8180 | return 1; | |
8181 | case 6: | |
8182 | return 2; | |
8183 | default: | |
8184 | return 0; | |
8185 | } | |
8186 | } | |
8187 | ||
8188 | ||
687629c2 AK |
8189 | /******************************************************************************* |
8190 | * Function: open_backup_targets | |
8191 | * Description: Function opens file descriptors for all devices given in | |
8192 | * info->devs | |
8193 | * Parameters: | |
8194 | * info : general array info | |
8195 | * raid_disks : number of disks | |
8196 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
8197 | * super : intel super for raid10 degradation check |
8198 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
8199 | * Returns: |
8200 | * 0 : success | |
8201 | * -1 : fail | |
8202 | ******************************************************************************/ | |
9a717282 AK |
8203 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
8204 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
8205 | { |
8206 | struct mdinfo *sd; | |
f627f5ad | 8207 | int i; |
9a717282 | 8208 | int opened = 0; |
f627f5ad AK |
8209 | |
8210 | for (i = 0; i < raid_disks; i++) | |
8211 | raid_fds[i] = -1; | |
687629c2 AK |
8212 | |
8213 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8214 | char *dn; | |
8215 | ||
8216 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
8217 | dprintf("disk is faulty!!\n"); | |
8218 | continue; | |
8219 | } | |
8220 | ||
8221 | if ((sd->disk.raid_disk >= raid_disks) || | |
8222 | (sd->disk.raid_disk < 0)) | |
8223 | continue; | |
8224 | ||
8225 | dn = map_dev(sd->disk.major, | |
8226 | sd->disk.minor, 1); | |
8227 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
8228 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
8229 | fprintf(stderr, "cannot open component\n"); | |
9a717282 | 8230 | continue; |
687629c2 | 8231 | } |
9a717282 AK |
8232 | opened++; |
8233 | } | |
8234 | /* check if maximum array degradation level is not exceeded | |
8235 | */ | |
8236 | if ((raid_disks - opened) > | |
8237 | imsm_get_allowed_degradation(info->new_level, | |
8238 | raid_disks, | |
8239 | super, dev)) { | |
8240 | fprintf(stderr, "Not enough disks can be opened.\n"); | |
8241 | close_targets(raid_fds, raid_disks); | |
8242 | return -2; | |
687629c2 AK |
8243 | } |
8244 | return 0; | |
8245 | } | |
8246 | ||
9e2d750d | 8247 | #ifndef MDASSEMBLE |
687629c2 AK |
8248 | /******************************************************************************* |
8249 | * Function: init_migr_record_imsm | |
8250 | * Description: Function inits imsm migration record | |
8251 | * Parameters: | |
8252 | * super : imsm internal array info | |
8253 | * dev : device under migration | |
8254 | * info : general array info to find the smallest device | |
8255 | * Returns: | |
8256 | * none | |
8257 | ******************************************************************************/ | |
8258 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
8259 | struct mdinfo *info) | |
8260 | { | |
8261 | struct intel_super *super = st->sb; | |
8262 | struct migr_record *migr_rec = super->migr_rec; | |
8263 | int new_data_disks; | |
8264 | unsigned long long dsize, dev_sectors; | |
8265 | long long unsigned min_dev_sectors = -1LLU; | |
8266 | struct mdinfo *sd; | |
8267 | char nm[30]; | |
8268 | int fd; | |
238c0a71 AK |
8269 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
8270 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 8271 | unsigned long long num_migr_units; |
3ef4403c | 8272 | unsigned long long array_blocks; |
687629c2 AK |
8273 | |
8274 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
8275 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
8276 | ||
8277 | /* only ascending reshape supported now */ | |
8278 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
8279 | ||
8280 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
8281 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
8282 | migr_rec->dest_depth_per_unit *= map_dest->blocks_per_strip; | |
238c0a71 | 8283 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
8284 | migr_rec->blocks_per_unit = |
8285 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
8286 | migr_rec->dest_depth_per_unit = | |
8287 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 8288 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
8289 | num_migr_units = |
8290 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
8291 | ||
8292 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
8293 | num_migr_units++; | |
8294 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
8295 | ||
8296 | migr_rec->post_migr_vol_cap = dev->size_low; | |
8297 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
8298 | ||
8299 | ||
8300 | /* Find the smallest dev */ | |
8301 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8302 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
8303 | fd = dev_open(nm, O_RDONLY); | |
8304 | if (fd < 0) | |
8305 | continue; | |
8306 | get_dev_size(fd, NULL, &dsize); | |
8307 | dev_sectors = dsize / 512; | |
8308 | if (dev_sectors < min_dev_sectors) | |
8309 | min_dev_sectors = dev_sectors; | |
8310 | close(fd); | |
8311 | } | |
8312 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
8313 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
8314 | ||
8315 | write_imsm_migr_rec(st); | |
8316 | ||
8317 | return; | |
8318 | } | |
8319 | ||
8320 | /******************************************************************************* | |
8321 | * Function: save_backup_imsm | |
8322 | * Description: Function saves critical data stripes to Migration Copy Area | |
8323 | * and updates the current migration unit status. | |
8324 | * Use restore_stripes() to form a destination stripe, | |
8325 | * and to write it to the Copy Area. | |
8326 | * Parameters: | |
8327 | * st : supertype information | |
aea93171 | 8328 | * dev : imsm device that backup is saved for |
687629c2 AK |
8329 | * info : general array info |
8330 | * buf : input buffer | |
687629c2 AK |
8331 | * length : length of data to backup (blocks_per_unit) |
8332 | * Returns: | |
8333 | * 0 : success | |
8334 | *, -1 : fail | |
8335 | ******************************************************************************/ | |
8336 | int save_backup_imsm(struct supertype *st, | |
8337 | struct imsm_dev *dev, | |
8338 | struct mdinfo *info, | |
8339 | void *buf, | |
687629c2 AK |
8340 | int length) |
8341 | { | |
8342 | int rv = -1; | |
8343 | struct intel_super *super = st->sb; | |
8344 | unsigned long long *target_offsets = NULL; | |
8345 | int *targets = NULL; | |
8346 | int i; | |
238c0a71 | 8347 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 8348 | int new_disks = map_dest->num_members; |
ab724b98 AK |
8349 | int dest_layout = 0; |
8350 | int dest_chunk; | |
d1877f69 | 8351 | unsigned long long start; |
238c0a71 | 8352 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
8353 | |
8354 | targets = malloc(new_disks * sizeof(int)); | |
8355 | if (!targets) | |
8356 | goto abort; | |
8357 | ||
7e45b550 AK |
8358 | for (i = 0; i < new_disks; i++) |
8359 | targets[i] = -1; | |
8360 | ||
687629c2 AK |
8361 | target_offsets = malloc(new_disks * sizeof(unsigned long long)); |
8362 | if (!target_offsets) | |
8363 | goto abort; | |
8364 | ||
d1877f69 | 8365 | start = info->reshape_progress * 512; |
687629c2 | 8366 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
8367 | target_offsets[i] = (unsigned long long) |
8368 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
8369 | /* move back copy area adderss, it will be moved forward |
8370 | * in restore_stripes() using start input variable | |
8371 | */ | |
8372 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
8373 | } |
8374 | ||
9a717282 AK |
8375 | if (open_backup_targets(info, new_disks, targets, |
8376 | super, dev)) | |
687629c2 AK |
8377 | goto abort; |
8378 | ||
68eb8bc6 | 8379 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
8380 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
8381 | ||
687629c2 AK |
8382 | if (restore_stripes(targets, /* list of dest devices */ |
8383 | target_offsets, /* migration record offsets */ | |
8384 | new_disks, | |
ab724b98 AK |
8385 | dest_chunk, |
8386 | map_dest->raid_level, | |
8387 | dest_layout, | |
8388 | -1, /* source backup file descriptor */ | |
8389 | 0, /* input buf offset | |
8390 | * always 0 buf is already offseted */ | |
d1877f69 | 8391 | start, |
687629c2 AK |
8392 | length, |
8393 | buf) != 0) { | |
8394 | fprintf(stderr, Name ": Error restoring stripes\n"); | |
8395 | goto abort; | |
8396 | } | |
8397 | ||
8398 | rv = 0; | |
8399 | ||
8400 | abort: | |
8401 | if (targets) { | |
9a717282 | 8402 | close_targets(targets, new_disks); |
687629c2 AK |
8403 | free(targets); |
8404 | } | |
8405 | free(target_offsets); | |
8406 | ||
8407 | return rv; | |
8408 | } | |
8409 | ||
8410 | /******************************************************************************* | |
8411 | * Function: save_checkpoint_imsm | |
8412 | * Description: Function called for current unit status update | |
8413 | * in the migration record. It writes it to disk. | |
8414 | * Parameters: | |
8415 | * super : imsm internal array info | |
8416 | * info : general array info | |
8417 | * Returns: | |
8418 | * 0: success | |
8419 | * 1: failure | |
0228d92c AK |
8420 | * 2: failure, means no valid migration record |
8421 | * / no general migration in progress / | |
687629c2 AK |
8422 | ******************************************************************************/ |
8423 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
8424 | { | |
8425 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
8426 | unsigned long long blocks_per_unit; |
8427 | unsigned long long curr_migr_unit; | |
8428 | ||
2e062e82 AK |
8429 | if (load_imsm_migr_rec(super, info) != 0) { |
8430 | dprintf("imsm: ERROR: Cannot read migration record " | |
8431 | "for checkpoint save.\n"); | |
8432 | return 1; | |
8433 | } | |
8434 | ||
f8b72ef5 AK |
8435 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
8436 | if (blocks_per_unit == 0) { | |
0228d92c AK |
8437 | dprintf("imsm: no migration in progress.\n"); |
8438 | return 2; | |
687629c2 | 8439 | } |
f8b72ef5 AK |
8440 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
8441 | /* check if array is alligned to copy area | |
8442 | * if it is not alligned, add one to current migration unit value | |
8443 | * this can happend on array reshape finish only | |
8444 | */ | |
8445 | if (info->reshape_progress % blocks_per_unit) | |
8446 | curr_migr_unit++; | |
687629c2 AK |
8447 | |
8448 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 8449 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
8450 | super->migr_rec->rec_status = __cpu_to_le32(state); |
8451 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
8452 | __cpu_to_le32(curr_migr_unit * |
8453 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
8454 | if (write_imsm_migr_rec(st) < 0) { |
8455 | dprintf("imsm: Cannot write migration record " | |
8456 | "outside backup area\n"); | |
8457 | return 1; | |
8458 | } | |
8459 | ||
8460 | return 0; | |
8461 | } | |
8462 | ||
276d77db AK |
8463 | /******************************************************************************* |
8464 | * Function: recover_backup_imsm | |
8465 | * Description: Function recovers critical data from the Migration Copy Area | |
8466 | * while assembling an array. | |
8467 | * Parameters: | |
8468 | * super : imsm internal array info | |
8469 | * info : general array info | |
8470 | * Returns: | |
8471 | * 0 : success (or there is no data to recover) | |
8472 | * 1 : fail | |
8473 | ******************************************************************************/ | |
8474 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
8475 | { | |
8476 | struct intel_super *super = st->sb; | |
8477 | struct migr_record *migr_rec = super->migr_rec; | |
8478 | struct imsm_map *map_dest = NULL; | |
8479 | struct intel_dev *id = NULL; | |
8480 | unsigned long long read_offset; | |
8481 | unsigned long long write_offset; | |
8482 | unsigned unit_len; | |
8483 | int *targets = NULL; | |
8484 | int new_disks, i, err; | |
8485 | char *buf = NULL; | |
8486 | int retval = 1; | |
8487 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
8488 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 8489 | char buffer[20]; |
6c3560c0 | 8490 | int skipped_disks = 0; |
276d77db AK |
8491 | |
8492 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
8493 | if (err < 1) | |
8494 | return 1; | |
8495 | ||
8496 | /* recover data only during assemblation */ | |
8497 | if (strncmp(buffer, "inactive", 8) != 0) | |
8498 | return 0; | |
8499 | /* no data to recover */ | |
8500 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
8501 | return 0; | |
8502 | if (curr_migr_unit >= num_migr_units) | |
8503 | return 1; | |
8504 | ||
8505 | /* find device during reshape */ | |
8506 | for (id = super->devlist; id; id = id->next) | |
8507 | if (is_gen_migration(id->dev)) | |
8508 | break; | |
8509 | if (id == NULL) | |
8510 | return 1; | |
8511 | ||
238c0a71 | 8512 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
8513 | new_disks = map_dest->num_members; |
8514 | ||
8515 | read_offset = (unsigned long long) | |
8516 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
8517 | ||
8518 | write_offset = ((unsigned long long) | |
8519 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
75b69ea4 | 8520 | __le32_to_cpu(map_dest->pba_of_lba0)) * 512; |
276d77db AK |
8521 | |
8522 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
8523 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
8524 | goto abort; | |
8525 | targets = malloc(new_disks * sizeof(int)); | |
8526 | if (!targets) | |
8527 | goto abort; | |
8528 | ||
9a717282 | 8529 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
f627f5ad AK |
8530 | fprintf(stderr, |
8531 | Name ": Cannot open some devices belonging to array.\n"); | |
8532 | goto abort; | |
8533 | } | |
276d77db AK |
8534 | |
8535 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
8536 | if (targets[i] < 0) { |
8537 | skipped_disks++; | |
8538 | continue; | |
8539 | } | |
276d77db AK |
8540 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
8541 | fprintf(stderr, | |
8542 | Name ": Cannot seek to block: %s\n", | |
8543 | strerror(errno)); | |
137debce AK |
8544 | skipped_disks++; |
8545 | continue; | |
276d77db | 8546 | } |
9ec11d1a | 8547 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
8548 | fprintf(stderr, |
8549 | Name ": Cannot read copy area block: %s\n", | |
8550 | strerror(errno)); | |
137debce AK |
8551 | skipped_disks++; |
8552 | continue; | |
276d77db AK |
8553 | } |
8554 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
8555 | fprintf(stderr, | |
8556 | Name ": Cannot seek to block: %s\n", | |
8557 | strerror(errno)); | |
137debce AK |
8558 | skipped_disks++; |
8559 | continue; | |
276d77db | 8560 | } |
9ec11d1a | 8561 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
8562 | fprintf(stderr, |
8563 | Name ": Cannot restore block: %s\n", | |
8564 | strerror(errno)); | |
137debce AK |
8565 | skipped_disks++; |
8566 | continue; | |
276d77db AK |
8567 | } |
8568 | } | |
8569 | ||
137debce AK |
8570 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
8571 | new_disks, | |
8572 | super, | |
8573 | id->dev)) { | |
6c3560c0 AK |
8574 | fprintf(stderr, |
8575 | Name ": Cannot restore data from backup." | |
8576 | " Too many failed disks\n"); | |
8577 | goto abort; | |
8578 | } | |
8579 | ||
befb629b AK |
8580 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
8581 | /* ignore error == 2, this can mean end of reshape here | |
8582 | */ | |
8583 | dprintf("imsm: Cannot write checkpoint to " | |
8584 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
8585 | } else | |
276d77db | 8586 | retval = 0; |
276d77db AK |
8587 | |
8588 | abort: | |
8589 | if (targets) { | |
8590 | for (i = 0; i < new_disks; i++) | |
8591 | if (targets[i]) | |
8592 | close(targets[i]); | |
8593 | free(targets); | |
8594 | } | |
8595 | free(buf); | |
8596 | return retval; | |
8597 | } | |
8598 | ||
2cda7640 ML |
8599 | static char disk_by_path[] = "/dev/disk/by-path/"; |
8600 | ||
8601 | static const char *imsm_get_disk_controller_domain(const char *path) | |
8602 | { | |
2cda7640 | 8603 | char disk_path[PATH_MAX]; |
96234762 LM |
8604 | char *drv=NULL; |
8605 | struct stat st; | |
2cda7640 | 8606 | |
96234762 LM |
8607 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
8608 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
8609 | if (stat(disk_path, &st) == 0) { | |
8610 | struct sys_dev* hba; | |
8611 | char *path=NULL; | |
8612 | ||
8613 | path = devt_to_devpath(st.st_rdev); | |
8614 | if (path == NULL) | |
8615 | return "unknown"; | |
8616 | hba = find_disk_attached_hba(-1, path); | |
8617 | if (hba && hba->type == SYS_DEV_SAS) | |
8618 | drv = "isci"; | |
8619 | else if (hba && hba->type == SYS_DEV_SATA) | |
8620 | drv = "ahci"; | |
8621 | else | |
8622 | drv = "unknown"; | |
8623 | dprintf("path: %s hba: %s attached: %s\n", | |
8624 | path, (hba) ? hba->path : "NULL", drv); | |
8625 | free(path); | |
8626 | if (hba) | |
8627 | free_sys_dev(&hba); | |
2cda7640 | 8628 | } |
96234762 | 8629 | return drv; |
2cda7640 ML |
8630 | } |
8631 | ||
78b10e66 N |
8632 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
8633 | { | |
8634 | char subdev_name[20]; | |
8635 | struct mdstat_ent *mdstat; | |
8636 | ||
8637 | sprintf(subdev_name, "%d", subdev); | |
8638 | mdstat = mdstat_by_subdev(subdev_name, container); | |
8639 | if (!mdstat) | |
8640 | return -1; | |
8641 | ||
8642 | *minor = mdstat->devnum; | |
8643 | free_mdstat(mdstat); | |
8644 | return 0; | |
8645 | } | |
8646 | ||
8647 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
8648 | struct geo_params *geo, | |
8649 | int *old_raid_disks) | |
8650 | { | |
694575e7 KW |
8651 | /* currently we only support increasing the number of devices |
8652 | * for a container. This increases the number of device for each | |
8653 | * member array. They must all be RAID0 or RAID5. | |
8654 | */ | |
78b10e66 N |
8655 | int ret_val = 0; |
8656 | struct mdinfo *info, *member; | |
8657 | int devices_that_can_grow = 0; | |
8658 | ||
8659 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
8660 | "st->devnum = (%i)\n", | |
8661 | st->devnum); | |
8662 | ||
8663 | if (geo->size != -1 || | |
8664 | geo->level != UnSet || | |
8665 | geo->layout != UnSet || | |
8666 | geo->chunksize != 0 || | |
8667 | geo->raid_disks == UnSet) { | |
8668 | dprintf("imsm: Container operation is allowed for " | |
8669 | "raid disks number change only.\n"); | |
8670 | return ret_val; | |
8671 | } | |
8672 | ||
8673 | info = container_content_imsm(st, NULL); | |
8674 | for (member = info; member; member = member->next) { | |
8675 | int result; | |
8676 | int minor; | |
8677 | ||
8678 | dprintf("imsm: checking device_num: %i\n", | |
8679 | member->container_member); | |
8680 | ||
d7d205bd | 8681 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
8682 | /* we work on container for Online Capacity Expansion |
8683 | * only so raid_disks has to grow | |
8684 | */ | |
8685 | dprintf("imsm: for container operation raid disks " | |
8686 | "increase is required\n"); | |
8687 | break; | |
8688 | } | |
8689 | ||
8690 | if ((info->array.level != 0) && | |
8691 | (info->array.level != 5)) { | |
8692 | /* we cannot use this container with other raid level | |
8693 | */ | |
690aae1a | 8694 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
8695 | " raid level (%i) detected\n", |
8696 | info->array.level); | |
8697 | break; | |
8698 | } else { | |
8699 | /* check for platform support | |
8700 | * for this raid level configuration | |
8701 | */ | |
8702 | struct intel_super *super = st->sb; | |
8703 | if (!is_raid_level_supported(super->orom, | |
8704 | member->array.level, | |
8705 | geo->raid_disks)) { | |
690aae1a | 8706 | dprintf("platform does not support raid%d with" |
78b10e66 N |
8707 | " %d disk%s\n", |
8708 | info->array.level, | |
8709 | geo->raid_disks, | |
8710 | geo->raid_disks > 1 ? "s" : ""); | |
8711 | break; | |
8712 | } | |
2a4a08e7 AK |
8713 | /* check if component size is aligned to chunk size |
8714 | */ | |
8715 | if (info->component_size % | |
8716 | (info->array.chunk_size/512)) { | |
8717 | dprintf("Component size is not aligned to " | |
8718 | "chunk size\n"); | |
8719 | break; | |
8720 | } | |
78b10e66 N |
8721 | } |
8722 | ||
8723 | if (*old_raid_disks && | |
8724 | info->array.raid_disks != *old_raid_disks) | |
8725 | break; | |
8726 | *old_raid_disks = info->array.raid_disks; | |
8727 | ||
8728 | /* All raid5 and raid0 volumes in container | |
8729 | * have to be ready for Online Capacity Expansion | |
8730 | * so they need to be assembled. We have already | |
8731 | * checked that no recovery etc is happening. | |
8732 | */ | |
8733 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
8734 | st->container_dev, | |
8735 | &minor); | |
8736 | if (result < 0) { | |
8737 | dprintf("imsm: cannot find array\n"); | |
8738 | break; | |
8739 | } | |
8740 | devices_that_can_grow++; | |
8741 | } | |
8742 | sysfs_free(info); | |
8743 | if (!member && devices_that_can_grow) | |
8744 | ret_val = 1; | |
8745 | ||
8746 | if (ret_val) | |
8747 | dprintf("\tContainer operation allowed\n"); | |
8748 | else | |
8749 | dprintf("\tError: %i\n", ret_val); | |
8750 | ||
8751 | return ret_val; | |
8752 | } | |
8753 | ||
8754 | /* Function: get_spares_for_grow | |
8755 | * Description: Allocates memory and creates list of spare devices | |
8756 | * avaliable in container. Checks if spare drive size is acceptable. | |
8757 | * Parameters: Pointer to the supertype structure | |
8758 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
8759 | * NULL if fail | |
8760 | */ | |
8761 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
8762 | { | |
78b10e66 | 8763 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 8764 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
8765 | } |
8766 | ||
8767 | /****************************************************************************** | |
8768 | * function: imsm_create_metadata_update_for_reshape | |
8769 | * Function creates update for whole IMSM container. | |
8770 | * | |
8771 | ******************************************************************************/ | |
8772 | static int imsm_create_metadata_update_for_reshape( | |
8773 | struct supertype *st, | |
8774 | struct geo_params *geo, | |
8775 | int old_raid_disks, | |
8776 | struct imsm_update_reshape **updatep) | |
8777 | { | |
8778 | struct intel_super *super = st->sb; | |
8779 | struct imsm_super *mpb = super->anchor; | |
8780 | int update_memory_size = 0; | |
8781 | struct imsm_update_reshape *u = NULL; | |
8782 | struct mdinfo *spares = NULL; | |
8783 | int i; | |
8784 | int delta_disks = 0; | |
bbd24d86 | 8785 | struct mdinfo *dev; |
78b10e66 N |
8786 | |
8787 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
8788 | geo->raid_disks); | |
8789 | ||
8790 | delta_disks = geo->raid_disks - old_raid_disks; | |
8791 | ||
8792 | /* size of all update data without anchor */ | |
8793 | update_memory_size = sizeof(struct imsm_update_reshape); | |
8794 | ||
8795 | /* now add space for spare disks that we need to add. */ | |
8796 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
8797 | ||
8798 | u = calloc(1, update_memory_size); | |
8799 | if (u == NULL) { | |
8800 | dprintf("error: " | |
8801 | "cannot get memory for imsm_update_reshape update\n"); | |
8802 | return 0; | |
8803 | } | |
8804 | u->type = update_reshape_container_disks; | |
8805 | u->old_raid_disks = old_raid_disks; | |
8806 | u->new_raid_disks = geo->raid_disks; | |
8807 | ||
8808 | /* now get spare disks list | |
8809 | */ | |
8810 | spares = get_spares_for_grow(st); | |
8811 | ||
8812 | if (spares == NULL | |
8813 | || delta_disks > spares->array.spare_disks) { | |
e14e5960 KW |
8814 | fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices " |
8815 | "for %s.\n", geo->dev_name); | |
e4c72d1d | 8816 | i = -1; |
78b10e66 N |
8817 | goto abort; |
8818 | } | |
8819 | ||
8820 | /* we have got spares | |
8821 | * update disk list in imsm_disk list table in anchor | |
8822 | */ | |
8823 | dprintf("imsm: %i spares are available.\n\n", | |
8824 | spares->array.spare_disks); | |
8825 | ||
bbd24d86 | 8826 | dev = spares->devs; |
78b10e66 | 8827 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
8828 | struct dl *dl; |
8829 | ||
bbd24d86 AK |
8830 | if (dev == NULL) |
8831 | break; | |
78b10e66 N |
8832 | u->new_disks[i] = makedev(dev->disk.major, |
8833 | dev->disk.minor); | |
8834 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
8835 | dl->index = mpb->num_disks; |
8836 | mpb->num_disks++; | |
bbd24d86 | 8837 | dev = dev->next; |
78b10e66 | 8838 | } |
78b10e66 N |
8839 | |
8840 | abort: | |
8841 | /* free spares | |
8842 | */ | |
8843 | sysfs_free(spares); | |
8844 | ||
d677e0b8 | 8845 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 8846 | if (i == delta_disks) { |
d677e0b8 | 8847 | dprintf(" OK\n"); |
78b10e66 N |
8848 | *updatep = u; |
8849 | return update_memory_size; | |
8850 | } | |
8851 | free(u); | |
d677e0b8 | 8852 | dprintf(" Error\n"); |
78b10e66 N |
8853 | |
8854 | return 0; | |
8855 | } | |
8856 | ||
48c5303a PC |
8857 | /****************************************************************************** |
8858 | * function: imsm_create_metadata_update_for_migration() | |
8859 | * Creates update for IMSM array. | |
8860 | * | |
8861 | ******************************************************************************/ | |
8862 | static int imsm_create_metadata_update_for_migration( | |
8863 | struct supertype *st, | |
8864 | struct geo_params *geo, | |
8865 | struct imsm_update_reshape_migration **updatep) | |
8866 | { | |
8867 | struct intel_super *super = st->sb; | |
8868 | int update_memory_size = 0; | |
8869 | struct imsm_update_reshape_migration *u = NULL; | |
8870 | struct imsm_dev *dev; | |
8871 | int previous_level = -1; | |
8872 | ||
8873 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
8874 | " New Level = %i\n", geo->level); | |
8875 | ||
8876 | /* size of all update data without anchor */ | |
8877 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
8878 | ||
8879 | u = calloc(1, update_memory_size); | |
8880 | if (u == NULL) { | |
8881 | dprintf("error: cannot get memory for " | |
8882 | "imsm_create_metadata_update_for_migration\n"); | |
8883 | return 0; | |
8884 | } | |
8885 | u->type = update_reshape_migration; | |
8886 | u->subdev = super->current_vol; | |
8887 | u->new_level = geo->level; | |
8888 | u->new_layout = geo->layout; | |
8889 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
8890 | u->new_disks[0] = -1; | |
4bba0439 | 8891 | u->new_chunksize = -1; |
48c5303a PC |
8892 | |
8893 | dev = get_imsm_dev(super, u->subdev); | |
8894 | if (dev) { | |
8895 | struct imsm_map *map; | |
8896 | ||
238c0a71 | 8897 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
8898 | if (map) { |
8899 | int current_chunk_size = | |
8900 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
8901 | ||
8902 | if (geo->chunksize != current_chunk_size) { | |
8903 | u->new_chunksize = geo->chunksize / 1024; | |
8904 | dprintf("imsm: " | |
8905 | "chunk size change from %i to %i\n", | |
8906 | current_chunk_size, u->new_chunksize); | |
8907 | } | |
48c5303a | 8908 | previous_level = map->raid_level; |
4bba0439 | 8909 | } |
48c5303a PC |
8910 | } |
8911 | if ((geo->level == 5) && (previous_level == 0)) { | |
8912 | struct mdinfo *spares = NULL; | |
8913 | ||
8914 | u->new_raid_disks++; | |
8915 | spares = get_spares_for_grow(st); | |
8916 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
8917 | free(u); | |
8918 | sysfs_free(spares); | |
8919 | update_memory_size = 0; | |
8920 | dprintf("error: cannot get spare device " | |
8921 | "for requested migration"); | |
8922 | return 0; | |
8923 | } | |
8924 | sysfs_free(spares); | |
8925 | } | |
8926 | dprintf("imsm: reshape update preparation : OK\n"); | |
8927 | *updatep = u; | |
8928 | ||
8929 | return update_memory_size; | |
8930 | } | |
8931 | ||
8dd70bce AK |
8932 | static void imsm_update_metadata_locally(struct supertype *st, |
8933 | void *buf, int len) | |
8934 | { | |
8935 | struct metadata_update mu; | |
8936 | ||
8937 | mu.buf = buf; | |
8938 | mu.len = len; | |
8939 | mu.space = NULL; | |
8940 | mu.space_list = NULL; | |
8941 | mu.next = NULL; | |
8942 | imsm_prepare_update(st, &mu); | |
8943 | imsm_process_update(st, &mu); | |
8944 | ||
8945 | while (mu.space_list) { | |
8946 | void **space = mu.space_list; | |
8947 | mu.space_list = *space; | |
8948 | free(space); | |
8949 | } | |
8950 | } | |
78b10e66 | 8951 | |
471bceb6 | 8952 | /*************************************************************************** |
694575e7 | 8953 | * Function: imsm_analyze_change |
471bceb6 KW |
8954 | * Description: Function analyze change for single volume |
8955 | * and validate if transition is supported | |
694575e7 KW |
8956 | * Parameters: Geometry parameters, supertype structure |
8957 | * Returns: Operation type code on success, -1 if fail | |
471bceb6 KW |
8958 | ****************************************************************************/ |
8959 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
8960 | struct geo_params *geo) | |
694575e7 | 8961 | { |
471bceb6 KW |
8962 | struct mdinfo info; |
8963 | int change = -1; | |
8964 | int check_devs = 0; | |
c21e737b | 8965 | int chunk; |
e91a3bad LM |
8966 | int devNumChange=0; |
8967 | int layout = -1; | |
471bceb6 KW |
8968 | |
8969 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
8970 | if ((geo->level != info.array.level) && |
8971 | (geo->level >= 0) && | |
8972 | (geo->level != UnSet)) { | |
8973 | switch (info.array.level) { | |
8974 | case 0: | |
8975 | if (geo->level == 5) { | |
b5347799 | 8976 | change = CH_MIGRATION; |
e13ce846 AK |
8977 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
8978 | fprintf(stderr, | |
8979 | Name " Error. Requested Layout " | |
8980 | "not supported (left-asymmetric layout " | |
8981 | "is supported only)!\n"); | |
8982 | change = -1; | |
8983 | goto analyse_change_exit; | |
8984 | } | |
e91a3bad | 8985 | layout = geo->layout; |
471bceb6 | 8986 | check_devs = 1; |
e91a3bad LM |
8987 | devNumChange = 1; /* parity disk added */ |
8988 | } else if (geo->level == 10) { | |
471bceb6 KW |
8989 | change = CH_TAKEOVER; |
8990 | check_devs = 1; | |
e91a3bad LM |
8991 | devNumChange = 2; /* two mirrors added */ |
8992 | layout = 0x102; /* imsm supported layout */ | |
471bceb6 | 8993 | } |
dfe77a9e KW |
8994 | break; |
8995 | case 1: | |
471bceb6 KW |
8996 | case 10: |
8997 | if (geo->level == 0) { | |
8998 | change = CH_TAKEOVER; | |
8999 | check_devs = 1; | |
e91a3bad LM |
9000 | devNumChange = -(geo->raid_disks/2); |
9001 | layout = 0; /* imsm raid0 layout */ | |
471bceb6 KW |
9002 | } |
9003 | break; | |
9004 | } | |
9005 | if (change == -1) { | |
9006 | fprintf(stderr, | |
9007 | Name " Error. Level Migration from %d to %d " | |
9008 | "not supported!\n", | |
9009 | info.array.level, geo->level); | |
9010 | goto analyse_change_exit; | |
9011 | } | |
9012 | } else | |
9013 | geo->level = info.array.level; | |
9014 | ||
9015 | if ((geo->layout != info.array.layout) | |
9016 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 9017 | change = CH_MIGRATION; |
471bceb6 KW |
9018 | if ((info.array.layout == 0) |
9019 | && (info.array.level == 5) | |
9020 | && (geo->layout == 5)) { | |
9021 | /* reshape 5 -> 4 */ | |
9022 | } else if ((info.array.layout == 5) | |
9023 | && (info.array.level == 5) | |
9024 | && (geo->layout == 0)) { | |
9025 | /* reshape 4 -> 5 */ | |
9026 | geo->layout = 0; | |
9027 | geo->level = 5; | |
9028 | } else { | |
9029 | fprintf(stderr, | |
9030 | Name " Error. Layout Migration from %d to %d " | |
9031 | "not supported!\n", | |
9032 | info.array.layout, geo->layout); | |
9033 | change = -1; | |
9034 | goto analyse_change_exit; | |
9035 | } | |
9036 | } else | |
9037 | geo->layout = info.array.layout; | |
9038 | ||
9039 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
9040 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 9041 | change = CH_MIGRATION; |
471bceb6 KW |
9042 | else |
9043 | geo->chunksize = info.array.chunk_size; | |
9044 | ||
c21e737b | 9045 | chunk = geo->chunksize / 1024; |
471bceb6 KW |
9046 | if (!validate_geometry_imsm(st, |
9047 | geo->level, | |
e91a3bad LM |
9048 | layout, |
9049 | geo->raid_disks + devNumChange, | |
c21e737b | 9050 | &chunk, |
471bceb6 KW |
9051 | geo->size, |
9052 | 0, 0, 1)) | |
9053 | change = -1; | |
9054 | ||
9055 | if (check_devs) { | |
9056 | struct intel_super *super = st->sb; | |
9057 | struct imsm_super *mpb = super->anchor; | |
9058 | ||
9059 | if (mpb->num_raid_devs > 1) { | |
9060 | fprintf(stderr, | |
9061 | Name " Error. Cannot perform operation on %s" | |
9062 | "- for this operation it MUST be single " | |
9063 | "array in container\n", | |
9064 | geo->dev_name); | |
9065 | change = -1; | |
9066 | } | |
9067 | } | |
9068 | ||
9069 | analyse_change_exit: | |
9070 | ||
9071 | return change; | |
694575e7 KW |
9072 | } |
9073 | ||
bb025c2f KW |
9074 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
9075 | { | |
9076 | struct intel_super *super = st->sb; | |
9077 | struct imsm_update_takeover *u; | |
9078 | ||
9079 | u = malloc(sizeof(struct imsm_update_takeover)); | |
9080 | if (u == NULL) | |
9081 | return 1; | |
9082 | ||
9083 | u->type = update_takeover; | |
9084 | u->subarray = super->current_vol; | |
9085 | ||
9086 | /* 10->0 transition */ | |
9087 | if (geo->level == 0) | |
9088 | u->direction = R10_TO_R0; | |
9089 | ||
0529c688 KW |
9090 | /* 0->10 transition */ |
9091 | if (geo->level == 10) | |
9092 | u->direction = R0_TO_R10; | |
9093 | ||
bb025c2f KW |
9094 | /* update metadata locally */ |
9095 | imsm_update_metadata_locally(st, u, | |
9096 | sizeof(struct imsm_update_takeover)); | |
9097 | /* and possibly remotely */ | |
9098 | if (st->update_tail) | |
9099 | append_metadata_update(st, u, | |
9100 | sizeof(struct imsm_update_takeover)); | |
9101 | else | |
9102 | free(u); | |
9103 | ||
9104 | return 0; | |
9105 | } | |
9106 | ||
78b10e66 N |
9107 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
9108 | int layout, int chunksize, int raid_disks, | |
41784c88 AK |
9109 | int delta_disks, char *backup, char *dev, |
9110 | int verbose) | |
78b10e66 | 9111 | { |
78b10e66 N |
9112 | int ret_val = 1; |
9113 | struct geo_params geo; | |
9114 | ||
9115 | dprintf("imsm: reshape_super called.\n"); | |
9116 | ||
71204a50 | 9117 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
9118 | |
9119 | geo.dev_name = dev; | |
694575e7 | 9120 | geo.dev_id = st->devnum; |
78b10e66 N |
9121 | geo.size = size; |
9122 | geo.level = level; | |
9123 | geo.layout = layout; | |
9124 | geo.chunksize = chunksize; | |
9125 | geo.raid_disks = raid_disks; | |
41784c88 AK |
9126 | if (delta_disks != UnSet) |
9127 | geo.raid_disks += delta_disks; | |
78b10e66 N |
9128 | |
9129 | dprintf("\tfor level : %i\n", geo.level); | |
9130 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
9131 | ||
9132 | if (experimental() == 0) | |
9133 | return ret_val; | |
9134 | ||
78b10e66 | 9135 | if (st->container_dev == st->devnum) { |
694575e7 KW |
9136 | /* On container level we can only increase number of devices. */ |
9137 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 9138 | int old_raid_disks = 0; |
6dc0be30 | 9139 | |
78b10e66 N |
9140 | if (imsm_reshape_is_allowed_on_container( |
9141 | st, &geo, &old_raid_disks)) { | |
9142 | struct imsm_update_reshape *u = NULL; | |
9143 | int len; | |
9144 | ||
9145 | len = imsm_create_metadata_update_for_reshape( | |
9146 | st, &geo, old_raid_disks, &u); | |
9147 | ||
ed08d51c AK |
9148 | if (len <= 0) { |
9149 | dprintf("imsm: Cannot prepare update\n"); | |
9150 | goto exit_imsm_reshape_super; | |
9151 | } | |
9152 | ||
8dd70bce AK |
9153 | ret_val = 0; |
9154 | /* update metadata locally */ | |
9155 | imsm_update_metadata_locally(st, u, len); | |
9156 | /* and possibly remotely */ | |
9157 | if (st->update_tail) | |
9158 | append_metadata_update(st, u, len); | |
9159 | else | |
ed08d51c | 9160 | free(u); |
8dd70bce | 9161 | |
694575e7 | 9162 | } else { |
e7ff7e40 AK |
9163 | fprintf(stderr, Name ": (imsm) Operation " |
9164 | "is not allowed on this container\n"); | |
694575e7 KW |
9165 | } |
9166 | } else { | |
9167 | /* On volume level we support following operations | |
471bceb6 KW |
9168 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
9169 | * - chunk size migration | |
9170 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
9171 | */ | |
9172 | struct intel_super *super = st->sb; | |
9173 | struct intel_dev *dev = super->devlist; | |
9174 | int change, devnum; | |
694575e7 | 9175 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
9176 | /* find requested device */ |
9177 | while (dev) { | |
19986c72 MB |
9178 | if (imsm_find_array_minor_by_subdev( |
9179 | dev->index, st->container_dev, &devnum) == 0 | |
9180 | && devnum == geo.dev_id) | |
471bceb6 KW |
9181 | break; |
9182 | dev = dev->next; | |
9183 | } | |
9184 | if (dev == NULL) { | |
9185 | fprintf(stderr, Name " Cannot find %s (%i) subarray\n", | |
9186 | geo.dev_name, geo.dev_id); | |
9187 | goto exit_imsm_reshape_super; | |
9188 | } | |
9189 | super->current_vol = dev->index; | |
694575e7 KW |
9190 | change = imsm_analyze_change(st, &geo); |
9191 | switch (change) { | |
471bceb6 | 9192 | case CH_TAKEOVER: |
bb025c2f | 9193 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 9194 | break; |
48c5303a PC |
9195 | case CH_MIGRATION: { |
9196 | struct imsm_update_reshape_migration *u = NULL; | |
9197 | int len = | |
9198 | imsm_create_metadata_update_for_migration( | |
9199 | st, &geo, &u); | |
9200 | if (len < 1) { | |
9201 | dprintf("imsm: " | |
9202 | "Cannot prepare update\n"); | |
9203 | break; | |
9204 | } | |
471bceb6 | 9205 | ret_val = 0; |
48c5303a PC |
9206 | /* update metadata locally */ |
9207 | imsm_update_metadata_locally(st, u, len); | |
9208 | /* and possibly remotely */ | |
9209 | if (st->update_tail) | |
9210 | append_metadata_update(st, u, len); | |
9211 | else | |
9212 | free(u); | |
9213 | } | |
9214 | break; | |
471bceb6 KW |
9215 | default: |
9216 | ret_val = 1; | |
694575e7 | 9217 | } |
694575e7 | 9218 | } |
78b10e66 | 9219 | |
ed08d51c | 9220 | exit_imsm_reshape_super: |
78b10e66 N |
9221 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
9222 | return ret_val; | |
9223 | } | |
2cda7640 | 9224 | |
eee67a47 AK |
9225 | /******************************************************************************* |
9226 | * Function: wait_for_reshape_imsm | |
9227 | * Description: Function writes new sync_max value and waits until | |
9228 | * reshape process reach new position | |
9229 | * Parameters: | |
9230 | * sra : general array info | |
eee67a47 AK |
9231 | * ndata : number of disks in new array's layout |
9232 | * Returns: | |
9233 | * 0 : success, | |
9234 | * 1 : there is no reshape in progress, | |
9235 | * -1 : fail | |
9236 | ******************************************************************************/ | |
ae9f01f8 | 9237 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
9238 | { |
9239 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
9240 | unsigned long long completed; | |
ae9f01f8 AK |
9241 | /* to_complete : new sync_max position */ |
9242 | unsigned long long to_complete = sra->reshape_progress; | |
9243 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 9244 | |
ae9f01f8 AK |
9245 | if (fd < 0) { |
9246 | dprintf("imsm: wait_for_reshape_imsm() " | |
9247 | "cannot open reshape_position\n"); | |
eee67a47 | 9248 | return 1; |
ae9f01f8 | 9249 | } |
eee67a47 | 9250 | |
ae9f01f8 AK |
9251 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
9252 | dprintf("imsm: wait_for_reshape_imsm() " | |
9253 | "cannot read reshape_position (no reshape in progres)\n"); | |
9254 | close(fd); | |
9255 | return 0; | |
9256 | } | |
eee67a47 | 9257 | |
ae9f01f8 AK |
9258 | if (completed > to_complete) { |
9259 | dprintf("imsm: wait_for_reshape_imsm() " | |
9260 | "wrong next position to set %llu (%llu)\n", | |
9261 | to_complete, completed); | |
9262 | close(fd); | |
9263 | return -1; | |
9264 | } | |
9265 | dprintf("Position set: %llu\n", position_to_set); | |
9266 | if (sysfs_set_num(sra, NULL, "sync_max", | |
9267 | position_to_set) != 0) { | |
9268 | dprintf("imsm: wait_for_reshape_imsm() " | |
9269 | "cannot set reshape position to %llu\n", | |
9270 | position_to_set); | |
9271 | close(fd); | |
9272 | return -1; | |
eee67a47 AK |
9273 | } |
9274 | ||
eee67a47 AK |
9275 | do { |
9276 | char action[20]; | |
9277 | fd_set rfds; | |
9278 | FD_ZERO(&rfds); | |
9279 | FD_SET(fd, &rfds); | |
a47e44fb AK |
9280 | select(fd+1, &rfds, NULL, NULL, NULL); |
9281 | if (sysfs_get_str(sra, NULL, "sync_action", | |
9282 | action, 20) > 0 && | |
9283 | strncmp(action, "reshape", 7) != 0) | |
9284 | break; | |
eee67a47 | 9285 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
9286 | dprintf("imsm: wait_for_reshape_imsm() " |
9287 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
9288 | close(fd); |
9289 | return 1; | |
9290 | } | |
eee67a47 AK |
9291 | } while (completed < to_complete); |
9292 | close(fd); | |
9293 | return 0; | |
9294 | ||
9295 | } | |
9296 | ||
b915c95f AK |
9297 | /******************************************************************************* |
9298 | * Function: check_degradation_change | |
9299 | * Description: Check that array hasn't become failed. | |
9300 | * Parameters: | |
9301 | * info : for sysfs access | |
9302 | * sources : source disks descriptors | |
9303 | * degraded: previous degradation level | |
9304 | * Returns: | |
9305 | * degradation level | |
9306 | ******************************************************************************/ | |
9307 | int check_degradation_change(struct mdinfo *info, | |
9308 | int *sources, | |
9309 | int degraded) | |
9310 | { | |
9311 | unsigned long long new_degraded; | |
9312 | sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
9313 | if (new_degraded != (unsigned long long)degraded) { | |
9314 | /* check each device to ensure it is still working */ | |
9315 | struct mdinfo *sd; | |
9316 | new_degraded = 0; | |
9317 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9318 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
9319 | continue; | |
9320 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
9321 | char sbuf[20]; | |
9322 | if (sysfs_get_str(info, | |
9323 | sd, "state", sbuf, 20) < 0 || | |
9324 | strstr(sbuf, "faulty") || | |
9325 | strstr(sbuf, "in_sync") == NULL) { | |
9326 | /* this device is dead */ | |
9327 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
9328 | if (sd->disk.raid_disk >= 0 && | |
9329 | sources[sd->disk.raid_disk] >= 0) { | |
9330 | close(sources[ | |
9331 | sd->disk.raid_disk]); | |
9332 | sources[sd->disk.raid_disk] = | |
9333 | -1; | |
9334 | } | |
9335 | new_degraded++; | |
9336 | } | |
9337 | } | |
9338 | } | |
9339 | } | |
9340 | ||
9341 | return new_degraded; | |
9342 | } | |
9343 | ||
10f22854 AK |
9344 | /******************************************************************************* |
9345 | * Function: imsm_manage_reshape | |
9346 | * Description: Function finds array under reshape and it manages reshape | |
9347 | * process. It creates stripes backups (if required) and sets | |
9348 | * checheckpoits. | |
9349 | * Parameters: | |
9350 | * afd : Backup handle (nattive) - not used | |
9351 | * sra : general array info | |
9352 | * reshape : reshape parameters - not used | |
9353 | * st : supertype structure | |
9354 | * blocks : size of critical section [blocks] | |
9355 | * fds : table of source device descriptor | |
9356 | * offsets : start of array (offest per devices) | |
9357 | * dests : not used | |
9358 | * destfd : table of destination device descriptor | |
9359 | * destoffsets : table of destination offsets (per device) | |
9360 | * Returns: | |
9361 | * 1 : success, reshape is done | |
9362 | * 0 : fail | |
9363 | ******************************************************************************/ | |
999b4972 N |
9364 | static int imsm_manage_reshape( |
9365 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 9366 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
9367 | int *fds, unsigned long long *offsets, |
9368 | int dests, int *destfd, unsigned long long *destoffsets) | |
9369 | { | |
10f22854 AK |
9370 | int ret_val = 0; |
9371 | struct intel_super *super = st->sb; | |
9372 | struct intel_dev *dv = NULL; | |
9373 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 9374 | struct imsm_map *map_src; |
10f22854 AK |
9375 | int migr_vol_qan = 0; |
9376 | int ndata, odata; /* [bytes] */ | |
9377 | int chunk; /* [bytes] */ | |
9378 | struct migr_record *migr_rec; | |
9379 | char *buf = NULL; | |
9380 | unsigned int buf_size; /* [bytes] */ | |
9381 | unsigned long long max_position; /* array size [bytes] */ | |
9382 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
9383 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
9384 | unsigned long long start_src; /* [bytes] */ |
9385 | unsigned long long start; /* [bytes] */ | |
9386 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 9387 | int degraded = 0; |
ab724b98 | 9388 | int source_layout = 0; |
10f22854 | 9389 | |
1ab242d8 | 9390 | if (!fds || !offsets || !sra) |
10f22854 AK |
9391 | goto abort; |
9392 | ||
9393 | /* Find volume during the reshape */ | |
9394 | for (dv = super->devlist; dv; dv = dv->next) { | |
9395 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
9396 | && dv->dev->vol.migr_state == 1) { | |
9397 | dev = dv->dev; | |
9398 | migr_vol_qan++; | |
9399 | } | |
9400 | } | |
9401 | /* Only one volume can migrate at the same time */ | |
9402 | if (migr_vol_qan != 1) { | |
9403 | fprintf(stderr, Name " : %s", migr_vol_qan ? | |
9404 | "Number of migrating volumes greater than 1\n" : | |
9405 | "There is no volume during migrationg\n"); | |
9406 | goto abort; | |
9407 | } | |
9408 | ||
238c0a71 | 9409 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
9410 | if (map_src == NULL) |
9411 | goto abort; | |
10f22854 | 9412 | |
238c0a71 AK |
9413 | ndata = imsm_num_data_members(dev, MAP_0); |
9414 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 9415 | |
7b1ab482 | 9416 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
9417 | old_data_stripe_length = odata * chunk; |
9418 | ||
9419 | migr_rec = super->migr_rec; | |
9420 | ||
10f22854 AK |
9421 | /* initialize migration record for start condition */ |
9422 | if (sra->reshape_progress == 0) | |
9423 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
9424 | else { |
9425 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
9426 | dprintf("imsm: cannot restart migration when data " | |
9427 | "are present in copy area.\n"); | |
9428 | goto abort; | |
9429 | } | |
9430 | } | |
10f22854 AK |
9431 | |
9432 | /* size for data */ | |
9433 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
9434 | /* extend buffer size for parity disk */ | |
9435 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9436 | /* add space for stripe aligment */ | |
9437 | buf_size += old_data_stripe_length; | |
9438 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
9439 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
9440 | goto abort; | |
9441 | } | |
9442 | ||
3ef4403c | 9443 | max_position = sra->component_size * ndata; |
68eb8bc6 | 9444 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
9445 | |
9446 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
9447 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
9448 | /* current reshape position [blocks] */ | |
9449 | unsigned long long current_position = | |
9450 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
9451 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
9452 | unsigned long long border; | |
9453 | ||
b915c95f AK |
9454 | /* Check that array hasn't become failed. |
9455 | */ | |
9456 | degraded = check_degradation_change(sra, fds, degraded); | |
9457 | if (degraded > 1) { | |
9458 | dprintf("imsm: Abort reshape due to degradation" | |
9459 | " level (%i)\n", degraded); | |
9460 | goto abort; | |
9461 | } | |
9462 | ||
10f22854 AK |
9463 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
9464 | ||
9465 | if ((current_position + next_step) > max_position) | |
9466 | next_step = max_position - current_position; | |
9467 | ||
92144abf | 9468 | start = current_position * 512; |
10f22854 AK |
9469 | |
9470 | /* allign reading start to old geometry */ | |
9471 | start_buf_shift = start % old_data_stripe_length; | |
9472 | start_src = start - start_buf_shift; | |
9473 | ||
9474 | border = (start_src / odata) - (start / ndata); | |
9475 | border /= 512; | |
9476 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
9477 | /* save critical stripes to buf | |
9478 | * start - start address of current unit | |
9479 | * to backup [bytes] | |
9480 | * start_src - start address of current unit | |
9481 | * to backup alligned to source array | |
9482 | * [bytes] | |
9483 | */ | |
9484 | unsigned long long next_step_filler = 0; | |
9485 | unsigned long long copy_length = next_step * 512; | |
9486 | ||
9487 | /* allign copy area length to stripe in old geometry */ | |
9488 | next_step_filler = ((copy_length + start_buf_shift) | |
9489 | % old_data_stripe_length); | |
9490 | if (next_step_filler) | |
9491 | next_step_filler = (old_data_stripe_length | |
9492 | - next_step_filler); | |
9493 | dprintf("save_stripes() parameters: start = %llu," | |
9494 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
9495 | "\tstart_in_buf_shift = %llu," | |
9496 | "\tnext_step_filler = %llu\n", | |
9497 | start, start_src, copy_length, | |
9498 | start_buf_shift, next_step_filler); | |
9499 | ||
9500 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
9501 | chunk, map_src->raid_level, |
9502 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
9503 | copy_length + |
9504 | next_step_filler + start_buf_shift, | |
9505 | buf)) { | |
9506 | dprintf("imsm: Cannot save stripes" | |
9507 | " to buffer\n"); | |
9508 | goto abort; | |
9509 | } | |
9510 | /* Convert data to destination format and store it | |
9511 | * in backup general migration area | |
9512 | */ | |
9513 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 9514 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
9515 | dprintf("imsm: Cannot save stripes to " |
9516 | "target devices\n"); | |
9517 | goto abort; | |
9518 | } | |
9519 | if (save_checkpoint_imsm(st, sra, | |
9520 | UNIT_SRC_IN_CP_AREA)) { | |
9521 | dprintf("imsm: Cannot write checkpoint to " | |
9522 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
9523 | goto abort; | |
9524 | } | |
8016a6d4 AK |
9525 | } else { |
9526 | /* set next step to use whole border area */ | |
9527 | border /= next_step; | |
9528 | if (border > 1) | |
9529 | next_step *= border; | |
10f22854 AK |
9530 | } |
9531 | /* When data backed up, checkpoint stored, | |
9532 | * kick the kernel to reshape unit of data | |
9533 | */ | |
9534 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
9535 | /* limit next step to array max position */ |
9536 | if (next_step > max_position) | |
9537 | next_step = max_position; | |
10f22854 AK |
9538 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
9539 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 9540 | sra->reshape_progress = next_step; |
10f22854 AK |
9541 | |
9542 | /* wait until reshape finish */ | |
ae9f01f8 | 9543 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
9544 | dprintf("wait_for_reshape_imsm returned error!\n"); |
9545 | goto abort; | |
9546 | } | |
10f22854 | 9547 | |
0228d92c AK |
9548 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
9549 | /* ignore error == 2, this can mean end of reshape here | |
9550 | */ | |
10f22854 AK |
9551 | dprintf("imsm: Cannot write checkpoint to " |
9552 | "migration record (UNIT_SRC_NORMAL)\n"); | |
9553 | goto abort; | |
9554 | } | |
9555 | ||
9556 | } | |
9557 | ||
9558 | /* return '1' if done */ | |
9559 | ret_val = 1; | |
9560 | abort: | |
9561 | free(buf); | |
9562 | abort_reshape(sra); | |
9563 | ||
9564 | return ret_val; | |
999b4972 | 9565 | } |
71204a50 | 9566 | #endif /* MDASSEMBLE */ |
999b4972 | 9567 | |
cdddbdbc DW |
9568 | struct superswitch super_imsm = { |
9569 | #ifndef MDASSEMBLE | |
9570 | .examine_super = examine_super_imsm, | |
9571 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 9572 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 9573 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
9574 | .detail_super = detail_super_imsm, |
9575 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 9576 | .write_init_super = write_init_super_imsm, |
0e600426 N |
9577 | .validate_geometry = validate_geometry_imsm, |
9578 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 9579 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 9580 | .detail_platform = detail_platform_imsm, |
33414a01 | 9581 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 9582 | .update_subarray = update_subarray_imsm, |
2b959fbf | 9583 | .load_container = load_container_imsm, |
71204a50 N |
9584 | .default_geometry = default_geometry_imsm, |
9585 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
9586 | .reshape_super = imsm_reshape_super, | |
9587 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 9588 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
9589 | #endif |
9590 | .match_home = match_home_imsm, | |
9591 | .uuid_from_super= uuid_from_super_imsm, | |
9592 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 9593 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
9594 | .update_super = update_super_imsm, |
9595 | ||
9596 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 9597 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
9598 | |
9599 | .compare_super = compare_super_imsm, | |
9600 | ||
9601 | .load_super = load_super_imsm, | |
bf5a934a | 9602 | .init_super = init_super_imsm, |
e683ca88 | 9603 | .store_super = store_super_imsm, |
cdddbdbc DW |
9604 | .free_super = free_super_imsm, |
9605 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 9606 | .container_content = container_content_imsm, |
cdddbdbc | 9607 | |
276d77db | 9608 | |
cdddbdbc | 9609 | .external = 1, |
4cce4069 | 9610 | .name = "imsm", |
845dea95 | 9611 | |
0e600426 | 9612 | #ifndef MDASSEMBLE |
845dea95 NB |
9613 | /* for mdmon */ |
9614 | .open_new = imsm_open_new, | |
ed9d66aa | 9615 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
9616 | .set_disk = imsm_set_disk, |
9617 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 9618 | .activate_spare = imsm_activate_spare, |
e8319a19 | 9619 | .process_update = imsm_process_update, |
8273f55e | 9620 | .prepare_update = imsm_prepare_update, |
0e600426 | 9621 | #endif /* MDASSEMBLE */ |
cdddbdbc | 9622 | }; |